Enhancing Disease Surveillance in Nigeria through Machine Learning: Opportunities, Challenges and Strategic Recommendations

Enhancing Disease Surveillance in Nigeria through Machine Learning: Opportunities, Challenges and Strategic Recommendations | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Systematic Review Enhancing Disease Surveillance in Nigeria through Machine Learning: Opportunities, Challenges and Strategic Recommendations Lukman Jibril Aliyu, Abbas Bashir Umar, Saifuddeen kamfut Sani, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8000523/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Disease surveillance is fundamental to public health, enabling timely outbreak detection, efficient resource allocation, and evidence-based policymaking. In Nigeria, the Integrated Disease Surveillance and Response (IDSR) framework, while structured, is hampered by inconsistent data quality, limited private sector participation, and infrastructural constraints. Machine Learning (ML), a powerful subset of Artificial Intelligence (AI), offers transformative potential through its capacity for predictive analytics, real-time data processing, and automated pattern recognition to enhance surveillance capabilities. Despite global advancements in ML for disease forecasting and syndromic surveillance, its adoption within Nigeria's IDSR system lags considerably. This paper addresses this critical gap by investigating how ML can overcome Nigeria-specific barriers, such as fragmented data systems and rural connectivity deficits. We provide enhanced technical depth on ML methodologies, comparing supervised and unsupervised learning, and detailing relevant architectures, including Decision Trees, Recurrent Neural Networks (RNNs), and Convolutional Neural Networks (CNNs), suited for time-series epidemiological forecasting. We present a methodological illustration of malaria surveillance in Northern Nigeria using synthetic data, benchmarking five ML models (Linear Regression, Decision Tree, Random Forest, Gradient Boosting, and Support Vector Machine) under temporal validation, and employing SHapley Additive exPlanations (SHAP) for robust model interpretability. A sensitivity analysis further examines the stability of model performance under coefficient perturbations. A benchmarking analysis compares Nigeria's ML adoption against Rwanda and Kenya. Finally, we propose a tiered strategic framework encompassing policy, infrastructure, and capacity-building recommendations, complemented by a cost-benefit perspective emphasizing potential Disability-Adjusted Life Years (DALYs) averted and significant economic returns, aiming to foster a more resilient and equitable public health system in Nigeria. Machine Learning Disease Surveillance Artificial Intelligence Public Health Outbreak Prediction Digital Health Explainable AI Nigeria Figures Figure 1 Figure 2 Figure 3 1.0 Introduction Disease surveillance, the systematic collection, analysis, and interpretation of health data, serves as a cornerstone of public health by enabling early outbreak detection, resource allocation, and policy formulation [1]. In Nigeria, traditional surveillance systems primarily rely on passive reporting mechanisms, often leading to delays in outbreak detection and response [2]. The Integrated Disease Surveillance and Response (IDSR) framework, adopted in 1998, marked a significant step toward a more structured surveillance approach [3]. However, its effectiveness is hindered by challenges such as inconsistent data quality, limited private sector participation, and infrastructural constraints, including inadequate digital health infrastructure and fragmented reporting channels. Artificial Intelligence (AI) is increasingly recognized as a transformative tool for addressing these challenges. Machine Learning (ML), a subset of AI, enables predictive analytics, real-time data processing, and automated pattern recognition, making it particularly useful for enhancing disease surveillance [4]. The application of ML in public health has been demonstrated in various contexts, including the prediction of epidemic outbreaks, syndromic surveillance, and antimicrobial resistance monitoring [5]. ML models have been successfully deployed in resource-limited settings to predict malaria outbreaks and detect emerging infectious diseases, highlighting their potential for Nigeria’s healthcare system. The COVID-19 pandemic underscored the critical role of ML in tracking disease spread, optimizing resource allocation, and improving diagnostic accuracy [6]. Digital/AI-assisted contact tracing in South Korea and Taiwan significantly reduced delays in case identification and quarantine, with Taiwan achieving a median response time of ~3 days from exposure to isolation, demonstrating substantial improvements over traditional manual tracing approaches [7,8]. In India, AI-driven syndromic surveillance flagged COVID-19 hotspots earlier before conventional reporting systems triggered alerts [9]. Within Africa, Rwanda’s AI-enhanced surveillance infrastructure enabled real-time tracking of COVID-19 case clusters, facilitating swift quarantine measures [10]. Despite these global advances, Nigeria’s healthcare system faces multiple barriers to fully adopting these innovations, including outdated health information systems, workforce capacity gaps, limited internet connectivity in rural areas, and uneven technological adoption across regions [2,11]. In contrast to prior studies that primarily focus on ML in high-income contexts or generalized African settings [6], this paper review addresses Nigeria’s unique structural challenges, including a pronounced urban-rural digital divide affecting approximately 60% of the population [11] and fragmented health data ecosystems, and proposes locally adaptable ML solutions to enhance IDSR efficacy. Data Sources and Review Methodology This study adopted a narrative review approach to explore the application of artificial intelligence (AI) and machine learning (ML) in disease surveillance systems, with a particular focus on Nigeria and other low- and middle-income countries (LMICs). A comprehensive literature search was conducted across multiple electronic databases, including PubMed, Scopus, Web of Science, and Google Scholar. Search terms combined keywords and Boolean operators such as “artificial intelligence,” “machine learning,” “disease surveillance,” “public health surveillance,” “Nigeria,” and “LMICs.” The search was limited to studies published between 2015 and 2025 to capture recent advancements in AI-driven surveillance technologies. A total of 87 sources were included in the final review. These comprised both peer-reviewed journal articles (approximately 69) and grey literature (approximately 18), including conference proceedings, technical reports, and policy documents. The inclusion of grey literature was intentional to capture emerging innovations, policy perspectives, and real-world implementations that may not yet be fully represented in peer-reviewed publications. Studies were included if they: Examined the application of AI or ML in disease surveillance, diagnosis, or outbreak prediction Focused on public health systems in Nigeria or comparable LMIC settings Provided empirical findings, technical insights, or policy-relevant discussions Exclusion criteria included: Studies not related to public health or disease surveillance Articles lacking sufficient methodological detail Non-English publications Data extraction focused on key variables such as study design, geographic setting, type of AI/ML approach used, data sources, and reported outcomes. A thematic synthesis approach was employed to identify recurring patterns, opportunities, and challenges related to AI integration in surveillance systems. 2.0 Current State of Machine Learning in Nigeria Nigeria's disease surveillance system is vital for early outbreak detection and management, transitioning from disjointed, paper-based reporting to a unified approach within the IDSR framework. Major epidemics, like the 1986/87 yellow fever outbreak, prompted the creation of formal notification systems, leading to the adoption of IDSR in 1998 to coordinate efforts and improve reporting methods [ 3 ]. Surveillance operates through a multi-tiered system where data flows from primary health facilities to Local Government Areas (LGAs), then to State Ministries of Health and the Nigeria Centre for Disease Control (NCDC) [ 12 ]. Disease Surveillance and Notification Officers (DSNOs) play a key role, often working under resource constraints, including insufficient funding, a shortage of trained personnel, and logistical challenges in remote areas. While public facilities follow reporting protocols, private sector participation remains limited, affecting data completeness and real-time disease tracking [ 13 ]. Historically, manual reporting led to delays and inconsistencies, but electronic systems, such as the electronic Integrated Disease Surveillance and Response (eIDSR) pilot, have improved reporting timeliness from 43% to 73%, with a near doubling of reporting facilities [ 14 ]. However, challenges persist, including incomplete documentation, lack of interoperability between health data systems, and disparities in adoption between urban and rural areas. Some regions have successfully embraced digital tools, while others continue to rely on outdated methods, leading to gaps in national-level disease monitoring and response [ 2 ]. ML applications in Nigeria’s surveillance system remain in the early stages. While pilot projects and research initiatives have demonstrated the feasibility of AI-powered disease forecasting and automated diagnostics, full-scale implementation is hindered by infrastructure gaps, unreliable internet connectivity, and low digital literacy among healthcare workers [ 15 ]. Moreover, issues such as ethical concerns, data privacy regulations, and limited investment in AI-driven health technologies further constrain the adoption of ML in disease surveillance [ 16 ]. Despite these challenges, there have been positive developments. Initiatives like the deployment of AI-enhanced diagnostic tools for tuberculosis detection and predictive models for malaria outbreaks have showcased the potential of ML in improving Nigeria’s public health surveillance system [ 17 ]. For instance, a 2022 pilot in Lagos using ML to predict malaria outbreaks achieved a 78% accuracy rate but was limited to urban centers with reliable internet, covering only 12% of the state’s health facilities [ 18 ]. Similarly, the Auto-Visual Acute Flaccid Paralysis Detection and Reporting (AVADAR) system increased polio case reporting by 45% across 10 states from 2017–2020, yet its reliance on mobile networks excluded 30% of rural areas [ 18 ]. These examples highlight scalability constraints tied to infrastructure and funding. Furthermore, collaborations between research institutions, technology firms, and government agencies are gradually paving the way for more structured AI integration in disease monitoring. Going forward, addressing the current limitations will require stronger policy frameworks, increased investment in health-tech infrastructure, and training programs to enhance AI literacy among healthcare professionals. Integrating ML with Nigeria’s existing digital health strategies can significantly improve disease surveillance, providing timely and accurate health intelligence to combat future outbreaks effectively. 3.0 Machine Learning Methodologies for Disease Surveillance 3.1 Supervised vs. Unsupervised Learning ML in disease surveillance can be divided into two major paradigms: Supervised learning involves training a model on labeled datasets (e.g., previous outbreaks). It is useful for case classification, syndromic pattern recognition, and predictive modeling for diseases like cholera and Lassa fever [ 3 ]. Unsupervised learning is applied where outcomes are unknown, allowing for clustering of cases or anomaly detection, especially useful for novel syndromes or underreported diseases. Both methods offer complementary benefits to Nigeria’s Integrated Disease Surveillance and Response (IDSR) strategy, enhancing early detection and prediction of infectious disease patterns [ 19 ]. 3.2 Relevant ML Architectures Several ML models have proven valuable for health surveillance: 3.2.1 Decision Trees & Random Forests Decision Trees (DTs) and Random Forests (RFs) are foundational classifiers in healthcare analytics due to their interpretability and robustness with structured data. They’re effective in predicting disease risk based on epidemiological variables and have been frequently used for fall detection, infection diagnosis, and chronic disease monitoring. For instance, Random Forests have been used to classify patients based on wearable sensor data, offering real-time health risk detection [ 20 ]. 3.2.2 Support Vector Machines (SVMs) SVMs are particularly suited to high-dimensional clinical and demographic data. Their margin-maximizing properties make them robust classifiers for binary and multi-class disease prediction tasks, such as early detection of infections from structured hospital records or real-time social media streams [ 21 ]. 3.2.3 Recurrent Neural Networks (RNNs) RNNs excel in modeling temporal health data such as daily syndromic surveillance or weekly IDSR reports. Their capacity for retaining sequential dependencies is useful in forecasting outbreaks and patient deterioration patterns over time [ 22 ]. 3.2.4 Convolutional Neural Networks (CNNs) Though CNNs originated in image recognition, they’ve been adapted for spatial epidemiology. By treating spatial disease data as “images,” CNNs can learn local transmission patterns and hotspot dynamics. CNNs have also been used to extract features from wearable sensor data and medical records in predictive health modeling [ 23 ]. These models form the backbone of contemporary health surveillance systems. Each architecture addresses unique data characteristics, from structured demographic inputs to temporal sequences and spatial epidemiological grids. When integrated, they provide comprehensive analytical capabilities. Moreover, advancements in Neural Networks (NNs) and Generative Adversarial Networks (GANs) further extend this potential, enabling simulation of disease spread scenarios, synthetic data generation for rare outbreaks, and the modeling of uncertainty in predictions [ 21 ]. Together, these models offer a robust analytical framework for extracting actionable insights from increasingly complex and high-dimensional epidemiological data landscapes. 3.3 Model Selection, Training-Validation Split, and Cross-Validation Model selection is a critical step in machine learning pipelines, especially in healthcare and epidemiological modeling. It often begins with simple, interpretable models such as logistic regression and progresses to more complex techniques like ensemble methods (e.g., Random Forests) or deep learning models. These methods are selected based on their predictive performance, interpretability, and appropriateness for the data characteristics. Exploratory approaches like learning curve-based cross-validation (LCCV) have been shown to streamline model selection by eliminating poorly performing models early, thus improving efficiency without sacrificing accuracy [ 24 ]. Training-validation splitting , commonly using an 80:20 ratio, helps evaluate a model’s ability to generalize to unseen data. A balanced split is vital: too little training data hampers learning, while too little validation data reduces evaluation reliability. Studies have found that optimal split ratios vary by dataset size, but the general principle holds, careful partitioning improves performance estimation and model robustness [ 25 ]. Cross-validation (CV) methods like k-fold CV, time-series CV, and bootstrap-based CV are crucial when data is scarce or disease events are rare. K-fold CV (often 5- or 10-fold) provides a more robust performance estimate by rotating training and validation sets, reducing the risk of overfitting to a particular split [ 26 ]. Time-series CV is especially important in epidemiology where data has temporal dependencies. Additionally, hybrid methods combining bootstrap and CV (e.g., kCV-B) enhance generalizability in deep learning applications by improving training data diversity and reducing overfitting [ 27 ] Such rigorous model evaluation strategies ensure robust, unbiased, and generalizable machine learning models, which is essential for applications in Nigeria’s complex and evolving epidemiological landscape. 3.4 Global AI-driven Disease Surveillance Systems: A Comparative Overview In addition to the supervised and unsupervised ML methods discussed, it is crucial to contextualize Nigeria’s disease surveillance approach by examining state-of-the-art AI-driven systems globally. Prominent examples include: 3.4.1 HealthMap : Utilizes real-time scraping of news media, social media, and official health reports worldwide to detect outbreaks early [ 27 ]. It employs natural language processing and anomaly detection algorithms to provide up-to-date global alerts. 3.4.2 ProMED-mail : A reporting platform that aggregates information from clinicians, media, and public health officials using expert review and crowd-sourced reports to monitor emerging infectious diseases worldwide [ 28 ]. Its approach combines manual curation with automated text processing. 3.4.3 Epidemic Intelligence from Open Sources (EIOS) by WHO : Integrates multiple open-source data streams, including news, social media, and official reports, using machine learning and data fusion techniques to support epidemic intelligence and risk assessment [ 29 ]. Compared to these systems, Nigeria’s IDSR framework currently relies heavily on manual reporting, limited private sector integration, and lacks scalable real-time analytics and diverse data sourcing. This overview highlights key gaps such as delayed reporting, limited automation for signal detection, and data fragmentation. Adoption of similar methodologies, including natural language processing and multi-source fusion, would significantly enhance the effectiveness and timeliness of surveillance in Nigeria. 3.5 Incorporating Advanced AI: Generative Models and Large Language Models Recent advances in large language models (LLMs) and generative AI offer promising opportunities to enhance disease surveillance. Applications include synthetic data generation to augment limited datasets, automated generation of epidemiological reports, and decision support tools that aid in complex outbreak management. While this study focuses on traditional machine learning approaches, primarily due to the current constraints in digital infrastructure and data availability in Nigeria, we recognize the transformative potential of these advanced methods. Future system updates could integrate LLMs and generative AI to improve data quality, automate routine tasks, and provide advanced interpretability and support for health decision-makers as infrastructure and data ecosystems mature. 4.0 Simulation-Based Illustration of a Malaria Surveillance Modeling Workflow To demonstrate the proposed framework, we present a simulation-based case study of malaria surveillance in Northern Nigeria. Due to limited access to high-resolution, longitudinal surveillance data, a synthetic dataset was constructed using plausible relationships between environmental variables (rainfall, temperature), health system access, and malaria incidence. 4.1 Synthetic Data Generation We generated a synthetic weekly dataset for malaria surveillance covering 10 representative Local Government Areas (LGAs) in Northern Nigeria over two years (2023–2024). Each record contains the year, week number, LGA name, malaria case count, total rainfall (mm), average temperature (°C), and a health access index. The data were designed to reflect regional patterns: for example, a pronounced rainy season mid-year (peaking around July–September) with largely dry conditions in the winter months, and consistently high temperatures (25–32°C) typical of tropical Sahel regions [ 28 ]. Each LGA was assigned a Health Access Index (0 to 1 scale) to represent local healthcare infrastructure and access (higher in urban centers, lower in rural areas), acknowledging the North’s limited infrastructure and digital divide. Because robust real-world data from many northern Nigerian districts are scarce mainly due to challenges like poor internet connectivity and limited digital infrastructure in rural areas [ 29 ], we employed synthetic data to enable this simulation without privacy or logistics concerns. Notably, synthetic data can augment or stand in for real health datasets, helping address data access and privacy issues in health research [ 30 ]. In our simulation, weekly malaria cases were modeled as a function of that week’s rainfall, temperature, and health access level, plus a minor random noise term to imitate unexplained variability. We assumed higher rainfall would lead to increased malaria cases following a short lag (due to mosquito breeding after rains) [ 28 ], while higher health access was assumed to reduce malaria incidence by improving prevention (e.g. bed net distribution) and treatment availability. Temperature was included as a secondary factor (malaria transmission is optimal in a moderate temperature range). This case study is designed to illustrate the modeling workflow, feature integration, and interpretability techniques, rather than to provide empirical evidence of predictive performance in real-world settings. 4.2 Pseudocode for Synthetic Malaria Case Simulation and Modeling Algorithm Synthetic Malaria Incidence Simulation and Prediction Step 1: Initialize Simulation Parameters Define the number of Local Government Areas (LGAs) = 10. Define the temporal span = 104 weeks (representing 2023–2024). Define feature ranges and distributions: Rainfall (mm per week): sample from Uniform [0, 300]. Temperature (°C): sample from Uniform[ 20 , 40 ]. Health Access Index (unitless): sample from Uniform[0.2, 1.0]. Step 2: Define Coefficients (epidemiologically motivated) Rainfall coefficient = + 0.15 (positive effect: higher rainfall increases malaria risk). Temperature coefficient = + 0.05 (small positive effect: higher temperature slightly increases malaria risk). Health Access coefficient = − 0.25 (negative effect: better access reduces malaria incidence). Noise term = Gaussian distribution with mean = 0 and standard deviation = 2 (represents unexplained weekly variability). Step 3: Generate Synthetic Dataset For each LGA i from 1 to 10: a. Assign a fixed Health Access Index value from Uniform[0.2, 1.0]. b. For each week t from 1 to 104: Sample rainfall value. Sample temperature value. Sample Gaussian noise term. Compute malaria cases using the function: malaria_cases = (0.15 × rainfall) + (0.05 × temperature) + (–0.25 × health_access) + noise v. Record malaria_cases for LGA i at week t. Save all records into a dataset containing columns: [LGA, Week, Rainfall, Temperature, Health Access, and Malaria Cases]. Step 4: Train Predictive Model Split dataset into: Training set: Weeks 1–52 (2023). Test set: Weeks 53–104 (2024). Train a Decision Tree regression model with: Maximum depth = 6. Input features: Rainfall, Temperature, Health Access. Output target: Malaria Cases. Justification for Model Selection: For the rationale behind Decision Tree model selection, see Section 4.3 . Step 5: Evaluate Model Predict malaria cases for the 2024 test set. Compute evaluation metrics: R² (coefficient of determination). RMSE (root mean square error). Example output: performance metrics are computed based on the selected validation design. Step 6: Interpret Model (Explainability) Apply SHAP (SHapley Additive exPlanations) to the trained Decision Tree. Compute local and global feature contributions. Analyze the relative importance of rainfall, temperature, and health access: Rainfall: primary positive driver of predicted cases. Health Access: strong negative driver. Temperature: weaker but consistent positive effect. Step 7: Export and Deployment Save the synthetic dataset (CSV), model file, and plots (predicted vs. actual, feature importance, SHAP summary). Integrate the model into public health early warning dashboards for malaria surveillance. Enable iterative updating with real-world surveillance data as they become available. 4.3 Predictive Modeling with Decision Trees We trained a Decision Tree regression model to predict weekly malaria case counts based on environmental and health access features. Decision Trees are well-suited for this task as they can capture nonlinear relationships and interactions among predictors [ 31 ]. Unlike traditional linear models, Decision Trees partition data into subsets based on feature thresholds, which is advantageous for discovering complex patterns in health surveillance data, for example, different rainfall–malaria relationships above versus below certain rainfall levels [ 31 ]. The full rationale for Decision Tree selection is presented here; subsequent sections referencing this choice cross-reference this section rather than repeat the justification. Model Selection Rationale Decision Tree (DT) algorithms were selected as the primary modeling approach for several reasons aligned with the resource-constrained context of Nigerian public health: Interpretability : Decision Trees produce transparent decision rules that can be readily understood by public health practitioners and policymakers. Each split in the tree corresponds to a clear threshold (e.g., rainfall > 50 mm), facilitating trust and clinical acceptance. Minimal Computational Requirements : Unlike deep learning approaches, DTs require limited computational resources, making them feasible for deployment in settings with constrained hardware and electricity availability. Suitability for Structured Data : DTs perform well with structured clinical and epidemiological data, requiring minimal preprocessing such as scaling or normalization. Appropriateness for LMIC Contexts : The model's simplicity, ease of training, and straightforward deployment make it particularly suitable for low- and middle-income country (LMIC) settings like Nigeria, where technical capacity for maintaining complex models may be limited. While other models such as Random Forests, Support Vector Machines, or Recurrent Neural Networks could offer potential accuracy improvements, thorough benchmarking is deferred to future work given current infrastructural and data limitations. Literature shows that Decision Trees provide competitive performance in similar epidemiological tasks with the added benefits of simplicity and lower computational demands [ 31 ]. Model Implementation The model used continuous input features including rainfall, temperature, and a health access index. It was trained on synthetic data from the first year (2023) and tested on data from the second year (2024) to evaluate temporal generalization. We selected a moderately deep tree with a maximum depth of 6 to balance model complexity while avoiding overfitting to noise. 4.4 Model Evaluation The model was evaluated using a temporally separated validation design, with data from 2023 used for training and data from 2024 reserved for testing. This approach was chosen to better reflect real-world deployment conditions, where models must generalize to future, unseen time periods. Under this temporal split, the decision tree model achieved an R² of approximately 0.66 and an RMSE of approximately 3.61 within the synthetic dataset. These metrics reflect the model’s ability to capture patterns embedded in the simulated data-generating process. However, because the outcome variable was constructed using predefined relationships with the predictors, these values should be interpreted strictly as internal simulation diagnostics, not as indicators of real-world predictive performance. The use of temporal validation, rather than random splitting, provides a more conservative and realistic assessment of model behavior in surveillance contexts. 4.4.1 Multi-Model Benchmarking Comparison To assess whether the Decision Tree model offers a reasonable interpretability-accuracy trade-off, five ML models were evaluated on the same synthetic dataset under identical temporal validation conditions (training: 2023; testing: 2024). Results are presented in Table 4 . The findings confirm that while ensemble methods (Random Forest and Gradient Boosting) and SVMs marginally outperform the Decision Tree on test R², the differences are modest. This supports the rationale for selecting the Decision Tree as the primary modeling approach in this illustration: its interpretability advantage is substantial, while its accuracy penalty is minimal within this synthetic dataset context. Table 4 Multi-Model Performance Comparison Under Temporal Validation (Synthetic Dataset) Model R 2 Train R 2 Test RMSE Test MAE Test Linear Regression 0.979 0.979 1.91 1.54 Decision Tree (DT)* 0.987 0.965 2.49 1.94 Random Forest (RF) 0.989 0.975 2.09 1.66 Gradient Boosting (GB) 0.993 0.975 2.10 1.69 Support Vector Machine (SVM) 0.978 0.979 1.93 1.55 *Selected model for this illustration. All models evaluated under identical temporal validation split (training: 2023; testing: 2024) on the synthetic dataset (seed = 42). DT = Decision Tree; RF = Random Forest; GB = Gradient Boosting; SVM = Support Vector Machine. Higher R² and lower RMSE/MAE indicate better fit. Ensemble models show marginal gains over DT at the cost of interpretability. To provide a clear and structured overview of the methodological steps undertaken in this case simulation, Fig. 1 illustrates the complete machine learning pipeline for synthetic malaria case prediction. This flowchart visually details the sequential processes, from the initial generation of the synthetic dataset to the final stages of model interpretation, deployment considerations, and iterative improvement. It serves as a visual guide to the algorithmic workflow described in this section, highlighting the interdependencies between data generation, predictive modeling, evaluation, and explainability. The overall algorithmic workflow for this simulation is outlined in the pseudocode below, detailing the key steps from data preparation to model interpretation. BEGIN 1. Synthetic Data Generation a. Define 10 representative LGAs in Northern Nigeria b. For each week in 2023–2024: - Generate synthetic features: • Rainfall (mm) • Temperature (°C) • Health Access Index (0–1) - Compute malaria case count using: malaria_cases = f(rainfall, temperature, health_access) + random_noise 2. Predictive Modeling a. Split dataset: - Training set: 2023 - Test set: 2024 b. Train decision tree regression model (max_depth = 6) - Inputs: rainfall, temperature, health access - Output: predicted malaria case count 3. Model Evaluation a. Predict weekly cases for 2024 b. Compute evaluation metrics: - R² (coefficient of determination) - RMSE (root mean squared error) 4. Model Interpretation (Explainability) a. Apply SHAP to trained model - Compute local and global feature contributions - Analyze importance of rainfall, temperature, and health access - Verify model reasoning aligns with domain knowledge 5. Result Export a. Save synthetic dataset (CSV) b. Save trained model and SHAP plots c. Document and publish code for reproducibility 6. Model Deployment a. Deploy the trained model to predict upcoming malaria trends b. Integrate with public health dashboards or early warning systems 7. Monitoring and Feedback a. Monitor predictions vs. actual malaria reports (if real data becomes available) b. Flag discrepancies and track model drift 8. Iterative Improvement a. Incorporate real-world data when available b. Retrain or fine-tune model with new surveillance data c. Re-evaluate performance and re-deploy updated model END 4.5 SHAP-Based Model Interpretability To ensure the model’s reasoning aligns with domain knowledge and to bolster trust in the findings, we applied SHapley Additive exPlanations (SHAP) for model interpretability to elucidate feature contributions in malaria case predictions [ 32 ]. SHAP is an approach to explain machine learning predictions by assigning each feature a contribution value for each individual prediction [ 32 ]. In essence, SHAP computes how much each input (rainfall, temperature, health access) “pushes” the model’s prediction up or down relative to a baseline prediction. This yields both local explanations (for a specific LGA-week, how did each feature affect the predicted cases?) and global insight into feature importance. We found that the SHAP values corroborated our expectations: rainfall and health access were the most influential features in the model’s decisions, while temperature had a comparatively smaller effect. For example, in weeks with high rainfall, SHAP values for the rainfall feature were strongly positive, indicating rainfall was driving the prediction of higher malaria cases. Conversely, LGAs with higher health access showed negative SHAP contributions for that feature, meaning the model predicted fewer cases than it would have if health access were low (all else being equal) – consistent with the idea that better healthcare access reduces malaria burden. Overall, rainfall emerged as a primary driver of model predictions (reflecting its role in mosquito breeding and seasonality) [ 28 ], and health access was almost equally influential, reflecting substantial differences in case counts between well-served and underserved communities. Temperature had a subtler influence on the model’s output, aligning with the notion that temperature in this region is generally always conducive to malaria transmission, with less week-to-week variability. The SHAP analysis thus provided an extra layer of confidence that the model was leveraging meaningful patterns: it highlighted, for instance, that an extremely low health access index in a given LGA contributed as much to high case predictions as a large increase in rainfall would. Such transparent explanations are crucial for deploying AI in healthcare, as they enable public health officials to verify that the model’s drivers make sense and to identify conditions associated with high-risk predictions. 4.6 Public Health Implications This case simulation demonstrates the potential of combining synthetic data and machine learning to enhance disease surveillance in low-resource settings. In northern Nigeria, where routine malaria reporting may be hampered by infrastructural challenges and incomplete data, a data-driven model could act as an early warning system. For instance, if exceptionally heavy rainfall is forecasted for a coming month, the model (once validated on real data) might predict a spike in malaria cases, prompting preemptive distribution of bed nets, insecticides, or malaria test kits to high-risk LGAs. Our decision tree model, while simplified, illustrates how key factors like weather and healthcare access can be synthesized into an actionable prediction. Importantly, the approach is interpretable, which is a critical requirement for adoption in public health contexts. Decision trees and their outputs can be readily understood by non-technical stakeholders [ 31 ], and our use of SHAP further ensures that the model’s reasoning is transparent and aligned with epidemiological understanding. This addresses a common concern that “black box” AI models might otherwise face resistance in the health sector. Additionally, by using synthetic data, we have shown a viable path to develop and test AI models without risking patient privacy or waiting for extensive data collection. Synthetic datasets can serve as a proxy when real data are scarce, helping to bridge the gap in settings where the digital divide limits real-time data availability [ 28 , 30 ]. The limitations inherent in a synthetic-data approach, including the assumption of stable relationships between climate variables and malaria incidence, the absence of human behavioral factors and vector control interventions, and the need for real-data recalibration before deployment, are discussed in detail in Section 4.7 . In practice, as Nigeria’s health system strengthens its health information infrastructure, such models could be incrementally updated with real data and used to support malaria control programs. While the synthetic case study demonstrates how environmental and health system variables can be integrated into an interpretable machine learning workflow, its primary contribution is conceptual rather than empirical. The results illustrate how such a model could be structured to support decision-making, such as identifying potential high-risk periods or location. However, this does not in any way establish real-world predictive accuracy. In practice, deployment would require training and validation on routine surveillance data, careful handling of missing data and reporting delays, and evaluation across diverse geographic and epidemiological contexts. 4.7 Limitations of the Simulation-Based Demonstration The case study presented in this work is based entirely on synthetic data, which introduces several important limitations. First, the outcome variable is generated using assumed relationships with predictor variables, meaning that model performance is partially determined by the structure imposed during simulation. As a result, the reported metrics do not reflect independent epidemiological validation. Second, the simulation does not fully capture key challenges present in real-world surveillance systems, including missing data, reporting delays, under-reporting, measurement error, and spatial heterogeneity across locations. Additionally, the model assumes stable relationships between predictors and outcomes, whereas real-world epidemiological processes may exhibit temporal drift and intervention effects. Third, the use of uniform coefficients across locations simplifies the underlying disease dynamics and may overestimate model generalizability. Accordingly, the results should be interpreted as a methodological illustration of a machine learning pipeline, rather than as evidence of predictive validity or operational readiness. 4.8 Sensitivity Analysis and Model Robustness Considerations Given that the predictive framework presented in this study is intended as a methodological illustration rather than a fully operational or validated model, the sensitivity analysis focuses on the effect of coefficient perturbations on Decision Tree performance under the same temporal validation design. The rainfall coefficient was varied across a plausible epidemiologically motivated range (± 33% of the base value of 0.15), while all other parameters remained constant. Results are presented in Table 5 . Table 5 Sensitivity Analysis-Decision Tree Performance under Rainfall Coefficient Perturbation Rainfall Coefficient % Change from Base R 2 Test RMSE Test 0.10 −33% 0.922 2.47 0.12 −20% 0.945 2.46 0.15 (base) 0% (reference) 0.966 2.43 0.18 + 20% 0.976 2.38 0.20 + 33% 0.980 2.42 All results from temporal validation (training: 2023; testing: 2024) on synthetic dataset (seed = 42). R² range across perturbations: 0.922–0.980, indicating robust model behavior under simulated coefficient uncertainty. Across all tested perturbations, the Decision Tree model maintained stable predictive performance, with R² values ranging from 0.922 to 0.980 and RMSE values remaining within a narrow band (2.38–2.47). This suggests that the model’s structure, informed by the feature thresholds learned from the training data, is reasonably robust to plausible variations in the assumed coefficient values. In practical implementations using real surveillance data, formal sensitivity analyses incorporating uncertainty in feature measurement and reporting completeness would be essential. 4.9 Roadmap for Real-World Retrospective Validation To transition from methodological illustration to operational utility, future work will focus on retrospective validation using real-world surveillance datasets. Potential data sources include Nigeria’s District Health Information System (DHIS2), the National Malaria Data Repository (NMDR), and complementary surveillance outputs from the Nigeria Centre for Disease Control (NCDC), alongside meteorological data. A retrospective panel dataset will be constructed at the state or local government area (LGA) level, combining weekly malaria case counts with environmental and health system indicators over multiple years. Model validation will prioritize temporally structured approaches, including holdout testing on future periods and rolling-origin evaluation, to better reflect deployment conditions. Performance will be assessed using metrics such as RMSE, MAE, R², and calibration measures, as well as classification-based metrics for outbreak detection where applicable. Additional analyses will examine robustness to missing data, reporting delays, and geographic variability. This phased validation approach will provide a more rigorous assessment of model applicability in real-world surveillance settings. 5.0 Opportunities for Using Machine Learning in Disease Surveillance in Nigeria Machine learning (ML) offers significant potential to enhance disease surveillance in Nigeria by improving data collection, analysis, and predictive capabilities. Several key opportunities have been identified in the literature. 5.1 Cost-Benefit and Economic Impact Analysis AI-driven disease surveillance systems have demonstrated significant economic benefits in various contexts. Operational efficiencies gained through automated data collection and analysis can reduce costs by up to 50%, as evidenced in deployments across India and Rwanda [ 33 , 34 ]. Furthermore, early outbreak detection via machine learning enables timely interventions, potentially lowering treatment and outbreak response expenses by 30–40%. These savings stem from reduced hospitalization rates, shorter outbreak durations, and optimized resource allocation. Applying these findings to Nigeria’s context, we anticipate similar cost reductions. For instance, by integrating predictive analytics into the IDSR framework, the government could realize substantial savings alongside improved health outcomes, making ML integration highly cost-effective and sustainable in the long-term. 5.2 Enhanced Data Collection and Reporting Machine learning (ML) has the potential to revolutionize disease surveillance by enhancing the efficiency and accuracy of data collection. AI-driven mobile applications can minimize human error by automatically capturing and reporting health data in real time. These applications facilitate seamless data transfer from local healthcare providers to centralized public health systems, enabling faster detection of disease outbreaks[ 35 ]. This approach has been successfully implemented in other developing nations, such as India, where AI-assisted data collection improved tuberculosis case detection rates, leading to earlier interventions and better patient outcomes [ 33 ]. ML algorithms can also analyze historical data to forecast potential disease outbreaks, allowing for timely interventions and efficient resource allocation. By identifying patterns in past outbreaks, ML models can predict the likelihood of future occurrences and help public health officials implement preventive measures [ 35 ]. A study in South Africa demonstrated how ML-based predictive models significantly improved early detection of malaria outbreaks, allowing for better vector control and resource distribution [ 36 ]. This model can be adapted for other diseases, such as HIV, malaria, and COVID-19, in regions with limited healthcare infrastructure in Nigeria. 5.3 Advanced Disease Modelling and Outbreak Detection ML techniques can be integrated with innovative surveillance methods to enhance outbreak detection. One such method is the analysis of wastewater samples using ML algorithms, which can provide early warnings of viral outbreaks [ 37 ]. In Brazil, wastewater surveillance detected COVID-19 traces with 85% sensitivity, enabling early interventions [ 38 ]. However, Nigeria’s fragmented sanitation infrastructure where only 32% of urban areas have sewer systems [ 11 ] limits direct applicability. Mobile-based sampling from community water sources could offer a viable adaptation, leveraging Nigeria’s 87% mobile penetration rate [ 15 ]. Additionally, AI-powered air sampling technologies can help identify and track the presence of airborne pathogens in high-risk areas [ 37 ]. Deep learning models have been applied in Kenya to monitor airborne tuberculosis pathogens in crowded urban areas, significantly improving early detection and reducing transmission risks [ 39 ]. These methods can be particularly useful in Nigeria, where densely populated urban centers and limited access to healthcare facilities increase the risk of undetected outbreaks. 5.4 Integration with Existing Surveillance Systems ML can augment existing disease surveillance efforts by providing AI-driven insights that assist healthcare professionals in identifying disease trends and making data-driven decisions [ 35 ]. In Rwanda, an ML-powered electronic medical record system improved real-time reporting and disease trend analysis, enabling public health authorities to respond to emerging threats more effectively [ 34 ]. Automated identification and reporting of notifiable diseases can help address challenges such as limited healthcare infrastructure and underreporting by healthcare workers, ultimately strengthening disease surveillance systems [ 40 ]. A notable example of successful implementation in a developing country is the Auto-Visual Acute Flaccid Paralysis Detection and Reporting (AVADAR) system, which has enhanced polio surveillance across Africa. AVADAR leverages mobile technology and artificial intelligence to support healthcare workers in detecting and promptly reporting suspected polio cases. This innovative approach has proven to be both cost-effective and efficient, enabling real-time data collection and improving the accuracy of surveillance efforts, thereby contributing to more effective disease control and eradication strategies [ 18 ]. By integrating ML with Nigeria’s existing Integrated Disease Surveillance and Response (IDSR) framework, the country can enhance the accuracy and timeliness of disease reporting 5.5 Public Health Applications beyond Infectious Diseases Beyond infectious diseases, ML can play a crucial role in addressing broader public health challenges. AI-driven models can analyze patterns in antimicrobial resistance, helping to guide treatment protocols and reduce the spread of resistant infections [ 41 ]. In Bangladesh, ML models were used to track and predict resistance patterns for common bacterial infections, informing policy decisions and improving antibiotic stewardship [ 42 ]. Predictive models can also aid in identifying risk factors for mental health disorders and informing therapeutic interventions [ 43 ]. AI-driven mental health assessment tools have been piloted in Ethiopia with promising results, offering a scalable solution for early diagnosis and treatment planning in resource-limited settings [ 44 ]. In Nigeria, where mental health services remain underdeveloped, integrating AI into mental health care could improve access to early diagnosis and intervention. 5.6 Workforce Training and Capacity Building An awareness and acceptance of AI among Nigerian healthcare professionals, targeted training programs are essential to enhance ML adoption. Training initiatives can equip healthcare workers with the skills needed to integrate AI-driven solutions into routine disease surveillance and patient care [ 45 ]. A similar approach in Uganda led to significant improvements in healthcare workers' ability to use AI tools for disease surveillance, ultimately improving health outcomes [ 46 ]. Collaboration between health institutions, technology companies, and academic researchers is also vital for driving innovation in AI-based disease surveillance [ 47 ]. In India, partnerships between universities and AI firms led to the successful deployment of ML models for predicting dengue outbreaks, demonstrating the potential of interdisciplinary cooperation [ 48 ]. Establishing similar collaborations in Nigeria could accelerate the adoption of AI-driven disease surveillance technologies and enhance the country’s ability to respond to public health challenges effectively. By leveraging ML for disease surveillance, Nigeria can significantly improve early outbreak detection, enhance data accuracy, and optimize resource allocation, ultimately strengthening its public health infrastructure and response capabilities. Table 1 Summary of Machine Learning Applications in Disease Surveillance in Nigeria and Beyond Study Authors Year ML Methodologies Used Geographic Focus Reported Health Outcomes [ 17 ] 2024 AI-driven hotspot mapping Southwestern Nigeria Improved tuberculosis case detection via active case finding [ 18 ] 2020 AI within AVADAR (mobile-based surveillance) Multiple African countries (including Nigeria) Increased acute flaccid paralysis/polio case reporting by ~ 45% [ 19 ] 2025 Event management + ML-supported surveillance Nigeria Strengthened outbreak detection and response coordination [ 28 ] 2020 Climate indices + ML modeling Tropical Africa (Nigeria focus) Predicted malaria and meningitis outbreaks using climate-health data [ 49 ] 2021 ML classification models (Decision Trees, SVM) Nigeria High accuracy in diagnosing malaria from symptoms [ 50 ] 2023 ML algorithms for pattern recognition Nigeria Early detection of Hepatitis B infections, including asymptomatic cases [ 51 ] 2024 Predictive analytics with ML Nigeria (Lassa fever) Early warning and outbreak forecasting of Lassa fever [ 52 ] 2019 ML algorithms (classification, regression) Nigeria Prediction of meningitis outbreaks with high accuracy [ 36 ] 2024 ML models for outbreak forecasting Nigeria Improved epidemic prediction and preparedness [ 53 ] 2024 Predictive analytics using ML Rural Nigeria Forecasting epidemic outbreaks in underserved areas [ 33 ] 2021 ML algorithm on health surveillance data France Estimated diabetes incidence from surveillance/EHR data [ 38 ] 2021 Wastewater surveillance + ML modeling Brazil Tracked SARS-CoV-2 trends at municipal level with high sensitivity [ 54 ] 2020 Regression-based ML models Rwanda Predicted out-of-pocket health expenditures from health surveillance data Table 2 highlights possible training areas, challenges, opportunities and solutions. Table 2 Opportunities and Challenges for ML in Nigeria’s Diseases Surveillance Opportunity Potential Impact Challenge Solution Enhanced Data Collection 30% faster reporting [ 35 ]. Poor rural connectivity Mobile-based tools Outbreak Detection 85% sensitivity [ 37 ]. Limited sanitation Community water sampling Workforce Training 50% AI literacy boost [ 45 ]. Lack of programs PPP-led workshops 6.0 Case Studies of ML in Nigerian Health Surveillance 6.1 Malaria Diagnosis through Symptom-Based ML Models Machine learning models trained on symptomatic patterns have been developed to enhance malaria diagnosis. The study assessed eight algorithms, with decision trees and support vector machines (SVMs) demonstrating superior performance. These models showed high predictive accuracy using symptoms alone, representing a particularly impactful innovation in Nigeria’s rural and resource-constrained areas where access to microscopy or rapid diagnostic tests is limited [ 49 ]. 6.2 ML-Enhanced Epidemic Surveillance at NCDC A pioneering initiative at the Nigeria Centre for Disease Control (NCDC) integrated machine learning (ML) algorithms and data analytics into epidemic surveillance. This pilot framework was deployed during COVID-19 and Lassa fever outbreaks, enabling early detection of epidemic signals. Notably, NCDC staff remarked that “we’ve moved from reactive firefighting to proactive detection,” reflecting the practical utility of ML-driven surveillance in strengthening public health preparedness [ 15 ]. 6.3 Hepatitis B Virus Prediction Using Routine Laboratory Data A study published in Scientific Reports trained machine learning (ML) algorithms on clinical datasets to predict Hepatitis B Virus (HBV) infections. The models effectively identified both symptomatic and asymptomatic cases, which is particularly valuable for pre-surgical screenings. Clinical collaborators at Jos University Teaching Hospital noted that “doctors were impressed by the model’s ability to flag asymptomatic carriers,” highlighting its potential for integration into hospital workflows across both rural and urban settings [ 50 ]. 6.4 Clinician Trust in AI for Infectious Disease Triage A review of Nigeria’s growing experience with AI-based diagnostic tools underscores how machine learning has begun to support triage decisions in overstretched clinical environments. A frontline clinician stated: “We’ve started trusting algorithms to support our triage decisions, especially when overwhelmed during Lassa and COVID surges.” This testimonial affirms the expanding role of ML in not only diagnosis but also in decision-support and care prioritization during public health emergencies [ 55 ]. 7.0 Challenges and Barriers to Implementing Machine Learning in Disease Surveillance in Nigeria Machine learning (ML) offers transformative potential for disease surveillance in Nigeria by improving data-driven decision-making, early outbreak detection, and public health interventions. However, several challenges hinder the effective implementation of ML-driven surveillance systems. These barriers span data quality, infrastructure, technical expertise, financial constraints, ethical concerns, and cultural resistance, all of which require targeted solutions [ 51 ]. Nigeria faces significant gaps in digital infrastructure, which affects the collection, storage, and analysis of health data. Many healthcare facilities lack electronic health records and rely on paper-based systems, making it difficult to train ML models effectively [ 56 ]. The deployment of ML-based surveillance systems requires robust digital infrastructure, including reliable internet access, stable electricity supply, and high-performance computing resources [ 39 ]. However, frequent power outages, poor internet connectivity, and inadequate technological resources in many healthcare facilities, particularly in rural areas, pose significant challenges. The lack of interoperability between existing health data systems further complicates data collection efforts, making it difficult to aggregate and analyze information in real-time [ 57 ]. For example, Nigeria’s health data systems lack adherence to global standards like HL7 or FHIR, with only 15% of facilities using interoperable electronic records [ 56 ]. This fragmentation delays data aggregation by up to 3 weeks, undermining ML’s real-time potential [ 2 ]. Adopting FHIR could reduce this lag to 48 hours, as seen in Kenya’s e-health rollout [ 39 ]. Health data in Nigeria is often fragmented, incomplete, or inconsistently recorded across different healthcare facilities. Manual record-keeping, lack of standardized electronic health records, and disparities in data collection between urban and rural areas exacerbate these issues [ 11 ]. ML models require large, high-quality datasets to function effectively, but the lack of centralized health databases impairs the training and validation of these models. Without reliable data, ML-driven insights may be inaccurate, leading to ineffective public health responses [ 58 ]. Moreover, poor data-sharing frameworks and institutional structures prevent seamless integration of surveillance data across various agencies, limiting the effectiveness of ML applications [ 59 ]. Implementing ML solutions in disease surveillance demands specialized knowledge in data science, artificial intelligence, and software engineering. However, Nigeria faces a shortage of professionals with the technical expertise needed to develop, deploy, and maintain ML-based health systems [ 2 ]. While interest in AI and data science is growing, the lack of structured training programs and educational opportunities in these fields remains a major gap. Additionally, many healthcare workers are unfamiliar with ML technologies, making it difficult to integrate these solutions into existing public health frameworks [ 60 ]. Without adequate technical capacity, even well-funded ML initiatives may fail due to a lack of personnel capable of maintaining and optimizing the systems. The high costs associated with ML implementation present another significant barrier. Developing, deploying, and maintaining ML-driven surveillance [ 11 ] systems require substantial financial investment in infrastructure, software, and workforce training. However, Nigeria's healthcare sector is underfunded, with limited budgets allocated for technological innovation. Given competing priorities, such as responding to immediate health crises and improving basic healthcare services, investments in ML-powered disease surveillance often take a back seat. Without sufficient funding, scaling ML solutions for nationwide disease surveillance remains a challenge [ 11 ]. International collaborations and public-private partnerships could help bridge this gap, but financial constraints remain a persistent challenge. The use of ML in disease surveillance raises concerns about patient privacy and data security. In Nigeria, the absence of comprehensive data protection laws and regulatory frameworks heightens these concerns. Issues surrounding data ownership, patient confidentiality, and informed consent must be addressed to ensure ethical AI deployment [ 61 ]. Additionally, algorithmic biases could result in disparities in healthcare delivery, disproportionately affecting certain populations. Establishing strong data governance policies is crucial to fostering trust and ensuring equitable use of ML in disease surveillance [ 62 ]. Nigeria’s National Data Protection Regulation (NDPR) offers a starting point, mandating consent and data anonymization, but enforcement remains weak, with only 10% compliance among health facilities [ 61 ]. Aligning ML deployment with NDPR, alongside regular audits, could mitigate privacy risks and reduce bias, ensuring equitable outcomes across urban and rural populations. Without clear regulatory oversight, ML applications in public health could lead to unintended consequences, such as discriminatory health policies or misuse of personal health data [ 63 ]. Most ML models are trained on datasets from high-income countries, which may not adequately capture the epidemiological patterns in Nigeria. This leads to biased predictions and reduced model effectiveness in local contexts. Without efforts to develop locally relevant datasets and ensure ML models are tailored to Nigeria’s unique healthcare landscape, the accuracy of disease predictions and outbreak surveillance could be compromised [ 53 ]. Inadequate representation of Nigerian-specific disease patterns in training datasets makes it difficult for ML models to detect emerging outbreaks effectively [ 64 ]. To improve generalizability, investments in local dataset curation and contextualized algorithm development are necessary. Resistance to adopting ML-driven surveillance systems is another significant hurdle. Many healthcare professionals and policymakers are hesitant to embrace new technologies due to a lack of awareness, fear of job displacement, or skepticism regarding ML's reliability [ 65 ]. Additionally, traditional disease surveillance methods are often perceived as more familiar and dependable. Low levels of digital literacy among healthcare workers and the general population further impede the widespread adoption of ML solutions [ 66 ]. Addressing this resistance requires awareness campaigns, stakeholder engagement, and targeted training programs to build confidence in ML technologies. Without proactive efforts to foster trust and acceptance, even the most advanced ML solutions may struggle to gain traction in Nigeria’s public health sector. 8.0 Ethical Considerations and Explainable AI in Disease Surveillance Explainable Artificial Intelligence (XAI) refers to a set of processes and methods that enable users to understand the results and outputs generated by AI/ML algorithms [ 67 ]. In the healthcare sector, where trust, accountability, and accuracy are critical, XAI plays a vital role in ensuring that AI predictions are interpretable to both clinicians and patients [ 68 ]. Black-box AI models are often met with skepticism by healthcare professionals; XAI enhances trust by making decision-making processes transparent, thereby fostering clinical acceptance [ 69 ]. Given that medical decisions can have life-or-death consequences, XAI supports ethical practices by enabling human validation of AI outputs [ 70 ], and many medical AI applications must comply with regulatory standards that mandate explainability [ 71 ]. The growing body of literature on XAI in disease prediction confirms its centrality to responsible AI deployment. A systematic literature review synthesized findings from 30 studies examining XAI’s evolving role in disease prediction, highlighting the effectiveness of SHAP and Local Interpretable Model-Agnostic Explanations (LIME) as the most widely used techniques [ 72 ]. This review underscored critical gaps including limited dataset diversity and reliance on single data modalities, especially relevant to Nigeria’s fragmented surveillance data landscape. Work on febrile disease diagnostics using data-driven XAI methodologies demonstrated that explainable diagnostic models can achieve clinically meaningful performance even with modest datasets, directly applicable to Nigeria’s resource-constrained setting [ 73 ]. In the context of tropical and infectious disease specifically, XAI has been applied to enhance the interpretability of malaria and typhoid diagnoses using large language model integration [ 74 ]. XAI applied to Parkinson’s disease prediction, demonstrating that SHAP clarifies the global importance of clinical features while LIME provides patient-specific explanations [ 75 ]. XAI frameworks also improve adoption of AI tools in clinical settings by reducing algorithmic resistance among healthcare professionals, a barrier particularly pronounced in Nigeria’s public health workforce [ 76 ]. XAI-enhanced ML models demonstrate superior interpretability-accuracy trade-offs compared to opaque black-box alternatives, reinforcing the case for XAI as a deployment standard [ 77 ]. SHAP builds on cooperative game theory to assign each feature a value reflecting its contribution to a specific prediction, offering both local explanations and global insights into model behavior [ 78 ] [ 32 ]. LIME takes a localized approach, approximating the behavior of complex ML models near an individual prediction using simpler interpretable models [ 79 ]. Both are critical in healthcare settings where interpretability informs both individual and population-level interventions [ 80 ]. In Nigeria’s evolving public health landscape, these tools hold particular promise. A public health officer in Edo or Ondo State, regions known for recurring Lassa fever outbreaks, could use LIME to examine why an AI-based surveillance system flagged a particular case as high-risk, revealing key contributors such as symptom profile, rodent infestation reports from SMS-based surveillance, and data on poor sanitation in the affected locality. On a broader scale, SHAP offers significant value in post-hoc analyses, helping policymakers identify which environmental and infrastructural factors consistently drive outbreaks across LGAs, guiding targeted policy responses accordingly 9.0 Comparing AI and ML Adoption in Public Health: Nigeria vs. Rwanda and Kenya Nigeria has made some effort in adopting AI and ML in public health. However, when benchmarked against Rwanda and Kenya, Nigeria lags in several critical areas. As shown in Table 5, key gaps include weak AI-health policy integration, fragmented data systems, limited funding, and slower reporting timelines. The comparative analysis is based on published literature and should be interpreted with caution, given differences in reporting systems, funding tracking, and data availability across countries [52,54,81–84]. Funding figures represent available estimates from cited literature rather than verified national accounts Table 5: Comparative Overview of AI/ML Adoption in Public Health-Nigeria, Rwanda, and Kenya Dimension Nigeria Rwanda Kenya National AI/Health Strategy NITDA AI Policy (2021); no specific health integration strategy Smart Rwanda Master Plan includes health sector AI targets [84] Digital Economy Blueprint integrates health AI [84] Estimated AI-Health Funding $10M; government-supported [84] >$5M; government + PPP [84] Key PPP Models Limited; few structured health AI PPPs Zipline drone logistics + ML for malaria prediction [54] IBM-Watson at Kenyatta Hospital; AFYA-Tek disease forecasting [83] Data Reporting Timeliness ~2–3 week lag in national data aggregation [52] 50% reduction in malaria case reporting time by 2022 [54] Sub-weekly outbreak prediction updates via AFYA-Tek [83] Capacity Building Programs No structured national program; emerging university initiatives Rwanda AI Academy (government-sponsored) [84] Nairobi AI Lab; university-industry partnerships [84] Nigeria’s key gaps, weak AI-health policy integration, fragmented data systems, limited dedicated funding, and fewer scalable AI hubs, are compounded by structural governance challenges. Despite these constraints, opportunities exist: leveraging the NITDA AI Policy to design a dedicated AI framework for health, partnering with private sector actors such as MTN and Flutterwave for local funding, and adapting Rwanda’s drone-AI model for rural healthcare logistics [82]. 10.0 Recommendations and future directions. Maximizing the potential of ML for disease surveillance in Nigeria requires a comprehensive strategy encompassing policy enhancements, workforce development, infrastructure upgrades, and active stakeholder collaboration. Recommendations are organized into three tiers based on feasibility, cost, and implementation readiness. Cost estimates are scenario-based illustrative figures drawn from comparable LMIC technology deployments; actual costs depend on procurement models, PPP structures, and phased scaling strategies. 10.1 Tier 1 - Foundational Reforms (Short-Term: Years 1–2) Tier 1 reforms address immediate, high-priority, lower-cost actions that can be implemented within existing institutional structures. These constitute the prerequisite foundations upon which more advanced AI integration depends: Strengthen eIDSR implementation: Equip DSNOs in priority states with low-cost digital reporting tools (< $ 5,000 per state), targeting a 30% improvement in reporting accuracy and reduction of the current 2–3 week data lag to under 72 hours. Improve interoperability standards: Mandate HL7/FHIR adoption in a phased rollout starting with tertiary and secondary facilities. Strengthen data governance: Enforce the NDPR across health facilities, conduct independent compliance audits, and establish clear policies for data anonymization. Establish a National AI-for-Health Working Group: Convene NCDC, Federal Ministry of Health, NITDA, and academic institutions to develop a dedicated ML-health integration roadmap. Integrate AI/ML into medical and public health curricula: Target 50% of graduating public health cohorts receiving foundational AI literacy training within two years. 10.2 Tier 2 - Infrastructure Scaling (Medium-Term: Years 3–5) Tier 2 actions require significant capital investment and PPP engagement, and are contingent upon achieving Tier 1 baseline interoperability and data maturity: Expand rural connectivity: Deploy ML-ready internet infrastructure through PPPs with telecoms, targeting a 50% increase in eIDSR coverage at an estimated $ 10–15M nationally (illustrative scenario). Cloud migration and centralized dashboards: Migrate disease surveillance data to secure cloud platforms with ML processing capabilities and develop real-time outbreak dashboards. Pilot ML models on real Nigerian data: Initiate retrospective validation studies for malaria, Lassa fever, and cholera using NCDC historical data, benchmarking Decision Trees, Random Forests, and gradient boosting. Implement measurable performance indicators: Track eIDSR reporting lag (target: 3 weeks → 72 hours), interoperability adoption rates (target: 50% of facilities), and ML temporal validation R² (target: ≥0.70 for priority diseases). 10.3 Tier 3 - Advanced AI Integration (Long-Term: Years 5–10) Tier 3 innovations are contingent upon achieving baseline data maturity, interoperability, and workforce capacity established in Tiers 1 and 2. Advanced technologies should not be pursued in parallel with foundational reforms but sequenced after their demonstrated success: Deploy cloud-based ML surveillance at national scale: Integrate validated models across all 36 states and FCT within eIDSR and NCDC platforms. Estimated infrastructure investment: $ 40–60M nationally over five years (illustrative scenario). Develop federated learning frameworks: Enable model training across 774 LGAs without centralizing sensitive patient records — contingent upon adequate state-level server infrastructure. Integrate genomic surveillance pipelines: Leverage AI-driven genomic analytics to track pathogen evolution and predict emerging disease strains. Establish cross-border AI surveillance collaboration: Work through ECOWAS and Africa CDC to create continent-wide AI-driven disease monitoring systems. Figure 3 below presents the proposed ML integration workflow for Nigeria’s IDSR system, illustrating the comprehensive system architecture from multi-source data ingestion through ML model development, explainable AI output, and continuous monitoring. 10.4 Economic Framing: DALYs Averted and Return on Investment Surveillance systems that enable early detection significantly reduce the burden of disease and are highly cost-effective. According to WHO benchmarks, interventions that avert DALYs at a cost below $ 150 per DALY are considered highly cost-effective in LMICs [ 85 ]. For illustrative purposes: an ML-based surveillance system preventing 100 outbreak-related deaths annually could conservatively save 3,000–4,000 DALYs per year, consistent with published estimates from zoonotic outbreak mitigation efforts in Nigeria [ 86 ]. WHO models suggest every $ 1 invested in surveillance may yield $ 5–7 in savings through reduced hospitalization, increased productivity, and avoided emergency expenditures [ 87 ]. It is essential to note that these projections are illustrative scenario-based estimates intended to inform the general cost-effectiveness framing of ML surveillance investments, rather than precise Nigeria-specific modeling. Actual DALYs averted and ROI figures will depend on disease burden trajectories, ML model accuracy on real data, implementation scale, and population-level factors. Nigeria-specific cost-effectiveness studies using real surveillance data are strongly recommended before large-scale funding decisions are made. 11.0 Conclusion Integrating ML into Nigeria’s disease surveillance system represents a pivotal pathway towards enhancing outbreak detection capabilities and bolstering public health resilience. This paper provided a structured examination of ML’s transformative potential, detailed key ML methodologies and architectures, presented a full methodological illustration using synthetic malaria surveillance data including a complete simulation algorithm and pseudocode, benchmarked five ML models (Linear Regression, Decision Tree, Random Forest, Gradient Boosting, and Support Vector Machine) under identical temporal validation conditions, and demonstrated model robustness through a rainfall coefficient sensitivity analysis (R² range: 0.922–0.980 across ± 33% perturbations). SHAP-based interpretability confirmed that rainfall and health access index were the dominant predictive features, consistent with epidemiological expectations. The paper also conducted a structured comparative analysis situating Nigeria’s AI adoption within the broader African landscape alongside Rwanda and Kenya. All findings remain illustrative and require real-world validation using NCDC or state-level surveillance data before deployment. With strategic reforms, implementing the proposed tiered integration roadmap, strengthening PPPs, and developing contextually validated ML models, Nigeria can build a proactive, equitable, and data-driven public health system capable of mitigating future health threats and advancing global health equity Abbreviations AI Artificial Intelligence AVADAR Auto—Visual Acute Flaccid Paralysis Detection and Reporting CNNs Convolutional Neural Networks CV Cross—Validation DALYs Disability—Adjusted Life Years DSNOs Disease Surveillance and Notification Officers eIDSR Electronic Integrated Disease Surveillance and Response EIOS Epidemic Intelligence from Open Sources FHIR Fast Healthcare Interoperability Resources GANs Generative Adversarial Networks HBV Hepatitis B Virus IDSR Integrated Disease Surveillance and Response LGAs Local Government Areas LIME Local Interpretable Model—Agnostic Explanations LLMs Large Language Models LMIC Low—and Middle—Income Country ML Machine Learning NCDC Nigeria Centre for Disease Control NDPR National Data Protection Regulation NITDA National Information Technology Development Agency PPPs Public—Private Partnerships RFs Random Forests RMSE Root Mean Square Error RNNs Recurrent Neural Networks ROI Return on Investment SHAP SHapley Additive exPlanations SVMs Support Vector Machines WBE Wastewater—Based Epidemiology WHO World Health Organization XAI Explainable Artificial Intelligence Declarations Funding This research received no specific funding from any public, commercial, or not-for-profit organization. The work was carried out with support from the authors’ respective institutions. Authors' Contributions Conceptualization & Design: L.J.A., A.B.U. Methodology & Simulation (Machine Learning): L.J.A. (Primary lead for the synthetic data model and analysis), A.B.U. Drafting Original Manuscript: A.B.U., L.J.A., S.K.S. Review & Editing (Public Health & Contextual Insights): H.Y., A.H.D., M.Z., U.A.H. Supervision: A.B.U., U.A.H. Final Approval: All authors read and approved the final manuscript. Clinical trial number Not applicable Consent for Publication Not applicable. This manuscript does not contain any images, videos, or personal data relating to individual participants. Competing Interests The authors declare that they have no competing interests (financial or non-financial) that could have influenced the work reported in this paper. References Hong R, Walker R, Hovan G, Henry L, Pescatore R. The Power of Public Health Surveillance. Del J Public Health. 2020 Jul 1;6(2):60. doi:10.32481/djph.2020.07.016 PubMed PMID: 34467112. Ibrahim LM, Stephen M, Okudo I, Kitgakka SM, Mamadu IN, Njai IF, et al. A rapid assessment of the implementation of integrated disease surveillance and response system in Northeast Nigeria, 2017. BMC Public Health. 2020 May 1;20(1):600. doi:10.1186/s12889-020-08707-4 Isere EE, Fatiregun AA, Ajayi IO. An overview of disease surveillance and notification system in Nigeria and the roles of clinicians in disease outbreak prevention and control. Niger Med J. 2015 Jun;56(3):161. doi:10.4103/0300-1652.160347 Rashid A B, Kausik MD A K. AI revolutionizing industries worldwide: A comprehensive overview of its diverse applications. Hybrid Adv. 2024;7:100277. doi:10.1016/j.hybadv.2024.100277. Anahtar MN, Yang JH, Kanjilal S. Applications of Machine Learning to the Problem of Antimicrobial Resistance: an Emerging Model for Translational Research. J Clin Microbiol. 2021 Jun 18;59(7):10.1128/jcm.01260-20. doi:10.1128/jcm.01260-20 Wang L, Zhang Y, Wang D, Tong X, Liu T, Zhang S, et al. Artificial Intelligence for COVID-19: A Systematic Review. Front Med. 2021 Sep 30;8. doi:10.3389/fmed.2021.704256 Jian SW, Cheng HY, Huang XT, Liu DP. Contact tracing with digital assistance in Taiwan’s COVID-19 outbreak response. Int J Infect Dis. 2020 Oct 6;101:348. doi:10.1016/j.ijid.2020.09.1483 PubMed PMID: 33035674. Kim J, Jo S, Cho S il. New framework to assess tracing and testing based on South Korea’s response to COVID-19. BMC Infect Dis. 2024 May 3;24(1):469. doi:10.1186/s12879-024-09363-4 Chellaiyan VG, Nirupama AY, Taneja N. Telemedicine in India: Where do we stand? J Family Med Prim Care. 2019 Jun;8(6):1872–1876. doi:10.4103/jfmpc.jfmpc_264_19 PubMed PMID: 31334148; PubMed Central PMCID: PMC6618203. Sangwa S, Placide M. Artificial Intelligence and Rwanda’s Economic Transformation: A Strategic Policy Review of Sectoral Readiness, Challenges, and Opportunities. Vol. 1. 2025 May 9;1:1–16. doi:10.5281/zenodo.15459202 Meribole EC, Makinde OA, Oyemakinde A, Oyediran KA, Atobatele A, Fadeyibi FA, et al. The Nigerian health information system policy review of 2014 : the need, content, expectations and progress. Health Inf Libr J. 2018 Dec;35(4):285–97. doi:10.1111/hir.12240 PubMed PMID: 30417971. Kusimo OC, Ugwu CI, Aduh U, Okoro CA. Implementing TB Surveillance in Nigeria: Best Practices, Challenges and Lessons Learnt. J Tuberc Res. 2020 Nov 9;8(4):199–208. doi:10.4236/jtr.2020.84018 Onwe FI, Okedo-Alex IN, Akamike IC, Igwe-Okomiso DO. Vertical disease programs and their effect on integrated disease surveillance and response: perspectives of epidemiologists and surveillance officers in Nigeria. Trop Dis Travel Med Vaccines. 2021 Oct 1;7(1):28. doi:10.1186/s40794-021-00152-4 Ibrahim LM, Okudo I, Stephen M, Ogundiran O, Pantuvo JS, Oyaole DR, et al. Electronic reporting of integrated disease surveillance and response: lessons learned from northeast, Nigeria, 2019. BMC Public Health. 2021. doi:10.1186/s12889-021-10957-9 Ogunboye I, Adebayo I, Anioke S, Egwuatu E, Ajala C, Awuah SB. Enhancing Nigeria’s health surveillance system: A data-driven approach to epidemic preparedness and response. World J Adv Res Rev. 2023 Oct 15;20:1352–69. doi:10.30574/wjarr.2023.20.1.2078 Akinpelu D, Akintola S. Navigating the Legal and Ethical Terrain of Artificial Intelligence in Enhancing Patient Safety in Nigeria. J Intellect Prop Inf Technol Law JIPIT. 2023. doi:10.52907/jipit.v3i1.261 Alege A, Hashmi S, Eneogu R, Meurrens V, Budts AL, Pedro M, et al. Effectiveness of Using AI-Driven Hotspot Mapping for Active Case Finding of Tuberculosis in Southwestern Nigeria. Trop Med Infect Dis. 2024 May;9(5):5. doi:10.3390/tropicalmed9050099 Ticha JM, Akpan GU, Paige LM, Senouci K, Stein A, Briand P, et al. Outcomes of the Deployment of the Auto-Visual Acute Flaccid Paralysis Detection and Reporting (AVADAR) System for Strengthening Polio Surveillance in Africa From 2017 to 2018: Evaluation Study. JMIR Public Health Surveill. 2020 Dec 2;6(4):e18950. doi:10.2196/18950 Elston J, Eteng WEO, Ihekweazu C, Oliver I, Aniaku E, Abubakar A, et al. Development and Implementation of a Public Health Event Management System, Nigeria, 2018–2024 - Volume 31, Number 1—January 2025 - Emerging Infectious Diseases journal - CDC [Internet]. doi:10.3201/eid3101.240379 Afuan L, Isnanto RR. A comparative study of machine learning algorithms for fall detection in technology-based healthcare system: Analyzing SVM, KNN, decision tree, random forest, LSTM, and CNN. E3S Web Conf. 2025;605:03051. doi:10.1051/e3sconf/202560503051 Damaševičius R, Jagatheesaperumal SK, Kandala RNVPS, Hussain S, Alizadehsani R, Gorriz JM. Deep learning for personalized health monitoring and prediction: A review. Comput Intell. 2024 Jun;40(3):e12682. doi:10.1111/coin.12682 Nisar DEM, Amin R, Shah NUH, Ghamdi MAA, Almotiri SH, Alruily M. Healthcare Techniques Through Deep Learning: Issues, Challenges and Opportunities. IEEE Access. 2021;9:98523–41. doi:10.1109/ACCESS.2021.3095312 Aldahiri A, Alrashed B, Hussain W. Trends in Using IoT with Machine Learning in Health Prediction System. Forecasting. 2021 Mar;3(1):1. doi:10.3390/forecast3010012 Mohr F, Rijn JN, Van JN. Towards Model Selection using Learning Curve Cross-Validation [Internet]. 2021 [cited 2025 Jun 11]. Available from: https://consensus.app/papers/towards-model-selection-using-learning-curve-mohr-rijn/fde46c822fc85b1f942a8bc210325480/ Xu Y, Goodacre R. On Splitting Training and Validation Set: A Comparative Study of Cross-Validation, Bootstrap and Systematic Sampling for Estimating the Generalization Performance of Supervised Learning. J Anal Test. 2018;2:249–62. doi:10.1007/s41664-018-0068-2 Yates LA, Aandahl Z, Richards S, Brook B. Cross validation for model selection: a primer with examples from ecology [Internet]. 2022 [cited 2025 Jun 11]. Available from: https://consensus.app/papers/cross-validation-for-model-selection-a-primer-with-yates-aandahl/76e95bf1055f507b959fcb32721ea21c/ Nurunnabi A, Teferle FN, Laefer D, Remondino F, Karas I, Li J. kCV-B: BOOTSTRAP WITH CROSS-VALIDATION FOR DEEP LEARNING MODEL DEVELOPMENT, ASSESSMENT AND SELECTION. Int Arch Photogramm Remote Sens Spat Inf Sci. 2022. doi:10.5194/isprs-archives-xlviii-4-w3-2022-111-2022 Ayanlade A, Nwayor IJ, Sergi C, Ayanlade OS, Di Carlo P, Jeje OD, et al. Early warning climate indices for malaria and meningitis in tropical ecological zones. Sci Rep. 2020 Aug 31;10(1):14303. doi:10.1038/s41598-020-71094-8 PubMed PMID: 32868821; PubMed Central PMCID: PMC7459128. Onuh S, Aiyeola SY, Tofade CT, Adesina LO. APPLICATIONS OF SPACE TECHNOLOGY FOR ACCELERATED INFRASTRUCTURES DEVELOPMENT FOR RURAL AREAS DEVELOPMENT IN NIGERIA. Arid Zone J Eng Technol Environ. 2019;15:21–34. Giuffrè M, Shung DL. Harnessing the power of synthetic data in healthcare: innovation, application, and privacy. NPJ Digit Med. 2023 Oct 9;6(1):186. doi:10.1038/s41746-023-00927-3 PubMed PMID: 37813960; PubMed Central PMCID: PMC10562365. Battista K, Diao L, Patte KA, Dubin JA, Leatherdale ST. Examining the use of decision trees in population health surveillance research: an application to youth mental health survey data in the COMPASS study. Health Promot Chronic Dis Prev Can Res Policy Pract. 2023 Feb;43(2):73–86. doi:10.24095/hpcdp.43.2.03 PubMed PMID: 36794824; PubMed Central PMCID: PMC10026612. Ranjbaran G, Recupero D. C-SHAP: A Hybrid Method for Fast and Efficient Interpretability. Appl Sci. 2025 Jan 1. doi:10.3390/APP15020672 Haneef R, Kab S, Hrzic R, Fuentes S, Fosse-Edorh S, Cosson E, et al. Use of artificial intelligence for public health surveillance: a case study to develop a machine Learning-algorithm to estimate the incidence of diabetes mellitus in France. Arch Public Health. 2021 Sep 22;79(1):168. doi:10.1186/s13690-021-00687-0 Ogwu MC, Izah SC. Innovations in Disease Surveillance and Monitoring. In: Ogwu MC, Izah SC, editors. Technological Innovations for Managing Tropical Diseases [Internet]. Cham: Springer Nature Switzerland; 2025 [cited 2025 Mar 2]. p. 83–108. Available from: https://doi.org/10.1007/978-3-031-82622-1_4 doi:10.1007/978-3-031-82622-1_4 Adigwe OP, Onavbavba G, Sanyaolu SE. Exploring the matrix: knowledge, perceptions and prospects of artificial intelligence and machine learning in Nigerian healthcare. Front Artif Intell. 2024 Jan 19;6. doi:10.3389/frai.2023.1293297 Ekundayo F. Using machine learning to predict disease outbreaks and enhance public health surveillance. 2024 Dec 14;794–811. doi:10.30574/wjarr.2024.24.3.3732 Abdeldayem OM, Dabbish AM, Habashy MM, Mostafa MK, Elhefnawy M, Amin L, et al. Viral outbreaks detection and surveillance using wastewater-based epidemiology, viral air sampling, and machine learning techniques: A comprehensive review and outlook. Sci Total Environ. 2022 Jan 10;803:149834. doi:10.1016/j.scitotenv.2021.149834 Prado T, Fumian TM, Mannarino CF, Resende PC, Motta FC, Eppinghaus ALF, et al. Wastewater-based epidemiology as a useful tool to track SARS-CoV-2 and support public health policies at municipal level in Brazil. Water Res. 2021 Mar 1;191:116810. doi:10.1016/j.watres.2021.116810 Mbunge E, Batani J. Application of deep learning and machine learning models to improve healthcare in sub-Saharan Africa: Emerging opportunities, trends and implications. Telemat Inform Rep. 2023 Sep 1;11:100097. doi:10.1016/j.teler.2023.100097 Lafond KE, Dalhatu I, Shinde V, Ekanem EE, Ahmed S, Peebles P, et al. Notifiable disease reporting among public sector physicians in Nigeria: a cross-sectional survey to evaluate possible barriers and identify best sources of information. BMC Health Serv Res. 2014 Nov 13;14:568. doi:10.1186/s12913-014-0568-3 PubMed PMID: 25391377; PubMed Central PMCID: PMC4233090. González A, Zanin M, Menasalvas E. Public Health and Epidemiology Informatics: Can Artificial Intelligence Help Future Global Challenges? An Overview of Antimicrobial Resistance and Impact of Climate Change in Disease Epidemiology. Yearb Med Inform. 2019 Aug 1;28:224–31. doi:10.1055/s-0039-1677910 Aytan-Aktug D, Clausen P, Bortolaia V, Aarestrup F, Lund O. Prediction of Acquired Antimicrobial Resistance for Multiple Bacterial Species Using Neural Networks. mSystems. 2020;5. doi:10.1128/mSystems.00774-19 Tai AMY, Albuquerque A, Carmona NE, Subramanieapillai M, Cha DS, Sheko M, et al. Machine learning and big data: Implications for disease modeling and therapeutic discovery in psychiatry. Artif Intell Med. 2019 Aug 1;99:101704. doi:10.1016/j.artmed.2019.101704 Gidey K, Beyene H, Haileslassie F, Berihu H. Apprentice bot model design and implementation for psychological clients’ therapy. Discov Appl Sci. 2025 Feb 24;7(3):170. doi:10.1007/s42452-025-06515-2 Oluwadiya K, Adeoti A, Nottidge T. Exploring Artificial Intelligence in the Nigerian Medical Educational Space: An Online Cross-sectional Study of Perceptions, Risks and Benefits among Students and Lecturers from Ten Universities. Niger Postgrad Med J. 2023 Nov 23;30:285. doi:10.4103/npmj.npmj_186_2 De-Arteaga M, Herlands W, Neill DB, Dubrawski A. Machine Learning for the Developing World. ACM Trans Manag Inf Syst. 2018 Jun 30;9(2):1–14. doi:10.1145/3210548 Isiaka AB, Anakwenze VN, Ilodinso CR, Anaukwu CG, Ezeokoli CMV, Noi SM, et al. Harnessing Artificial Intelligence for Early Detection and Management of Infectious Disease Outbreaks. Int J Innov Res Dev. 2024. doi:10.24940/ijird/2024/v13/i2/feb24016 Naseem M, Akhund R, Arshad H, Ibrahim MT. Exploring the Potential of Artificial Intelligence and Machine Learning to Combat COVID-19 and Existing Opportunities for LMIC: A Scoping Review. J Prim Care Community Health. 2020 Jan 1;11:2150132720963634. doi:10.1177/2150132720963634 Okagbue HI, Oguntunde PE, Obasi ECM, Adamu PI, Opanuga AA. Diagnosing malaria from some symptoms: a machine learning approach and public health implications. Health Technol. 2021 Jan 1;11(1):23–37. doi:10.1007/s12553-020-00488-5 Ajuwon BI, Richardson A, Roper K, Sheel M, Audu R, Salako BL, et al. The development of a machine learning algorithm for early detection of viral hepatitis B infection in Nigerian patients. Sci Rep. 2023 Feb 24;13(1):3244. doi:10.1038/s41598-023-30440-2 Quezada DA, Akwafuo S, Halyal S. Harnessing Machine Learning for Predictive Analytics: A Case Study of Lassa Fever Outbreaks in Nigeria. In: 2024 10th International Conference on Control, Decision and Information Technologies (CoDIT) [Internet]. 2024 [cited 2025 Feb 22]. p. 377–82. Available from: https://ieeexplore.ieee.org/document/10708095 doi:10.1109/CoDIT62066.2024.10708095 Tian X, Xu Q, Wang Y. Prediction of Meningitis Outbreaks in Nigeria Using Machine Learning Algorithms. Proc 2019 2nd Artif Intell Cloud Comput Conf. 2019. doi:10.1145/3375959.3375984 Nwankwo E, Emeihe E, Ajegbile M, Olaboye J, Maha C. Artificial Intelligence in predictive analytics for epidemic outbreaks in rural populations. Int J Life Sci Res Arch. 2024 Aug 8;07:78–094. doi:10.53771/ijlsra.2024.7.1.0062 Muremyi R, Haughton D, Kabano I, Niragire F. Prediction of out-of-pocket health expenditures in Rwanda using machine learning techniques. Pan Afr Med J. 2020;37. doi:10.11604/pamj.2020.37.357.27287 Davies G, Mudele O, Wright-Ajayi B, Mosaku L, Moneke K, Shaba D, et al. Clinical Case Reports International Enhancing Infectious Disease Management in Nigeria: The Role of Artificial Intelligence in Diagnosis and Treatment Review Article. Clin Case Rep Int. 2024 Mar 18;8:3. Umar AB, Sani SK, Aliyu LJ, Hassan M, Imam M, Haruna UA, et al. Enhancing primary healthcare delivery in Nigeria through the adoption of advanced technologies. Narra X. 2024 Dec 31;2(3):3. doi:10.52225/narrax.v2i3.180 Olaronke I, Gambo I, Ikono R, Olaleke J. Interoperability in Nigeria Healthcare System: The Ways Forward. Int J Inf Eng Electron Bus. 2013 Oct 1;5:16–23. doi:10.5815/ijieeb.2013.04.03 Adewole KS, Alozie E, Olagunju H, Faruk N, Aliyu RY, Imoize AL, et al. A systematic review and meta-data analysis of clinical data repositories in Africa and beyond: recent development, challenges, and future directions. Discov Data. 2024 Jun 26;2(1):8. doi:10.1007/s44248-024-00012-4 Jussen-Lengersdorf I, Schweihoff J, Möller F. Tensions in Inter-Organizational Data Sharing: Findings from Literature and Practice. 2023. doi:10.1109/CBI58679.2023.10187530 Achieng M, Ogundaini O. Big Data Analytics for Integrated Infectious Disease Surveillance in sub-Saharan Africa. SA J Inf Manag. 2024 Jan 30;26:1–11. doi:10.4102/sajim.v26i1.1668 Akintola S. Navigating the Legal and Ethical Terrain of Artificial Intelligence in Enhancing Patient Safety in Nigeria [Internet]. 2023 Nov 27 [cited 2025 Feb 22]. Available from: https://www.academia.edu/124846767/Navigating_the_Legal_and_Ethical_Terrain_of_Artificial_Intelligence_in_Enhancing_Patient_Safety_in_Nigeria Olayiwola K, Opele J. PRACTICE OF HEALTH INFORMATION MANAGEMENT IN NIGERIA: OPPORTUNITIES, CHALLENGES, AND FUTURE DIRECTIONS. Vol. 16. 2024 May 30;16:2023. Prabhakaran V, Martin D. Participatory Machine Learning Using Community-Based System Dynamics. Health Hum Rights. 2020 Dec;22(2):71–4. PubMed PMID: 33390696; PubMed Central PMCID: PMC7762892. Olumade TJ, Adesanya OA, Fred-Akintunwsa IJ, Babalola DO, Oguzie JU, Ogunsanya OA, et al. Infectious disease outbreak preparedness and response in Nigeria: history, limitations and recommendations for global health policy and practice. AIMS Public Health. 2020 Sep 22;7(4):736–57. doi:10.3934/publichealth.2020057 PubMed PMID: 33294478; PubMed Central PMCID: PMC7719556. Li X, Abangbila L. Resistance to medical artificial intelligence: Integrating AI awareness, AI risks, and displacement of responsibility. J Infrastruct Policy Dev. 2024 Oct 17;8(11):11. doi:10.24294/jipd.v8i11.7923 Mumtaz H, Riaz MH, Wajid H, Saqib M, Zeeshan MH, Khan SE, et al. Current challenges and potential solutions to the use of digital health technologies in evidence generation: a narrative review. Front Digit Health. 2023 Sep 28;5:1203945. doi:10.3389/fdgth.2023.1203945 PubMed PMID: 37840685; PubMed Central PMCID: PMC10568450. Sewada R, Jangid A, Kumar P, Mishra N. Explainable Artificial Intelligence (XAI). Int J Food Nutr Sci. 2023. doi:10.48047/ijfans/v12/i1/271 Chaddad A, Peng J, Xu J, Bouridane A. Survey of Explainable AI Techniques in Healthcare. Sensors. 2023;23. doi:10.3390/s23020634 Varghese A, Varghese J, Biju J, Thomas R, Thomas M. Explainable AI in Healthcare Applications. Int Res J Adv Eng Manag IRJAEM. 2024. doi:10.47392/irjaem.2024.0545 Cina G, Rober TE, Goedhart R, Birbil I. Why we do need Explainable AI for Healthcare. ArXiv. 2022;abs/2206.15363. doi:10.48550/arXiv.2206.15363 Amann J, Blasimme A, Vayena E, Frey D, Madai VI, the Precise4Q consortium. Explainability for artificial intelligence in healthcare: a multidisciplinary perspective. BMC Med Inform Decis Mak. 2020 Nov 30;20(1):310. doi:10.1186/s12911-020-01332-6 Alkhanbouli R, Matar Abdulla Almadhaani H, Alhosani F, Simsekler MCE. The role of explainable artificial intelligence in disease prediction: a systematic literature review and future research directions. BMC Med Inform Decis Mak. 2025 Mar 4;25(1):110. doi:10.1186/s12911-025-02944-6 PubMed PMID: 40038704; PubMed Central PMCID: PMC11877768. Amannah C, Attai KF, Uzoka FM. A Data-Driven Intelligent Methodology for Developing Explainable Diagnostic Model for Febrile Diseases. Algorithms. 2025 Apr;18(4):190. doi:10.3390/a18040190 Attai K, Ekpenyong M, Amannah C, Asuquo D, Ajuga P, Obot O, et al. Enhancing the Interpretability of Malaria and Typhoid Diagnosis with Explainable AI and Large Language Models. Trop Med Infect Dis. 2024 Sep 16;9(9):216. doi:10.3390/tropicalmed9090216 PubMed PMID: 39330905; PubMed Central PMCID: PMC11436130. Esan AO, Olawade DB, Soladoye AA, Omodunbi BA, Adeyanju IA, Aderinto N. Explainable AI for Parkinson’s disease prediction: A machine learning approach with interpretable models. Curr Res Transl Med. 2025 Dec;73(4):103541. doi:10.1016/j.retram.2025.103541 PubMed PMID: 40945155. Attai KF, Amannah C, Ekpenyong M, Asuquo DE, Akputu OK, Obot OU, et al. Developing an Explainable Artificial Intelligence System for the Mobile-Based Diagnosis of Febrile Diseases Using Random Forest, LIME, and GPT. Healthc Inform Res. 2025 Apr 30;31(2):125–35. doi:10.4258/hir.2025.31.2.125 Attai KF, Amannah C, Ekpenyong M, Baadel S, Obot O, Asuquo D, et al. Predicting Predisposition to Tropical Diseases in Female Adults Using Risk Factors: An Explainable-Machine Learning Approach. Information. 2025 Jul;16(7):520. doi:10.3390/info16070520 Ahmed A, Oluomachi E, Abdullah A, Tochukwu N. Enhancing Data Privacy In Wireless Sensor Networks: Investigating Techniques And Protocols To Protect Privacy Of Data Transmitted Over Wireless Sensor Networks In Critical Applications Of Healthcare And National Security. Int J Netw Secur Its Appl. 2024 Mar 29;16(2):47–63. doi:10.5121/ijnsa.2024.16204 Salih AM, Raisi‐Estabragh Z, Galazzo IB, Radeva P, Petersen SE, Lekadir K, et al. A Perspective on Explainable Artificial Intelligence Methods: SHAP and LIME. Adv Intell Syst. 2025 Jan;7(1):2400304. doi:10.1002/aisy.202400304 Mujawar S, Sayed M, Chaware A. Empowering Decision Making in Healthcare - A Comparative Analysis of eXplainable AI Techniques Using SHAP and LIME for Cancer Patients Data. In: 2024 3rd Edition of IEEE Delhi Section Flagship Conference (DELCON) [Internet]. 2024 [cited 2025 May 19]. p. 1–5. Available from: https://ieeexplore.ieee.org/abstract/document/10867049 doi:10.1109/DELCON64804.2024.10867049 Ibrahim AM, Ahmed MM, Musa SS, Haruna UA, Hamid MR, Okesanya O, et al. Leveraging AI for early cholera detection and response: transforming public health surveillance in Nigeria. Explor Digit Health Technol. 2025. doi:10.37349/edht.2025.101140 Olayinka TA, Waghid Z, Mataka TW, Adegoke O. Demystifying Lesotho, Rwandan and Nigerian educators’ viewpoints on smart technologies supporting AI in higher education. Educ Inf Technol. 2024;29:24285–307. doi:10.1007/s10639-024-12820-x Puijenbroek E van, Barry A, Khaemba C, Ntirenganya L, Gebreyesus TD, Fimbo A, et al. Short-Term Training, a Useful Approach for Sustainable Pharmacovigilance Knowledge Development in Tanzania, Kenya, Ethiopia and Rwanda. Drug Saf. 2024;47:1193–202. doi:10.1007/s40264-024-01469-7 Townsend B, Sihlahla I, Naidoo M, Naidoo S, Donnelly D, Thaldar D. Mapping the regulatory landscape of AI in healthcare in Africa. Front Pharmacol. 2023;14. doi:10.3389/fphar.2023.1214422 Daroudi R, Sari AA, Nahvijou A, Faramarzi A. Cost per DALY averted in low, middle- and high-income countries: evidence from the global burden of disease study to estimate the cost-effectiveness thresholds. Cost Eff Resour Alloc CE. 2021;19. doi:10.1186/s12962-021-00260-0 Sanni AO, Jonker A, Were V, Fasanmi O, Adebowale O, Shittu A, et al. Cost-effectiveness of One Health intervention to reduce risk of human exposure and infection with non-typhoidal salmonellosis (NTS) in Nigeria. One Health. 2024;18. doi:10.1016/j.onehlt.2024.100703 Neumann P, Anderson JE, Panzer A, Pope E, D’Cruz BN, Kim DD, et al. Comparing the cost-per-QALYs gained and cost-per-DALYs averted literatures. Gates Open Res. 2018;2. doi:10.12688/gatesopenres.12786.2 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 21 Apr, 2026 Reviewers agreed at journal 21 Apr, 2026 Reviewers invited by journal 01 Apr, 2026 Submission checks completed at journal 28 Mar, 2026 First submitted to journal 25 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8000523","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Systematic Review","associatedPublications":[],"authors":[{"id":628961871,"identity":"d1c7f2f1-37ce-45b8-9f19-bad4f6822426","order_by":0,"name":"Lukman Jibril Aliyu","email":"","orcid":"","institution":"Zipline","correspondingAuthor":false,"prefix":"","firstName":"Lukman","middleName":"Jibril","lastName":"Aliyu","suffix":""},{"id":628961872,"identity":"46716a34-2d45-493b-9255-94df87727181","order_by":1,"name":"Abbas Bashir Umar","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABEUlEQVRIiWNgGAWjYDACCR4GBsYGBgYDECfB4L8ciD7wgGgtHyqYjcFaEojVwjjjDHNiA9g6PDr4Z/ce/Phzh529OXvzsc+8bWzp88MOPwTaYien24DDkjvnkqV5zyQn7uw5ljybt40nd+PtNAOglmRjswM4rLmRYyDN2MacYHAjx5iZt00id+PsBJCWA4nbcGiRB6r8+bOt3t7g/vvPQC0G6Yaz0z/g1QI03EyCt+0w44YbPMxA7yckyEvn4LfF8M65NGveM8cTN5xJMwYG8gHDDdI5BQcSDHD7Re527+GbP3dU2xscP/wYGJUH5OVnp2/+8KHCTg6n9zGdClZpQKxyEJBvIEX1KBgFo2AUjAQAAETlaH8hZXqOAAAAAElFTkSuQmCC","orcid":"","institution":"Ahmadu Bello University","correspondingAuthor":true,"prefix":"","firstName":"Abbas","middleName":"Bashir","lastName":"Umar","suffix":""},{"id":628961873,"identity":"b33efc3c-eb23-47ed-bf92-e306c04f96ff","order_by":2,"name":"Saifuddeen kamfut Sani","email":"","orcid":"","institution":"National Agency for Food and Drug Administration and control","correspondingAuthor":false,"prefix":"","firstName":"Saifuddeen","middleName":"kamfut","lastName":"Sani","suffix":""},{"id":628961874,"identity":"167bd884-0e59-4064-9235-a263a9a124d1","order_by":3,"name":"Usman Abubakar Haruna","email":"","orcid":"","institution":"Ahmadu Bello University","correspondingAuthor":false,"prefix":"","firstName":"Usman","middleName":"Abubakar","lastName":"Haruna","suffix":""},{"id":628961875,"identity":"4fe3d771-68ed-458a-aa7f-5480a38355de","order_by":4,"name":"Ahmed Haruna Danjuma","email":"","orcid":"","institution":"Society for Family Health","correspondingAuthor":false,"prefix":"","firstName":"Ahmed","middleName":"Haruna","lastName":"Danjuma","suffix":""},{"id":628961876,"identity":"83ea0376-ec4e-49ed-8ec1-99c004c97066","order_by":5,"name":"Mubarak Zubairu","email":"","orcid":"","institution":"UNICEF","correspondingAuthor":false,"prefix":"","firstName":"Mubarak","middleName":"","lastName":"Zubairu","suffix":""},{"id":628961877,"identity":"c4844581-2efa-4b6a-8de3-1f8aff23ba87","order_by":6,"name":"Hafsat Yahaya","email":"","orcid":"","institution":"Specialist Hospital","correspondingAuthor":false,"prefix":"","firstName":"Hafsat","middleName":"","lastName":"Yahaya","suffix":""}],"badges":[],"createdAt":"2025-10-31 17:08:18","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8000523/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8000523/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108609517,"identity":"3d870b69-681b-401b-9923-d6ba0c367ac7","added_by":"auto","created_at":"2026-05-06 12:52:47","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":60487,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart for machine learning pipeline (Using Synthetic Data)\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8000523/v1/732e815522b2a010fe9ad4c0.jpg"},{"id":108609518,"identity":"51062a44-e4b6-44a0-b513-d4bd58d99716","added_by":"auto","created_at":"2026-05-06 12:52:47","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":45577,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSHAP Summary of Feature Effects on Malaria Predictions\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8000523/v1/dd9fd7d93b938433aa0f5685.jpg"},{"id":109067839,"identity":"1dfea85a-0005-4c54-a557-8c140fba0299","added_by":"auto","created_at":"2026-05-12 10:01:33","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":195212,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eProposed ML Integration Workflow for Nigeria’s IDSR System\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8000523/v1/cc1e54b1914e8075d585a8b0.jpg"},{"id":109069281,"identity":"95c708ff-e432-490e-aae9-92174191fb19","added_by":"auto","created_at":"2026-05-12 10:22:17","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":857059,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8000523/v1/b9d036d6-c457-4352-b1a9-ed2f62c25833.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Enhancing Disease Surveillance in Nigeria through Machine Learning: Opportunities, Challenges and Strategic Recommendations","fulltext":[{"header":"1.0 Introduction","content":"\u003cp\u003eDisease surveillance, the systematic collection, analysis, and interpretation of health data, serves as a cornerstone of public health by enabling early outbreak detection, resource allocation, and policy formulation\u0026nbsp;[1]. In Nigeria, traditional surveillance systems primarily rely on passive reporting mechanisms, often leading to delays in outbreak detection and response\u0026nbsp;[2]. The Integrated Disease Surveillance and Response (IDSR) framework, adopted in 1998, marked a significant step toward a more structured surveillance approach\u0026nbsp;[3]. However, its effectiveness is hindered by challenges such as inconsistent data quality, limited private sector participation, and infrastructural constraints, including inadequate digital health infrastructure and fragmented reporting channels.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eArtificial Intelligence (AI) is increasingly recognized as a transformative tool for addressing these challenges. Machine Learning (ML), a subset of AI, enables predictive analytics, real-time data processing, and automated pattern recognition, making it particularly useful for enhancing disease surveillance\u0026nbsp;[4]. The application of ML in public health has been demonstrated in various contexts, including the prediction of epidemic outbreaks, syndromic surveillance, and antimicrobial resistance monitoring\u0026nbsp;[5]. ML models have been successfully deployed in resource-limited settings to predict malaria outbreaks and detect emerging infectious diseases, highlighting their potential for Nigeria\u0026rsquo;s healthcare system.\u003c/p\u003e\n\u003cp\u003eThe COVID-19 pandemic underscored the critical role of ML in tracking disease spread, optimizing resource allocation, and improving diagnostic accuracy\u0026nbsp;[6].\u0026nbsp;Digital/AI-assisted contact tracing in South Korea and Taiwan significantly reduced delays in case identification and quarantine, with Taiwan achieving a median response time of ~3 days from exposure to isolation, demonstrating substantial improvements over traditional manual tracing approaches [7,8]. In India, AI-driven syndromic surveillance flagged COVID-19 hotspots earlier before conventional reporting systems triggered alerts [9]. Within Africa, Rwanda\u0026rsquo;s AI-enhanced surveillance infrastructure enabled real-time tracking of COVID-19 case clusters, facilitating swift quarantine measures [10]. Despite these global advances, Nigeria\u0026rsquo;s healthcare system faces multiple barriers to fully adopting these innovations, including outdated health information systems, workforce capacity gaps, limited internet connectivity in rural areas, and uneven technological adoption across regions [2,11]. In contrast to prior studies that primarily focus on ML in high-income contexts or generalized African settings [6], this paper review addresses Nigeria\u0026rsquo;s unique structural challenges, including a pronounced urban-rural digital divide affecting approximately 60% of the population [11] and fragmented health data ecosystems, and proposes locally adaptable ML solutions to enhance IDSR efficacy.\u003c/p\u003e\n\u003col\u003e\n \u003cli\u003e\u003cstrong\u003eData Sources and Review Methodology\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThis study adopted a narrative review approach to explore the application of artificial intelligence (AI) and machine learning (ML) in disease surveillance systems, with a particular focus on Nigeria and other low- and middle-income countries (LMICs). A comprehensive literature search was conducted across multiple electronic databases, including PubMed, Scopus, Web of Science, and Google Scholar.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eSearch terms combined keywords and Boolean operators such as \u003cem\u003e\u0026ldquo;artificial intelligence,\u0026rdquo; \u0026ldquo;machine learning,\u0026rdquo; \u0026ldquo;disease surveillance,\u0026rdquo; \u0026ldquo;public health surveillance,\u0026rdquo; \u0026ldquo;Nigeria,\u0026rdquo;\u003c/em\u003e and \u003cem\u003e\u0026ldquo;LMICs.\u0026rdquo;\u003c/em\u003e The search was limited to studies published between 2015 and 2025 to capture recent advancements in AI-driven surveillance technologies.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eA total of 87 sources were included in the final review. These comprised both peer-reviewed journal articles (approximately 69) and grey literature (approximately 18), including conference proceedings, technical reports, and policy documents. The inclusion of grey literature was intentional to capture emerging innovations, policy perspectives, and real-world implementations that may not yet be fully represented in peer-reviewed publications.\u003c/p\u003e\n\u003cp\u003eStudies were included if they:\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003eExamined the application of AI or ML in disease surveillance, diagnosis, or outbreak prediction\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eFocused on public health systems in Nigeria or comparable LMIC settings\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eProvided empirical findings, technical insights, or policy-relevant discussions\u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eExclusion criteria included:\u003c/p\u003e\n\u003cul type=\"disc\"\u003e\n \u003cli\u003eStudies not related to public health or disease surveillance\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eArticles lacking sufficient methodological detail\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eNon-English publications\u0026nbsp;\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eData extraction focused on key variables such as study design, geographic setting, type of AI/ML approach used, data sources, and reported outcomes. A thematic synthesis approach was employed to identify recurring patterns, opportunities, and challenges related to AI integration in surveillance systems.\u003c/p\u003e"},{"header":"2.0 Current State of Machine Learning in Nigeria","content":"\u003cp\u003eNigeria's disease surveillance system is vital for early outbreak detection and management, transitioning from disjointed, paper-based reporting to a unified approach within the IDSR framework. Major epidemics, like the 1986/87 yellow fever outbreak, prompted the creation of formal notification systems, leading to the adoption of IDSR in 1998 to coordinate efforts and improve reporting methods [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSurveillance operates through a multi-tiered system where data flows from primary health facilities to Local Government Areas (LGAs), then to State Ministries of Health and the Nigeria Centre for Disease Control (NCDC) [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Disease Surveillance and Notification Officers (DSNOs) play a key role, often working under resource constraints, including insufficient funding, a shortage of trained personnel, and logistical challenges in remote areas. While public facilities follow reporting protocols, private sector participation remains limited, affecting data completeness and real-time disease tracking [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHistorically, manual reporting led to delays and inconsistencies, but electronic systems, such as the electronic Integrated Disease Surveillance and Response (eIDSR) pilot, have improved reporting timeliness from 43% to 73%, with a near doubling of reporting facilities [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. However, challenges persist, including incomplete documentation, lack of interoperability between health data systems, and disparities in adoption between urban and rural areas. Some regions have successfully embraced digital tools, while others continue to rely on outdated methods, leading to gaps in national-level disease monitoring and response [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eML applications in Nigeria\u0026rsquo;s surveillance system remain in the early stages. While pilot projects and research initiatives have demonstrated the feasibility of AI-powered disease forecasting and automated diagnostics, full-scale implementation is hindered by infrastructure gaps, unreliable internet connectivity, and low digital literacy among healthcare workers [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Moreover, issues such as ethical concerns, data privacy regulations, and limited investment in AI-driven health technologies further constrain the adoption of ML in disease surveillance [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDespite these challenges, there have been positive developments. Initiatives like the deployment of AI-enhanced diagnostic tools for tuberculosis detection and predictive models for malaria outbreaks have showcased the potential of ML in improving Nigeria\u0026rsquo;s public health surveillance system [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. For instance, a 2022 pilot in Lagos using ML to predict malaria outbreaks achieved a 78% accuracy rate but was limited to urban centers with reliable internet, covering only 12% of the state\u0026rsquo;s health facilities [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Similarly, the Auto-Visual Acute Flaccid Paralysis Detection and Reporting (AVADAR) system increased polio case reporting by 45% across 10 states from 2017\u0026ndash;2020, yet its reliance on mobile networks excluded 30% of rural areas [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. These examples highlight scalability constraints tied to infrastructure and funding. Furthermore, collaborations between research institutions, technology firms, and government agencies are gradually paving the way for more structured AI integration in disease monitoring.\u003c/p\u003e \u003cp\u003eGoing forward, addressing the current limitations will require stronger policy frameworks, increased investment in health-tech infrastructure, and training programs to enhance AI literacy among healthcare professionals. Integrating ML with Nigeria\u0026rsquo;s existing digital health strategies can significantly improve disease surveillance, providing timely and accurate health intelligence to combat future outbreaks effectively.\u003c/p\u003e"},{"header":"3.0 Machine Learning Methodologies for Disease Surveillance","content":"\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Supervised vs. Unsupervised Learning\u003c/h2\u003e \u003cp\u003eML in disease surveillance can be divided into two major paradigms:\u003c/p\u003e \u003cp\u003e \u003cb\u003eSupervised learning\u003c/b\u003e involves training a model on labeled datasets (e.g., previous outbreaks). It is useful for case classification, syndromic pattern recognition, and predictive modeling for diseases like cholera and Lassa fever [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cb\u003eUnsupervised learning\u003c/b\u003e is applied where outcomes are unknown, allowing for clustering of cases or anomaly detection, especially useful for novel syndromes or underreported diseases.\u003c/p\u003e \u003cp\u003eBoth methods offer complementary benefits to Nigeria\u0026rsquo;s Integrated Disease Surveillance and Response (IDSR) strategy, enhancing early detection and prediction of infectious disease patterns [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Relevant ML Architectures\u003c/h2\u003e \u003cp\u003eSeveral ML models have proven valuable for health surveillance:\u003c/p\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003e3.2.1 Decision Trees \u0026amp; Random Forests\u003c/h2\u003e \u003cp\u003eDecision Trees (DTs) and Random Forests (RFs) are foundational classifiers in healthcare analytics due to their interpretability and robustness with structured data. They\u0026rsquo;re effective in predicting disease risk based on epidemiological variables and have been frequently used for fall detection, infection diagnosis, and chronic disease monitoring. For instance, Random Forests have been used to classify patients based on wearable sensor data, offering real-time health risk detection [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003e3.2.2 Support Vector Machines (SVMs)\u003c/h2\u003e \u003cp\u003eSVMs are particularly suited to high-dimensional clinical and demographic data. Their margin-maximizing properties make them robust classifiers for binary and multi-class disease prediction tasks, such as early detection of infections from structured hospital records or real-time social media streams [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003e3.2.3 Recurrent Neural Networks (RNNs)\u003c/h2\u003e \u003cp\u003eRNNs excel in modeling temporal health data such as daily syndromic surveillance or weekly IDSR reports. Their capacity for retaining sequential dependencies is useful in forecasting outbreaks and patient deterioration patterns over time [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e3.2.4 Convolutional Neural Networks (CNNs)\u003c/h2\u003e \u003cp\u003eThough CNNs originated in image recognition, they\u0026rsquo;ve been adapted for spatial epidemiology. By treating spatial disease data as \u0026ldquo;images,\u0026rdquo; CNNs can learn local transmission patterns and hotspot dynamics. CNNs have also been used to extract features from wearable sensor data and medical records in predictive health modeling [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThese models form the backbone of contemporary health surveillance systems. Each architecture addresses unique data characteristics, from structured demographic inputs to temporal sequences and spatial epidemiological grids. When integrated, they provide comprehensive analytical capabilities. Moreover, advancements in Neural Networks (NNs) and Generative Adversarial Networks (GANs) further extend this potential, enabling simulation of disease spread scenarios, synthetic data generation for rare outbreaks, and the modeling of uncertainty in predictions [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. Together, these models offer a robust analytical framework for extracting actionable insights from increasingly complex and high-dimensional epidemiological data landscapes.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Model Selection, Training-Validation Split, and Cross-Validation\u003c/h2\u003e \u003cp\u003eModel selection is a critical step in machine learning pipelines, especially in healthcare and epidemiological modeling. It often begins with simple, interpretable models such as logistic regression and progresses to more complex techniques like ensemble methods (e.g., Random Forests) or deep learning models. These methods are selected based on their predictive performance, interpretability, and appropriateness for the data characteristics. Exploratory approaches like learning curve-based cross-validation (LCCV) have been shown to streamline model selection by eliminating poorly performing models early, thus improving efficiency without sacrificing accuracy [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cb\u003eTraining-validation splitting\u003c/b\u003e, commonly using an 80:20 ratio, helps evaluate a model\u0026rsquo;s ability to generalize to unseen data. A balanced split is vital: too little training data hampers learning, while too little validation data reduces evaluation reliability. Studies have found that optimal split ratios vary by dataset size, but the general principle holds, careful partitioning improves performance estimation and model robustness [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cb\u003eCross-validation (CV)\u003c/b\u003e methods like k-fold CV, time-series CV, and bootstrap-based CV are crucial when data is scarce or disease events are rare. K-fold CV (often 5- or 10-fold) provides a more robust performance estimate by rotating training and validation sets, reducing the risk of overfitting to a particular split [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Time-series CV is especially important in epidemiology where data has temporal dependencies. Additionally, hybrid methods combining bootstrap and CV (e.g., kCV-B) enhance generalizability in deep learning applications by improving training data diversity and reducing overfitting [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eSuch rigorous model evaluation strategies ensure robust, unbiased, and generalizable machine learning models, which is essential for applications in Nigeria\u0026rsquo;s complex and evolving epidemiological landscape.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Global AI-driven Disease Surveillance Systems: A Comparative Overview\u003c/h2\u003e \u003cp\u003eIn addition to the supervised and unsupervised ML methods discussed, it is crucial to contextualize Nigeria\u0026rsquo;s disease surveillance approach by examining state-of-the-art AI-driven systems globally. Prominent examples include:\u003c/p\u003e \u003cp\u003e \u003cb\u003e3.4.1 HealthMap\u003c/b\u003e: Utilizes real-time scraping of news media, social media, and official health reports worldwide to detect outbreaks early [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. It employs natural language processing and anomaly detection algorithms to provide up-to-date global alerts.\u003c/p\u003e \u003cp\u003e \u003cb\u003e3.4.2 ProMED-mail\u003c/b\u003e: A reporting platform that aggregates information from clinicians, media, and public health officials using expert review and crowd-sourced reports to monitor emerging infectious diseases worldwide [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Its approach combines manual curation with automated text processing.\u003c/p\u003e \u003cp\u003e \u003cb\u003e3.4.3 Epidemic Intelligence from Open Sources (EIOS) by WHO\u003c/b\u003e: Integrates multiple open-source data streams, including news, social media, and official reports, using machine learning and data fusion techniques to support epidemic intelligence and risk assessment [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eCompared to these systems, Nigeria\u0026rsquo;s IDSR framework currently relies heavily on manual reporting, limited private sector integration, and lacks scalable real-time analytics and diverse data sourcing. This overview highlights key gaps such as delayed reporting, limited automation for signal detection, and data fragmentation. Adoption of similar methodologies, including natural language processing and multi-source fusion, would significantly enhance the effectiveness and timeliness of surveillance in Nigeria.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Incorporating Advanced AI: Generative Models and Large Language Models\u003c/h2\u003e \u003cp\u003eRecent advances in large language models (LLMs) and generative AI offer promising opportunities to enhance disease surveillance. Applications include synthetic data generation to augment limited datasets, automated generation of epidemiological reports, and decision support tools that aid in complex outbreak management.\u003c/p\u003e \u003cp\u003eWhile this study focuses on traditional machine learning approaches, primarily due to the current constraints in digital infrastructure and data availability in Nigeria, we recognize the transformative potential of these advanced methods. Future system updates could integrate LLMs and generative AI to improve data quality, automate routine tasks, and provide advanced interpretability and support for health decision-makers as infrastructure and data ecosystems mature.\u003c/p\u003e \u003c/div\u003e"},{"header":"4.0 Simulation-Based Illustration of a Malaria Surveillance Modeling Workflow","content":"\u003cp\u003eTo demonstrate the proposed framework, we present a simulation-based case study of malaria surveillance in Northern Nigeria. Due to limited access to high-resolution, longitudinal surveillance data, a synthetic dataset was constructed using plausible relationships between environmental variables (rainfall, temperature), health system access, and malaria incidence.\u003c/p\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Synthetic Data Generation\u003c/h2\u003e \u003cp\u003eWe generated a synthetic weekly dataset for malaria surveillance covering 10 representative Local Government Areas (LGAs) in Northern Nigeria over two years (2023\u0026ndash;2024). Each record contains the year, week number, LGA name, malaria case count, total rainfall (mm), average temperature (\u0026deg;C), and a health access index. The data were designed to reflect regional patterns: for example, a pronounced rainy season mid-year (peaking around July\u0026ndash;September) with largely dry conditions in the winter months, and consistently high temperatures (25\u0026ndash;32\u0026deg;C) typical of tropical Sahel regions [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Each LGA was assigned a Health Access Index (0 to 1 scale) to represent local healthcare infrastructure and access (higher in urban centers, lower in rural areas), acknowledging the North\u0026rsquo;s limited infrastructure and digital divide. Because robust real-world data from many northern Nigerian districts are scarce mainly due to challenges like poor internet connectivity and limited digital infrastructure in rural areas [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e], we employed synthetic data to enable this simulation without privacy or logistics concerns. Notably, synthetic data can augment or stand in for real health datasets, helping address data access and privacy issues in health research [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. In our simulation, weekly malaria cases were modeled as a function of that week\u0026rsquo;s rainfall, temperature, and health access level, plus a minor random noise term to imitate unexplained variability. We assumed higher rainfall would lead to increased malaria cases following a short lag (due to mosquito breeding after rains) [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e], while higher health access was assumed to reduce malaria incidence by improving prevention (e.g. bed net distribution) and treatment availability. Temperature was included as a secondary factor (malaria transmission is optimal in a moderate temperature range). This case study is designed to illustrate the modeling workflow, feature integration, and interpretability techniques, rather than to provide empirical evidence of predictive performance in real-world settings.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Pseudocode for Synthetic Malaria Case Simulation and Modeling\u003c/h2\u003e \u003cp\u003e \u003cstrong\u003eAlgorithm\u003c/strong\u003e \u003cp\u003e \u003cb\u003eSynthetic Malaria Incidence Simulation and Prediction\u003c/b\u003e \u003c/p\u003e \u003c/p\u003e \u003cp\u003eStep 1: Initialize Simulation Parameters\u003c/p\u003e \u003cp\u003eDefine the number of Local Government Areas (LGAs)\u0026thinsp;=\u0026thinsp;10.\u003c/p\u003e \u003cp\u003eDefine the temporal span\u0026thinsp;=\u0026thinsp;104 weeks (representing 2023\u0026ndash;2024).\u003c/p\u003e \u003cp\u003eDefine feature ranges and distributions:\u003c/p\u003e \u003cp\u003eRainfall (mm per week): sample from Uniform [0, 300].\u003c/p\u003e \u003cp\u003eTemperature (\u0026deg;C): sample from Uniform[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHealth Access Index (unitless): sample from Uniform[0.2, 1.0].\u003c/p\u003e \u003cp\u003eStep 2: Define Coefficients (epidemiologically motivated)\u003c/p\u003e \u003cp\u003eRainfall coefficient\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.15 (positive effect: higher rainfall increases malaria risk).\u003c/p\u003e \u003cp\u003eTemperature coefficient\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.05 (small positive effect: higher temperature slightly increases malaria risk).\u003c/p\u003e \u003cp\u003eHealth Access coefficient = \u0026minus;\u0026thinsp;0.25 (negative effect: better access reduces malaria incidence).\u003c/p\u003e \u003cp\u003eNoise term\u0026thinsp;=\u0026thinsp;Gaussian distribution with mean\u0026thinsp;=\u0026thinsp;0 and standard deviation\u0026thinsp;=\u0026thinsp;2 (represents unexplained weekly variability).\u003c/p\u003e \u003cp\u003eStep 3: Generate Synthetic Dataset\u003c/p\u003e \u003cp\u003eFor each LGA i from 1 to 10:\u003c/p\u003e \u003cp\u003ea. Assign a fixed Health Access Index value from Uniform[0.2, 1.0].\u003c/p\u003e \u003cp\u003eb. For each week t from 1 to 104:\u003c/p\u003e \u003cp\u003e \u003col style=\"list-style-type:lower-roman;\"\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eSample rainfall value.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eSample temperature value.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eSample Gaussian noise term.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eCompute malaria cases using the function:\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003emalaria_cases = (0.15 \u0026times; rainfall) + (0.05 \u0026times; temperature) + (\u0026ndash;0.25 \u0026times; health_access) + noise\u003c/p\u003e \u003cp\u003ev. Record malaria_cases for LGA i at week t.\u003c/p\u003e \u003cp\u003eSave all records into a dataset containing columns: [LGA, Week, Rainfall, Temperature, Health Access, and Malaria Cases].\u003c/p\u003e \u003cp\u003eStep 4: Train Predictive Model\u003c/p\u003e \u003cp\u003eSplit dataset into:\u003c/p\u003e \u003cp\u003eTraining set: Weeks 1\u0026ndash;52 (2023).\u003c/p\u003e \u003cp\u003eTest set: Weeks 53\u0026ndash;104 (2024).\u003c/p\u003e \u003cp\u003eTrain a Decision Tree regression model with:\u003c/p\u003e \u003cp\u003eMaximum depth\u0026thinsp;=\u0026thinsp;6.\u003c/p\u003e \u003cp\u003eInput features: Rainfall, Temperature, Health Access.\u003c/p\u003e \u003cp\u003eOutput target: Malaria Cases.\u003c/p\u003e \u003cp\u003eJustification for Model Selection:\u003c/p\u003e \u003cp\u003e \u003cem\u003eFor the rationale behind Decision Tree model selection, see\u003c/em\u003e Section \u003cspan refid=\"Sec16\" class=\"InternalRef\"\u003e4.3\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eStep 5: Evaluate Model\u003c/p\u003e \u003cp\u003ePredict malaria cases for the 2024 test set.\u003c/p\u003e \u003cp\u003eCompute evaluation metrics:\u003c/p\u003e \u003cp\u003eR\u0026sup2; (coefficient of determination).\u003c/p\u003e \u003cp\u003eRMSE (root mean square error).\u003c/p\u003e \u003cp\u003eExample output: performance metrics are computed based on the selected validation design.\u003c/p\u003e \u003cp\u003eStep 6: Interpret Model (Explainability)\u003c/p\u003e \u003cp\u003eApply SHAP (SHapley Additive exPlanations) to the trained Decision Tree.\u003c/p\u003e \u003cp\u003eCompute local and global feature contributions.\u003c/p\u003e \u003cp\u003eAnalyze the relative importance of rainfall, temperature, and health access:\u003c/p\u003e \u003cp\u003eRainfall: primary positive driver of predicted cases.\u003c/p\u003e \u003cp\u003eHealth Access: strong negative driver.\u003c/p\u003e \u003cp\u003eTemperature: weaker but consistent positive effect.\u003c/p\u003e \u003cp\u003eStep 7: Export and Deployment\u003c/p\u003e \u003cp\u003eSave the synthetic dataset (CSV), model file, and plots (predicted vs. actual, feature importance, SHAP summary).\u003c/p\u003e \u003cp\u003eIntegrate the model into public health early warning dashboards for malaria surveillance.\u003c/p\u003e \u003cp\u003eEnable iterative updating with real-world surveillance data as they become available.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Predictive Modeling with Decision Trees\u003c/h2\u003e \u003cp\u003eWe trained a Decision Tree regression model to predict weekly malaria case counts based on environmental and health access features. Decision Trees are well-suited for this task as they can capture nonlinear relationships and interactions among predictors [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. Unlike traditional linear models, Decision Trees partition data into subsets based on feature thresholds, which is advantageous for discovering complex patterns in health surveillance data, for example, different rainfall\u0026ndash;malaria relationships above versus below certain rainfall levels [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. The full rationale for Decision Tree selection is presented here; subsequent sections referencing this choice cross-reference this section rather than repeat the justification.\u003c/p\u003e \u003cp\u003e \u003cb\u003eModel Selection Rationale\u003c/b\u003e \u003c/p\u003e \u003cp\u003eDecision Tree (DT) algorithms were selected as the primary modeling approach for several reasons aligned with the resource-constrained context of Nigerian public health:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eInterpretability\u003c/b\u003e: Decision Trees produce transparent decision rules that can be readily understood by public health practitioners and policymakers. Each split in the tree corresponds to a clear threshold (e.g., rainfall\u0026thinsp;\u0026gt;\u0026thinsp;50 mm), facilitating trust and clinical acceptance.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eMinimal Computational Requirements\u003c/b\u003e: Unlike deep learning approaches, DTs require limited computational resources, making them feasible for deployment in settings with constrained hardware and electricity availability.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eSuitability for Structured Data\u003c/b\u003e: DTs perform well with structured clinical and epidemiological data, requiring minimal preprocessing such as scaling or normalization.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eAppropriateness for LMIC Contexts\u003c/b\u003e: The model's simplicity, ease of training, and straightforward deployment make it particularly suitable for low- and middle-income country (LMIC) settings like Nigeria, where technical capacity for maintaining complex models may be limited.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eWhile other models such as Random Forests, Support Vector Machines, or Recurrent Neural Networks could offer potential accuracy improvements, thorough benchmarking is deferred to future work given current infrastructural and data limitations. Literature shows that Decision Trees provide competitive performance in similar epidemiological tasks with the added benefits of simplicity and lower computational demands [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cb\u003eModel Implementation\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe model used continuous input features including rainfall, temperature, and a health access index. It was trained on synthetic data from the first year (2023) and tested on data from the second year (2024) to evaluate temporal generalization. We selected a moderately deep tree with a maximum depth of 6 to balance model complexity while avoiding overfitting to noise.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Model Evaluation\u003c/h2\u003e \u003cp\u003eThe model was evaluated using a temporally separated validation design, with data from 2023 used for training and data from 2024 reserved for testing. This approach was chosen to better reflect real-world deployment conditions, where models must generalize to future, unseen time periods.\u003c/p\u003e \u003cp\u003eUnder this temporal split, the decision tree model achieved an R\u0026sup2; of approximately 0.66 and an RMSE of approximately 3.61 within the synthetic dataset.\u003c/p\u003e \u003cp\u003eThese metrics reflect the model\u0026rsquo;s ability to capture patterns embedded in the simulated data-generating process. However, because the outcome variable was constructed using predefined relationships with the predictors, these values should be interpreted strictly as internal simulation diagnostics, not as indicators of real-world predictive performance.\u003c/p\u003e \u003cp\u003eThe use of temporal validation, rather than random splitting, provides a more conservative and realistic assessment of model behavior in surveillance contexts.\u003c/p\u003e \u003cdiv id=\"Sec18\" class=\"Section3\"\u003e \u003ch2\u003e4.4.1 Multi-Model Benchmarking Comparison\u003c/h2\u003e \u003cp\u003eTo assess whether the Decision Tree model offers a reasonable interpretability-accuracy trade-off, five ML models were evaluated on the same synthetic dataset under identical temporal validation conditions (training: 2023; testing: 2024). Results are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e4\u003c/span\u003e. The findings confirm that while ensemble methods (Random Forest and Gradient Boosting) and SVMs marginally outperform the Decision Tree on test R\u0026sup2;, the differences are modest. This supports the rationale for selecting the Decision Tree as the primary modeling approach in this illustration: its interpretability advantage is substantial, while its accuracy penalty is minimal within this synthetic dataset context.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMulti-Model Performance Comparison Under Temporal Validation (Synthetic Dataset)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e Train\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e Test\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRMSE Test\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMAE Test\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLinear Regression\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.979\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.979\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.54\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDecision Tree (DT)*\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0.987\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.965\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e2.49\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e1.94\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRandom Forest (RF)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.989\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.975\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.09\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.66\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGradient Boosting (GB)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.993\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.975\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.69\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSupport Vector Machine (SVM)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.978\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.979\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.55\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003e*Selected model for this illustration. All models evaluated under identical temporal validation split (training: 2023; testing: 2024) on the synthetic dataset (seed\u0026thinsp;=\u0026thinsp;42). DT\u0026thinsp;=\u0026thinsp;Decision Tree; RF\u0026thinsp;=\u0026thinsp;Random Forest; GB\u0026thinsp;=\u0026thinsp;Gradient Boosting; SVM\u0026thinsp;=\u0026thinsp;Support Vector Machine. Higher R\u0026sup2; and lower RMSE/MAE indicate better fit. Ensemble models show marginal gains over DT at the cost of interpretability.\u003c/em\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo provide a clear and structured overview of the methodological steps undertaken in this case simulation, Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the complete machine learning pipeline for synthetic malaria case prediction. This flowchart visually details the sequential processes, from the initial generation of the synthetic dataset to the final stages of model interpretation, deployment considerations, and iterative improvement. It serves as a visual guide to the algorithmic workflow described in this section, highlighting the interdependencies between data generation, predictive modeling, evaluation, and explainability.\u003c/p\u003e \u003cp\u003eThe overall algorithmic workflow for this simulation is outlined in the pseudocode below, detailing the key steps from data preparation to model interpretation.\u003c/p\u003e\u003cp\u003eBEGIN\u003c/p\u003e\n\u003cp\u003e1. Synthetic Data Generation\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; a. Define 10 representative LGAs in Northern Nigeria\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; b. For each week in 2023–2024:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Generate synthetic features:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; • Rainfall (mm)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; • Temperature (°C)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; • Health Access Index (0–1)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Compute malaria case count using:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; malaria_cases = f(rainfall, temperature, health_access) + random_noise\u003c/p\u003e\n\u003cp\u003e2. Predictive Modeling\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; a. Split dataset:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Training set: 2023\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Test set: 2024\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; b. Train decision tree regression model (max_depth = 6)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Inputs: rainfall, temperature, health access\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Output: predicted malaria case count\u003c/p\u003e\n\u003cp\u003e3. Model Evaluation\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; a. Predict weekly cases for 2024\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; b. Compute evaluation metrics:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - R² (coefficient of determination)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - RMSE (root mean squared error)\u003c/p\u003e\n\u003cp\u003e4. Model Interpretation (Explainability)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; a. Apply SHAP to trained model\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Compute local and global feature contributions\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Analyze importance of rainfall, temperature, and health access\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; - Verify model reasoning aligns with domain knowledge\u003c/p\u003e\n\u003cp\u003e5. Result Export\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; a. Save synthetic dataset (CSV)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; b. Save trained model and SHAP plots\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; c. Document and publish code for reproducibility\u003c/p\u003e\n\u003cp\u003e6. Model Deployment\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; a. Deploy the trained model to predict upcoming malaria trends\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; b. Integrate with public health dashboards or early warning systems\u003c/p\u003e\n\u003cp\u003e7. Monitoring and Feedback\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; a. Monitor predictions vs. actual malaria reports (if real data becomes available)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; b. Flag discrepancies and track model drift\u003c/p\u003e\n\u003cp\u003e8. Iterative Improvement\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; a. Incorporate real-world data when available\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; b. Retrain or fine-tune model with new surveillance data\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; c. Re-evaluate performance and re-deploy updated model\u003c/p\u003e\n\u003cp\u003eEND\u003c/p\u003e\u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e4.5 SHAP-Based Model Interpretability\u003c/h2\u003e \u003cp\u003eTo ensure the model\u0026rsquo;s reasoning aligns with domain knowledge and to bolster trust in the findings, we applied SHapley Additive exPlanations (SHAP) for model interpretability to elucidate feature contributions in malaria case predictions [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. SHAP is an approach to explain machine learning predictions by assigning each feature a contribution value for each individual prediction [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. In essence, SHAP computes how much each input (rainfall, temperature, health access) \u0026ldquo;pushes\u0026rdquo; the model\u0026rsquo;s prediction up or down relative to a baseline prediction. This yields both local explanations (for a specific LGA-week, how did each feature affect the predicted cases?) and global insight into feature importance. We found that the SHAP values corroborated our expectations: rainfall and health access were the most influential features in the model\u0026rsquo;s decisions, while temperature had a comparatively smaller effect. For example, in weeks with high rainfall, SHAP values for the rainfall feature were strongly positive, indicating rainfall was driving the prediction of higher malaria cases. Conversely, LGAs with higher health access showed negative SHAP contributions for that feature, meaning the model predicted fewer cases than it would have if health access were low (all else being equal) \u0026ndash; consistent with the idea that better healthcare access reduces malaria burden. Overall, rainfall emerged as a primary driver of model predictions (reflecting its role in mosquito breeding and seasonality) [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e], and health access was almost equally influential, reflecting substantial differences in case counts between well-served and underserved communities. Temperature had a subtler influence on the model\u0026rsquo;s output, aligning with the notion that temperature in this region is generally always conducive to malaria transmission, with less week-to-week variability. The SHAP analysis thus provided an extra layer of confidence that the model was leveraging meaningful patterns: it highlighted, for instance, that an extremely low health access index in a given LGA contributed as much to high case predictions as a large increase in rainfall would. Such transparent explanations are crucial for deploying AI in healthcare, as they enable public health officials to verify that the model\u0026rsquo;s drivers make sense and to identify conditions associated with high-risk predictions.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e4.6 Public Health Implications\u003c/h2\u003e \u003cp\u003eThis case simulation demonstrates the potential of combining synthetic data and machine learning to enhance disease surveillance in low-resource settings. In northern Nigeria, where routine malaria reporting may be hampered by infrastructural challenges and incomplete data, a data-driven model could act as an early warning system. For instance, if exceptionally heavy rainfall is forecasted for a coming month, the model (once validated on real data) might predict a spike in malaria cases, prompting preemptive distribution of bed nets, insecticides, or malaria test kits to high-risk LGAs. Our decision tree model, while simplified, illustrates how key factors like weather and healthcare access can be synthesized into an actionable prediction. Importantly, the approach is interpretable, which is a critical requirement for adoption in public health contexts. Decision trees and their outputs can be readily understood by non-technical stakeholders [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], and our use of SHAP further ensures that the model\u0026rsquo;s reasoning is transparent and aligned with epidemiological understanding. This addresses a common concern that \u0026ldquo;black box\u0026rdquo; AI models might otherwise face resistance in the health sector. Additionally, by using synthetic data, we have shown a viable path to develop and test AI models without risking patient privacy or waiting for extensive data collection. Synthetic datasets can serve as a proxy when real data are scarce, helping to bridge the gap in settings where the digital divide limits real-time data availability [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. The limitations inherent in a synthetic-data approach, including the assumption of stable relationships between climate variables and malaria incidence, the absence of human behavioral factors and vector control interventions, and the need for real-data recalibration before deployment, are discussed in detail in Section \u003cspan refid=\"Sec21\" class=\"InternalRef\"\u003e4.7\u003c/span\u003e. In practice, as Nigeria\u0026rsquo;s health system strengthens its health information infrastructure, such models could be incrementally updated with real data and used to support malaria control programs.\u003c/p\u003e \u003cp\u003eWhile the synthetic case study demonstrates how environmental and health system variables can be integrated into an interpretable machine learning workflow, its primary contribution is conceptual rather than empirical.\u003c/p\u003e \u003cp\u003eThe results illustrate how such a model could be structured to support decision-making, such as identifying potential high-risk periods or location. However, this does not in any way establish real-world predictive accuracy.\u003c/p\u003e \u003cp\u003eIn practice, deployment would require training and validation on routine surveillance data, careful handling of missing data and reporting delays, and evaluation across diverse geographic and epidemiological contexts.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e4.7 Limitations of the Simulation-Based Demonstration\u003c/h2\u003e \u003cp\u003eThe case study presented in this work is based entirely on synthetic data, which introduces several important limitations. First, the outcome variable is generated using assumed relationships with predictor variables, meaning that model performance is partially determined by the structure imposed during simulation. As a result, the reported metrics do not reflect independent epidemiological validation.\u003c/p\u003e \u003cp\u003eSecond, the simulation does not fully capture key challenges present in real-world surveillance systems, including missing data, reporting delays, under-reporting, measurement error, and spatial heterogeneity across locations. Additionally, the model assumes stable relationships between predictors and outcomes, whereas real-world epidemiological processes may exhibit temporal drift and intervention effects.\u003c/p\u003e \u003cp\u003eThird, the use of uniform coefficients across locations simplifies the underlying disease dynamics and may overestimate model generalizability.\u003c/p\u003e \u003cp\u003eAccordingly, the results should be interpreted as a methodological illustration of a machine learning pipeline, rather than as evidence of predictive validity or operational readiness.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003e4.8 Sensitivity Analysis and Model Robustness Considerations\u003c/h2\u003e \u003cp\u003eGiven that the predictive framework presented in this study is intended as a methodological illustration rather than a fully operational or validated model, the sensitivity analysis focuses on the effect of coefficient perturbations on Decision Tree performance under the same temporal validation design. The rainfall coefficient was varied across a plausible epidemiologically motivated range (\u0026plusmn;\u0026thinsp;33% of the base value of 0.15), while all other parameters remained constant. Results are presented in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e5\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSensitivity Analysis-Decision Tree Performance under Rainfall Coefficient Perturbation\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRainfall Coefficient\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e% Change from Base\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e Test\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRMSE Test\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;33%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.922\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.47\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;20%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.945\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.46\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003e0.15 (base)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e0% (reference)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e0.966\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e2.43\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e+\u0026thinsp;20%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.976\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.38\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e+\u0026thinsp;33%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.980\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.42\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eAll results from temporal validation (training: 2023; testing: 2024) on synthetic dataset (seed\u0026thinsp;=\u0026thinsp;42). R\u0026sup2; range across perturbations: 0.922\u0026ndash;0.980, indicating robust model behavior under simulated coefficient uncertainty.\u003c/em\u003e \u003c/p\u003e \u003cp\u003eAcross all tested perturbations, the Decision Tree model maintained stable predictive performance, with R\u0026sup2; values ranging from 0.922 to 0.980 and RMSE values remaining within a narrow band (2.38\u0026ndash;2.47). This suggests that the model\u0026rsquo;s structure, informed by the feature thresholds learned from the training data, is reasonably robust to plausible variations in the assumed coefficient values. In practical implementations using real surveillance data, formal sensitivity analyses incorporating uncertainty in feature measurement and reporting completeness would be essential.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e\u003cb\u003e4.9 Roadmap for Real-World Retrospective Validation\u003c/b\u003e\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"BlockQuote\"\u003e \u003cp\u003eTo transition from methodological illustration to operational utility, future work will focus on retrospective validation using real-world surveillance datasets. Potential data sources include Nigeria\u0026rsquo;s District Health Information System (DHIS2), the National Malaria Data Repository (NMDR), and complementary surveillance outputs from the Nigeria Centre for Disease Control (NCDC), alongside meteorological data.\u003c/p\u003e \u003cp\u003eA retrospective panel dataset will be constructed at the state or local government area (LGA) level, combining weekly malaria case counts with environmental and health system indicators over multiple years.\u003c/p\u003e \u003cp\u003eModel validation will prioritize temporally structured approaches, including holdout testing on future periods and rolling-origin evaluation, to better reflect deployment conditions. Performance will be assessed using metrics such as RMSE, MAE, R\u0026sup2;, and calibration measures, as well as classification-based metrics for outbreak detection where applicable.\u003c/p\u003e \u003cp\u003eAdditional analyses will examine robustness to missing data, reporting delays, and geographic variability. This phased validation approach will provide a more rigorous assessment of model applicability in real-world surveillance settings.\u003c/p\u003e \u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"5.0 Opportunities for Using Machine Learning in Disease Surveillance in Nigeria","content":"\u003cp\u003eMachine learning (ML) offers significant potential to enhance disease surveillance in Nigeria by improving data collection, analysis, and predictive capabilities. Several key opportunities have been identified in the literature.\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Cost-Benefit and Economic Impact Analysis\u003c/h2\u003e \u003cp\u003eAI-driven disease surveillance systems have demonstrated significant economic benefits in various contexts. Operational efficiencies gained through automated data collection and analysis can reduce costs by up to 50%, as evidenced in deployments across India and Rwanda [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Furthermore, early outbreak detection via machine learning enables timely interventions, potentially lowering treatment and outbreak response expenses by 30\u0026ndash;40%. These savings stem from reduced hospitalization rates, shorter outbreak durations, and optimized resource allocation. Applying these findings to Nigeria\u0026rsquo;s context, we anticipate similar cost reductions. For instance, by integrating predictive analytics into the IDSR framework, the government could realize substantial savings alongside improved health outcomes, making ML integration highly cost-effective and sustainable in the long-term.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Enhanced Data Collection and Reporting\u003c/h2\u003e \u003cp\u003eMachine learning (ML) has the potential to revolutionize disease surveillance by enhancing the efficiency and accuracy of data collection. AI-driven mobile applications can minimize human error by automatically capturing and reporting health data in real time. These applications facilitate seamless data transfer from local healthcare providers to centralized public health systems, enabling faster detection of disease outbreaks[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. This approach has been successfully implemented in other developing nations, such as India, where AI-assisted data collection improved tuberculosis case detection rates, leading to earlier interventions and better patient outcomes [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. ML algorithms can also analyze historical data to forecast potential disease outbreaks, allowing for timely interventions and efficient resource allocation. By identifying patterns in past outbreaks, ML models can predict the likelihood of future occurrences and help public health officials implement preventive measures [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. A study in South Africa demonstrated how ML-based predictive models significantly improved early detection of malaria outbreaks, allowing for better vector control and resource distribution [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. This model can be adapted for other diseases, such as HIV, malaria, and COVID-19, in regions with limited healthcare infrastructure in Nigeria.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section2\"\u003e \u003ch2\u003e5.3 Advanced Disease Modelling and Outbreak Detection\u003c/h2\u003e \u003cp\u003eML techniques can be integrated with innovative surveillance methods to enhance outbreak detection. One such method is the analysis of wastewater samples using ML algorithms, which can provide early warnings of viral outbreaks [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. In Brazil, wastewater surveillance detected COVID-19 traces with 85% sensitivity, enabling early interventions [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. However, Nigeria\u0026rsquo;s fragmented sanitation infrastructure where only 32% of urban areas have sewer systems [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] limits direct applicability. Mobile-based sampling from community water sources could offer a viable adaptation, leveraging Nigeria\u0026rsquo;s 87% mobile penetration rate [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Additionally, AI-powered air sampling technologies can help identify and track the presence of airborne pathogens in high-risk areas [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. Deep learning models have been applied in Kenya to monitor airborne tuberculosis pathogens in crowded urban areas, significantly improving early detection and reducing transmission risks [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. These methods can be particularly useful in Nigeria, where densely populated urban centers and limited access to healthcare facilities increase the risk of undetected outbreaks.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003e5.4 Integration with Existing Surveillance Systems\u003c/h2\u003e \u003cp\u003eML can augment existing disease surveillance efforts by providing AI-driven insights that assist healthcare professionals in identifying disease trends and making data-driven decisions [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. In Rwanda, an ML-powered electronic medical record system improved real-time reporting and disease trend analysis, enabling public health authorities to respond to emerging threats more effectively [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAutomated identification and reporting of notifiable diseases can help address challenges such as limited healthcare infrastructure and underreporting by healthcare workers, ultimately strengthening disease surveillance systems [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. A notable example of successful implementation in a developing country is the Auto-Visual Acute Flaccid Paralysis Detection and Reporting (AVADAR) system, which has enhanced polio surveillance across Africa. AVADAR leverages mobile technology and artificial intelligence to support healthcare workers in detecting and promptly reporting suspected polio cases. This innovative approach has proven to be both cost-effective and efficient, enabling real-time data collection and improving the accuracy of surveillance efforts, thereby contributing to more effective disease control and eradication strategies [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. By integrating ML with Nigeria\u0026rsquo;s existing Integrated Disease Surveillance and Response (IDSR) framework, the country can enhance the accuracy and timeliness of disease reporting\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003e5.5 Public Health Applications beyond Infectious Diseases\u003c/h2\u003e \u003cp\u003eBeyond infectious diseases, ML can play a crucial role in addressing broader public health challenges. AI-driven models can analyze patterns in antimicrobial resistance, helping to guide treatment protocols and reduce the spread of resistant infections [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. In Bangladesh, ML models were used to track and predict resistance patterns for common bacterial infections, informing policy decisions and improving antibiotic stewardship [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e].\u003c/p\u003e \u003cp\u003ePredictive models can also aid in identifying risk factors for mental health disorders and informing therapeutic interventions [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]. AI-driven mental health assessment tools have been piloted in Ethiopia with promising results, offering a scalable solution for early diagnosis and treatment planning in resource-limited settings [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. In Nigeria, where mental health services remain underdeveloped, integrating AI into mental health care could improve access to early diagnosis and intervention.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section2\"\u003e \u003ch2\u003e5.6 Workforce Training and Capacity Building\u003c/h2\u003e \u003cp\u003eAn awareness and acceptance of AI among Nigerian healthcare professionals, targeted training programs are essential to enhance ML adoption. Training initiatives can equip healthcare workers with the skills needed to integrate AI-driven solutions into routine disease surveillance and patient care [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. A similar approach in Uganda led to significant improvements in healthcare workers' ability to use AI tools for disease surveillance, ultimately improving health outcomes [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eCollaboration between health institutions, technology companies, and academic researchers is also vital for driving innovation in AI-based disease surveillance [\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. In India, partnerships between universities and AI firms led to the successful deployment of ML models for predicting dengue outbreaks, demonstrating the potential of interdisciplinary cooperation [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. Establishing similar collaborations in Nigeria could accelerate the adoption of AI-driven disease surveillance technologies and enhance the country\u0026rsquo;s ability to respond to public health challenges effectively.\u003c/p\u003e \u003cp\u003eBy leveraging ML for disease surveillance, Nigeria can significantly improve early outbreak detection, enhance data accuracy, and optimize resource allocation, ultimately strengthening its public health infrastructure and response capabilities.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of Machine Learning Applications in Disease Surveillance in Nigeria and Beyond\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStudy Authors\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYear\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eML Methodologies Used\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eGeographic Focus\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eReported Health Outcomes\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAI-driven hotspot mapping\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSouthwestern Nigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eImproved tuberculosis case detection via active case finding\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAI within AVADAR (mobile-based surveillance)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMultiple African countries (including Nigeria)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eIncreased acute flaccid paralysis/polio case reporting by ~\u0026thinsp;45%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2025\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEvent management\u0026thinsp;+\u0026thinsp;ML-supported surveillance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eStrengthened outbreak detection and response coordination\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClimate indices\u0026thinsp;+\u0026thinsp;ML modeling\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTropical Africa (Nigeria focus)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePredicted malaria and meningitis outbreaks using climate-health data\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eML classification models (Decision Trees, SVM)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eHigh accuracy in diagnosing malaria from symptoms\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2023\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eML algorithms for pattern recognition\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eEarly detection of Hepatitis B infections, including asymptomatic cases\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePredictive analytics with ML\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNigeria (Lassa fever)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eEarly warning and outbreak forecasting of Lassa fever\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eML algorithms (classification, regression)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePrediction of meningitis outbreaks with high accuracy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eML models for outbreak forecasting\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eImproved epidemic prediction and preparedness\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePredictive analytics using ML\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRural Nigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eForecasting epidemic outbreaks in underserved areas\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eML algorithm on health surveillance data\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eFrance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eEstimated diabetes incidence from surveillance/EHR data\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eWastewater surveillance\u0026thinsp;+\u0026thinsp;ML modeling\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBrazil\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTracked SARS-CoV-2 trends at municipal level with high sensitivity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e[\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRegression-based ML models\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRwanda\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePredicted out-of-pocket health expenditures from health surveillance data\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e2\u003c/span\u003e highlights possible training areas, challenges, opportunities and solutions.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eOpportunities and Challenges for ML in Nigeria\u0026rsquo;s Diseases Surveillance\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOpportunity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePotential Impact\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eChallenge\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSolution\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEnhanced Data Collection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30% faster reporting [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e].\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePoor rural connectivity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMobile-based tools\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOutbreak Detection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e85% sensitivity [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e].\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLimited sanitation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCommunity water sampling\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWorkforce Training\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e50% AI literacy boost [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e].\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLack of programs\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePPP-led workshops\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"6.0 Case Studies of ML in Nigerian Health Surveillance","content":"\u003cdiv id=\"Sec32\" class=\"Section2\"\u003e \u003ch2\u003e6.1 Malaria Diagnosis through Symptom-Based ML Models\u003c/h2\u003e \u003cp\u003eMachine learning models trained on symptomatic patterns have been developed to enhance malaria diagnosis. The study assessed eight algorithms, with decision trees and support vector machines (SVMs) demonstrating superior performance. These models showed high predictive accuracy using symptoms alone, representing a particularly impactful innovation in Nigeria\u0026rsquo;s rural and resource-constrained areas where access to microscopy or rapid diagnostic tests is limited [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec33\" class=\"Section2\"\u003e \u003ch2\u003e6.2 ML-Enhanced Epidemic Surveillance at NCDC\u003c/h2\u003e \u003cp\u003eA pioneering initiative at the Nigeria Centre for Disease Control (NCDC) integrated machine learning (ML) algorithms and data analytics into epidemic surveillance. This pilot framework was deployed during COVID-19 and Lassa fever outbreaks, enabling early detection of epidemic signals. Notably, NCDC staff remarked that \u0026ldquo;we\u0026rsquo;ve moved from reactive firefighting to proactive detection,\u0026rdquo; reflecting the practical utility of ML-driven surveillance in strengthening public health preparedness [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec34\" class=\"Section2\"\u003e \u003ch2\u003e6.3 Hepatitis B Virus Prediction Using Routine Laboratory Data\u003c/h2\u003e \u003cp\u003eA study published in Scientific Reports trained machine learning (ML) algorithms on clinical datasets to predict Hepatitis B Virus (HBV) infections. The models effectively identified both symptomatic and asymptomatic cases, which is particularly valuable for pre-surgical screenings. Clinical collaborators at Jos University Teaching Hospital noted that \u0026ldquo;doctors were impressed by the model\u0026rsquo;s ability to flag asymptomatic carriers,\u0026rdquo; highlighting its potential for integration into hospital workflows across both rural and urban settings [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec35\" class=\"Section2\"\u003e \u003ch2\u003e6.4 Clinician Trust in AI for Infectious Disease Triage\u003c/h2\u003e \u003cp\u003eA review of Nigeria\u0026rsquo;s growing experience with AI-based diagnostic tools underscores how machine learning has begun to support triage decisions in overstretched clinical environments. A frontline clinician stated: \u0026ldquo;We\u0026rsquo;ve started trusting algorithms to support our triage decisions, especially when overwhelmed during Lassa and COVID surges.\u0026rdquo; This testimonial affirms the expanding role of ML in not only diagnosis but also in decision-support and care prioritization during public health emergencies [\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e"},{"header":"7.0 Challenges and Barriers to Implementing Machine Learning in Disease Surveillance in Nigeria","content":"\u003cp\u003eMachine learning (ML) offers transformative potential for disease surveillance in Nigeria by improving data-driven decision-making, early outbreak detection, and public health interventions. However, several challenges hinder the effective implementation of ML-driven surveillance systems. These barriers span data quality, infrastructure, technical expertise, financial constraints, ethical concerns, and cultural resistance, all of which require targeted solutions [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eNigeria faces significant gaps in digital infrastructure, which affects the collection, storage, and analysis of health data. Many healthcare facilities lack electronic health records and rely on paper-based systems, making it difficult to train ML models effectively [\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e]. The deployment of ML-based surveillance systems requires robust digital infrastructure, including reliable internet access, stable electricity supply, and high-performance computing resources [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. However, frequent power outages, poor internet connectivity, and inadequate technological resources in many healthcare facilities, particularly in rural areas, pose significant challenges. The lack of interoperability between existing health data systems further complicates data collection efforts, making it difficult to aggregate and analyze information in real-time [\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e]. For example, Nigeria\u0026rsquo;s health data systems lack adherence to global standards like HL7 or FHIR, with only 15% of facilities using interoperable electronic records [\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e]. This fragmentation delays data aggregation by up to 3 weeks, undermining ML\u0026rsquo;s real-time potential [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Adopting FHIR could reduce this lag to 48 hours, as seen in Kenya\u0026rsquo;s e-health rollout [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHealth data in Nigeria is often fragmented, incomplete, or inconsistently recorded across different healthcare facilities. Manual record-keeping, lack of standardized electronic health records, and disparities in data collection between urban and rural areas exacerbate these issues [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. ML models require large, high-quality datasets to function effectively, but the lack of centralized health databases impairs the training and validation of these models. Without reliable data, ML-driven insights may be inaccurate, leading to ineffective public health responses [\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e]. Moreover, poor data-sharing frameworks and institutional structures prevent seamless integration of surveillance data across various agencies, limiting the effectiveness of ML applications [\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eImplementing ML solutions in disease surveillance demands specialized knowledge in data science, artificial intelligence, and software engineering. However, Nigeria faces a shortage of professionals with the technical expertise needed to develop, deploy, and maintain ML-based health systems [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. While interest in AI and data science is growing, the lack of structured training programs and educational opportunities in these fields remains a major gap. Additionally, many healthcare workers are unfamiliar with ML technologies, making it difficult to integrate these solutions into existing public health frameworks [\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e]. Without adequate technical capacity, even well-funded ML initiatives may fail due to a lack of personnel capable of maintaining and optimizing the systems.\u003c/p\u003e \u003cp\u003eThe high costs associated with ML implementation present another significant barrier. Developing, deploying, and maintaining ML-driven surveillance [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] systems require substantial financial investment in infrastructure, software, and workforce training. However, Nigeria's healthcare sector is underfunded, with limited budgets allocated for technological innovation. Given competing priorities, such as responding to immediate health crises and improving basic healthcare services, investments in ML-powered disease surveillance often take a back seat. Without sufficient funding, scaling ML solutions for nationwide disease surveillance remains a challenge [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. International collaborations and public-private partnerships could help bridge this gap, but financial constraints remain a persistent challenge.\u003c/p\u003e \u003cp\u003eThe use of ML in disease surveillance raises concerns about patient privacy and data security. In Nigeria, the absence of comprehensive data protection laws and regulatory frameworks heightens these concerns. Issues surrounding data ownership, patient confidentiality, and informed consent must be addressed to ensure ethical AI deployment [\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e]. Additionally, algorithmic biases could result in disparities in healthcare delivery, disproportionately affecting certain populations. Establishing strong data governance policies is crucial to fostering trust and ensuring equitable use of ML in disease surveillance [\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e]. Nigeria\u0026rsquo;s National Data Protection Regulation (NDPR) offers a starting point, mandating consent and data anonymization, but enforcement remains weak, with only 10% compliance among health facilities [\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e]. Aligning ML deployment with NDPR, alongside regular audits, could mitigate privacy risks and reduce bias, ensuring equitable outcomes across urban and rural populations. Without clear regulatory oversight, ML applications in public health could lead to unintended consequences, such as discriminatory health policies or misuse of personal health data [\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eMost ML models are trained on datasets from high-income countries, which may not adequately capture the epidemiological patterns in Nigeria. This leads to biased predictions and reduced model effectiveness in local contexts. Without efforts to develop locally relevant datasets and ensure ML models are tailored to Nigeria\u0026rsquo;s unique healthcare landscape, the accuracy of disease predictions and outbreak surveillance could be compromised [\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e]. Inadequate representation of Nigerian-specific disease patterns in training datasets makes it difficult for ML models to detect emerging outbreaks effectively [\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e]. To improve generalizability, investments in local dataset curation and contextualized algorithm development are necessary.\u003c/p\u003e \u003cp\u003eResistance to adopting ML-driven surveillance systems is another significant hurdle. Many healthcare professionals and policymakers are hesitant to embrace new technologies due to a lack of awareness, fear of job displacement, or skepticism regarding ML's reliability [\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e]. Additionally, traditional disease surveillance methods are often perceived as more familiar and dependable. Low levels of digital literacy among healthcare workers and the general population further impede the widespread adoption of ML solutions [\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e]. Addressing this resistance requires awareness campaigns, stakeholder engagement, and targeted training programs to build confidence in ML technologies. Without proactive efforts to foster trust and acceptance, even the most advanced ML solutions may struggle to gain traction in Nigeria\u0026rsquo;s public health sector.\u003c/p\u003e"},{"header":"8.0 Ethical Considerations and Explainable AI in Disease Surveillance","content":"\u003cp\u003eExplainable Artificial Intelligence (XAI) refers to a set of processes and methods that enable users to understand the results and outputs generated by AI/ML algorithms [\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e]. In the healthcare sector, where trust, accountability, and accuracy are critical, XAI plays a vital role in ensuring that AI predictions are interpretable to both clinicians and patients [\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e]. Black-box AI models are often met with skepticism by healthcare professionals; XAI enhances trust by making decision-making processes transparent, thereby fostering clinical acceptance [\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e]. Given that medical decisions can have life-or-death consequences, XAI supports ethical practices by enabling human validation of AI outputs [\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e], and many medical AI applications must comply with regulatory standards that mandate explainability [\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe growing body of literature on XAI in disease prediction confirms its centrality to responsible AI deployment. A systematic literature review synthesized findings from 30 studies examining XAI\u0026rsquo;s evolving role in disease prediction, highlighting the effectiveness of SHAP and Local Interpretable Model-Agnostic Explanations (LIME) as the most widely used techniques [\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e]. This review underscored critical gaps including limited dataset diversity and reliance on single data modalities, especially relevant to Nigeria\u0026rsquo;s fragmented surveillance data landscape. Work on febrile disease diagnostics using data-driven XAI methodologies demonstrated that explainable diagnostic models can achieve clinically meaningful performance even with modest datasets, directly applicable to Nigeria\u0026rsquo;s resource-constrained setting [\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e]. In the context of tropical and infectious disease specifically, XAI has been applied to enhance the interpretability of malaria and typhoid diagnoses using large language model integration [\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eXAI applied to Parkinson\u0026rsquo;s disease prediction, demonstrating that SHAP clarifies the global importance of clinical features while LIME provides patient-specific explanations [\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e]. XAI frameworks also improve adoption of AI tools in clinical settings by reducing algorithmic resistance among healthcare professionals, a barrier particularly pronounced in Nigeria\u0026rsquo;s public health workforce [\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e]. XAI-enhanced ML models demonstrate superior interpretability-accuracy trade-offs compared to opaque black-box alternatives, reinforcing the case for XAI as a deployment standard [\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSHAP builds on cooperative game theory to assign each feature a value reflecting its contribution to a specific prediction, offering both local explanations and global insights into model behavior [\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e] [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. LIME takes a localized approach, approximating the behavior of complex ML models near an individual prediction using simpler interpretable models [\u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e]. Both are critical in healthcare settings where interpretability informs both individual and population-level interventions [\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn Nigeria\u0026rsquo;s evolving public health landscape, these tools hold particular promise. A public health officer in Edo or Ondo State, regions known for recurring Lassa fever outbreaks, could use LIME to examine why an AI-based surveillance system flagged a particular case as high-risk, revealing key contributors such as symptom profile, rodent infestation reports from SMS-based surveillance, and data on poor sanitation in the affected locality. On a broader scale, SHAP offers significant value in post-hoc analyses, helping policymakers identify which environmental and infrastructural factors consistently drive outbreaks across LGAs, guiding targeted policy responses accordingly\u003c/p\u003e"},{"header":"9.0 Comparing AI and ML Adoption in Public Health: Nigeria vs. Rwanda and Kenya","content":"\u003cp\u003eNigeria has made some effort in adopting AI and ML in public health. However, when benchmarked against Rwanda and Kenya, Nigeria lags in several critical areas. \u003cstrong\u003eAs shown in Table 5, key gaps include weak AI-health policy integration, fragmented data systems, limited funding, and slower reporting timelines.\u003c/strong\u003e The comparative analysis is based on published literature and should be interpreted with caution, given differences in reporting systems, funding tracking, and data availability across countries [52,54,81\u0026ndash;84]. Funding figures represent available estimates from cited literature rather than verified national accounts\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 5: Comparative Overview of AI/ML Adoption in Public Health-Nigeria, Rwanda, and Kenya\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDimension\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNigeria\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRwanda\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eKenya\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNational AI/Health Strategy\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eNITDA AI Policy (2021); no specific health integration strategy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eSmart Rwanda Master Plan includes health sector AI targets \u0026nbsp; \u0026nbsp; [84]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eDigital Economy Blueprint integrates health AI [84]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eEstimated AI-Health Funding\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u0026lt;$1M; primarily donor-driven[84]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u0026gt;$10M; government-supported [84]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u0026gt;$5M; government + PPP [84]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eKey PPP Models\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eLimited; few structured health AI PPPs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eZipline drone logistics + ML for malaria prediction [54]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eIBM-Watson at Kenyatta Hospital; AFYA-Tek disease forecasting [83]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eData Reporting Timeliness\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e~2\u0026ndash;3 week lag in national data aggregation [52]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e50% reduction in malaria case reporting time by 2022 [54]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eSub-weekly outbreak prediction updates via AFYA-Tek [83]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCapacity Building Programs\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eNo structured national program; emerging university initiatives\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eRwanda AI Academy (government-sponsored) [84]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 150px;\"\u003e\n \u003cp\u003eNairobi AI Lab; university-industry partnerships [84]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eNigeria\u0026rsquo;s key gaps, weak AI-health policy integration, fragmented data systems, limited dedicated funding, and fewer scalable AI hubs, are compounded by structural governance challenges. Despite these constraints, opportunities exist: leveraging the NITDA AI Policy to design a dedicated AI framework for health, partnering with private sector actors such as MTN and Flutterwave for local funding, and adapting Rwanda\u0026rsquo;s drone-AI model for rural healthcare logistics [82].\u003c/p\u003e"},{"header":"10.0 Recommendations and future directions.","content":"\u003cp\u003eMaximizing the potential of ML for disease surveillance in Nigeria requires a comprehensive strategy encompassing policy enhancements, workforce development, infrastructure upgrades, and active stakeholder collaboration. Recommendations are organized into three tiers based on feasibility, cost, and implementation readiness. Cost estimates are scenario-based illustrative figures drawn from comparable LMIC technology deployments; actual costs depend on procurement models, PPP structures, and phased scaling strategies.\u003c/p\u003e \u003cdiv id=\"Sec40\" class=\"Section2\"\u003e \u003ch2\u003e10.1 Tier 1 - Foundational Reforms (Short-Term: Years 1\u0026ndash;2)\u003c/h2\u003e \u003cp\u003eTier 1 reforms address immediate, high-priority, lower-cost actions that can be implemented within existing institutional structures. These constitute the prerequisite foundations upon which more advanced AI integration depends:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eStrengthen eIDSR implementation: Equip DSNOs in priority states with low-cost digital reporting tools (\u0026lt;\u003cspan\u003e$\u003c/span\u003e5,000 per state), targeting a 30% improvement in reporting accuracy and reduction of the current 2\u0026ndash;3 week data lag to under 72 hours.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eImprove interoperability standards: Mandate HL7/FHIR adoption in a phased rollout starting with tertiary and secondary facilities.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eStrengthen data governance: Enforce the NDPR across health facilities, conduct independent compliance audits, and establish clear policies for data anonymization.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eEstablish a National AI-for-Health Working Group: Convene NCDC, Federal Ministry of Health, NITDA, and academic institutions to develop a dedicated ML-health integration roadmap.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eIntegrate AI/ML into medical and public health curricula: Target 50% of graduating public health cohorts receiving foundational AI literacy training within two years.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec41\" class=\"Section2\"\u003e \u003ch2\u003e10.2 Tier 2 - Infrastructure Scaling (Medium-Term: Years 3\u0026ndash;5)\u003c/h2\u003e \u003cp\u003eTier 2 actions require significant capital investment and PPP engagement, and are contingent upon achieving Tier 1 baseline interoperability and data maturity:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eExpand rural connectivity: Deploy ML-ready internet infrastructure through PPPs with telecoms, targeting a 50% increase in eIDSR coverage at an estimated \u003cspan\u003e$\u003c/span\u003e10\u0026ndash;15M nationally (illustrative scenario).\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eCloud migration and centralized dashboards: Migrate disease surveillance data to secure cloud platforms with ML processing capabilities and develop real-time outbreak dashboards.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003ePilot ML models on real Nigerian data: Initiate retrospective validation studies for malaria, Lassa fever, and cholera using NCDC historical data, benchmarking Decision Trees, Random Forests, and gradient boosting.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eImplement measurable performance indicators: Track eIDSR reporting lag (target: 3 weeks \u0026rarr; 72 hours), interoperability adoption rates (target: 50% of facilities), and ML temporal validation R\u0026sup2; (target: \u0026ge;0.70 for priority diseases).\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec42\" class=\"Section2\"\u003e \u003ch2\u003e10.3 Tier 3 - Advanced AI Integration (Long-Term: Years 5\u0026ndash;10)\u003c/h2\u003e \u003cp\u003eTier 3 innovations are contingent upon achieving baseline data maturity, interoperability, and workforce capacity established in Tiers 1 and 2. Advanced technologies should not be pursued in parallel with foundational reforms but sequenced after their demonstrated success:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eDeploy cloud-based ML surveillance at national scale: Integrate validated models across all 36 states and FCT within eIDSR and NCDC platforms. Estimated infrastructure investment: \u003cspan\u003e$\u003c/span\u003e40\u0026ndash;60M nationally over five years (illustrative scenario).\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eDevelop federated learning frameworks: Enable model training across 774 LGAs without centralizing sensitive patient records \u0026mdash; contingent upon adequate state-level server infrastructure.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eIntegrate genomic surveillance pipelines: Leverage AI-driven genomic analytics to track pathogen evolution and predict emerging disease strains.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eEstablish cross-border AI surveillance collaboration: Work through ECOWAS and Africa CDC to create continent-wide AI-driven disease monitoring systems.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e below presents the proposed ML integration workflow for Nigeria\u0026rsquo;s IDSR system, illustrating the comprehensive system architecture from multi-source data ingestion through ML model development, explainable AI output, and continuous monitoring.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec43\" class=\"Section2\"\u003e \u003ch2\u003e10.4 Economic Framing: DALYs Averted and Return on Investment\u003c/h2\u003e \u003cp\u003eSurveillance systems that enable early detection significantly reduce the burden of disease and are highly cost-effective. According to WHO benchmarks, interventions that avert DALYs at a cost below \u003cspan\u003e$\u003c/span\u003e150 per DALY are considered highly cost-effective in LMICs [\u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e85\u003c/span\u003e]. For illustrative purposes: an ML-based surveillance system preventing 100 outbreak-related deaths annually could conservatively save 3,000\u0026ndash;4,000 DALYs per year, consistent with published estimates from zoonotic outbreak mitigation efforts in Nigeria [\u003cspan citationid=\"CR86\" class=\"CitationRef\"\u003e86\u003c/span\u003e]. WHO models suggest every \u003cspan\u003e$\u003c/span\u003e1 invested in surveillance may yield \u003cspan\u003e$\u003c/span\u003e5\u0026ndash;7 in savings through reduced hospitalization, increased productivity, and avoided emergency expenditures [\u003cspan citationid=\"CR87\" class=\"CitationRef\"\u003e87\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIt is essential to note that these projections are illustrative scenario-based estimates intended to inform the general cost-effectiveness framing of ML surveillance investments, rather than precise Nigeria-specific modeling. Actual DALYs averted and ROI figures will depend on disease burden trajectories, ML model accuracy on real data, implementation scale, and population-level factors. Nigeria-specific cost-effectiveness studies using real surveillance data are strongly recommended before large-scale funding decisions are made.\u003c/p\u003e \u003c/div\u003e"},{"header":"11.0 Conclusion","content":"\u003cp\u003eIntegrating ML into Nigeria\u0026rsquo;s disease surveillance system represents a pivotal pathway towards enhancing outbreak detection capabilities and bolstering public health resilience. This paper provided a structured examination of ML\u0026rsquo;s transformative potential, detailed key ML methodologies and architectures, presented a full methodological illustration using synthetic malaria surveillance data including a complete simulation algorithm and pseudocode, benchmarked five ML models (Linear Regression, Decision Tree, Random Forest, Gradient Boosting, and Support Vector Machine) under identical temporal validation conditions, and demonstrated model robustness through a rainfall coefficient sensitivity analysis (R\u0026sup2; range: 0.922\u0026ndash;0.980 across \u0026plusmn;\u0026thinsp;33% perturbations). SHAP-based interpretability confirmed that rainfall and health access index were the dominant predictive features, consistent with epidemiological expectations. The paper also conducted a structured comparative analysis situating Nigeria\u0026rsquo;s AI adoption within the broader African landscape alongside Rwanda and Kenya. All findings remain illustrative and require real-world validation using NCDC or state-level surveillance data before deployment. With strategic reforms, implementing the proposed tiered integration roadmap, strengthening PPPs, and developing contextually validated ML models, Nigeria can build a proactive, equitable, and data-driven public health system capable of mitigating future health threats and advancing global health equity\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eAI\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eArtificial Intelligence\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eAVADAR\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAuto\u0026mdash;Visual Acute Flaccid Paralysis Detection and Reporting\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eCNNs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eConvolutional Neural Networks\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eCV\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCross\u0026mdash;Validation\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eDALYs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDisability\u0026mdash;Adjusted Life Years\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eDSNOs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDisease Surveillance and Notification Officers\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eeIDSR\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eElectronic Integrated Disease Surveillance and Response\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eEIOS\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eEpidemic Intelligence from Open Sources\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eFHIR\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eFast Healthcare Interoperability Resources\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eGANs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGenerative Adversarial Networks\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eHBV\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHepatitis B Virus\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eIDSR\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eIntegrated Disease Surveillance and Response\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eLGAs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLocal Government Areas\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eLIME\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLocal Interpretable Model\u0026mdash;Agnostic Explanations\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eLLMs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLarge Language Models\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eLMIC\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLow\u0026mdash;and Middle\u0026mdash;Income Country\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eML\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMachine Learning\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eNCDC\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNigeria Centre for Disease Control\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eNDPR\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNational Data Protection Regulation\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eNITDA\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNational Information Technology Development Agency\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003ePPPs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePublic\u0026mdash;Private Partnerships\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eRFs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRandom Forests\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eRMSE\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRoot Mean Square Error\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eRNNs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRecurrent Neural Networks\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eROI\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eReturn on Investment\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eSHAP\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSHapley Additive exPlanations\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eSVMs\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSupport Vector Machines\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eWBE\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eWastewater\u0026mdash;Based Epidemiology\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eWHO\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eWorld Health Organization\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e\u003cb\u003eXAI\u003c/b\u003e\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eExplainable Artificial Intelligence\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research received no specific funding from any public, commercial, or not-for-profit organization. The work was carried out with support from the authors\u0026rsquo; respective institutions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConceptualization \u0026amp; Design: L.J.A., A.B.U.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMethodology \u0026amp; Simulation (Machine Learning):\u0026nbsp;L.J.A. (Primary lead for the synthetic data model and analysis), A.B.U.\u003c/p\u003e\n\u003cp\u003eDrafting Original Manuscript: A.B.U., L.J.A., S.K.S.\u003c/p\u003e\n\u003cp\u003eReview \u0026amp; Editing (Public Health \u0026amp; Contextual Insights): H.Y., A.H.D., M.Z., U.A.H.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSupervision: A.B.U., U.A.H.\u003c/p\u003e\n\u003cp\u003eFinal Approval: All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for Publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable. This manuscript does not contain any images, videos, or personal data relating to individual participants.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests (financial or non-financial) that could have influenced the work reported in this paper.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eHong R, Walker R, Hovan G, Henry L, Pescatore R. The Power of Public Health Surveillance. Del J Public Health. 2020 Jul 1;6(2):60. doi:10.32481/djph.2020.07.016 PubMed PMID: 34467112.\u003c/li\u003e\n\u003cli\u003eIbrahim LM, Stephen M, Okudo I, Kitgakka SM, Mamadu IN, Njai IF, et al. A rapid assessment of the implementation of integrated disease surveillance and response system in Northeast Nigeria, 2017. BMC Public Health. 2020 May 1;20(1):600. doi:10.1186/s12889-020-08707-4\u003c/li\u003e\n\u003cli\u003eIsere EE, Fatiregun AA, Ajayi IO. An overview of disease surveillance and notification system in Nigeria and the roles of clinicians in disease outbreak prevention and control. Niger Med J. 2015 Jun;56(3):161. doi:10.4103/0300-1652.160347\u003c/li\u003e\n\u003cli\u003eRashid A B, Kausik MD A K. \u003cem\u003eAI revolutionizing industries worldwide: A comprehensive overview of its diverse applications.\u003c/em\u003e \u003cstrong\u003eHybrid Adv.\u003c/strong\u003e2024;7:100277. doi:10.1016/j.hybadv.2024.100277.\u003c/li\u003e\n\u003cli\u003eAnahtar MN, Yang JH, Kanjilal S. Applications of Machine Learning to the Problem of Antimicrobial Resistance: an Emerging Model for Translational Research. J Clin Microbiol. 2021 Jun 18;59(7):10.1128/jcm.01260-20. doi:10.1128/jcm.01260-20\u003c/li\u003e\n\u003cli\u003eWang L, Zhang Y, Wang D, Tong X, Liu T, Zhang S, et al. Artificial Intelligence for COVID-19: A Systematic Review. Front Med. 2021 Sep 30;8. doi:10.3389/fmed.2021.704256\u003c/li\u003e\n\u003cli\u003eJian SW, Cheng HY, Huang XT, Liu DP. Contact tracing with digital assistance in Taiwan\u0026rsquo;s COVID-19 outbreak response. Int J Infect Dis. 2020 Oct 6;101:348. doi:10.1016/j.ijid.2020.09.1483 PubMed PMID: 33035674.\u003c/li\u003e\n\u003cli\u003eKim J, Jo S, Cho S il. New framework to assess tracing and testing based on South Korea\u0026rsquo;s response to COVID-19. BMC Infect Dis. 2024 May 3;24(1):469. doi:10.1186/s12879-024-09363-4\u003c/li\u003e\n\u003cli\u003eChellaiyan VG, Nirupama AY, Taneja N. Telemedicine in India: Where do we stand? J Family Med Prim Care. 2019 Jun;8(6):1872\u0026ndash;1876. doi:10.4103/jfmpc.jfmpc_264_19 PubMed PMID: 31334148; PubMed Central PMCID: PMC6618203.\u003c/li\u003e\n\u003cli\u003eSangwa S, Placide M. Artificial Intelligence and Rwanda\u0026rsquo;s Economic Transformation: A Strategic Policy Review of Sectoral Readiness, Challenges, and Opportunities. Vol. 1. 2025 May 9;1:1\u0026ndash;16. doi:10.5281/zenodo.15459202\u003c/li\u003e\n\u003cli\u003eMeribole EC, Makinde OA, Oyemakinde A, Oyediran KA, Atobatele A, Fadeyibi FA, et al. The Nigerian health information system policy review of 2014 : the need, content, expectations and progress. Health Inf Libr J. 2018 Dec;35(4):285\u0026ndash;97. doi:10.1111/hir.12240 PubMed PMID: 30417971.\u003c/li\u003e\n\u003cli\u003eKusimo OC, Ugwu CI, Aduh U, Okoro CA. Implementing TB Surveillance in Nigeria: Best Practices, Challenges and Lessons Learnt. J Tuberc Res. 2020 Nov 9;8(4):199\u0026ndash;208. doi:10.4236/jtr.2020.84018\u003c/li\u003e\n\u003cli\u003eOnwe FI, Okedo-Alex IN, Akamike IC, Igwe-Okomiso DO. Vertical disease programs and their effect on integrated disease surveillance and response: perspectives of epidemiologists and surveillance officers in Nigeria. Trop Dis Travel Med Vaccines. 2021 Oct 1;7(1):28. doi:10.1186/s40794-021-00152-4\u003c/li\u003e\n\u003cli\u003eIbrahim LM, Okudo I, Stephen M, Ogundiran O, Pantuvo JS, Oyaole DR, et al. Electronic reporting of integrated disease surveillance and response: lessons learned from northeast, Nigeria, 2019. BMC Public Health. 2021. doi:10.1186/s12889-021-10957-9\u003c/li\u003e\n\u003cli\u003eOgunboye I, Adebayo I, Anioke S, Egwuatu E, Ajala C, Awuah SB. Enhancing Nigeria\u0026rsquo;s health surveillance system: A data-driven approach to epidemic preparedness and response. World J Adv Res Rev. 2023 Oct 15;20:1352\u0026ndash;69. doi:10.30574/wjarr.2023.20.1.2078\u003c/li\u003e\n\u003cli\u003eAkinpelu D, Akintola S. Navigating the Legal and Ethical Terrain of Artificial Intelligence in Enhancing Patient Safety in Nigeria. J Intellect Prop Inf Technol Law JIPIT. 2023. doi:10.52907/jipit.v3i1.261\u003c/li\u003e\n\u003cli\u003eAlege A, Hashmi S, Eneogu R, Meurrens V, Budts AL, Pedro M, et al. Effectiveness of Using AI-Driven Hotspot Mapping for Active Case Finding of Tuberculosis in Southwestern Nigeria. Trop Med Infect Dis. 2024 May;9(5):5. doi:10.3390/tropicalmed9050099\u003c/li\u003e\n\u003cli\u003eTicha JM, Akpan GU, Paige LM, Senouci K, Stein A, Briand P, et al. Outcomes of the Deployment of the Auto-Visual Acute Flaccid Paralysis Detection and Reporting (AVADAR) System for Strengthening Polio Surveillance in Africa From 2017 to 2018: Evaluation Study. JMIR Public Health Surveill. 2020 Dec 2;6(4):e18950. doi:10.2196/18950\u003c/li\u003e\n\u003cli\u003eElston J, Eteng WEO, Ihekweazu C, Oliver I, Aniaku E, Abubakar A, et al. Development and Implementation of a Public Health Event Management System, Nigeria, 2018\u0026ndash;2024 - Volume 31, Number 1\u0026mdash;January 2025 - Emerging Infectious Diseases journal - CDC [Internet]. doi:10.3201/eid3101.240379\u003c/li\u003e\n\u003cli\u003eAfuan L, Isnanto RR. A comparative study of machine learning algorithms for fall detection in technology-based healthcare system: Analyzing SVM, KNN, decision tree, random forest, LSTM, and CNN. E3S Web Conf. 2025;605:03051. doi:10.1051/e3sconf/202560503051\u003c/li\u003e\n\u003cli\u003eDama\u0026scaron;evičius R, Jagatheesaperumal SK, Kandala RNVPS, Hussain S, Alizadehsani R, Gorriz JM. Deep learning for personalized health monitoring and prediction: A review. Comput Intell. 2024 Jun;40(3):e12682. doi:10.1111/coin.12682\u003c/li\u003e\n\u003cli\u003eNisar DEM, Amin R, Shah NUH, Ghamdi MAA, Almotiri SH, Alruily M. Healthcare Techniques Through Deep Learning: Issues, Challenges and Opportunities. IEEE Access. 2021;9:98523\u0026ndash;41. doi:10.1109/ACCESS.2021.3095312\u003c/li\u003e\n\u003cli\u003eAldahiri A, Alrashed B, Hussain W. Trends in Using IoT with Machine Learning in Health Prediction System. Forecasting. 2021 Mar;3(1):1. doi:10.3390/forecast3010012\u003c/li\u003e\n\u003cli\u003eMohr F, Rijn JN, Van JN. Towards Model Selection using Learning Curve Cross-Validation [Internet]. 2021 [cited 2025 Jun 11]. Available from: https://consensus.app/papers/towards-model-selection-using-learning-curve-mohr-rijn/fde46c822fc85b1f942a8bc210325480/\u003c/li\u003e\n\u003cli\u003eXu Y, Goodacre R. On Splitting Training and Validation Set: A Comparative Study of Cross-Validation, Bootstrap and Systematic Sampling for Estimating the Generalization Performance of Supervised Learning. J Anal Test. 2018;2:249\u0026ndash;62. doi:10.1007/s41664-018-0068-2\u003c/li\u003e\n\u003cli\u003eYates LA, Aandahl Z, Richards S, Brook B. Cross validation for model selection: a primer with examples from ecology [Internet]. 2022 [cited 2025 Jun 11]. Available from: https://consensus.app/papers/cross-validation-for-model-selection-a-primer-with-yates-aandahl/76e95bf1055f507b959fcb32721ea21c/\u003c/li\u003e\n\u003cli\u003eNurunnabi A, Teferle FN, Laefer D, Remondino F, Karas I, Li J. kCV-B: BOOTSTRAP WITH CROSS-VALIDATION FOR DEEP LEARNING MODEL DEVELOPMENT, ASSESSMENT AND SELECTION. Int Arch Photogramm Remote Sens Spat Inf Sci. 2022. doi:10.5194/isprs-archives-xlviii-4-w3-2022-111-2022\u003c/li\u003e\n\u003cli\u003eAyanlade A, Nwayor IJ, Sergi C, Ayanlade OS, Di Carlo P, Jeje OD, et al. Early warning climate indices for malaria and meningitis in tropical ecological zones. Sci Rep. 2020 Aug 31;10(1):14303. doi:10.1038/s41598-020-71094-8 PubMed PMID: 32868821; PubMed Central PMCID: PMC7459128.\u003c/li\u003e\n\u003cli\u003eOnuh S, Aiyeola SY, Tofade CT, Adesina LO. APPLICATIONS OF SPACE TECHNOLOGY FOR ACCELERATED INFRASTRUCTURES DEVELOPMENT FOR RURAL AREAS DEVELOPMENT IN NIGERIA. Arid Zone J Eng Technol Environ. 2019;15:21\u0026ndash;34.\u003c/li\u003e\n\u003cli\u003eGiuffr\u0026egrave; M, Shung DL. Harnessing the power of synthetic data in healthcare: innovation, application, and privacy. NPJ Digit Med. 2023 Oct 9;6(1):186. doi:10.1038/s41746-023-00927-3 PubMed PMID: 37813960; PubMed Central PMCID: PMC10562365.\u003c/li\u003e\n\u003cli\u003eBattista K, Diao L, Patte KA, Dubin JA, Leatherdale ST. Examining the use of decision trees in population health surveillance research: an application to youth mental health survey data in the COMPASS study. Health Promot Chronic Dis Prev Can Res Policy Pract. 2023 Feb;43(2):73\u0026ndash;86. doi:10.24095/hpcdp.43.2.03 PubMed PMID: 36794824; PubMed Central PMCID: PMC10026612.\u003c/li\u003e\n\u003cli\u003eRanjbaran G, Recupero D. C-SHAP: A Hybrid Method for Fast and Efficient Interpretability. Appl Sci. 2025 Jan 1. doi:10.3390/APP15020672\u003c/li\u003e\n\u003cli\u003eHaneef R, Kab S, Hrzic R, Fuentes S, Fosse-Edorh S, Cosson E, et al. Use of artificial intelligence for public health surveillance: a case study to develop a machine Learning-algorithm to estimate the incidence of diabetes mellitus in France. Arch Public Health. 2021 Sep 22;79(1):168. doi:10.1186/s13690-021-00687-0\u003c/li\u003e\n\u003cli\u003eOgwu MC, Izah SC. Innovations in Disease Surveillance and Monitoring. In: Ogwu MC, Izah SC, editors. Technological Innovations for Managing Tropical Diseases [Internet]. Cham: Springer Nature Switzerland; 2025 [cited 2025 Mar 2]. p. 83\u0026ndash;108. Available from: https://doi.org/10.1007/978-3-031-82622-1_4 doi:10.1007/978-3-031-82622-1_4\u003c/li\u003e\n\u003cli\u003eAdigwe OP, Onavbavba G, Sanyaolu SE. Exploring the matrix: knowledge, perceptions and prospects of artificial intelligence and machine learning in Nigerian healthcare. Front Artif Intell. 2024 Jan 19;6. doi:10.3389/frai.2023.1293297\u003c/li\u003e\n\u003cli\u003eEkundayo F. Using machine learning to predict disease outbreaks and enhance public health surveillance. 2024 Dec 14;794\u0026ndash;811. doi:10.30574/wjarr.2024.24.3.3732\u003c/li\u003e\n\u003cli\u003eAbdeldayem OM, Dabbish AM, Habashy MM, Mostafa MK, Elhefnawy M, Amin L, et al. Viral outbreaks detection and surveillance using wastewater-based epidemiology, viral air sampling, and machine learning techniques: A comprehensive review and outlook. Sci Total Environ. 2022 Jan 10;803:149834. doi:10.1016/j.scitotenv.2021.149834\u003c/li\u003e\n\u003cli\u003ePrado T, Fumian TM, Mannarino CF, Resende PC, Motta FC, Eppinghaus ALF, et al. Wastewater-based epidemiology as a useful tool to track SARS-CoV-2 and support public health policies at municipal level in Brazil. Water Res. 2021 Mar 1;191:116810. doi:10.1016/j.watres.2021.116810\u003c/li\u003e\n\u003cli\u003eMbunge E, Batani J. Application of deep learning and machine learning models to improve healthcare in sub-Saharan Africa: Emerging opportunities, trends and implications. Telemat Inform Rep. 2023 Sep 1;11:100097. doi:10.1016/j.teler.2023.100097\u003c/li\u003e\n\u003cli\u003eLafond KE, Dalhatu I, Shinde V, Ekanem EE, Ahmed S, Peebles P, et al. Notifiable disease reporting among public sector physicians in Nigeria: a cross-sectional survey to evaluate possible barriers and identify best sources of information. BMC Health Serv Res. 2014 Nov 13;14:568. doi:10.1186/s12913-014-0568-3 PubMed PMID: 25391377; PubMed Central PMCID: PMC4233090.\u003c/li\u003e\n\u003cli\u003eGonz\u0026aacute;lez A, Zanin M, Menasalvas E. Public Health and Epidemiology Informatics: Can Artificial Intelligence Help Future Global Challenges? An Overview of Antimicrobial Resistance and Impact of Climate Change in Disease Epidemiology. Yearb Med Inform. 2019 Aug 1;28:224\u0026ndash;31. doi:10.1055/s-0039-1677910\u003c/li\u003e\n\u003cli\u003eAytan-Aktug D, Clausen P, Bortolaia V, Aarestrup F, Lund O. Prediction of Acquired Antimicrobial Resistance for Multiple Bacterial Species Using Neural Networks. mSystems. 2020;5. doi:10.1128/mSystems.00774-19\u003c/li\u003e\n\u003cli\u003eTai AMY, Albuquerque A, Carmona NE, Subramanieapillai M, Cha DS, Sheko M, et al. Machine learning and big data: Implications for disease modeling and therapeutic discovery in psychiatry. Artif Intell Med. 2019 Aug 1;99:101704. doi:10.1016/j.artmed.2019.101704\u003c/li\u003e\n\u003cli\u003eGidey K, Beyene H, Haileslassie F, Berihu H. Apprentice bot model design and implementation for psychological clients\u0026rsquo; therapy. Discov Appl Sci. 2025 Feb 24;7(3):170. doi:10.1007/s42452-025-06515-2\u003c/li\u003e\n\u003cli\u003eOluwadiya K, Adeoti A, Nottidge T. Exploring Artificial Intelligence in the Nigerian Medical Educational Space: An Online Cross-sectional Study of Perceptions, Risks and Benefits among Students and Lecturers from Ten Universities. Niger Postgrad Med J. 2023 Nov 23;30:285. doi:10.4103/npmj.npmj_186_2\u003c/li\u003e\n\u003cli\u003eDe-Arteaga M, Herlands W, Neill DB, Dubrawski A. Machine Learning for the Developing World. ACM Trans Manag Inf Syst. 2018 Jun 30;9(2):1\u0026ndash;14. doi:10.1145/3210548\u003c/li\u003e\n\u003cli\u003eIsiaka AB, Anakwenze VN, Ilodinso CR, Anaukwu CG, Ezeokoli CMV, Noi SM, et al. Harnessing Artificial Intelligence for Early Detection and Management of Infectious Disease Outbreaks. Int J Innov Res Dev. 2024. doi:10.24940/ijird/2024/v13/i2/feb24016\u003c/li\u003e\n\u003cli\u003eNaseem M, Akhund R, Arshad H, Ibrahim MT. Exploring the Potential of Artificial Intelligence and Machine Learning to Combat COVID-19 and Existing Opportunities for LMIC: A Scoping Review. J Prim Care Community Health. 2020 Jan 1;11:2150132720963634. doi:10.1177/2150132720963634\u003c/li\u003e\n\u003cli\u003eOkagbue HI, Oguntunde PE, Obasi ECM, Adamu PI, Opanuga AA. Diagnosing malaria from some symptoms: a machine learning approach and public health implications. Health Technol. 2021 Jan 1;11(1):23\u0026ndash;37. doi:10.1007/s12553-020-00488-5\u003c/li\u003e\n\u003cli\u003eAjuwon BI, Richardson A, Roper K, Sheel M, Audu R, Salako BL, et al. The development of a machine learning algorithm for early detection of viral hepatitis B infection in Nigerian patients. Sci Rep. 2023 Feb 24;13(1):3244. doi:10.1038/s41598-023-30440-2\u003c/li\u003e\n\u003cli\u003eQuezada DA, Akwafuo S, Halyal S. Harnessing Machine Learning for Predictive Analytics: A Case Study of Lassa Fever Outbreaks in Nigeria. In: 2024 10th International Conference on Control, Decision and Information Technologies (CoDIT) [Internet]. 2024 [cited 2025 Feb 22]. p. 377\u0026ndash;82. Available from: https://ieeexplore.ieee.org/document/10708095 doi:10.1109/CoDIT62066.2024.10708095\u003c/li\u003e\n\u003cli\u003eTian X, Xu Q, Wang Y. Prediction of Meningitis Outbreaks in Nigeria Using Machine Learning Algorithms. Proc 2019 2nd Artif Intell Cloud Comput Conf. 2019. doi:10.1145/3375959.3375984\u003c/li\u003e\n\u003cli\u003eNwankwo E, Emeihe E, Ajegbile M, Olaboye J, Maha C. Artificial Intelligence in predictive analytics for epidemic outbreaks in rural populations. Int J Life Sci Res Arch. 2024 Aug 8;07:78\u0026ndash;094. doi:10.53771/ijlsra.2024.7.1.0062\u003c/li\u003e\n\u003cli\u003eMuremyi R, Haughton D, Kabano I, Niragire F. Prediction of out-of-pocket health expenditures in Rwanda using machine learning techniques. Pan Afr Med J. 2020;37. doi:10.11604/pamj.2020.37.357.27287\u003c/li\u003e\n\u003cli\u003eDavies G, Mudele O, Wright-Ajayi B, Mosaku L, Moneke K, Shaba D, et al. Clinical Case Reports International Enhancing Infectious Disease Management in Nigeria: The Role of Artificial Intelligence in Diagnosis and Treatment Review Article. Clin Case Rep Int. 2024 Mar 18;8:3.\u003c/li\u003e\n\u003cli\u003eUmar AB, Sani SK, Aliyu LJ, Hassan M, Imam M, Haruna UA, et al. Enhancing primary healthcare delivery in Nigeria through the adoption of advanced technologies. Narra X. 2024 Dec 31;2(3):3. doi:10.52225/narrax.v2i3.180\u003c/li\u003e\n\u003cli\u003eOlaronke I, Gambo I, Ikono R, Olaleke J. Interoperability in Nigeria Healthcare System: The Ways Forward. Int J Inf Eng Electron Bus. 2013 Oct 1;5:16\u0026ndash;23. doi:10.5815/ijieeb.2013.04.03\u003c/li\u003e\n\u003cli\u003eAdewole KS, Alozie E, Olagunju H, Faruk N, Aliyu RY, Imoize AL, et al. A systematic review and meta-data analysis of clinical data repositories in Africa and beyond: recent development, challenges, and future directions. Discov Data. 2024 Jun 26;2(1):8. doi:10.1007/s44248-024-00012-4\u003c/li\u003e\n\u003cli\u003eJussen-Lengersdorf I, Schweihoff J, M\u0026ouml;ller F. Tensions in Inter-Organizational Data Sharing: Findings from Literature and Practice. 2023. doi:10.1109/CBI58679.2023.10187530\u003c/li\u003e\n\u003cli\u003eAchieng M, Ogundaini O. Big Data Analytics for Integrated Infectious Disease Surveillance in sub-Saharan Africa. SA J Inf Manag. 2024 Jan 30;26:1\u0026ndash;11. doi:10.4102/sajim.v26i1.1668\u003c/li\u003e\n\u003cli\u003eAkintola S. Navigating the Legal and Ethical Terrain of Artificial Intelligence in Enhancing Patient Safety in Nigeria [Internet]. 2023 Nov 27 [cited 2025 Feb 22]. Available from: https://www.academia.edu/124846767/Navigating_the_Legal_and_Ethical_Terrain_of_Artificial_Intelligence_in_Enhancing_Patient_Safety_in_Nigeria\u003c/li\u003e\n\u003cli\u003eOlayiwola K, Opele J. PRACTICE OF HEALTH INFORMATION MANAGEMENT IN NIGERIA: OPPORTUNITIES, CHALLENGES, AND FUTURE DIRECTIONS. Vol. 16. 2024 May 30;16:2023.\u003c/li\u003e\n\u003cli\u003ePrabhakaran V, Martin D. Participatory Machine Learning Using Community-Based System Dynamics. Health Hum Rights. 2020 Dec;22(2):71\u0026ndash;4. PubMed PMID: 33390696; PubMed Central PMCID: PMC7762892.\u003c/li\u003e\n\u003cli\u003eOlumade TJ, Adesanya OA, Fred-Akintunwsa IJ, Babalola DO, Oguzie JU, Ogunsanya OA, et al. Infectious disease outbreak preparedness and response in Nigeria: history, limitations and recommendations for global health policy and practice. AIMS Public Health. 2020 Sep 22;7(4):736\u0026ndash;57. doi:10.3934/publichealth.2020057 PubMed PMID: 33294478; PubMed Central PMCID: PMC7719556.\u003c/li\u003e\n\u003cli\u003eLi X, Abangbila L. Resistance to medical artificial intelligence: Integrating AI awareness, AI risks, and displacement of responsibility. J Infrastruct Policy Dev. 2024 Oct 17;8(11):11. doi:10.24294/jipd.v8i11.7923\u003c/li\u003e\n\u003cli\u003eMumtaz H, Riaz MH, Wajid H, Saqib M, Zeeshan MH, Khan SE, et al. Current challenges and potential solutions to the use of digital health technologies in evidence generation: a narrative review. Front Digit Health. 2023 Sep 28;5:1203945. doi:10.3389/fdgth.2023.1203945 PubMed PMID: 37840685; PubMed Central PMCID: PMC10568450.\u003c/li\u003e\n\u003cli\u003eSewada R, Jangid A, Kumar P, Mishra N. Explainable Artificial Intelligence (XAI). Int J Food Nutr Sci. 2023. doi:10.48047/ijfans/v12/i1/271\u003c/li\u003e\n\u003cli\u003eChaddad A, Peng J, Xu J, Bouridane A. Survey of Explainable AI Techniques in Healthcare. Sensors. 2023;23. doi:10.3390/s23020634\u003c/li\u003e\n\u003cli\u003eVarghese A, Varghese J, Biju J, Thomas R, Thomas M. Explainable AI in Healthcare Applications. Int Res J Adv Eng Manag IRJAEM. 2024. doi:10.47392/irjaem.2024.0545\u003c/li\u003e\n\u003cli\u003eCina G, Rober TE, Goedhart R, Birbil I. Why we do need Explainable AI for Healthcare. ArXiv. 2022;abs/2206.15363. doi:10.48550/arXiv.2206.15363\u003c/li\u003e\n\u003cli\u003eAmann J, Blasimme A, Vayena E, Frey D, Madai VI, the Precise4Q consortium. Explainability for artificial intelligence in healthcare: a multidisciplinary perspective. BMC Med Inform Decis Mak. 2020 Nov 30;20(1):310. doi:10.1186/s12911-020-01332-6\u003c/li\u003e\n\u003cli\u003eAlkhanbouli R, Matar Abdulla Almadhaani H, Alhosani F, Simsekler MCE. The role of explainable artificial intelligence in disease prediction: a systematic literature review and future research directions. BMC Med Inform Decis Mak. 2025 Mar 4;25(1):110. doi:10.1186/s12911-025-02944-6 PubMed PMID: 40038704; PubMed Central PMCID: PMC11877768.\u003c/li\u003e\n\u003cli\u003eAmannah C, Attai KF, Uzoka FM. A Data-Driven Intelligent Methodology for Developing Explainable Diagnostic Model for Febrile Diseases. Algorithms. 2025 Apr;18(4):190. doi:10.3390/a18040190\u003c/li\u003e\n\u003cli\u003eAttai K, Ekpenyong M, Amannah C, Asuquo D, Ajuga P, Obot O, et al. Enhancing the Interpretability of Malaria and Typhoid Diagnosis with Explainable AI and Large Language Models. Trop Med Infect Dis. 2024 Sep 16;9(9):216. doi:10.3390/tropicalmed9090216 PubMed PMID: 39330905; PubMed Central PMCID: PMC11436130.\u003c/li\u003e\n\u003cli\u003eEsan AO, Olawade DB, Soladoye AA, Omodunbi BA, Adeyanju IA, Aderinto N. Explainable AI for Parkinson\u0026rsquo;s disease prediction: A machine learning approach with interpretable models. Curr Res Transl Med. 2025 Dec;73(4):103541. doi:10.1016/j.retram.2025.103541 PubMed PMID: 40945155.\u003c/li\u003e\n\u003cli\u003eAttai KF, Amannah C, Ekpenyong M, Asuquo DE, Akputu OK, Obot OU, et al. Developing an Explainable Artificial Intelligence System for the Mobile-Based Diagnosis of Febrile Diseases Using Random Forest, LIME, and GPT. Healthc Inform Res. 2025 Apr 30;31(2):125\u0026ndash;35. doi:10.4258/hir.2025.31.2.125\u003c/li\u003e\n\u003cli\u003eAttai KF, Amannah C, Ekpenyong M, Baadel S, Obot O, Asuquo D, et al. Predicting Predisposition to Tropical Diseases in Female Adults Using Risk Factors: An Explainable-Machine Learning Approach. Information. 2025 Jul;16(7):520. doi:10.3390/info16070520\u003c/li\u003e\n\u003cli\u003eAhmed A, Oluomachi E, Abdullah A, Tochukwu N. Enhancing Data Privacy In Wireless Sensor Networks: Investigating Techniques And Protocols To Protect Privacy Of Data Transmitted Over Wireless Sensor Networks In Critical Applications Of Healthcare And National Security. Int J Netw Secur Its Appl. 2024 Mar 29;16(2):47\u0026ndash;63. doi:10.5121/ijnsa.2024.16204\u003c/li\u003e\n\u003cli\u003eSalih AM, Raisi‐Estabragh Z, Galazzo IB, Radeva P, Petersen SE, Lekadir K, et al. A Perspective on Explainable Artificial Intelligence Methods: SHAP and LIME. Adv Intell Syst. 2025 Jan;7(1):2400304. doi:10.1002/aisy.202400304\u003c/li\u003e\n\u003cli\u003eMujawar S, Sayed M, Chaware A. Empowering Decision Making in Healthcare - A Comparative Analysis of eXplainable AI Techniques Using SHAP and LIME for Cancer Patients Data. In: 2024 3rd Edition of IEEE Delhi Section Flagship Conference (DELCON) [Internet]. 2024 [cited 2025 May 19]. p. 1\u0026ndash;5. Available from: https://ieeexplore.ieee.org/abstract/document/10867049 doi:10.1109/DELCON64804.2024.10867049\u003c/li\u003e\n\u003cli\u003eIbrahim AM, Ahmed MM, Musa SS, Haruna UA, Hamid MR, Okesanya O, et al. Leveraging AI for early cholera detection and response: transforming public health surveillance in Nigeria. Explor Digit Health Technol. 2025. doi:10.37349/edht.2025.101140\u003c/li\u003e\n\u003cli\u003eOlayinka TA, Waghid Z, Mataka TW, Adegoke O. Demystifying Lesotho, Rwandan and Nigerian educators\u0026rsquo; viewpoints on smart technologies supporting AI in higher education. Educ Inf Technol. 2024;29:24285\u0026ndash;307. doi:10.1007/s10639-024-12820-x\u003c/li\u003e\n\u003cli\u003ePuijenbroek E van, Barry A, Khaemba C, Ntirenganya L, Gebreyesus TD, Fimbo A, et al. Short-Term Training, a Useful Approach for Sustainable Pharmacovigilance Knowledge Development in Tanzania, Kenya, Ethiopia and Rwanda. Drug Saf. 2024;47:1193\u0026ndash;202. doi:10.1007/s40264-024-01469-7\u003c/li\u003e\n\u003cli\u003eTownsend B, Sihlahla I, Naidoo M, Naidoo S, Donnelly D, Thaldar D. Mapping the regulatory landscape of AI in healthcare in Africa. Front Pharmacol. 2023;14. doi:10.3389/fphar.2023.1214422\u003c/li\u003e\n\u003cli\u003eDaroudi R, Sari AA, Nahvijou A, Faramarzi A. Cost per DALY averted in low, middle- and high-income countries: evidence from the global burden of disease study to estimate the cost-effectiveness thresholds. Cost Eff Resour Alloc CE. 2021;19. doi:10.1186/s12962-021-00260-0\u003c/li\u003e\n\u003cli\u003eSanni AO, Jonker A, Were V, Fasanmi O, Adebowale O, Shittu A, et al. Cost-effectiveness of One Health intervention to reduce risk of human exposure and infection with non-typhoidal salmonellosis (NTS) in Nigeria. One Health. 2024;18. doi:10.1016/j.onehlt.2024.100703\u003c/li\u003e\n\u003cli\u003eNeumann P, Anderson JE, Panzer A, Pope E, D\u0026rsquo;Cruz BN, Kim DD, et al. Comparing the cost-per-QALYs gained and cost-per-DALYs averted literatures. Gates Open Res. 2018;2. doi:10.12688/gatesopenres.12786.2\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-artificial-intelligence","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [BMC Artificial Intelligence](https://bmcartificialintel.biomedcentral.com)","snPcode":"44398","submissionUrl":"https://submission.nature.com/new-submission/44398/3","title":"BMC Artificial Intelligence","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Machine Learning, Disease Surveillance, Artificial Intelligence, Public Health, Outbreak Prediction, Digital Health, Explainable AI, Nigeria","lastPublishedDoi":"10.21203/rs.3.rs-8000523/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8000523/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDisease surveillance is fundamental to public health, enabling timely outbreak detection, efficient resource allocation, and evidence-based policymaking. In Nigeria, the Integrated Disease Surveillance and Response (IDSR) framework, while structured, is hampered by inconsistent data quality, limited private sector participation, and infrastructural constraints. Machine Learning (ML), a powerful subset of Artificial Intelligence (AI), offers transformative potential through its capacity for predictive analytics, real-time data processing, and automated pattern recognition to enhance surveillance capabilities. Despite global advancements in ML for disease forecasting and syndromic surveillance, its adoption within Nigeria's IDSR system lags considerably. This paper addresses this critical gap by investigating how ML can overcome Nigeria-specific barriers, such as fragmented data systems and rural connectivity deficits. We provide enhanced technical depth on ML methodologies, comparing supervised and unsupervised learning, and detailing relevant architectures, including Decision Trees, Recurrent Neural Networks (RNNs), and Convolutional Neural Networks (CNNs), suited for time-series epidemiological forecasting. We present a methodological illustration of malaria surveillance in Northern Nigeria using synthetic data, benchmarking five ML models (Linear Regression, Decision Tree, Random Forest, Gradient Boosting, and Support Vector Machine) under temporal validation, and employing SHapley Additive exPlanations (SHAP) for robust model interpretability. A sensitivity analysis further examines the stability of model performance under coefficient perturbations. A benchmarking analysis compares Nigeria's ML adoption against Rwanda and Kenya. Finally, we propose a tiered strategic framework encompassing policy, infrastructure, and capacity-building recommendations, complemented by a cost-benefit perspective emphasizing potential Disability-Adjusted Life Years (DALYs) averted and significant economic returns, aiming to foster a more resilient and equitable public health system in Nigeria.\u003c/p\u003e","manuscriptTitle":"Enhancing Disease Surveillance in Nigeria through Machine Learning: Opportunities, Challenges and Strategic Recommendations","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-06 12:52:43","doi":"10.21203/rs.3.rs-8000523/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-04-21T20:04:47+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"301964946037586657712299543714383991818","date":"2026-04-21T19:54:54+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-01T12:10:16+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-28T08:29:14+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Artificial Intelligence","date":"2026-03-25T08:59:37+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-artificial-intelligence","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [BMC Artificial Intelligence](https://bmcartificialintel.biomedcentral.com)","snPcode":"44398","submissionUrl":"https://submission.nature.com/new-submission/44398/3","title":"BMC Artificial Intelligence","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"7e9c97bc-3c69-4cb7-b061-24f831030775","owner":[],"postedDate":"May 6th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-06T12:52:43+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-06 12:52:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8000523","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8000523","identity":"rs-8000523","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}