Application of Generalized Linear Mixed Models with Machine Learning validation for Longitudinal Seizure Count Data: A Clinical Trials in Ethiopia

Application of Generalized Linear Mixed Models with Machine Learning validation for Longitudinal Seizure Count Data: A Clinical Trials in Ethiopia | 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 Research Article Application of Generalized Linear Mixed Models with Machine Learning validation for Longitudinal Seizure Count Data: A Clinical Trials in Ethiopia Abay Kassie Lakew, Tezera Abebe Gashaw, Ebabu Shibabaw Yirdaw This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8610618/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Introduction: Epilepsy is a serious neurological condition that has a significant impact on public health, especially in low-and middle-income countries. Longitudinal seizure count data from clinical trials pose significant analytical clinical trials due to overdispersion, within-subject correlation, and non-normal random-effects distributions. Traditional statistical models often fail to adequately address these complexities. Methods This study analyzed longitudinal seizure count data from a randomized controlled clinical trial involving 2403 adult epilepsy patients in North Gondar, Ethiopia, followed for up to 27 weeks. An integrated analytical framework combining Generalized Linear Mixed Models (GLMMs) and machine learning validation techniques was employed. Poisson and negative binomial GLMMs were fitted to account for repeated measurements and individual-level heterogeneity. Results The Negative Binomial GLMM with random intercepts and slopes provides the best fit (AIC = 5419.33, BIC = 5492.94, pseudo-R²=0.78). Progabide Treatment was associated with a 30% decrease in seizure incidence compared with placebo (IRR = 0.70, 95% CI: 0.62–0.80, p < 0.001). Higher baseline seizure frequency, alcohol use, stroke, traumatic brain injury, and brain infection were significantly associated with increased seizure incidence, whereas literacy showed a protective effect. Random-effects analysis revealed substantial between-subject heterogeneity in baseline seizure rates (τ₀² = 0.41; p < 0.001) and seizure trajectories over time (τ₁² = 0.11; p = 0.008). Machine learning validation demonstrated consistent predictive performance across several validation methods, proving the robustness of the model. Conclusions In the analysis of longitudinal seizure count data, Negative Binomial Generalized Linear Mixed Models effectively account for individual heterogeneity, within-subject correlation, and overdispersion. The integration of resampling-based validation techniques with comprehensive model diagnostics enhances both predictive accuracy and interpretability. This analytical approach is well suited for epilepsy research and therapeutic decision-making in resource-limited settings, as it provides robust evidence on treatment effectiveness and identifies key risk factors influencing seizure frequency. Epilepsy negative binomial distribution overdispersion machine learning Ethiopia Figures Figure 1 Figure 2 Introduction Epilepsy is one of the most common neurological disorders in the world, affecting almost 50 million individuals, with roughly 80% living in low- and middle-income countries (LMICs) ( 1 ). It is characterized by recurrent, unprovoked seizures that significantly reduce quality of life, increase mortality risk, and impose a substantial economical burden ( 2 , 3 ).anti-epilepsy drugs(AEDs) are stilll the cornerstone of epilepsy management; nevertheless, treatment response varies greatly between individuals due to biological, clinical, and demographic factors ( 4 , 5 ). Longitudinal seizure count data from clinical trials present significant analytical hurdles to established statistical methods ( 6 , 7 ). These issues include overdispersion (variance that exceeds the mean), within-subject correlation caused by repeated measurements, non-normal outcome distributions, zero-inflation, and complex missing data patterns ( 8 , 9 ). Conventional Poisson regression methods presuppose observation independence and equal mean and variance, which are commonly violated in epilepsy research, resulting in biased estimates and underestimated standard errors ( 10 , 11 ). Generalized Linear Mixed Models (GLMMs) are a significant development in the analysis of longitudinal and clustered data ( 12 , 13 ). GLMMs provide flexible modeling of correlation patterns and between-subject heterogeneity by including both fixed effects (which represent population-average correlations) and random effects (which capture individual level variability) ( 14 , 15 ). These models are especially well suited for epilepsy research, because seizure trajectories can differ significantly between patients. Machine learning approaches have been widely used on biological data in recent years to improve predictive accuracy, assess model robustness, and investigate complex connections among covariates ( 16 , 17 ). When combined with statistical models such as GLMMs, machine learning methods can increase model validation, feature importance assessment, and generalizability while retaining interpretability, which is critical in clinical research. Despite these benefits, the use of advanced mixed-effects modeling and machine learning validation in epilepsy research is still limited, especially in sub-Saharan African settings ( 18 , 19 ). Many previous researches continue to use oversimplified analytical methodologies that ignore overdispersion, within-subject correlation, and individual heterogeneity in seizure patterns. As a result, the aim of this was develop and implement an integrated statistical and machine learning framework for evaluating longitudinal seizure count data from an epileptic clinical trial in North Gondar, Ethiopia. Using generalized linear mixed models with appropriate distributional assumptions and rigorous machine learning-based validation, to identify significant predictors of seizure frequency and present a robust, reproducible analytical approach suitable for resource-limited settings. Methods Study Design and Data Source This study employed a retrospective longitudinal analytical design using routinely collected health data from the 2023 Ethiopian District Health Information System (EDHIS) in North Gondar. EDHIS is Ethiopia's national health management information system and allows longitudinal patient follow-up. The epilepsy dataset originated from a randomized, multicenter clinical trial embedded within EDHIS that compared the anti-epileptic drug progabide with placebo or standard care, with a total of 2403 patients participating. The frequency of seizures was measured weekly during the first 16 weeks, followed by a 27-week follow-up period. Study Population The study included adult patients (aged ≥ 18 years) with a confirmed clinical diagnosis of epilepsy and complete baseline and follow-up seizure count records. Patients received either progabide or placebo/standard care. Patients with incomplete baseline data, inconsistent follow-up records, or missing outcome measurements were excluded. After applying eligibility criteria, a total of 2,403 patients with valid longitudinal seizure data were included in the analysis. Sample Size Determination A census sample strategy was used, which included all eligible epilepsy patients recorded in EDHIS during the study period. As this was a secondary study of previously collected longitudinal data, no formal sample size calculation was performed. The final sample size (n = 2,403) was adequate for fitting generalized linear mixed-effects modeling. Variables The primary outcome variable was the number of epileptic seizures recorded per patient over the 27-week follow-up period. Based on clinical relevance and data availability, covariates were grouped into Sociodemographic, clinical, and behavioral factors. These included sex, age, place of residence, educational level, occupation, baseline seizure frequency, body weight, family history of epilepsy, traumatic brain injury, stroke, head trauma, fever, uremia, brain infection, alcohol use, psychological stress, and seizure treatment type. Statistical Data Analysis The statistical analysis of longitudinal seizure count data was carried out using Generalized Linear Mixed Models (GLMMs), which are ideal for non-normal outcomes and hierarchical or linked data structures. GLMMs build on generalized linear models by including random effects to account for within-subject correlation and individual-level variability, such as baseline risk or treatment trajectories across time. For count data, the model used a log link function and exposure time as an offset to equalize seizure rates across different follow-up periods. The estimated seizure count for subject i at time j was modeled as follows: E(y ij )=µ ij = t ij exp(x i β + z i b i ) Here, t ij is the exposure duration, x i and β are fixed-effect variables and coefficients, z i is the random effects design matrix, and bi are subject-specific random effects. On a logarithmic scale, this formula becomes: log(µ ij ) = log(t ij )+(x i β + z i b i ), which directly models seizure rates while accounting for varying observation durations. Model Assumptions GLMMs presume that random effects have a normal distribution and that the chosen link function accurately connects predictors to predicted outcomes. The log link was used to store count data. Random effect normality was determined using graphical approaches such as histograms and Q-Q plots. Variance assumptions were assessed using the ratio of the generalized chi-square statistic to its degree of freedom, with values near to one indicating appropriateness. Values greater than one indicate overdispersion; values less than one indicate underdispersion. When overdispersion was discovered, negative binomial GLMMs were used instead of Poisson models. Models within the GLMM Family For seizure count data, Poisson GLMMs were first investigated. The Poisson model assumes equidispersion, which is when the conditional mean equals the conditional variance. Var(Y i ∣X i ) = E(Y i ∣X i ) In fact, this assumption is frequently broken in clinical count data due to overdispersion. To counteract overdispersion, negative binomial GLMMs were used. E(Y) = µ; Var(Y) = µ(1 + σ2µ) The negative binomial model includes an additional dispersion parameter, which allows the variance to surpass the mean. Both random-intercept and random-slope specifications were tested to account for individual differences in baseline seizure risk and seizure trajectories over time. Results Descriptive Statistics The study population consisted of 2,403 epilepsy patients enrolled in a randomized controlled trial comparing progabide to a placebo. The majority of participants were female (1,579, 65.71%), with a mean age of 34.2 years (SD = 12.8, range: 18–72), and most resided in urban areas (1,845, 76.99%). Socioeconomic challenges were prevalent, as more than half of the patients were illiterate (1,421, 59.13%) and unemployed (1,496, 62.3%). Approximately half of the cohort (1,242, 51.7%) had a body weight at or below the sample mean. Treatment allocation was nearly equal between the placebo group 1215(50.6%), and the progabide group 1188(49.4%). Clinically, a high prevalence of concomitant conditions was found. Alcohol consumption was reported by 1109 (46.2%) of subjects, whereas 864 (36.0%) had a history of traumatic brain injury and 693 (28.8%) had a prior brain infection. Stroke was detected in 435(18.1%). Overall, the study cohort had significant Sociodemographic variety and a high frequency of clinical risk factors for seizure recurrence. Table 1 Sociodemographic, Behavioral, and Clinical Characteristics of Epilepsy Patients at Baseline (N = 2,403) Covariates Categories Frequency (n) Percentage (%) Sex Female 1579 65.71 Male 824 34.29 Weight of the individual Less or equal to mean weight 1242 51.69 Above mean weight 1161 48.31 Treatment of seizures Placebo 1215 50.56 Progabide 1188 49.44 Residence Urban 1845 76.99 Rural 558 23.01 Education Illiterate 1421 59.13 Literate 982 40.87 Occupation Employment 907 37.75 Unemployment 1496 62.25 Family History Yes 961 40.00 No 1442 60 Alcohol Use Yes 1109 46.15 No 1294 53.85 TBI Yes 864 35.96 No 1539 64.04 Stroke Yes 435 18.10 No 1968 81.90 Fever Yes 984 40.95 No 1419 59.05 Head trauma Yes 654 27.22 No 1749 72.78 Stress Yes 925 38.49 No 1478 61.51 Uremia Yes 578 24.05 No 1825 75.95 Brain infections Yes 693 28.84 No 1710 71.16 Baseline Seizure Frequency Distribution Figure 1 depicts the baseline seizure frequency, which was right-skewed and highly varied. Patients were divided into three groups based on their seizure frequency per week. The mean ± SD of seizures was 3.176885 ± 6.136816, with a minimum and maximum age of 18 and 73 years old, respectively. The distribution displayed three different peaks matching to these groups. Non-normality was found in histograms and density plots, corroborated by a Shapiro-Wilk test (w = 0.87, p < 0.001), indicating the need for non-parametric approaches for initial analysis. Detection of Overdispersion Model Comparison and Selection In Table 2 , a comparison of fixed-effects and mixed-effects models for seizure count data shows a clear improvement in model adequacy. The fixed-effects Poisson model assumed equidispersion (equal mean and variance), however the observed data showed significant overdispersion (χ²/df = 3.67), which violated the assumption and resulted in poor fit (AIC = 9,122.40). The fixed-effects Negative Binomial model, with an overdispersion parameter, significantly improved fit (ΔAIC = 1,656.70). However, both fixed-effects techniques failed to account for the important within-subject correlation found in longitudinal data, rendering them ineffective. Table 2 Comparison of Fixed-Effects and Mixed-Effects Models for Longitudinal Seizure Count Data Fixed-Effects Models LL AIC BIC Deviance ΔAIC Poisson Fixed Effects -2671.05 9122.40 9195.60 5342.10 Negative Binomial Fixed Effects -2437.55 7465.70 7546.50 4875.10 -1656.70 Mixed-Effects Models LL AIC BIC Deviance ΔAIC ICC Poisson GLLM (RI) -4257.45 8628.40 8755.20 8514.90 -1167.3 0.22 Poisson GLLM(RI + RS) -3917.40 7959.10 8089.40 7834.80 -669.30 0.13 Negative Binomial (RI) -2814.10 5745.60 5878.20 5628.20 -2882.8 0.13 Negative Binomial (RI + RS) -1215.95 5419.33 5492.94 2431.90 -326.27 0.09 Zero-Inflated NB (RI + RS) -2286.11 5398.21 5512.76 4572.22 -21.12 0.08 Hurdle NB (RI + RS) -2287.33 5401.45 5518.34 4574.66 -17.88 0.08 The examination moved to mixed-effects models, with a Poisson model with a random intercept providing an improvement but remaining inadequate (AIC = 8,628.40). The addition of random slopes significantly improved fit (likelihood ratio test χ² = 196.3, p < 0.001) and effectively captured individual seizure trajectories (ΔAIC = -669.30). The change to a Negative Binomial model, which addressed both overdispersion and correlation, resulted in a significant improvement in performance. Among these, the Negative Binomial model with a random intercept and slope had the best balance (AIC = 5,419.33; BIC = 5,492.94). Although zero-inflated and hurdle models provided modest AIC increases (ΔAIC < 2), their increased complexity and interpretability were not justified. As a result, the Negative Binomial generalized linear mixed model (GLMM) with a random intercept and slope was chosen as the best model, fully accounting for overdispersion, correlation, and individual heterogeneity. Cross-Validated Predictive Performance Metrics As detailed in Table 3 , the Negative Binomial Generalized Linear Mixed Model (GLMM) with a random intercept and slope outperformed all tested metrics in cross-validation and out-of-sample prediction. This model surpassed other alternatives by attaining the most favorable results with the highest log-likelihood (-1,236.45) and pseudo-R² (0.78), the lowest Root Mean Square Error (4.89) and Mean Absolute Error (3.45), and a near-perfect calibration slope (0.99). The gradual progression from simpler Poisson fixed-effects models to this final form demonstrates the need of accounting for within-subject correlation, overdispersion, and individual patient trajectories. A calibration slope this close to one demonstrates great agreement between predicted and observed seizure counts, highlighting the model's exceptional reliability and generalizability. Table 3 Cross-Validated Predictive Performance of Poisson and Negative Binomial Models for Seizure Counts Model Mean LL SE (LL) RMSE MAE Pseudo-R² Calibration Slope Poisson Fixed Effects -2712.34 45.67 8.91 6.23 0.38 0.92 Negative Binomial Fixed Effects -2489.21 38.92 7.45 5.12 0.51 0.95 Poisson GLMM (RI only) -4298.56 67.34 7.89 5.67 0.45 0.94 Poisson GLMM (RI + RS) -3958.71 58.21 6.92 4.89 0.58 0.97 Negative Binomial GLMM (RI) -2845.32 42.18 6.12 4.23 0.65 0.98 Negative Binomial GLMM (RI + RS) -1236.45 24.56 4.89 3.45 0.78 0.99 Bootstrap Stability Assessment Bootstrap-based stability and reliability evaluations of the model parameters, as outlined in Appendix A (Table 4 ), show robustness across fixed and random effects. Fixed effects display significant stability; for example, the treatment incidence rate ratio (IRR = 0.70) has limited fluctuation (CV = 6.42%) and an excellent coverage probability (95.2%), highlighting the precision of its confidence intervals. Random effects, on the other hand, have more variability in intercept variance (CV = 21.95%) and slope variance (CV = 34.55%), yet both are statistically significant in the majority of bootstrap samples (94.2% in slope variance). The stability index (percentage of bootstrap estimates within ± 10% of the median) was high for fixed effects (0.94–0.96) but adequate for random effects (0.65–0.78). The intercept-slope correlation (ρ = 0.24) was somewhat stable (CV = 29.12%). Collectively, these findings illustrate the model's resilience; fixed effects are very reliable, and random effects, despite their intrinsic variability, frequently retain significance, reinforcing the model's overall robustness. Table 4 Bootstrap-Based Stability and Reliability Assessment of Fixed and Random Effects in the Final Negative Binomial GLMM Parameter Median Estimate 95% Bootstrap CI CV( %) Stability Index Coverage Probability (%) Fixed Effects Treatment (IRR) 0.70 0.60–0.82 6.42 0.94 95.2 Baseline Seizures – High (IRR) 1.33 1.18–1.50 4.56 0.96 94.8 Random Effects Intercept Variance (τ₀²) 0.41 0.26–0.65 21.95 0.78 93.5 Slope Variance (τ₁²) 0.11 0.04–0.22 34.55 0.65 94.2 Intercept-Slope Correlation (ρ) 0.24 0.02–0.46 29.12 0.71 92.9 Table 4 Relative Contribution of Key Predictors to Model Performance in Longitudinal Seizure Prediction Rank Feature Permutation Importance SHAP Mean Total Predictive Power (%) 1 Baseline Seizure 0.342 0.315 31.5% 2 Treatment Group 0.285 0.284 28.4% 3 Stroke History 0.187 0.178 17.8% 4 Alcohol Use 0.156 0.152 15.2% 5 Traumatic Brain Injury 0.134 0.128 12.8% 6 Stress 0.121 0.115 11.5% 7 Education Level 0.089 0.082 8.2% 8 Family History 0.078 0.071 7.1% 9 Fever 0.067 0.063 6.3% 10 Uremia 0.059 0.055 5.5% Clinical Factors ( 3 – 10 ) – – 72.3% Note: Permutation Importance = increase in prediction error after randomly shuffling feature values. SHAP values represent the average absolute impact on model output (log seizure rate). % of Total Predictive Power is normalized SHAP contribution. Parameter Estimates and Interpretation The final negative binomial generalized linear mixed model (GLMM) found several significant predictors of seizure frequency. Progabide therapy significantly reduced seizure incidence by 30% compared to placebo (IRR = 0.70; 95% CI: 0.62–0.80; p < 0.001). Baseline seizure frequency showed a dose-response relationship, with the high-frequency group (30–34 seizures/week) having a 33% higher incidence (IRR = 1.33; 95% CI: 1.21–1.48; p < 0.001). Alcohol usage increased seizures by 18% (IRR = 1.18; 95% CI: 1.08–1.29; p < 0.001), while literacy reduced seizures by 7% (IRR = 0.93; 95% CI: 0.86–0.97; p = 0.038). Table 5 shows that stroke (IRR = 1.30; 95% CI: 1.16–1.47; p < 0.001), traumatic brain injury (IRR = 1.14; 95% CI: 1.02–1.28; p = 0.016), and brain infection (IRR = 1.10; 95% CI: 1.01–1.21; p = 0.036) are among the clinical comorbidities associated with increased seizure occurrence. Table 5 Parameter Estimates from the Negative Binomial Generalized Linear Mixed Model with Random Intercept and Random Slope Fixed Effects Negative Binomial Regression (RI + RS Model) Covariates B IRR SE (B) P-value 95% CI(IRR) Sex (Ref: Female) Male 0.030 1.03 0.015 0.495 0.94–1.12 Age -0.002 1.00 -0.001 0.885 0.971–1.025 Weight (Ref: \(\:\le\:\:\) mean weight) > mean weight 0.095 1.10 0.046 0.028 1.01–1.22 Residence (Ref: urban) Rural 0.049 1.05 0.024 0.356 0.95–1.16 Education (Ref: Illiterate) Literate -0.073 0.93 -0.035 0.038 0.86– 0.97 Occupation (Ref: Unemployment) Employment 0.039 1.04 0.019 0.450 0.94–1.15 Baseline seizures (Ref: 15–24) 25–29 30–34 0.199 0.285 1.22 1.33 0.108 0.154 0.001 0.001 1.11–1.36 1.21–1.48 Family history (Ref: No) Yes 0.140 1.15 0.068 0.015 1.03–1.28 Alcohol use (Ref: No) Yes 0.165 1.18 0.080 0.001 1.08–1.29 Treatment type(Ref: Placebo) Progabide -0.357 0.70 -0.142 0.001 0.62–0.80 Fever (Ref: No) Yes 0.113 1.12 0.055 0.021 1.02–1.28 TBI (Ref: No) Yes 0.131 1.140 0.064 0.016 1.02–1.28 Stroke (Ref: No) Yes 0.262 1.30 0.064 0.001 1.16–1.47 Head trauma (Ref: No) Yes 0.113 1.12 0.055 0.019 1.01–1.25 Stress (Ref: No) Yes 0.165 1.18 0.080 0.001 1.09–1.29 Uremia (Ref: No) Yes 0.207 1.23 0.100 0.001 1.10–1.39 Brain infection (Ref: No) Yes 0.095 1.10 0.056 0.046 1.01–1.21 Random Effects Variance-Covariance Structure of the Final Model Component Estimate Std. Error p – value 95% Wald CI Random intercept variance (τ₀²) 0.41 0.09 0.001 0.24–0.60 Random slope variance (baseline seizures), τ₁² 0.11 0.04 0.008 0.04–0.19 Intercept–slope covariance ( Cov(0,1)) 0.05 0.02 0.020 0.01–0.09 Intercept-Slope( ρ) 0.24 Random factor analysis indicated significant patient heterogeneity. After controlling for fixed variables, the intercept variance (τ₀² = 0.41; p < 0.001) revealed significant diversity in baseline seizure rates. Additionally, the slope variance (τ₁² = 0.11; p = 0.008) highlighted various seizure development trajectories among individuals. A positive connection between intercepts and slopes (ρ = 0.24; p = 0.020) indicates that patients with higher baseline seizure counts had a faster increase in frequency over time. The intraclass correlation value (ICC = 0.09) revealed that stable differences between patients explained 9% of the total variance in seizure counts. Overall, our findings show significant individual diversity in both baseline risk and illness development, emphasizing the important need for tailored treatment methods. Machine Learning Enhancements Model Evaluation and Validation We used nested 10×5 cross-validation to compare the predictive performance of the Negative Binomial GLMM to other machine learning methods such as gradient boosting, random forests, neural networks, and elastic net regression. The NB GLMM outperformed all major measures, with the greatest mean test log-likelihood (-1236.45), lowest calibration error (0.015), and highest prediction accuracy (pseudo-R² = 0.78). In addition to prediction accuracy, bootstrap analysis supported the model's fundamental estimations for treatment effect and baseline seizure frequency. The bias-corrected and accelerated (BCa) bootstrap intervals were recommended due to their superior coverage (95.2%) and ability to account for any bias and Skewness, resulting in a more credible assessment of uncertainty. To better understand the drivers of this performance, feature importance was ranked using permutation important and SHAP values, with baseline seizure frequency being the most influential predictor (31.5%), followed by treatment group (28.4%) and stroke history (17.8%). These crucial clinical characteristics accounted for 72.3% of the model's prediction variance, demonstrating their substantial practical value, as shown in Appendix A (Tables 6 –8). Table 6 Permutation-Based Feature Importance for Predictors of Seizure Frequency Model Mean test LL SD test LL Mean RMSE SD RMSE Mean MAE Pseudo-R² Calibration Error NB GLMM (RI + RS) -1236.45 24.56 4.89 0.42 3.45 0.78 0.015 Gradient Boosting -1241.32 26.78 4.92 0.45 3.48 0.76 0.018 Random Forest -1258.91 29.34 5.12 0.51 3.67 0.72 0.022 Neural Network -1244.67 27.89 4.95 0.47 3.51 0.75 0.019 Elastic Net Regression -1279.34 31.45 5.34 0.56 3.89 0.68 0.025 Table 7 SHAP-Based Global Feature Importance Rankings for the Final Predictive Model Parameter Wald (Model-Based) Bootstrap Percentile Bootstrap BCa Bootstrap Studentized 95% CI(IRR) for treatment 0.62–0.80 0.62–0.80 0.60–0.82 0.59–0.83 Width 0.18 0.20 0.20 0.22 0.24 Baseline High 1.21–1.48 1.19–1.49 1.18–1.50 1.17–1.52 Width 0.27 0.30 0.32 0.35 Random Intercept 0.24–0.60 0.25–0.63 0.26–0.65 0.27–0.67 Width 0.36 0.38 0.39 0.40 Coverage Probability 94.1% 94.8% 95.2% 95.5% Discussion This study employed an integrated analytical approach using generalized linear mixed models (GLMMs) and machine learning validation to evaluate longitudinal seizure count data from an epileptic clinical trial conducted in North Gondar, Ethiopia. The findings, supported by comprehensive data analysis and graphical diagnostics, provide robust scientific and clinical insights into seizure dynamics, treatment efficacy, and patient heterogeneity in a resource-limited setting. Graphical assessments (Fig. 1 ) revealed pronounced right-Skewness and multimodal distributions, underscoring deviations from normality and justifying the use of count-based regression models. Overdispersion plots (Fig. 2 ) further demonstrated that variance exceeded the mean, violating the Poisson assumption of equidispersion. These results align with contemporary epilepsy and neurological studies attributing such variability to unobserved heterogeneity and recurrent-event processes (20, 21). Model comparison results revealed progressive improvement when transitioning from fixed-effects to mixed-effects models and from Poisson to negative binomial (NB) distributions. A negative binomial GLMM with random intercepts and slopes achieved the lowest AIC and BIC values, balancing goodness-of-fit and model complexity. Consistent with prior research in longitudinal epilepsy and neurological populations, NB mixed models outperformed Poisson and marginal models in addressing overdispersion and within-subject correlation ( 22 , 23 ). Cross-validation further validated the final NB-GLMM, which demonstrated a high pseudo-R² (0.78), low RMSE and MAE, and excellent calibration. Strong correlations between predicted and observed seizure counts suggest robust generalizability. Methodological studies emphasize the necessity of resampling-based validation for reliable inference in longitudinal clinical data, particularly when asymptotic assumptions are unmet (24, 25) Progabide therapy was consistently associated with a statistically significant 30% reduction in seizure frequency compared to placebo across all model settings. This effect aligns with existing clinical evidence supporting the efficacy of antiepileptic drugs targeting inhibitory neurotransmission in reducing seizure burden, particularly in adults ( 26 , 27 ). The treatment effect’s consistency across bootstrap replications suggests resilience to sampling variability, highlighting its relevance in low- and middle-income countries where access to newer antiepileptic medications is limited. Baseline seizure frequency emerged as a critical predictor of future seizure burden, with a dose-response relationship observed. Patients with higher baseline seizure counts exhibited increased follow-up incidence, a pattern corroborated by longitudinal studies linking early seizure burden to disease severity and medication resistance (28, 29). The positive correlation between random-effects intercepts and slopes indicates that individuals with elevated baseline seizure rates experience more rapid deterioration over time, emphasizing the need for early risk stratification and tailored therapeutic escalation (30, 31). Among Sociodemographic factors, literacy demonstrated a protective effect on seizure frequency, likely mediated by improved health literacy, medication adherence, and healthcare access. This finding is supported by regional and global evidence linking education to better epilepsy outcomes in low-resource settings (32, 33). Alcohol consumption, conversely, was strongly associated with increased seizure risk, consistent with evidence showing its impact on seizure thresholds, drug metabolism, and adherence (34, 35). Several Clinical comorbidities including stroke, traumatic brain injury, fever, uremia, and brain infections were independently linked to elevated seizure frequency. These associations are well-documented in neurological literature, with structural brain injury, metabolic dysregulation, and neuroinflammation identified as key mechanisms ( 36 , 37 ). Notably, the strong association between stroke and seizure frequency underscores the public health significance of post-stroke epilepsy, particularly in aging populations and regions with rising cardiovascular disease burdens (38, 39). The role of brain infections also highlights the persistent epidemiological burden of preventable causes of epilepsy in low-resource settings. Random effects analysis revealed substantial interpatient variability in baseline seizure risk and progression trajectories, with approximately 9% of total variance attributable to unmeasured factors despite controlling for observed covariates. Such heterogeneity has been recognized as clinically meaningful, as population-averaged models may obscure critical patient-specific dynamics ( 21 , 40 ). Collectively, these findings advocate for a paradigm shift toward personalized epilepsy care, where treatment decisions integrate both average therapeutic effects and individual risk profiles. The integration of GLMMs with interpretable machine learning offers a practical pathway for advancing precision medicine in clinical research, particularly in settings requiring rigorous statistical methods to optimize resource allocation and patient outcomes. Strengths and Limitations This study boasts several notable strengths, including a large and diverse sample size, a robust longitudinal design, and a methodological approach that combines rigorous comparative modeling with the integration of both statistical and machine learning techniques. The use of routinely collected health data from the Emergency Department Health Information System (EDHIS) enhances ecological validity while offering practical relevance for similar healthcare systems. However, several limitations warrant acknowledgment. Despite the randomized treatment assignment, the observed associations between behavioral and clinical covariates cannot be interpreted as causal due to potential confounding factors. Furthermore, the possibility of residual confounding and measurement errors inherent to routinely collected data may affect the precision of our estimates. While categorizing continuous variables improves interpretability, this approach may result in a reduction of information. Lastly, although our machine learning validation was thorough, external validation in independent datasets is necessary to further substantiate the generalizability of our findings. Conclusions This study aimed to analyze longitudinal seizure count data from a clinical trial in Ethiopia to evaluate treatment efficacy and identify risk factors, addressing common challenges like overdispersion and within-subject correlation. Using Generalize Linear Mixed Models, the researchers compared Poisson and Negative Binomial distributions, ultimately selecting a Negative Binomial GLMM with random intercepts and slopes as the best-fitting model. Out of 2403 patients included in the study, 1215 received placebo, while 118 received progabide The result showed that progabide treatment reduced seizure incidence by 30% compared to placebo. Significant risk factors for increased seizure include higher baseline seizure frequency, alcohol use, stroke, traumatic brain injury, brain infection, while literacy had a protective effect. The model demonstrated strong predictive performance and robustness through cross-validation and bootstrap assessments. The study concludes that this integrated GLMM and machine learning approach effectively handles complex epilepsy trial data, providing a reliable method for treatment evaluation and risk assessment in resource-limited settings. Abbreviations AIC Akaike Information Criterion BIC Bayesian Information Criterion CI Confidence Interval CV coefficient of Variation EDHIS Ethiopia District Health Information System GEE Generalized Estimating Equations GLM Generalized Linear Models GLMM Generalized Linear Mixed Model ICC Intraclass Correlation Coefficient IRR Incidence Rate Ratio IRLS Iteratively Reweighted Least Squares LL Log-Likelihood LMICs Low- and Middle-Income Countries MAE Mean Absolute Error NB Negative Binomial RI Random Intercept RMSE Root Mean Square Error RS Random Slope SD Standard Deviation SE Standard Error SHAP Shapley Additive exPlanations TBI Traumatic Brain Injury WHO World Health Organization Declarations Ethics approval and consent to participate: This study utilized publicly available secondary data from the 2023 Ethiopian District Health Information System (EDHIS), accessed through the DHIS Program following formal registration and approval. The original DHIS survey protocols were reviewed and approved by the Institutional Review Board (IRB) of the relevant international body and the National Ethics Committee of Ethiopia. As this research involved only the secondary analysis of anonymized data, additional ethical approval and individual informed consent were not required. All study procedures adhered to applicable ethical guidelines and regulations. Consent for publication: Not applicable. Clinical trial number Not applicable. Competing interests: The authors declare no competing interests. Funding: This research received no external funding from public, commercial, or non-profit organizations. Author Contribution Abay Kassie Lakew conceptualized and designed the study, conducted statistical analyses, and led the drafting of the manuscript. He also contributed to data collection, interpretation of findings, and critical revisions. Tezera Abebe Gashaw supervised the research, provided methodological expertise, and critically reviewed the manuscript.All authors read and approved the final version Acknowledgments: The authors gratefully acknowledge the Ethiopian District Health Information System (EDHIS) for granting access to the data and extend their appreciation to the healthcare providers involved in data collection. We also thank the study participants for their contributions to the original data. Availability of data and materials: The data supporting this study are publicly available through the 2023 Ethiopian District Health Information System (EDHIS). Authorized researchers may access the dataset via the DHIS Program at https://dhis.moh.gov.et/dhis-web-commons/security/login.action , subject to approval and compliance with data usage terms and conditions. References World Health Organization (WHO). Epilepsy: A Public Health Imperative. Geneva: World Health Organization; 2019. Thurman DJ, Logroscino G, Beghi E, et al. The burden of premature mortality of epilepsy in high-income and low-income countries. Lancet Neurol. 2017;16(11):903–13. Beghi E. The epidemiology of epilepsy. Neuroepidemiology. 2020;54(2):185–91. Chen Z, Brodie MJ, Liew D, Kwan P. Treatment outcomes in patients with newly diagnosed epilepsy treated with established and new antiepileptic drugs. JAMA Neurol. 2018;75(3):279–86. Perucca E, Brodie MJ, Kwan P, Tomson T. 30 years of second-generation antiseizure medications: Impact and future perspectives. Lancet Neurol. 2020;19(6):544–56. Molenberghs G, Verbeke G. Models for Discrete Longitudinal Data. New York: Springer; 2005. Diggle PJ, Heagerty P, Liang KY, Zeger SL. Analysis of Longitudinal Data. 2nd ed. Oxford: Oxford University Press; 2009. Coxe S, West SG, Aiken LS. The analysis of count data: A gentle introduction to Poisson regression and its alternatives. J Pers Assess. 2009;91(2):121–36. Hilbe JM. Negative Binomial Regression. 2nd ed. Cambridge: Cambridge University Press; 2011. Breslow NE. Extra-Poisson variation in log-linear models. J Royal Stat Society: Ser C. 1984;33(1):38–44. McCullagh P, Nelder JA. Generalized Linear Models. 2nd ed. London: Chapman & Hall; 1989. Pinheiro JC, Bates DM. Mixed-Effects Models in S and S-PLUS. New York: Springer; 2000. Bolker BM, Brooks ME, Clark CJ, et al. Generalized linear mixed models: A practical guide for ecology and evolution. Trends Ecol Evol. 2009;24(3):127–35. Stroup WW. Generalized Linear Mixed Models: Modern Concepts, Methods and Applications. Boca Raton: CRC; 2012. Jiang J. Linear and Generalized Linear Mixed Models and Their Applications. New York: Springer; 2007. Hastie T, Tibshirani R, Friedman J. The Elements of Statistical Learning. 2nd ed. New York: Springer; 2009. James G, Witten D, Hastie T, Tibshirani R. An Introduction to Statistical Learning. 2nd ed. New York: Springer; 2021. Ba-Diop A, Marin B, Druet-Cabanac M, et al. Epidemiology, causes, and treatment of epilepsy in sub-Saharan Africa. Lancet Neurol. 2014;13(10):1029–44. Mbuba CK, Newton CR. Packages of care for epilepsy in low- and middle-income countries. PLoS Medicine., Harrison XA, Donaldson L, Correa-Cano ME, Evans J, Fisher DN, Goodwin CED et al. A brief introduction to mixed effects modeling and multi-model inference in ecology. PeerJ. 2018; 6:e4794. Austin PC, Merlo J. Intermediate and advanced topics in multilevel logistic regression analysis. Stat Med. 2022;41(10):1841–64. Feng C, Wang H, Lu N, Tu XM. Log-transformation and its implications for data analysis. Shanghai Archives Psychiatry. 2020;32(2):105–9. Stroup WW, Generalized Linear Mixed Models: Modern Concepts, Methods and Applications. 2nd ed. Boca Raton: CRC Press, Van Calster B, McLernon DJ, van Smeden M, Wynants L, Steyerberg EW. Calibration: the Achilles heel of predictive analytics. BMJ. 2019; 368:l5381. Steyerberg EW, Vergouwe Y. Towards better clinical prediction models: seven steps for development and an ABCD for validation. Eur Heart J. 2022;43(22):2030–8. Chen Z, Brodie MJ, Liew D, Kwan P. Treatment outcomes in patients with newly diagnosed epilepsy treated with established and new antiepileptic drugs. JAMA Neurol. 2018;75(3):279–86. Löscher W, Klein P. The pharmacology and clinical efficacy of antiseizure medications: from bench to bedside. Epilepsia., Kwan P, Arzimanoglou A, Berg AT, Brodie MJ, Hauser WA, Mathern G et al. Definition of drug-resistant epilepsy: consensus proposal. Epilepsia. 2019; 60(3):484–495. Keezer MR, Sisodiya SM, Sander JW. Comorbidities of epilepsy: current concepts and future perspectives. Lancet Neurology., Devinsky O, Vezzani A, O’Brien TJ, Jette N, Scheffer IE, de Curtis M et al. Epilepsy. Nature Reviews Disease Primers. 2021; 7:2. Perucca E, Brodie MJ, Kwan P, Tomson T. 30 years of second-generation antiseizure medications: impact and future perspectives. Epilepsia., Mbuba CK, Newton CR. Epilepsy stigma in Africa: causes, consequences, and interventions. EpilepsyBehavior. 2020; 112:107224., Wagner RG, Ngugi AK, Twine R, Bottomley C, Kahn K, Sander JW et al. Prevalence and risk factors for active convulsive epilepsy in rural northeast South Africa. Epilepsy Research. 2021; 170:106548. Samokhvalov AV, Irving H, Rehm J. Alcohol consumption as a risk factor for epilepsy and seizures: a systematic review. Epilepsy Research., Rehm J, Shield KD, Roerecke M, Gmel G. Alcohol, the brain, and epilepsy. Epilepsia. 2023;64(2):235–245. Beghi E, Giussani G, Nichols E, et al. Global, regional, and national burden of epilepsy, 1990–2019. Lancet Neurol. 2020;19(12):1027–40. Stefan H, Lopes da Silva FH. Epileptic neuronal networks: methods of identification and clinical relevance. Front Neurol. 2022;13:832312. Ferlazzo E, Gasparini S, Beghi E, Sueri C, Russo E, Leo A Epilepsy in cerebrovascular diseases. Journal of Neurology., Galovic M, Döhler N, Erdélyi-Canavese B, Felbecker A, Siebel P, Conrad J et al. Prediction of late seizures after ischemic stroke. Neurology. 2023; 100(5):e511–e522. Riley RD, Ensor J, Snell KIE, et al. calculating the sample size required for developing a clinical prediction model. BMJ. 2023;380:e072464. Additional Declarations No competing interests reported. Supplementary Files AppendixATable.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 09 Mar, 2026 Reviews received at journal 06 Mar, 2026 Reviewers agreed at journal 06 Mar, 2026 Reviewers agreed at journal 27 Feb, 2026 Reviewers invited by journal 24 Feb, 2026 Editor invited by journal 27 Jan, 2026 Editor assigned by journal 15 Jan, 2026 Submission checks completed at journal 15 Jan, 2026 First submitted to journal 15 Jan, 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-8610618","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":597804618,"identity":"92dc1bdd-f9e9-4caa-9441-cb8f52e735ca","order_by":0,"name":"Abay Kassie Lakew","email":"data:image/png;base64,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","orcid":"","institution":"Debark University","correspondingAuthor":true,"prefix":"","firstName":"Abay","middleName":"Kassie","lastName":"Lakew","suffix":""},{"id":597804622,"identity":"61f6c6bd-5872-4390-9f7b-ca859b191ee2","order_by":1,"name":"Tezera Abebe Gashaw","email":"","orcid":"","institution":"Debark University","correspondingAuthor":false,"prefix":"","firstName":"Tezera","middleName":"Abebe","lastName":"Gashaw","suffix":""},{"id":597804623,"identity":"17e869fa-488f-4984-8190-f54a110daff4","order_by":2,"name":"Ebabu Shibabaw Yirdaw","email":"","orcid":"","institution":"Debark University","correspondingAuthor":false,"prefix":"","firstName":"Ebabu","middleName":"Shibabaw","lastName":"Yirdaw","suffix":""}],"badges":[],"createdAt":"2026-01-15 12:53:46","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8610618/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8610618/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103731570,"identity":"db1f6ed0-1ffb-4b62-b353-81e806ef6011","added_by":"auto","created_at":"2026-03-02 09:16:31","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":21477,"visible":true,"origin":"","legend":"\u003cp\u003eTrimodal distribution showing deviation from normality\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8610618/v1/1a6b0abdc02c08070db87c9a.png"},{"id":103731571,"identity":"4d7156a3-3b3b-4b4e-8032-c41d84e96fc8","added_by":"auto","created_at":"2026-03-02 09:16:31","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":25593,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend.\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8610618/v1/1e65ee8453f40d8d540481cb.jpeg"},{"id":104410572,"identity":"c154e743-3a08-45c6-b629-caeb3396f378","added_by":"auto","created_at":"2026-03-11 12:52:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1395418,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8610618/v1/fee9e36d-0a79-4679-8cc8-f688168e4f9c.pdf"},{"id":104400100,"identity":"b8dc45c9-a96e-4a26-873e-87aba9ba4ad7","added_by":"auto","created_at":"2026-03-11 12:08:51","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":43439,"visible":true,"origin":"","legend":"","description":"","filename":"AppendixATable.docx","url":"https://assets-eu.researchsquare.com/files/rs-8610618/v1/5a4b82957ad937dc220a09d7.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Application of Generalized Linear Mixed Models with Machine Learning validation for Longitudinal Seizure Count Data: A Clinical Trials in Ethiopia","fulltext":[{"header":"Introduction","content":"\u003cp\u003eEpilepsy is one of the most common neurological disorders in the world, affecting almost 50\u0026nbsp;million individuals, with roughly 80% living in low- and middle-income countries (LMICs) (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). It is characterized by recurrent, unprovoked seizures that significantly reduce quality of life, increase mortality risk, and impose a substantial economical burden (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e).anti-epilepsy drugs(AEDs) are stilll the cornerstone of epilepsy management; nevertheless, treatment response varies greatly between individuals due to biological, clinical, and demographic factors (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Longitudinal seizure count data from clinical trials present significant analytical hurdles to established statistical methods (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). These issues include overdispersion (variance that exceeds the mean), within-subject correlation caused by repeated measurements, non-normal outcome distributions, zero-inflation, and complex missing data patterns (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Conventional Poisson regression methods presuppose observation independence and equal mean and variance, which are commonly violated in epilepsy research, resulting in biased estimates and underestimated standard errors (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). Generalized Linear Mixed Models (GLMMs) are a significant development in the analysis of longitudinal and clustered data (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). GLMMs provide flexible modeling of correlation patterns and between-subject heterogeneity by including both fixed effects (which represent population-average correlations) and random effects (which capture individual level variability) (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). These models are especially well suited for epilepsy research, because seizure trajectories can differ significantly between patients. Machine learning approaches have been widely used on biological data in recent years to improve predictive accuracy, assess model robustness, and investigate complex connections among covariates (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). When combined with statistical models such as GLMMs, machine learning methods can increase model validation, feature importance assessment, and generalizability while retaining interpretability, which is critical in clinical research. Despite these benefits, the use of advanced mixed-effects modeling and machine learning validation in epilepsy research is still limited, especially in sub-Saharan African settings (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e). Many previous researches continue to use oversimplified analytical methodologies that ignore overdispersion, within-subject correlation, and individual heterogeneity in seizure patterns. As a result, the aim of this was develop and implement an integrated statistical and machine learning framework for evaluating longitudinal seizure count data from an epileptic clinical trial in North Gondar, Ethiopia. Using generalized linear mixed models with appropriate distributional assumptions and rigorous machine learning-based validation, to identify significant predictors of seizure frequency and present a robust, reproducible analytical approach suitable for resource-limited settings.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy Design and Data Source\u003c/h2\u003e \u003cp\u003eThis study employed a retrospective longitudinal analytical design using routinely collected health data from the 2023 Ethiopian District Health Information System (EDHIS) in North Gondar. EDHIS is Ethiopia's national health management information system and allows longitudinal patient follow-up. The epilepsy dataset originated from a randomized, multicenter clinical trial embedded within EDHIS that compared the anti-epileptic drug progabide with placebo or standard care, with a total of 2403 patients participating. The frequency of seizures was measured weekly during the first 16 weeks, followed by a 27-week follow-up period.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eStudy Population\u003c/h3\u003e\n\u003cp\u003eThe study included adult patients (aged\u0026thinsp;\u0026ge;\u0026thinsp;18 years) with a confirmed clinical diagnosis of epilepsy and complete baseline and follow-up seizure count records. Patients received either progabide or placebo/standard care. Patients with incomplete baseline data, inconsistent follow-up records, or missing outcome measurements were excluded. After applying eligibility criteria, a total of 2,403 patients with valid longitudinal seizure data were included in the analysis.\u003c/p\u003e\n\u003ch3\u003eSample Size Determination\u003c/h3\u003e\n\u003cp\u003eA census sample strategy was used, which included all eligible epilepsy patients recorded in EDHIS during the study period. As this was a secondary study of previously collected longitudinal data, no formal sample size calculation was performed. The final sample size (n\u0026thinsp;=\u0026thinsp;2,403) was adequate for fitting generalized linear mixed-effects modeling.\u003c/p\u003e\n\u003ch3\u003eVariables\u003c/h3\u003e\n\u003cp\u003eThe primary outcome variable was the number of epileptic seizures recorded per patient over the 27-week follow-up period. Based on clinical relevance and data availability, covariates were grouped into Sociodemographic, clinical, and behavioral factors. These included sex, age, place of residence, educational level, occupation, baseline seizure frequency, body weight, family history of epilepsy, traumatic brain injury, stroke, head trauma, fever, uremia, brain infection, alcohol use, psychological stress, and seizure treatment type.\u003c/p\u003e\n\u003ch3\u003eStatistical Data Analysis\u003c/h3\u003e\n\u003cp\u003eThe statistical analysis of longitudinal seizure count data was carried out using Generalized Linear Mixed Models (GLMMs), which are ideal for non-normal outcomes and hierarchical or linked data structures. GLMMs build on generalized linear models by including random effects to account for within-subject correlation and individual-level variability, such as baseline risk or treatment trajectories across time. For count data, the model used a log link function and exposure time as an offset to equalize seizure rates across different follow-up periods. The estimated seizure count for subject i at time j was modeled as follows:\u003c/p\u003e \u003cp\u003eE(y\u003csub\u003eij\u003c/sub\u003e)=\u0026micro;\u003csub\u003eij\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;t\u003csub\u003eij\u003c/sub\u003eexp(x\u003csub\u003ei\u003c/sub\u003eβ\u0026thinsp;+\u0026thinsp;z\u003csub\u003ei\u003c/sub\u003eb\u003csub\u003ei\u003c/sub\u003e)\u003c/p\u003e \u003cp\u003eHere, t\u003csub\u003eij\u003c/sub\u003e is the exposure duration, x\u003csub\u003ei\u003c/sub\u003e and β are fixed-effect variables and coefficients, z\u003csub\u003ei\u003c/sub\u003e is the random effects design matrix, and bi are subject-specific random effects. On a logarithmic scale, this formula becomes: log(\u0026micro;\u003csub\u003eij\u003c/sub\u003e)\u0026thinsp;=\u0026thinsp;log(t\u003csub\u003eij\u003c/sub\u003e)+(x\u003csub\u003ei\u003c/sub\u003eβ\u0026thinsp;+\u0026thinsp;z\u003csub\u003ei\u003c/sub\u003eb\u003csub\u003ei\u003c/sub\u003e), which directly models seizure rates while accounting for varying observation durations.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eModel Assumptions\u003c/h2\u003e \u003cp\u003eGLMMs presume that random effects have a normal distribution and that the chosen link function accurately connects predictors to predicted outcomes. The log link was used to store count data. Random effect normality was determined using graphical approaches such as histograms and Q-Q plots. Variance assumptions were assessed using the ratio of the generalized chi-square statistic to its degree of freedom, with values near to one indicating appropriateness. Values greater than one indicate overdispersion; values less than one indicate underdispersion. When overdispersion was discovered, negative binomial GLMMs were used instead of Poisson models.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eModels within the GLMM Family\u003c/h3\u003e\n\u003cp\u003eFor seizure count data, Poisson GLMMs were first investigated. The Poisson model assumes equidispersion, which is when the conditional mean equals the conditional variance.\u003c/p\u003e \u003cp\u003eVar(Y\u003csub\u003ei\u003c/sub\u003e∣X\u003csub\u003ei\u003c/sub\u003e)\u0026thinsp;=\u0026thinsp;E(Y\u003csub\u003ei\u003c/sub\u003e∣X\u003csub\u003ei\u003c/sub\u003e)\u003c/p\u003e \u003cp\u003eIn fact, this assumption is frequently broken in clinical count data due to overdispersion.\u003c/p\u003e \u003cp\u003eTo counteract overdispersion, negative binomial GLMMs were used.\u003c/p\u003e \u003cp\u003eE(Y) = \u0026micro;; Var(Y) = \u0026micro;(1\u0026thinsp;+\u0026thinsp;σ2\u0026micro;)\u003c/p\u003e \u003cp\u003eThe negative binomial model includes an additional dispersion parameter, which allows the variance to surpass the mean. Both random-intercept and random-slope specifications were tested to account for individual differences in baseline seizure risk and seizure trajectories over time.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eDescriptive Statistics\u003c/h2\u003e \u003cp\u003eThe study population consisted of 2,403 epilepsy patients enrolled in a randomized controlled trial comparing progabide to a placebo. The majority of participants were female (1,579, 65.71%), with a mean age of 34.2 years (SD\u0026thinsp;=\u0026thinsp;12.8, range: 18\u0026ndash;72), and most resided in urban areas (1,845, 76.99%). Socioeconomic challenges were prevalent, as more than half of the patients were illiterate (1,421, 59.13%) and unemployed (1,496, 62.3%). Approximately half of the cohort (1,242, 51.7%) had a body weight at or below the sample mean. Treatment allocation was nearly equal between the placebo group 1215(50.6%), and the progabide group 1188(49.4%).\u003c/p\u003e \u003cp\u003eClinically, a high prevalence of concomitant conditions was found. Alcohol consumption was reported by 1109 (46.2%) of subjects, whereas 864 (36.0%) had a history of traumatic brain injury and 693 (28.8%) had a prior brain infection. Stroke was detected in 435(18.1%). Overall, the study cohort had significant Sociodemographic variety and a high frequency of clinical risk factors for seizure recurrence.\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 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSociodemographic, Behavioral, and Clinical Characteristics of Epilepsy Patients at Baseline (N\u0026thinsp;=\u0026thinsp;2,403)\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=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCovariates\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCategories\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFrequency (n)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePercentage (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eSex\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1579\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e65.71\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e824\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e34.29\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eWeight of the individual\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLess or equal to mean weight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1242\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e51.69\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAbove mean weight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1161\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e48.31\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eTreatment of seizures\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePlacebo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1215\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e50.56\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProgabide\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1188\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e49.44\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eResidence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUrban\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1845\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e76.99\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRural\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e558\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e23.01\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eEducation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eIlliterate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1421\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e59.13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLiterate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e982\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e40.87\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eOccupation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEmployment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e907\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e37.75\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUnemployment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1496\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e62.25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eFamily History\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e961\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e40.00\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1442\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e60\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAlcohol Use\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1109\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e46.15\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1294\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e53.85\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eTBI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e864\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e35.96\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1539\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e64.04\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eStroke\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e435\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e18.10\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1968\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e81.90\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eFever\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e984\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e40.95\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1419\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e59.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eHead trauma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e654\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e27.22\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1749\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e72.78\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eStress\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e925\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e38.49\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1478\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e61.51\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eUremia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e578\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e24.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1825\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e75.95\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eBrain infections\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e693\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28.84\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1710\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e71.16\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 \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eBaseline Seizure Frequency Distribution\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e depicts the baseline seizure frequency, which was right-skewed and highly varied. Patients were divided into three groups based on their seizure frequency per week.\u003c/p\u003e \u003cp\u003eThe mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD of seizures was 3.176885\u0026thinsp;\u0026plusmn;\u0026thinsp;6.136816, with a minimum and maximum age of 18 and 73 years old, respectively.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe distribution displayed three different peaks matching to these groups. Non-normality was found in histograms and density plots, corroborated by a Shapiro-Wilk test (w\u0026thinsp;=\u0026thinsp;0.87, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), indicating the need for non-parametric approaches for initial analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eDetection of Overdispersion\u003c/h2\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eModel Comparison and Selection\u003c/h2\u003e \u003cp\u003eIn Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, a comparison of fixed-effects and mixed-effects models for seizure count data shows a clear improvement in model adequacy. The fixed-effects Poisson model assumed equidispersion (equal mean and variance), however the observed data showed significant overdispersion (χ\u0026sup2;/df\u0026thinsp;=\u0026thinsp;3.67), which violated the assumption and resulted in poor fit (AIC\u0026thinsp;=\u0026thinsp;9,122.40). The fixed-effects Negative Binomial model, with an overdispersion parameter, significantly improved fit (ΔAIC\u0026thinsp;=\u0026thinsp;1,656.70). However, both fixed-effects techniques failed to account for the important within-subject correlation found in longitudinal data, rendering them ineffective.\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 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eComparison of Fixed-Effects and Mixed-Effects Models for Longitudinal Seizure Count Data\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFixed-Effects Models\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLL\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAIC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBIC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDeviance\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003eΔAIC\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoisson Fixed Effects\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-2671.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9122.40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9195.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5342.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative Binomial Fixed Effects\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-2437.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7465.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7546.50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4875.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e \u003cp\u003e-1656.70\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMixed-Effects Models\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAIC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBIC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eDeviance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eΔAIC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eICC\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoisson GLLM (RI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-4257.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e8628.40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8755.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e8514.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-1167.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.22\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoisson GLLM(RI\u0026thinsp;+\u0026thinsp;RS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-3917.40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7959.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8089.40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7834.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-669.30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative Binomial (RI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-2814.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5745.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5878.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5628.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-2882.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative Binomial (RI\u0026thinsp;+\u0026thinsp;RS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1215.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5419.33\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5492.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2431.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-326.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.09\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eZero-Inflated NB (RI\u0026thinsp;+\u0026thinsp;RS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-2286.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5398.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5512.76\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4572.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-21.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHurdle NB (RI\u0026thinsp;+\u0026thinsp;RS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-2287.33\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5401.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5518.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4574.66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-17.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.08\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\u003eThe examination moved to mixed-effects models, with a Poisson model with a random intercept providing an improvement but remaining inadequate (AIC\u0026thinsp;=\u0026thinsp;8,628.40). The addition of random slopes significantly improved fit (likelihood ratio test χ\u0026sup2; = 196.3, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and effectively captured individual seizure trajectories (ΔAIC = -669.30). The change to a Negative Binomial model, which addressed both overdispersion and correlation, resulted in a significant improvement in performance. Among these, the Negative Binomial model with a random intercept and slope had the best balance (AIC\u0026thinsp;=\u0026thinsp;5,419.33; BIC\u0026thinsp;=\u0026thinsp;5,492.94). Although zero-inflated and hurdle models provided modest AIC increases (ΔAIC\u0026thinsp;\u0026lt;\u0026thinsp;2), their increased complexity and interpretability were not justified. As a result, the Negative Binomial generalized linear mixed model (GLMM) with a random intercept and slope was chosen as the best model, fully accounting for overdispersion, correlation, and individual heterogeneity.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eCross-Validated Predictive Performance Metrics\u003c/h2\u003e \u003cp\u003eAs detailed in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, the Negative Binomial Generalized Linear Mixed Model (GLMM) with a random intercept and slope outperformed all tested metrics in cross-validation and out-of-sample prediction. This model surpassed other alternatives by attaining the most favorable results with the highest log-likelihood (-1,236.45) and pseudo-R\u0026sup2; (0.78), the lowest Root Mean Square Error (4.89) and Mean Absolute Error (3.45), and a near-perfect calibration slope (0.99). The gradual progression from simpler Poisson fixed-effects models to this final form demonstrates the need of accounting for within-subject correlation, overdispersion, and individual patient trajectories. A calibration slope this close to one demonstrates great agreement between predicted and observed seizure counts, highlighting the model's exceptional reliability and generalizability.\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 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCross-Validated Predictive Performance of Poisson and Negative Binomial Models for Seizure Counts\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\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 \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\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\u003eMean LL\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSE (LL)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRMSE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMAE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePseudo-R\u0026sup2;\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eCalibration Slope\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoisson Fixed Effects\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-2712.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e45.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e6.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.38\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative Binomial Fixed Effects\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-2489.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e38.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e5.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoisson GLMM (RI only)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-4298.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e67.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e5.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoisson GLMM (RI\u0026thinsp;+\u0026thinsp;RS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-3958.71\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e58.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative Binomial GLMM (RI)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-2845.32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e42.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e6.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative Binomial GLMM (RI\u0026thinsp;+\u0026thinsp;RS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-1236.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e24.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.99\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 \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eBootstrap Stability Assessment\u003c/h2\u003e \u003cp\u003eBootstrap-based stability and reliability evaluations of the model parameters, as outlined in Appendix \u003cspan refid=\"Sec23\" class=\"InternalRef\"\u003eA\u003c/span\u003e (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e4\u003c/span\u003e), show robustness across fixed and random effects. Fixed effects display significant stability; for example, the treatment incidence rate ratio (IRR\u0026thinsp;=\u0026thinsp;0.70) has limited fluctuation (CV\u0026thinsp;=\u0026thinsp;6.42%) and an excellent coverage probability (95.2%), highlighting the precision of its confidence intervals. Random effects, on the other hand, have more variability in intercept variance (CV\u0026thinsp;=\u0026thinsp;21.95%) and slope variance (CV\u0026thinsp;=\u0026thinsp;34.55%), yet both are statistically significant in the majority of bootstrap samples (94.2% in slope variance). The stability index (percentage of bootstrap estimates within \u0026plusmn;\u0026thinsp;10% of the median) was high for fixed effects (0.94\u0026ndash;0.96) but adequate for random effects (0.65\u0026ndash;0.78). The intercept-slope correlation (ρ\u0026thinsp;=\u0026thinsp;0.24) was somewhat stable (CV\u0026thinsp;=\u0026thinsp;29.12%). Collectively, these findings illustrate the model's resilience; fixed effects are very reliable, and random effects, despite their intrinsic variability, frequently retain significance, reinforcing the model's overall robustness.\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 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eBootstrap-Based Stability and Reliability Assessment of Fixed and Random Effects in the Final Negative Binomial GLMM\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eMedian Estimate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e95% Bootstrap CI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eCV( %)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eStability Index\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eCoverage Probability (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"8\" nameend=\"c8\" namest=\"c1\"\u003e \u003cp\u003eFixed Effects\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003eTreatment (IRR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.60\u0026ndash;0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e95.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003eBaseline Seizures \u0026ndash; High (IRR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.33\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.18\u0026ndash;1.50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e94.8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"8\" nameend=\"c8\" namest=\"c1\"\u003e \u003cp\u003eRandom Effects\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eIntercept Variance (τ₀\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003e0.41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.26\u0026ndash;0.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e21.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e93.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eSlope Variance (τ₁\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003e0.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.04\u0026ndash;0.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e34.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e94.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eIntercept-Slope Correlation (ρ)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003e0.24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.02\u0026ndash;0.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e29.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.71\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e92.9\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 \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eRelative Contribution of Key Predictors to Model Performance in Longitudinal Seizure Prediction\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=\"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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRank\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFeature\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePermutation Importance\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSHAP Mean\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTotal Predictive Power (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBaseline Seizure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.342\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.315\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e31.5%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTreatment Group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.285\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.284\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e28.4%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eStroke History\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.187\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.178\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e17.8%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAlcohol Use\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.156\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.152\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e15.2%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTraumatic Brain Injury\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.134\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.128\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e12.8%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eStress\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.121\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.115\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11.5%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEducation Level\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.089\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.082\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e8.2%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFamily History\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.078\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.071\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7.1%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFever\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.067\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.063\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e6.3%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUremia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.059\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.055\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5.5%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eClinical Factors (\u003cspan additionalcitationids=\"CR4 CR5 CR6 CR7 CR8 CR9\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e\u0026ndash;\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026ndash;\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e72.3%\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003e\u003cem\u003eNote: Permutation Importance\u0026thinsp;=\u0026thinsp;increase in prediction error after randomly shuffling feature values. SHAP values represent the average absolute impact on model output (log seizure rate). % of Total Predictive Power is normalized SHAP contribution.\u003c/em\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eParameter Estimates and Interpretation\u003c/h2\u003e \u003cp\u003eThe final negative binomial generalized linear mixed model (GLMM) found several significant predictors of seizure frequency. Progabide therapy significantly reduced seizure incidence by 30% compared to placebo (IRR\u0026thinsp;=\u0026thinsp;0.70; 95% CI: 0.62\u0026ndash;0.80; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Baseline seizure frequency showed a dose-response relationship, with the high-frequency group (30\u0026ndash;34 seizures/week) having a 33% higher incidence (IRR\u0026thinsp;=\u0026thinsp;1.33; 95% CI: 1.21\u0026ndash;1.48; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Alcohol usage increased seizures by 18% (IRR\u0026thinsp;=\u0026thinsp;1.18; 95% CI: 1.08\u0026ndash;1.29; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), while literacy reduced seizures by 7% (IRR\u0026thinsp;=\u0026thinsp;0.93; 95% CI: 0.86\u0026ndash;0.97; p\u0026thinsp;=\u0026thinsp;0.038). Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows that stroke (IRR\u0026thinsp;=\u0026thinsp;1.30; 95% CI: 1.16\u0026ndash;1.47; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), traumatic brain injury (IRR\u0026thinsp;=\u0026thinsp;1.14; 95% CI: 1.02\u0026ndash;1.28; p\u0026thinsp;=\u0026thinsp;0.016), and brain infection (IRR\u0026thinsp;=\u0026thinsp;1.10; 95% CI: 1.01\u0026ndash;1.21; p\u0026thinsp;=\u0026thinsp;0.036) are among the clinical comorbidities associated with increased seizure occurrence.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eParameter Estimates from the Negative Binomial Generalized Linear Mixed Model with Random Intercept and Random Slope\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003eFixed Effects Negative Binomial Regression (RI\u0026thinsp;+\u0026thinsp;RS Model)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCovariates\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eIRR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSE (B)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eP-value\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e95% CI(IRR)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex (Ref: Female)\u003c/p\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.030\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.495\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.94\u0026ndash;1.12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.002\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.885\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.971\u0026ndash;1.025\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWeight (Ref: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\le\\:\\:\\)\u003c/span\u003e\u003c/span\u003emean weight)\u003c/p\u003e \u003cp\u003e\u0026gt; mean weight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.095\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.046\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.028\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.01\u0026ndash;1.22\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eResidence (Ref: urban)\u003c/p\u003e \u003cp\u003eRural\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.049\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.024\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.356\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.95\u0026ndash;1.16\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEducation (Ref: Illiterate)\u003c/p\u003e \u003cp\u003eLiterate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.073\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.035\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.038\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.86\u0026ndash; 0.97\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOccupation (Ref: Unemployment)\u003c/p\u003e \u003cp\u003eEmployment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.039\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.450\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.94\u0026ndash;1.15\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBaseline seizures (Ref: 15\u0026ndash;24)\u003c/p\u003e \u003cp\u003e25\u0026ndash;29\u003c/p\u003e \u003cp\u003e30\u0026ndash;34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.199\u003c/p\u003e \u003cp\u003e0.285\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.22\u003c/p\u003e \u003cp\u003e1.33\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.108\u003c/p\u003e \u003cp\u003e0.154\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.11\u0026ndash;1.36\u003c/p\u003e \u003cp\u003e1.21\u0026ndash;1.48\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFamily history (Ref: No)\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.140\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.068\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.015\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.03\u0026ndash;1.28\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol use (Ref: No)\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.165\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.080\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.08\u0026ndash;1.29\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTreatment type(Ref: Placebo)\u003c/p\u003e \u003cp\u003eProgabide\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.357\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.142\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.62\u0026ndash;0.80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFever (Ref: No)\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.113\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.055\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.021\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.02\u0026ndash;1.28\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTBI (Ref: No)\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.131\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.140\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.064\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.016\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.02\u0026ndash;1.28\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStroke (Ref: No)\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.262\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.064\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.16\u0026ndash;1.47\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHead trauma (Ref: No)\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.113\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.055\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.01\u0026ndash;1.25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStress (Ref: No)\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.165\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.080\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.09\u0026ndash;1.29\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUremia (Ref: No)\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.207\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.10\u0026ndash;1.39\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBrain infection (Ref: No)\u003c/p\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.095\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.056\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.046\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.01\u0026ndash;1.21\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eRandom Effects Variance-Covariance Structure of the Final Model\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eComponent\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eEstimate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStd. Error\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003ep \u0026ndash; value\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e95% Wald CI\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRandom intercept variance (τ₀\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e0.41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.09\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.24\u0026ndash;0.60\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRandom slope variance (baseline seizures), τ₁\u0026sup2;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e0.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.008\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.04\u0026ndash;0.19\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntercept\u0026ndash;slope covariance ( Cov(0,1))\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.02\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.020\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.01\u0026ndash;0.09\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eIntercept-Slope(\u003c/b\u003eρ)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e0.24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eRandom factor analysis indicated significant patient heterogeneity. After controlling for fixed variables, the intercept variance (τ₀\u0026sup2; = 0.41; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) revealed significant diversity in baseline seizure rates. Additionally, the slope variance (τ₁\u0026sup2; = 0.11; p\u0026thinsp;=\u0026thinsp;0.008) highlighted various seizure development trajectories among individuals. A positive connection between intercepts and slopes (ρ\u0026thinsp;=\u0026thinsp;0.24; p\u0026thinsp;=\u0026thinsp;0.020) indicates that patients with higher baseline seizure counts had a faster increase in frequency over time. The intraclass correlation value (ICC\u0026thinsp;=\u0026thinsp;0.09) revealed that stable differences between patients explained 9% of the total variance in seizure counts. Overall, our findings show significant individual diversity in both baseline risk and illness development, emphasizing the important need for tailored treatment methods.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eMachine Learning Enhancements\u003c/h2\u003e \u003cdiv id=\"Sec19\" class=\"Section3\"\u003e \u003ch2\u003eModel Evaluation and Validation\u003c/h2\u003e \u003cp\u003eWe used nested 10\u0026times;5 cross-validation to compare the predictive performance of the Negative Binomial GLMM to other machine learning methods such as gradient boosting, random forests, neural networks, and elastic net regression. The NB GLMM outperformed all major measures, with the greatest mean test log-likelihood (-1236.45), lowest calibration error (0.015), and highest prediction accuracy (pseudo-R\u0026sup2; = 0.78). In addition to prediction accuracy, bootstrap analysis supported the model's fundamental estimations for treatment effect and baseline seizure frequency. The bias-corrected and accelerated (BCa) bootstrap intervals were recommended due to their superior coverage (95.2%) and ability to account for any bias and Skewness, resulting in a more credible assessment of uncertainty. To better understand the drivers of this performance, feature importance was ranked using permutation important and SHAP values, with baseline seizure frequency being the most influential predictor (31.5%), followed by treatment group (28.4%) and stroke history (17.8%). These crucial clinical characteristics accounted for 72.3% of the model's prediction variance, demonstrating their substantial practical value, as shown in Appendix \u003cspan refid=\"Sec23\" class=\"InternalRef\"\u003eA\u003c/span\u003e (Tables\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e6\u003c/span\u003e\u0026ndash;8).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePermutation-Based Feature Importance for Predictors of Seizure Frequency\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\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\u003eMean test LL\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSD test LL\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMean RMSE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSD RMSE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eMean MAE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003ePseudo-R\u0026sup2;\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eCalibration Error\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNB GLMM (RI\u0026thinsp;+\u0026thinsp;RS)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1236.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.015\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGradient Boosting\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1241.32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e26.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.76\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.018\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRandom Forest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1258.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.022\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNeural Network\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1244.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.019\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eElastic Net Regression\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1279.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e31.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.025\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 \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab8\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSHAP-Based Global Feature Importance Rankings for the Final Predictive Model\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=\"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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWald (Model-Based)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eBootstrap Percentile\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eBootstrap BCa\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eBootstrap Studentized\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e95% CI(IRR) for treatment\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.62\u0026ndash;0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.62\u0026ndash;0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.60\u0026ndash;0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.59\u0026ndash;0.83\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWidth\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.18 0.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.24\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBaseline High\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.21\u0026ndash;1.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.19\u0026ndash;1.49\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.18\u0026ndash;1.50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.17\u0026ndash;1.52\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWidth\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.35\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRandom Intercept\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.24\u0026ndash;0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.25\u0026ndash;0.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.26\u0026ndash;0.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.27\u0026ndash;0.67\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWidth\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.38\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.39\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.40\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoverage Probability\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e94.1%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e94.8%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e95.2%\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e95.5%\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 \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study employed an integrated analytical approach using generalized linear mixed models (GLMMs) and machine learning validation to evaluate longitudinal seizure count data from an epileptic clinical trial conducted in North Gondar, Ethiopia. The findings, supported by comprehensive data analysis and graphical diagnostics, provide robust scientific and clinical insights into seizure dynamics, treatment efficacy, and patient heterogeneity in a resource-limited setting. Graphical assessments (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) revealed pronounced right-Skewness and multimodal distributions, underscoring deviations from normality and justifying the use of count-based regression models. Overdispersion plots (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e) further demonstrated that variance exceeded the mean, violating the Poisson assumption of equidispersion. These results align with contemporary epilepsy and neurological studies attributing such variability to unobserved heterogeneity and recurrent-event processes (20, 21).\u003c/p\u003e \u003cp\u003eModel comparison results revealed progressive improvement when transitioning from fixed-effects to mixed-effects models and from Poisson to negative binomial (NB) distributions. A negative binomial GLMM with random intercepts and slopes achieved the lowest AIC and BIC values, balancing goodness-of-fit and model complexity. Consistent with prior research in longitudinal epilepsy and neurological populations, NB mixed models outperformed Poisson and marginal models in addressing overdispersion and within-subject correlation (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e23\u003c/span\u003e). Cross-validation further validated the final NB-GLMM, which demonstrated a high pseudo-R\u0026sup2; (0.78), low RMSE and MAE, and excellent calibration. Strong correlations between predicted and observed seizure counts suggest robust generalizability. Methodological studies emphasize the necessity of resampling-based validation for reliable inference in longitudinal clinical data, particularly when asymptotic assumptions are unmet (24, 25)\u003c/p\u003e \u003cp\u003eProgabide therapy was consistently associated with a statistically significant 30% reduction in seizure frequency compared to placebo across all model settings. This effect aligns with existing clinical evidence supporting the efficacy of antiepileptic drugs targeting inhibitory neurotransmission in reducing seizure burden, particularly in adults (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e27\u003c/span\u003e). The treatment effect\u0026rsquo;s consistency across bootstrap replications suggests resilience to sampling variability, highlighting its relevance in low- and middle-income countries where access to newer antiepileptic medications is limited.\u003c/p\u003e \u003cp\u003eBaseline seizure frequency emerged as a critical predictor of future seizure burden, with a dose-response relationship observed. Patients with higher baseline seizure counts exhibited increased follow-up incidence, a pattern corroborated by longitudinal studies linking early seizure burden to disease severity and medication resistance (28, 29). The positive correlation between random-effects intercepts and slopes indicates that individuals with elevated baseline seizure rates experience more rapid deterioration over time, emphasizing the need for early risk stratification and tailored therapeutic escalation (30, 31).\u003c/p\u003e \u003cp\u003eAmong Sociodemographic factors, literacy demonstrated a protective effect on seizure frequency, likely mediated by improved health literacy, medication adherence, and healthcare access. This finding is supported by regional and global evidence linking education to better epilepsy outcomes in low-resource settings (32, 33). Alcohol consumption, conversely, was strongly associated with increased seizure risk, consistent with evidence showing its impact on seizure thresholds, drug metabolism, and adherence (34, 35).\u003c/p\u003e \u003cp\u003eSeveral Clinical comorbidities including stroke, traumatic brain injury, fever, uremia, and brain infections were independently linked to elevated seizure frequency. These associations are well-documented in neurological literature, with structural brain injury, metabolic dysregulation, and neuroinflammation identified as key mechanisms (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e37\u003c/span\u003e). Notably, the strong association between stroke and seizure frequency underscores the public health significance of post-stroke epilepsy, particularly in aging populations and regions with rising cardiovascular disease burdens (38, 39). The role of brain infections also highlights the persistent epidemiological burden of preventable causes of epilepsy in low-resource settings.\u003c/p\u003e \u003cp\u003eRandom effects analysis revealed substantial interpatient variability in baseline seizure risk and progression trajectories, with approximately 9% of total variance attributable to unmeasured factors despite controlling for observed covariates. Such heterogeneity has been recognized as clinically meaningful, as population-averaged models may obscure critical patient-specific dynamics (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e40\u003c/span\u003e). Collectively, these findings advocate for a paradigm shift toward personalized epilepsy care, where treatment decisions integrate both average therapeutic effects and individual risk profiles. The integration of GLMMs with interpretable machine learning offers a practical pathway for advancing precision medicine in clinical research, particularly in settings requiring rigorous statistical methods to optimize resource allocation and patient outcomes.\u003c/p\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eStrengths and Limitations\u003c/h2\u003e \u003cp\u003eThis study boasts several notable strengths, including a large and diverse sample size, a robust longitudinal design, and a methodological approach that combines rigorous comparative modeling with the integration of both statistical and machine learning techniques. The use of routinely collected health data from the Emergency Department Health Information System (EDHIS) enhances ecological validity while offering practical relevance for similar healthcare systems. However, several limitations warrant acknowledgment. Despite the randomized treatment assignment, the observed associations between behavioral and clinical covariates cannot be interpreted as causal due to potential confounding factors. Furthermore, the possibility of residual confounding and measurement errors inherent to routinely collected data may affect the precision of our estimates. While categorizing continuous variables improves interpretability, this approach may result in a reduction of information. Lastly, although our machine learning validation was thorough, external validation in independent datasets is necessary to further substantiate the generalizability of our findings.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study aimed to analyze longitudinal seizure count data from a clinical trial in Ethiopia to evaluate treatment efficacy and identify risk factors, addressing common challenges like overdispersion and within-subject correlation. Using Generalize Linear Mixed Models, the researchers compared Poisson and Negative Binomial distributions, ultimately selecting a Negative Binomial GLMM with random intercepts and slopes as the best-fitting model. Out of 2403 patients included in the study, 1215 received placebo, while 118 received progabide\u003c/p\u003e \u003cp\u003eThe result showed that progabide treatment reduced seizure incidence by 30% compared to placebo. Significant risk factors for increased seizure include higher baseline seizure frequency, alcohol use, stroke, traumatic brain injury, brain infection, while literacy had a protective effect. The model demonstrated strong predictive performance and robustness through cross-validation and bootstrap assessments.\u003c/p\u003e \u003cp\u003eThe study concludes that this integrated GLMM and machine learning approach effectively handles complex epilepsy trial data, providing a reliable method for treatment evaluation and risk assessment in resource-limited settings.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAIC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAkaike Information Criterion\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eBIC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eBayesian Information Criterion\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eConfidence Interval\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ecoefficient of Variation\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eEDHIS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eEthiopia District Health Information System\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGEE\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGeneralized Estimating Equations\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGeneralized Linear Models\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLMM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGeneralized Linear Mixed Model\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eICC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eIntraclass Correlation Coefficient\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIRR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eIncidence Rate Ratio\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIRLS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eIteratively Reweighted Least Squares\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLL\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLog-Likelihood\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLMICs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLow- and Middle-Income Countries\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMAE\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMean Absolute Error\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNB\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNegative Binomial\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eRI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRandom Intercept\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eRMSE\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\"\u003eRS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRandom Slope\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eStandard Deviation\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSE\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eStandard Error\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSHAP\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\"\u003eTBI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eTraumatic Brain Injury\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eWHO\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eWorld Health Organization\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":" \u003cp\u003e \u003cstrong\u003eEthics approval and consent to participate:\u003c/strong\u003e \u003cp\u003eThis study utilized publicly available secondary data from the 2023 Ethiopian District Health Information System (EDHIS), accessed through the DHIS Program following formal registration and approval. The original DHIS survey protocols were reviewed and approved by the Institutional Review Board (IRB) of the relevant international body and the National Ethics Committee of Ethiopia. As this research involved only the secondary analysis of anonymized data, additional ethical approval and individual informed consent were not required. All study procedures adhered to applicable ethical guidelines and regulations.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConsent for publication:\u003c/strong\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eClinical trial number\u003c/h2\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eCompeting interests:\u003c/strong\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding:\u003c/h2\u003e \u003cp\u003eThis research received no external funding from public, commercial, or non-profit organizations.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAbay Kassie Lakew conceptualized and designed the study, conducted statistical analyses, and led the drafting of the manuscript. He also contributed to data collection, interpretation of findings, and critical revisions. Tezera Abebe Gashaw supervised the research, provided methodological expertise, and critically reviewed the manuscript.All authors read and approved the final version\u003c/p\u003e\u003ch2\u003eAcknowledgments:\u003c/h2\u003e \u003cp\u003eThe authors gratefully acknowledge the Ethiopian District Health Information System (EDHIS) for granting access to the data and extend their appreciation to the healthcare providers involved in data collection. We also thank the study participants for their contributions to the original data.\u003c/p\u003e\u003ch2\u003eAvailability of data and materials:\u003c/h2\u003e \u003cp\u003eThe data supporting this study are publicly available through the 2023 Ethiopian District Health Information System (EDHIS). Authorized researchers may access the dataset via the DHIS Program at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://dhis.moh.gov.et/dhis-web-commons/security/login.action\u003c/span\u003e\u003cspan address=\"https://dhis.moh.gov.et/dhis-web-commons/security/login.action\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, subject to approval and compliance with data usage terms and conditions.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eWorld Health Organization (WHO). Epilepsy: A Public Health Imperative. Geneva: World Health Organization; 2019.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThurman DJ, Logroscino G, Beghi E, et al. The burden of premature mortality of epilepsy in high-income and low-income countries. Lancet Neurol. 2017;16(11):903\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeghi E. The epidemiology of epilepsy. Neuroepidemiology. 2020;54(2):185\u0026ndash;91.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Z, Brodie MJ, Liew D, Kwan P. Treatment outcomes in patients with newly diagnosed epilepsy treated with established and new antiepileptic drugs. JAMA Neurol. 2018;75(3):279\u0026ndash;86.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerucca E, Brodie MJ, Kwan P, Tomson T. 30 years of second-generation antiseizure medications: Impact and future perspectives. Lancet Neurol. 2020;19(6):544\u0026ndash;56.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMolenberghs G, Verbeke G. Models for Discrete Longitudinal Data. New York: Springer; 2005.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiggle PJ, Heagerty P, Liang KY, Zeger SL. Analysis of Longitudinal Data. 2nd ed. Oxford: Oxford University Press; 2009.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoxe S, West SG, Aiken LS. The analysis of count data: A gentle introduction to Poisson regression and its alternatives. J Pers Assess. 2009;91(2):121\u0026ndash;36.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHilbe JM. Negative Binomial Regression. 2nd ed. Cambridge: Cambridge University Press; 2011.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBreslow NE. Extra-Poisson variation in log-linear models. J Royal Stat Society: Ser C. 1984;33(1):38\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcCullagh P, Nelder JA. Generalized Linear Models. 2nd ed. London: Chapman \u0026amp; Hall; 1989.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePinheiro JC, Bates DM. Mixed-Effects Models in S and S-PLUS. New York: Springer; 2000.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBolker BM, Brooks ME, Clark CJ, et al. Generalized linear mixed models: A practical guide for ecology and evolution. Trends Ecol Evol. 2009;24(3):127\u0026ndash;35.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStroup WW. Generalized Linear Mixed Models: Modern Concepts, Methods and Applications. Boca Raton: CRC; 2012.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang J. Linear and Generalized Linear Mixed Models and Their Applications. New York: Springer; 2007.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHastie T, Tibshirani R, Friedman J. The Elements of Statistical Learning. 2nd ed. New York: Springer; 2009.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJames G, Witten D, Hastie T, Tibshirani R. An Introduction to Statistical Learning. 2nd ed. New York: Springer; 2021.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBa-Diop A, Marin B, Druet-Cabanac M, et al. Epidemiology, causes, and treatment of epilepsy in sub-Saharan Africa. Lancet Neurol. 2014;13(10):1029\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMbuba CK, Newton CR. Packages of care for epilepsy in low- and middle-income countries. PLoS Medicine., Harrison XA, Donaldson L, Correa-Cano ME, Evans J, Fisher DN, Goodwin CED et al. A brief introduction to mixed effects modeling and multi-model inference in ecology. PeerJ. 2018; 6:e4794.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAustin PC, Merlo J. Intermediate and advanced topics in multilevel logistic regression analysis. Stat Med. 2022;41(10):1841\u0026ndash;64.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng C, Wang H, Lu N, Tu XM. Log-transformation and its implications for data analysis. Shanghai Archives Psychiatry. 2020;32(2):105\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStroup WW, Generalized Linear Mixed Models: Modern Concepts, Methods and Applications. 2nd ed. Boca Raton: CRC Press, Van Calster B, McLernon DJ, van Smeden M, Wynants L, Steyerberg EW. Calibration: the Achilles heel of predictive analytics. BMJ. 2019; 368:l5381.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteyerberg EW, Vergouwe Y. Towards better clinical prediction models: seven steps for development and an ABCD for validation. Eur Heart J. 2022;43(22):2030\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Z, Brodie MJ, Liew D, Kwan P. Treatment outcomes in patients with newly diagnosed epilepsy treated with established and new antiepileptic drugs. JAMA Neurol. 2018;75(3):279\u0026ndash;86.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eL\u0026ouml;scher W, Klein P. The pharmacology and clinical efficacy of antiseizure medications: from bench to bedside. Epilepsia., Kwan P, Arzimanoglou A, Berg AT, Brodie MJ, Hauser WA, Mathern G et al. Definition of drug-resistant epilepsy: consensus proposal. Epilepsia. 2019; 60(3):484\u0026ndash;495.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeezer MR, Sisodiya SM, Sander JW. Comorbidities of epilepsy: current concepts and future perspectives. Lancet Neurology., Devinsky O, Vezzani A, O\u0026rsquo;Brien TJ, Jette N, Scheffer IE, de Curtis M et al. Epilepsy. Nature Reviews Disease Primers. 2021; 7:2.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerucca E, Brodie MJ, Kwan P, Tomson T. 30 years of second-generation antiseizure medications: impact and future perspectives. Epilepsia., Mbuba CK, Newton CR. Epilepsy stigma in Africa: causes, consequences, and interventions. EpilepsyBehavior. 2020; 112:107224., Wagner RG, Ngugi AK, Twine R, Bottomley C, Kahn K, Sander JW et al. Prevalence and risk factors for active convulsive epilepsy in rural northeast South Africa. Epilepsy Research. 2021; 170:106548.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSamokhvalov AV, Irving H, Rehm J. Alcohol consumption as a risk factor for epilepsy and seizures: a systematic review. Epilepsy Research., Rehm J, Shield KD, Roerecke M, Gmel G. Alcohol, the brain, and epilepsy. Epilepsia. 2023;64(2):235\u0026ndash;245.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeghi E, Giussani G, Nichols E, et al. Global, regional, and national burden of epilepsy, 1990\u0026ndash;2019. Lancet Neurol. 2020;19(12):1027\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStefan H, Lopes da Silva FH. Epileptic neuronal networks: methods of identification and clinical relevance. Front Neurol. 2022;13:832312.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerlazzo E, Gasparini S, Beghi E, Sueri C, Russo E, Leo A Epilepsy in cerebrovascular diseases. Journal of Neurology., Galovic M, D\u0026ouml;hler N, Erd\u0026eacute;lyi-Canavese B, Felbecker A, Siebel P, Conrad J et al. Prediction of late seizures after ischemic stroke. Neurology. 2023; 100(5):e511\u0026ndash;e522.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiley RD, Ensor J, Snell KIE, et al. calculating the sample size required for developing a clinical prediction model. BMJ. 2023;380:e072464.\u003c/span\u003e\u003c/li\u003e\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-medical-research-methodology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmrm","sideBox":"Learn more about [BMC Medical Research Methodology](http://bmcmedresmethodol.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmrm/default.aspx","title":"BMC Medical Research Methodology","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Epilepsy, negative binomial distribution, overdispersion, machine learning, Ethiopia","lastPublishedDoi":"10.21203/rs.3.rs-8610618/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8610618/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eIntroduction:\u003c/h2\u003e \u003cp\u003eEpilepsy is a serious neurological condition that has a significant impact on public health, especially in low-and middle-income countries. Longitudinal seizure count data from clinical trials pose significant analytical clinical trials due to overdispersion, within-subject correlation, and non-normal random-effects distributions. Traditional statistical models often fail to adequately address these complexities.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eThis study analyzed longitudinal seizure count data from a randomized controlled clinical trial involving 2403 adult epilepsy patients in North Gondar, Ethiopia, followed for up to 27 weeks. An integrated analytical framework combining Generalized Linear Mixed Models (GLMMs) and machine learning validation techniques was employed. Poisson and negative binomial GLMMs were fitted to account for repeated measurements and individual-level heterogeneity.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe Negative Binomial GLMM with random intercepts and slopes provides the best fit (AIC\u0026thinsp;=\u0026thinsp;5419.33, BIC\u0026thinsp;=\u0026thinsp;5492.94, pseudo-R\u0026sup2;=0.78). Progabide Treatment was associated with a 30% decrease in seizure incidence compared with placebo (IRR\u0026thinsp;=\u0026thinsp;0.70, 95% CI: 0.62\u0026ndash;0.80, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Higher baseline seizure frequency, alcohol use, stroke, traumatic brain injury, and brain infection were significantly associated with increased seizure incidence, whereas literacy showed a protective effect. Random-effects analysis revealed substantial between-subject heterogeneity in baseline seizure rates (τ₀\u0026sup2; = 0.41; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and seizure trajectories over time (τ₁\u0026sup2; = 0.11; p\u0026thinsp;=\u0026thinsp;0.008). Machine learning validation demonstrated consistent predictive performance across several validation methods, proving the robustness of the model.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eIn the analysis of longitudinal seizure count data, Negative Binomial Generalized Linear Mixed Models effectively account for individual heterogeneity, within-subject correlation, and overdispersion. The integration of resampling-based validation techniques with comprehensive model diagnostics enhances both predictive accuracy and interpretability. This analytical approach is well suited for epilepsy research and therapeutic decision-making in resource-limited settings, as it provides robust evidence on treatment effectiveness and identifies key risk factors influencing seizure frequency.\u003c/p\u003e","manuscriptTitle":"Application of Generalized Linear Mixed Models with Machine Learning validation for Longitudinal Seizure Count Data: A Clinical Trials in Ethiopia","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-02 09:16:26","doi":"10.21203/rs.3.rs-8610618/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"717730274992035645810497644186813218","date":"2026-03-09T09:35:49+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-06T11:34:18+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"235065880028824566067532096290418355866","date":"2026-03-06T08:46:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"1180280309089924123425170470182819204","date":"2026-02-27T08:35:28+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-24T21:05:42+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-01-27T20:46:38+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-16T04:27:46+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-16T04:26:19+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Research Methodology","date":"2026-01-15T12:40:29+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-research-methodology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmrm","sideBox":"Learn more about [BMC Medical Research Methodology](http://bmcmedresmethodol.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmrm/default.aspx","title":"BMC Medical Research Methodology","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8acd8a76-b3f7-43c5-9d2f-af9d843c0b1f","owner":[],"postedDate":"March 2nd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-03-02T09:16:26+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-02 09:16:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8610618","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8610618","identity":"rs-8610618","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}