A cluster analysis of neonates using clinical signs of possible serious bacterial infection at hospital admission in Kenya: A retrospective multicentre cohort study

A cluster analysis of neonates using clinical signs of possible serious bacterial infection at hospital admission in Kenya: A retrospective multicentre cohort study | 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 A cluster analysis of neonates using clinical signs of possible serious bacterial infection at hospital admission in Kenya: A retrospective multicentre cohort study Timothy Tuti, Tabitha Muema, Mike English, Jalemba Aluvaala This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9480999/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Neonatal sepsis remains a major cause of mortality in Sub-Saharan Africa (SSA). Despite presenting with considerable clinical heterogeneity, suspected cases are managed uniformly with broad-spectrum antibiotics. Typical data-driven approaches developed in high-resource settings to identify clinically meaningful phenotypes and support management of neonatal sepsis are largely ungeneralisable to typical SSA public hospital settings, due to inclusion of variables that are largely unavailable at admission. This study’s objective was to identify sepsis clusters using signs of possible Serious Bacterial Infection (pSBI) readily available at the time of admission, and to assess the clusters performance in predicting mortality. Methods We conducted unsupervised model-based cluster analysis using Latent Class Analysis based on pSBI data collected at admission. All in-born neonates < 28 days old admitted to 21 Kenyan hospitals between January 2022 and December 2024 with ≥ 1 pSBI sign/symptom at admission were eligible for inclusion. We further explored the external validity of this clustering approach on new patient populations, and assessed the ability of the identified clusters to accurately predict in-hospital mortality compared to the World Health Organization neonatal sepsis severity classification guidelines. Results Five clusters of minimal, low, moderate, substantial and critical mortality risk were identified from development dataset with 33094 patients from eight hospitals. The models had an accuracy, positive predictive value and specificity of at least 83.16% (82.72% to 83.62%), 81.02% (80.58% to 81.45%) and 86.91% (86.61% to 87.23%) respectively in predicting cluster membership of 23704 patients in the external validation dataset admitted to thirteen different hospitals. From an internal-external cross-validation approach of the in-hospital mortality risk, the model-based clustering approach had discrimination (AUROC) of 0.867 (0.863 to 0.871) and calibration intercept and slope of -0.004 (-0.031 to 0.023) and 0.996 (0.979 to 1.014) respectively, outperforming the WHO sepsis severity classification whose discrimination was 0.721 (0.715 to 0.727) and calibration intercept and slope being 0.018 (-0.005 to 0.041) and 1.015 (0.986 to 1.043) respectively. Conclusion The identified clusters can complement clinicians’ judgement in assessing risk among neonates with sepsis at admission. Future work evaluating the utility of these clusters and potential differences in treatment response across clusters are therefore recommended to help strengthen the case for more targeted, risk-based neonatal sepsis management. neonates newborn sepsis hospital unsupervised learning Sub-Sahara Africa Figures Figure 1 Figure 2 Figure 3 Author summary (Lay summary) Why was this study done? Neonatal sepsis, a major cause of mortality in children aged <28 days, is a heterogenous syndrome with variation in clinical presentation and mortality outcomes. Defining distinct phenotypes from a limited set of clinical signs easy to use by clinicians (usually busy junior clinicians) may facilitate better emergency triaging, more targeted treatment and improve patient outcomes. This research sought to address the sepsis variation by grouping neonates based on clinical signs readily available on admission. The focus on non-microbiology variables is due to limited microbiology results in typical SSA public hospital settings at admission. What did the researchers do and find? We applied a model-based clustering approach that uses probabilistic framework that accounts for uncertainty to assign each patient a probability of belonging to each cluster given the presence/absence of clinical signs at admission. We identified five clinically relevant clusters that effectively stratified neonatal sepsis into sub-populations with differing risks of mortality. These clusters demonstrated superior discrimination and calibration in predicting in-hospital mortality when compared to both the WHO expert-based and typical distance-based clustering approaches. What do these findings mean? The five clusters capture the heterogeneity in neonatal sepsis at admission and reliably stratify sepsis into groups predictive of in-hospital mortality at admission This work’s findings lay the foundation for future work exploring the association of cluster membership with microbiological results and use of emulated trials to inform antibiotic use strategies. This work’s findings also lay the foundation for future research into the utility and usability of the identified clusters in risk stratification at hospital admission in typical SSA hospital settings. Background The most vulnerable time for a child’s survival are the first 28 days of life – the neonatal period, with 2.3 million children dying during this period globally, with a child from Sub-Saharan Africa (SSA) being 10 times more likely to die than a child from a high-income country [ 1 ]. Sepsis, which is a dysregulated immune response to infection that leads to acute organ dysfunction, is a leading contributor to burden of disease in neonates in SSA, both as primary cause of death and as a frequent contributor [ 2 , 3 ], and carries a high risk of death even when care is provided promptly [ 4 ]. In newborns, sepsis occurs in two forms: early-onset neonatal sepsis, which occurs within the first 72 hours of life, often linked to complications during birth, and late-onset neonatal sepsis, which develops between after 72 hours of life [ 5 ]. These infections are more common in premature or low-birth-weight babies, and their rates are higher in regions with limited or poor quality healthcare [ 6 ]. One-sixth of neonates treated for sepsis face significant challenges including functional limitation, cognitive impairment, and mental health disorders after recovery with 40% of sepsis patients requiring rehospitalisation within 90 days of discharge [ 3 ]. Many of these neonatal sepsis deaths in SSA are preventable with early detection and proper treatment. To achieve this healthcare providers rely on a high index of suspicion of sepsis (in the absence of laboratory confirmation) based on signs of possible Serious Bacterial Infection (pSBI) like poor feeding, lethargy, temperature instability, or respiratory distress [ 7 , 8 ]. While laboratory tests such as blood cultures to identify pathogens, complete blood counts to assess the immune response, and urine cultures for accurate sepsis diagnosis are crucial for effective treatment [ 9 ], in many health facilities in SSA, these diagnostic tests are unavailable at the time of admission [ 10 ]. There is no consensus on the clinical definition of neonatal sepsis where microbiological confirmations are unavailable [ 11 , 12 ]. It is a heterogeneous syndrome representing a diverse group of patients, ranging from neonates with minor infections that will resolve quickly to those with severe, life-threatening sepsis. Using broad-spectrum, empiric antibiotic treatment for all suspected cases of neonatal sepsis may not be ideal, as it can lead to unnecessary prolonged antibiotic use in non-infected infants, potentially causing harm [ 4 ]. Tailoring treatment based on specific diagnoses and risk of mortality could improve outcomes and reduce overuse of antibiotics [ 13 ] . The challenge lies in developing alternative effective strategies for early diagnosis and treatment of neonatal sepsis in environments such as SSA where microbiological investigation or laboratory testing is either unavailable or inaccessible [ 10 ]. Different combinations of pSBI signs may naturally cluster into previously undescribed subsets or phenotypes that may have different risks for the outcome and may respond differently to treatments. Identification of distinct clinical phenotypes may allow more precise therapy and improve care [ 4 ]. However, many guidelines and hospital protocols continue to recommend a one-size-fits-all approach; recommending the same initial inpatient treatment and follow up [ 4 ]. Efforts to address the variability in neonatal sepsis have included the WHO’s triaging guidelines for pSBI [ 4 , 14 ] as illustrated in supplementary table 1. Other approaches have included use of (1) cluster analysis of neonates based on clinical and laboratory data from electronic health records (EHRs) collected within the first six hours of the patients’ hospital stay, combined with serum biomarkers[ 15 ], (2) latent class analysis (LCA) to identify molecular phenotypes in sepsis patients using both clinical and protein biomarker data [ 16 ], and even used self-organizing maps to identify sepsis groups susceptible to multiple organ dysfunction syndrome based on their age and sequential organ failure assessment score [ 17 ]. Most of these approaches rely on clinical variables that are typically not collected for neonates in SSA at admission, given the limited laboratory capacity and lack of point-of-care diagnostics at scale, making findings from inclusion of such variables and subsequent clusters, largely ungeneralisable to many SSA hospital settings [ 4 , 10 ]. Using routinely collected data that is available in typical SSA hospital settings, the aim of this study is to identify clusters of neonates at time of admission that can be candidates for appropriate interventions to maximise patient outcomes given the resource constraints. More specifically, the objectives of this study were: To use unsupervised model-based cluster analysis to group and assign neonates into clinically meaningful clusters based on key clinical features (i.e. pSBI) from data routinely collected at admission in many SSA health facilities. To explores the external validity of the model-based clustering approach to different patient populations in similar clinical contexts. To assesses the identified clusters ability to predict in-hospital mortality and compares this with the “Gold standard”, the World Health Organization (WHO) classification of neonatal sepsis severity. Methods Ethics and reporting The reporting of this study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines, which provide a set of recommendations for transparent and comprehensive reporting of observational studies using cohort, case-control, or cross-sectional designs [ 18 ]. The reporting also follows the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines, which is a set of recommendations for the reporting of studies developing, validating, or updating prediction models for prognostic purposes [ 19 ]. The Scientific and Ethics Review Unit of the Kenya Medical Research Institute (KEMRI) approved the collection of the de-identified data that provides the basis for this study as part of the Clinical Information Network (CIN) and is run in partnership with the Ministry of Health and the participating hospitals. We elaborate on the CIN in the study design and settings section. Individual consent for retrospective access to the de-identified anonymised patient records was not required and was waived by KEMRI with the authors having no access to the information that could identify the patients. Study design and participants This is a retrospective cohort analysis utilizing data from the Clinical Information Network (CIN). The CIN collects standardized routine admission and discharge data from newborn units (NBU) across 21 public county hospitals in 14 out of the 47 counties in Kenya with a detailed description of the methods of data collection and management is provided elsewhere [ 20 ]. In brief, neonatal admission data are recorded using paper forms such as Neonatal Admission Records (NAR), treatment sheets, continuous monitoring charts, supplementary forms etc. These documents provide a structured checklist for typically junior clinicians [ 21 ], covering nine essential domains: demographics, admission details, maternal history, presenting complaints, cardinal signs, physical examinations, nursing monitoring, discharge status and supportive care. Each hospital has a clerk who then extracts data from these typical hospital forms into a Research Electronic Data Capture (REDCap) database immediately after patient death or discharge [ 22 ]. The study participants included inborn neonates aged 0–28 days admitted into newborn units between 01st January 2022 and 31st December 2024 with at least one pSBI sign/symptom at admission, regardless of whether they had a neonatal sepsis admission diagnosis, or were prescribed first-line intravenous antibiotics (e.g. Crystalline Penicillin and Gentamicin) at admission [ 23 ]. Data from 8 hospitals (n = 33094) was used as model derivation dataset, with the external geographical validation dataset coming from 13 hospitals (n = 23704); Hospitals in the derivation dataset represent large hospitals with ≥ 1000 NBU admissions per year, with the validation dataset having hospitals with < 1000 NBU admissions per year. Exclusion of out-born neonates was to rule out participants with community-acquired pSBI and/or those that might had outpatient antibiotic treatments [ 23 ]. Outcome The primary objective of this study is to identify clinically meaningful sub-populations of neonates with pSBI signs at admission based on key clinical features, derived using unsupervised clustering techniques fitted on the development dataset (n = 33094 admissions from 8 hospitals). The selected pSBI clinical variables included signs of local infection, fever, hypothermia, difficulty feeding, difficulty breathing, convulsions, apnoea, floppy, indrawing, grunting, crackles, jaundice, slow capillary refill, central cyanosis, irritable, hypoxic, and bulging fontanelle [ 14 ]. The second objective’s outcome was the predicted cluster membership in the external dataset (n = 23704 admissions from 13 hospitals) from a parsimonious classification model compared to the derived cluster membership from using the original model-based clustering approach from objective 1. For the third objective, the outcome of interest was mortality at discharge. Predictors This study analysed two sets of predictors corresponding to the study objectives. For the first and second objective, clinical variables listed in the section above which are commonly associated with sepsis/pSBI [ 14 ] were used to identify and characterise distinct clusters of neonates. Additionally, objective two adopted feature selection based on feature importance to identify a parsimonious set of pSBI predictors for cluster membership classification [ 24 ]. The identified clusters from objective one are the primary predictor variables for objective three, when assessing their association with mortality outcome; To ensure robust estimation, adjustments were made for potential confounders, neonate’s age and birth weight, given their well-established impact on length of stay and mortality [ 25 – 27 ]. Missing data Previous studies using CIN data showed that missing at random (MAR) is a reasonable assumption for our dataset [ 28 ]; For objective one, the model-based clustering approach adopts a Full Information Maximum Likelihood (FIML) approach where the cluster memberships are computed using all available information while taking the missingness into account under the MAR assumption, therefore no imputation was required [ 29 ]. For objective three, with the MAR assumption, multiple imputation using the chained equation approach was separately applied for both the development and validation datasets [ 30 ] in predicting in-hospital mortality outcome. Sample size For the first objective, given clustering is an unsupervised statistical learning technique that does not rely on predefined sample sizes but rather extracted patterns from the available data [ 31 ], coupled with a large development dataset (n = 33094 admissions), a formal sample size calculation was deemed unnecessary. All eligible neonatal records were thus included to maximize the robustness of the clustering approach. For the second objective, we based our sample size calculation on previously reported mortality outcome prevalence of between 9% − 14% with R-squared values of 0.453 derived from previously-developed mortality risk prediction models in similar neonatal patient populations [ 32 ]. Using the pmsampsize library in R [ 33 ] and assuming a maximum of six clinically-meaningfully clusters identified [ 34 , 35 ] after including age (in days) and birth weight, the required sample size for the mortality risk model development and validation would be 260 patients with 24 deaths; making both our development dataset with 33094 admissions (with 9768 deaths) and validation dataset with 23704 admissions (with 4180 deaths) satisfactory for clinical prediction model development and external validation. Statistical analysis methods Clustering approaches for neonates with pSBI managed for sepsis Latent Class Analysis (LCA) which is a Finite Mixture Modelling approach was used since it offered a model-based (i.e. probabilistic) approach to derive clusters [ 29 , 36 ]; LCA adopts Full Information Maximum Likelihood (FIML) approach making use of observations with missing data in the cluster estimation process [ 29 , 36 , 37 ]. LCA uses the distribution of the dataset to assess probabilities that certain patients are members of certain clusters. Best model-fit was determined by evaluating different metrics from LCA approach summarised in supplementary table two; In summary the most probable number of clusters is determined by clear delineation (i.e. entropy thresholds > 0.8), the smallest cluster size having considerable number of admissions, with the cluster model able to accurately predicting class membership for individual patients (i.e. Average Latent Class Posterior Probability (ALCPP) thresholds > 0.85) (Supplementary table two). The final LCA model used to generate the clusters in the development dataset was also used to also predict the cluster membership of the validation dataset. As part of sensitivity analysis, we compared the model-based clustering approach to a distance-based unsupervised machine learning technique of agglomerative hierarchical cluster analysis (HCA) that groups patients with similar pSBI into clusters without imposing a specific sequential order on them [ 38 , 39 ]. Agglomerative HCA was selected for sensitivity analysis due to its efficiency in handling mixed data types and its ability to generate clinically meaningful clusters that are stable and easy to interpret [ 40 ]. Given our dataset consists of both continuous and categorical variables, standard Euclidean distance measures suited for numerical data, were unsuitable [ 41 ]; Gower’s distance which is well suited for mixed data types was used instead [ 42 , 43 ], together with Ward’s minimum variance method as the linkage criterion to group neonates into meaningful clusters by minimising the total within-cluster variance [ 44 ]. The optimal number of clusters from HCA approach was determined using the silhouette score [ 45 ]. The Silhouette Score evaluated how well individual data points fit within their assigned clusters relative to the nearest alternative cluster, and ranges from − 1 to 1 with higher scores indicating better clustering [ 45 ]. Parsimonious classification model for cluster-membership For objective two, we used eXtreme Gradient Boosting (xgboost), a machine learning technique with implicit features selection that is robust to missing data patterns, to identify a subset of pSBI features based on feature importance to predict cluster membership [ 46 , 47 ]. Feature selection was based on SHAP (SHapley Additive exPlanations) values where the SHAP values derived from xgboost model which explain how the variables contribute to the clinical model's mortality prediction, showing if the variable pushed the prediction higher or lower than the average [ 48 ]. The parsimonious multivariate logistic regression model was developed from the development dataset (n = 33094 admissions from 8 hospitals) and externally validated on the validation dataset (n = 23704 admissions from 13 hospitals). For the process of model validation, the predicted class-membership from the parsimonious model (which was determined based on which cluster the patient had the highest probability of being a member of) was compared to class-membership from the original model-based clustering model. Typical classification metrics of accuracy, sensitivity, specificity, positive predictive value (PPV) etc. were used to report performance of the parsimonious classification model to correctly cluster patients into the model-based clusters. Logistic regression for in-hospital mortality For objective three, Logistic regression without variable selection was used together with the imputed datasets with parameter estimates being combined using Rubin’s rule. The number of imputation datasets to use was determined from the integer value of the percentage of patients in the derivation dataset that had one or more missing values, rounded upwards. To examine heterogeneity in model performance while incorporating geographical external validation, we compared the logistic regression models internal-external cross-validation performance where we omitted one hospital at a time using it as the validation dataset, built the model on the remaining hospitals, and evaluated model’s discrimination and calibration performance on the hospital left out. We repeated this process with each iteration using a different hospital as the validation data source [ 49 ]. The predictive performance of in-hospital mortality by the model-based and hierarchical clustering approaches was then compared to the WHO’s pSBI expert-based clustering approach (i.e. “Gold standard”)(Supplementary Table 1). Model performance was assessed using calibration and discrimination performance metrics detailed in supplementary table 3, with the confidence intervals for both c-statistic and calibration slope and intercept, calculated through bootstrapping (i.e., iterative sampling with replacement). Results Description of participants Between 01st January 2022 and 31st December 2024, 73140 neonates aged 0–28 days were admitted to newborn units across the 21 CIN hospitals. For analysis, 59302/73140 (81.08%) of the admissions had at least one pSBI sign/symptom and were eligible for inclusion in the study (Supplementary Fig. 1) with patterns of pSBI signs/symptoms illustrated in Supplementary Fig. 2; Of the 59302 admissions with ≥ 1 pSBI sign/symptom, 2504/59302 (4.22%) had signs of systematic missingness (i.e. ≥7/19 (≥ 35%) pSBI signs not documented; Within the CIN, this level of missingness is indicative of missing documentation due to patients with either severe symptoms, in emergency situations or paper documents removed from patient files; The patients with systematic missingness were excluded due to high likelihood of introducing bias if included in subsequent analysis due to violating the missing at random (MAR) assumption. The neonatal admissions in CIN during this period were divided geographically by hospitals into a development dataset from 8 CIN hospitals (n = 33094/56798, 58.27%), and a validation set from the remaining 13 CIN hospitals (n = 23704/56798, 41.73%). Excluding sepsis admission diagnoses, respiratory distress syndrome, birth asphyxia, and meconium aspiration were the most common admission diagnoses in the included patients (Supplementary Fig. 3). Table 1 illustrates the model-based clustering fit statistics in the development dataset. The identified clusters are of patients with pSBI managed for sepsis grouped to maximise similarity in pSBI characteristics. From the model fit statistics described in supplementary table 2, the maximum number of clusters with clear delineation (i.e. entropy thresholds > 0.8), whose smallest cluster size had arguably considerable number of admissions, and whose cluster model was able to accurately predicting class membership for individual patients (i.e. Average Latent Class Posterior Probability (ALCPP) thresholds > 0.85), is five (Table 1 ). Table 1 Model fit indices from Latent Class Analysis Fit Indices1 Clusters Parameters LL 2 AIC 3 BIC 4 aBIC 5 Entropy 6 Smallest Cluster Size ALCPP 7 LMR P-value 8 2 53 -367766 735639 736084 735916 0.870 10792 (32.61%) 0.957 < 0.001 3 79 -360415 720989 721653 721402 0.785 7162 (21.64%) 0.897 < 0.001 4 105 -357058 714327 715210 714876 0.752 4551 (13.75%) 0.857 < 0.001 5 131 -351102 702466 703568 703151 0.803 3368 (10.18%) 0.883 < 0.001 6 157 -350771 701857 703177 702678 0.799 1754 (5.3%) 0.844 < 0.001 1 Clustering mixture model performance metrics from MPLUS 7.1 software 2 Log-Likelihood 3 Akaike Information Criteria (AIC) 4 Bayesian Information Criteria (BIC) 5 Sample-Size adjusted Bayesian Information Criteria (aBIC) 6 How clear the class delineation is; How well the classes separate the population into distinct subgroups of patients 7 Average Latent Class Posterior Probability (ALCPP). The average probability of the class model accurately predicting class membership for individuals 8 Lo-Mendell-Rubin (LMR) p-value from likelihood ratio test for K-1 versus K classes Figure 2 illustrates the difference in the probability of pSBI signs and symptoms given cluster membership. The pSBI signs of floppy, tachypnoea, grunting, hypothermia, indrawing, hypoxia, difficulty in feeding and difficulty breathing appear to be useful in distinguishing different clusters (Fig. 2). There is substantive variation with in-hospital mortality based on cluster membership, with the minimal risk cluster having the least mortality cases (0.98%) and critical risk having the highest mortality cases (67.8%). The same pattern is present in the validation dataset, with 3.06% and 61.76% case fatality rate between the minimal risk cluster and critical risk cluster (Table 2 ). Patients in the critical risk cluster had low gestational age in weeks and low birth weight while patients in the substantial risk cluster had normal gestational age and birth weight but both clusters had at least 4 pSBI signs/symptoms and substantially high mortality rates of > 40% in both the development and validation datasets (Table 2 ). Admissions in the minimal and low mortality risk clusters had 2 (IQR: 1–3) pSBI signs/symptoms with antibiotics prescribed in ≥ 45% of the cases within each of these clusters in both the development and validation datasets. Table 2 Characteristics of patients included in model development and external validation Indicator Development (n = 33094) 1 Validation (n = 23704) 1 Minimal Risk n = 3368 (10.18%) Low Risk n = 12907 (39%) Moderate Risk n = 8879 (26.83%) Substantial Risk n = 4130 (12.48%) Critical Risk n = 3810 (11.51%) Minimal Risk n = 3763 (15.87%) Low Risk n = 8333 (35.15%) Moderate Risk n = 5841 (24.64%) Substantial Risk n = 2940 (12.4%) Critical Risk n = 2827 (11.93%) Neonatal Essential Signs & Symptoms: Binary Jaundice 1510 (44.83%) 21 (0.16%) 94 (1.06%) 99 (2.4%) 53 (1.39%) 1630 (43.32%) 14 (0.17%) 149 (2.55%) 178 (6.05%) 55 (1.95%) Crackles 29 (0.86%) 469 (3.63%) 304 (3.42%) 993 (24.04%) 298 (7.82%) 51 (1.36%) 394 (4.73%) 154 (2.64%) 751 (25.54%) 199 (7.04%) Bulging Fontanelle 57 (1.69%) 156 (1.21%) 101 (1.14%) 88 (2.13%) 40 (1.05%) 97 (2.58%) 149 (1.79%) 71 (1.22%) 102 (3.47%) 35 (1.24%) Local Infection 141 (4.19%) 95 (0.74%) 38 (0.43%) 70 (1.69%) 10 (0.26%) 318 (8.45%) 75 (0.9%) 39 (0.67%) 99 (3.37%) 10 (0.35%) Apnoea 6 (0.18%) 88 (0.68%) 190 (2.14%) 656 (15.88%) 620 (16.27%) 20 (0.53%) 135 (1.62%) 250 (4.28%) 654 (22.24%) 618 (21.86%) Difficulty Eating 675 (20.04%) 2892 (22.41%) 1982 (22.32%) 2284 (55.3%) 1667 (43.75%) 925 (24.58%) 1577 (18.92%) 1295 (22.17%) 1837 (62.48%) 1335 (47.22%) Difficulty Breathing 75 (2.23%) 5997 (46.46%) 5193 (58.49%) 3949 (95.62%) 3095 (81.23%) 172 (4.57%) 3341 (40.09%) 2858 (48.93%) 2762 (93.95%) 2142 (75.77%) Convulsions 107 (3.18%) 991 (7.68%) 39 (0.44%) 739 (17.89%) 29 (0.76%) 261 (6.94%) 973 (11.68%) 45 (0.77%) 780 (26.53%) 33 (1.17%) Floppy 656 (19.48%) 5616 (43.51%) 5101 (57.45%) 3183 (77.07%) 3175 (83.33%) 779 (20.7%) 2937 (35.25%) 2984 (51.09%) 2141 (72.82%) 2210 (78.17%) Indrawing 52 (1.54%) 279 (2.16%) 842 (9.48%) 1427 (34.55%) 1192 (31.29%) 33 (0.88%) 118 (1.42%) 359 (6.15%) 748 (25.44%) 748 (26.46%) Grunting 15 (0.45%) 675 (5.23%) 1722 (19.39%) 2378 (57.58%) 1514 (39.74%) 67 (1.78%) 408 (4.9%) 715 (12.24%) 1323 (45%) 797 (28.19%) Slow Capillary Refill 232 (6.89%) 812 (6.29%) 391 (4.4%) 407 (9.85%) 243 (6.38%) 298 (7.92%) 582 (6.98%) 410 (7.02%) 339 (11.53%) 287 (10.15%) Central Cyanosis 22 (0.65%) 248 (1.92%) 310 (3.49%) 536 (12.98%) 444 (11.65%) 43 (1.14%) 216 (2.59%) 194 (3.32%) 360 (12.24%) 276 (9.76%) Irritability 592 (17.58%) 774 (6%) 236 (2.66%) 562 (13.61%) 136 (3.57%) 852 (22.64%) 639 (7.67%) 254 (4.35%) 534 (18.16%) 104 (3.68%) Tachypnoea 358 (10.63%) 4957 (38.41%) 3415 (38.46%) 2077 (50.29%) 1344 (35.28%) 569 (15.12%) 3296 (39.55%) 1968 (33.69%) 1276 (43.4%) 846 (29.93%) Hypoxic 82 (2.43%) 1883 (14.59%) 1296 (14.6%) 1421 (34.41%) 1033 (27.11%) 149 (3.96%) 1136 (13.63%) 874 (14.96%) 935 (31.8%) 675 (23.88%) Fever (Examined) 1267 (37.62%) 406 (3.15%) 123 (1.39%) 188 (4.55%) 25 (0.66%) 1605 (42.65%) 452 (5.42%) 215 (3.68%) 370 (12.59%) 45 (1.59%) Hypothermia 17 (0.5%) 437 (3.39%) 837 (9.43%) 474 (11.48%) 744 (19.53%) 31 (0.82%) 490 (5.88%) 787 (13.47%) 366 (12.45%) 803 (28.4%) Abnormal Heart Rate 121 (3.59%) 418 (3.24%) 291 (3.28%) 383 (9.27%) 305 (8.01%) 208 (5.53%) 401 (4.81%) 309 (5.29%) 364 (12.38%) 289 (10.22%) Male 1896 (56.29%) 7814 (60.54%) 4602 (51.83%) 2580 (62.47%) 2043 (53.62%) 2132 (56.66%) 4967 (59.61%) 2890 (49.48%) 1754 (59.66%) 1455 (51.47%) Neonatal Essential Signs & Symptoms: Continuous pSBI signs 1 (IQR: 1–2) 2 (IQR: 1–3) 2 (IQR: 1–3) 5 (IQR: 4–6) 4 (IQR: 3–5) 2 (IQR: 1–3) 2 (IQR: 1–3) 2 (IQR: 1–3) 5 (IQR: 4–6) 4 (IQR: 3–5) Gestational Age 38 (IQR: 38–40) 38 (IQR: 38–40) 34 (IQR: 32–35) 38 (IQR: 38–40) 28 (IQR: 27–30) 38 (IQR: 38–40) 39 (IQR: 38–40) 33 (IQR: 32–35) 38 (IQR: 38–40) 28 (IQR: 27–30) Birth Weight 3.1 (IQR: 2.8–3.45) 3.1 (IQR: 2.79–3.435) 1.86 (IQR: 1.6–2.1) 2.99 (IQR: 2.6–3.3) 1.17 (IQR: 0.94–1.4) 3.1 (IQR: 2.78–3.5) 3.1 (IQR: 2.785–3.46) 1.8 (IQR: 1.54–2.08) 3 (IQR: 2.6-3.3525) 1.2 (IQR: 1-1.47) Age 3 (IQR: 2–4) 1 (IQR: 1–1) 1 (IQR: 1–1) 1 (IQR: 1–1) 1 (IQR: 1–1) 3 (IQR: 2–5) 1 (IQR: 1–1) 1 (IQR: 1–1) 1 (IQR: 1–2) 1 (IQR: 1–1) Temperature 37.4 (IQR: 36.7–38.4) 36.7 (IQR: 36.3–36.9) 36.5 (IQR: 36.1–36.8) 36.5 (IQR: 36.1–36.8) 36.3 (IQR: 35.675–36.7) 37.8 (IQR: 36.8–38.5) 36.7 (IQR: 36.2–37.1) 36.5 (IQR: 36-36.9) 36.7 (IQR: 36.1–37.3) 36.2 (IQR: 35.2–36.7) Respiratory Rate 49 (IQR: 45–56) 57 (IQR: 48–66) 57 (IQR: 49–65) 62 (IQR: 50–71) 56 (IQR: 46–65) 49 (IQR: 44–57) 56 (IQR: 48–67) 55 (IQR: 46–64) 60 (IQR: 48–70) 53 (IQR: 43–63) Heart Rate 138 (IQR: 127–146) 138 (IQR: 127–148) 139 (IQR: 128–149) 140 (IQR: 125–152) 140 (IQR: 124–152) 139 (IQR: 126–150) 138 (IQR: 125–149) 141 (IQR: 127–152) 142 (IQR: 124–156) 140 (IQR: 124–153) Oxygen Saturation 95 (IQR: 94–97) 95 (IQR: 92–97) 95 (IQR: 92–97) 92 (IQR: 84–96) 93 (IQR: 87–96) 95 (IQR: 93–97) 95 (IQR: 92–98) 95 (IQR: 92–98) 92 (IQR: 83–96) 94 (IQR: 87–97) Neonatal Essential Interventions Oxygen Prescribed 928 (27.55%) 7035 (54.51%) 5776 (65.05%) 3018 (73.08%) 2907 (76.3%) 398 (10.58%) 3702 (44.43%) 3033 (51.93%) 2093 (71.19%) 1978 (69.97%) Fluids Prescribed 103 (3.06%) 5109 (39.58%) 4159 (46.84%) 1973 (47.77%) 2038 (53.49%) 72 (1.91%) 1298 (15.58%) 1016 (17.39%) 398 (13.54%) 603 (21.33%) Feeds Prescribed 627 (18.62%) 1527 (11.83%) 817 (9.2%) 232 (5.62%) 178 (4.67%) 201 (5.34%) 312 (3.74%) 242 (4.14%) 99 (3.37%) 64 (2.26%) Antibiotics Prescribed 2166 (64.31%) 8641 (66.95%) 6116 (68.88%) 2827 (68.45%) 2672 (70.13%) 1789 (47.54%) 3806 (45.67%) 2362 (40.44%) 1241 (42.21%) 1066 (37.71%) Phenobarbital Prescribed 105 (3.12%) 1423 (11.03%) 166 (1.87%) 976 (23.63%) 51 (1.34%) 273 (7.25%) 1355 (16.26%) 115 (1.97%) 787 (26.77%) 46 (1.63%) Neonatal Outcomes Length Of Stay 5 (IQR: 3–7) 5 (IQR: 3–7) 12 (IQR: 5–23) 4 (IQR: 1–8) 4 (IQR: 1–27) 5 (IQR: 3–7) 5 (IQR: 3–7) 12 (IQR: 5–22) 3 (IQR: 1–8) 5 (IQR: 1–29) Died 33 (0.98%) 369 (2.86%) 849 (9.56%) 1754 (42.47%) 2583 (67.8%) 115 (3.06%) 355 (4.26%) 524 (8.97%) 1441 (49.01%) 1745 (61.73%) 1 Continuous variable summaries given as median (Interquartile range). Discrete variable summaries given as proportions External validation of a parsimonious model-based clustering approach The pSBI signs and symptoms used in model-based clustering were 19, excluding birth weight and gestational age and the mortality outcome. From feature selection using xgboost, using we subjectively judged 11 (52%) of pSBI signs and symptoms to be sufficiently capable of assigning the patients into the mortality risk clusters using a multivariate logistic model with the outcome as the clusters based on values (Table 3 , Supplementary Fig. 4). From Table 3 , the cluster membership of any neonate with a pSBI sign/symptom can be determined based on which cluster the patient had the highest probability of being a member of, with the cluster-specific probabilities computed from the sum of the logits based on the presence/absence of the sign or symptom. The parsimonious model performance in predicting class membership in the external validation dataset showed relatively high accuracy, specificity, sensitivity, negative- and positive predictive performance across all clusters (Table 4 ). Table 3 Parsimonious multivariate logistic model for determining neonatal cluster membership based on pSBI signs and symptoms1 Indicator Minimal Low Moderate Substantial Critical Beta SE p-value Beta SE p-value Beta SE p-value Beta SE p-value Beta SE p-value Intercept 4.936 0.146 < 0.001 −6.060 0.181 < 0.001 −2.202 0.092 < 0.001 −7.620 0.165 < 0.001 −22.123 0.783 < 0.001 Birth Weight 1 0.499 0.058 < 0.001 1.260 0.035 < 0.001 −1.692 0.034 < 0.001 0.645 0.050 < 0.001 −4.861 0.198 < 0.001 Gestational Age 2 0.206 0.017 < 0.001 0.508 0.010 < 0.001 −0.017 0.006 0.004 0.601 0.015 < 0.001 −1.807 0.059 < 0.001 Early Onset Age 3 −5.876 0.134 < 0.001 6.592 0.183 < 0.001 1.531 0.094 < 0.001 −1.861 0.131 < 0.001 −0.262 0.370 0.479 Difficulty Breathing −4.215 0.166 < 0.001 −0.015 0.040 0.698 −0.045 0.033 0.177 3.578 0.117 < 0.001 1.804 0.145 < 0.001 Floppy −2.246 0.082 < 0.001 −0.261 0.040 < 0.001 −0.218 0.034 < 0.001 1.728 0.071 < 0.001 2.106 0.160 < 0.001 Grunting −2.633 0.414 < 0.001 −2.109 0.059 < 0.001 −0.048 0.042 0.256 2.627 0.069 < 0.001 1.542 0.135 < 0.001 Hypoxic −1.885 0.174 < 0.001 −0.450 0.050 < 0.001 −0.508 0.042 < 0.001 1.087 0.066 < 0.001 1.483 0.144 < 0.001 Difficulty Feeding −0.314 0.085 < 0.001 −0.675 0.044 < 0.001 −0.572 0.037 < 0.001 1.295 0.059 < 0.001 1.502 0.130 < 0.001 Tachypnoea −1.998 0.092 < 0.001 0.161 0.041 < 0.001 0.104 0.033 0.002 0.508 0.062 < 0.001 −0.176 0.118 0.134 Indrawing −1.732 0.287 < 0.001 −2.145 0.086 < 0.001 −0.854 0.053 < 0.001 1.941 0.076 < 0.001 1.956 0.156 < 0.001 Crackles −2.264 0.323 < 0.001 −1.254 0.078 < 0.001 −0.571 0.074 < 0.001 1.901 0.084 < 0.001 1.410 0.233 < 0.001 Hypothermia −2.817 0.374 < 0.001 −1.031 0.079 < 0.001 −0.649 0.054 < 0.001 1.421 0.098 < 0.001 1.404 0.152 < 0.001 Apnoea −2.788 0.642 < 0.001 −2.653 0.139 < 0.001 −1.624 0.090 < 0.001 2.006 0.105 < 0.001 3.641 0.248 < 0.001 Note : 1 Birth weight centered at 2.5 kgs; 2 Gestational age centered at 37 weeks; 3 Neonate’s age in ≤ 72 hours Table 4 Performance of the parsimonious multivariate logistic models in predicting cluster membership in the external validation dataset Metric Minimal Risk Low Risk Moderate Risk Substantial Risk Critical Risk Balanced Accuracy 83.16 (82.72 to 83.62) 91.87 (91.68 to 92.05) 93.94 (93.71 to 94.19) 85.21 (84.68 to 85.7) 95.36 (95 to 95.71) F1 Score 78.01 (77.36 to 78.68) 88.22 (87.93 to 88.49) 91.1 (90.78 to 91.42) 78.72 (77.92 to 79.47) 93.14 (92.69 to 93.59) Negative Predictive Value 94.14 (93.96 to 94.32) 97.94 (97.8 to 98.08) 96.96 (96.81 to 97.13) 96.32 (96.17 to 96.47) 98.9 (98.81 to 98.99) Positive Predictive Value 92.72 (92.2 to 93.26) 81.02 (80.58 to 81.45) 91.65 (91.21 to 92.03) 87.05 (86.15 to 87.98) 95.02 (94.5 to 95.59) Recall 67.33 (66.43 to 68.24) 96.83 (96.61 to 97.04) 90.55 (90.08 to 91.03) 71.86 (70.77 to 72.85) 91.32 (90.62 to 92.03) Sensitivity 67.33 (66.43 to 68.24) 96.83 (96.61 to 97.04) 90.55 (90.08 to 91.03) 71.86 (70.77 to 72.85) 91.32 (90.62 to 92.03) Specificity 99 (98.93 to 99.08) 86.91 (86.61 to 87.23) 97.33 (97.19 to 97.47) 98.57 (98.47 to 98.67) 99.39 (99.33 to 99.46) Note : The 95% confidence interval was determined through computing the metrics in each of the 1000 iterations of sampling with replacement of 10% of the external validation dataset Mortality risk prediction performance based on clusters Figure 3 compares the performance of the three clustering approaches in predicting mortality risk. Overall, 33004/33094 (99%) admissions of the development dataset and 23601/23704 (99%) of the validation dataset were eligible for this comparison; The omitted admissions were excluded since their clinical signs and symptoms did not align with any of the predefined WHO clusters (Supplementary Table 1), coupled with variable missingness, posing a challenge in assigning them to a WHO cluster. The c-statistic (discrimination) for the WHO expert-based clustering approach was 0.721 (95% CI: 0.715 to 0.727) with a calibration intercept of 0.018 (95% CI: -0.005 to 0.041) and calibration slope of 1.015 (95% CI: 0.986 to 1.043). The c-statistic (discrimination) for the distance-based clustering approach was 0.721 (95% CI: 0.714 to 0.727) while its calibration intercept was − 0.002 (95% CI: -0.025 to 0.021) and calibration slope was 0.987 (95% CI: 0.96 to 1.014). The WHO and distance-based clustering approach calibration curves both showed substantive variances across the mortality risk spectrum with over-estimation of mortality at the lowest predicted risks and under-estimation at moderate to high risks of mortality (Fig. 3). The model-based clustering approach had the highest c-statistic (discrimination) of 0.867 (95% CI: 0.863 to 0.871), with a calibration intercept of -0.004 (95% CI: -0.031 to 0.023) and a calibration slope of 0.996 (95% CI: 0.979 to 1.043). The model-based clustering approach’s calibration curve showed close and relatively better alignment between the predicted risks and observed outcomes across the probability range compared to the alternative clustering approaches (Fig. 3). Discussion Summary of findings This study used a pragmatic approach with a limited set of signs easy to use by clinicians (usually busy junior clinicians) to identify residual heterogeneity in the neonates admitted with signs indicative of sepsis, that can be used to inform to inform antibiotic use treatment decisions and strategies. From applying model-based clustering approaches on routinely collected neonatal data at hospital admission, we identified four clinically distinct clusters of neonates with pSBI managed for sepsis, labelled based on the mortality risk in each cluster: minimal risk, moderate risk, substantial risk and critical risk (Table 2 , Fig. 2). From external validation, we further demonstrate that the model-based clusters outperform distance-based and WHO expert-based clustering approaches in predicting in-hospital mortality based on discrimination and calibration statistics (Fig. 3). Among the 19 pSBI signs/symptoms collected at admission and used in the initial model-based clustering, a subset of 11 pSBI signs and symptoms were sufficient to determine cluster membership of a patient and subsequently the associated mortality risk from pSBI at admission (Table 4 , Table 5). Hypothermia, indrawing, floppy, tachypnoea, grunting, hypoxia, difficulty in feeding and difficulty breathing were key in distinguishing the clusters. These signs were mostly prevalent in the substantial and critical risk groups with tachypnoea being more common in the substantial risk cluster while hypothermia and indrawing were more common in the critical risk cluster. The critical risk cluster also comprised of neonates with low gestation age and birthweight while those in substantial risk cluster had normal gestation age and birthweight yet still faced substantially high mortality rate. The moderate risk cluster on the hand, included neonates with low birth weight but demonstrated a relatively lower rate of mortality risk, highlighting the heterogeneity in mortality risk beyond birth-related factors. Comparison to other findings The findings of this study contribute to the growing evidence of how unsupervised machine learning methods might identify meaningful phenotypes in neonatal sepsis populations. For instance, Seymour et al. and Jang et al. [ 4 , 15 ] employed k-means clustering and identified up to four clusters. Sinha et al. used model-based clustering and derived two clusters, unlike our findings. These studies used patients with different characteristics and clinical variables whose availability at the time of admission varied considerably. Unlike this current study which used routinely collected clinical data from low-resource neonatal units in an SSA country and relied only on clinical signs and symptoms, some of the other prior studies were conducted in high-income settings and included a set of variables combining both clinical and laboratory parameters that are typically unavailable at neonatal admission such as white blood cell count, platelet count, C-reactive protein, premature neutrophil count and serum levels. Furthermore, these studies neither directly assessed the external validity of the clustering approach to classify new patients into the identified clusters nor did they provide the probability of patients having a combination of signs and symptoms given their cluster membership. These methodological differences highlight how selection of variables and validation approaches influence findings across studies. The observed differences in mortality risk among the clusters in this study and how the risk differs based on the probability of pSBI in each cluster agree with findings from Puri et al [ 50 ] although our findings support five distinct neonatal sepsis clusters. Hypothermia, among the signs previously identified as strongly predictive of death, was particularly common in the critical risk cluster of our study which had the highest mortality risk. Implications of findings The clusters identified through model-based unsupervised learning approach reliably stratify neonatal sepsis into patient sub-populations predictive of in-hospital mortality at admission, outperforming the WHO expert-based and distance-based clustering methods as exemplified by the reported discrimination and calibration statistics (Fig. 3). This work shows that there is residual heterogeneity in the neonatal sepsis in-hospital admissions linked to mortality as an outcome. The improved mortality risk stratification exemplified by our model-based clustering approach provides a foundation for more tailored clinical decision making and potential development of phenotype-specific treatment guidelines that does not neglect the typical multimorbid nature of neonatal sepsis admissions in routine SSA hospitals. Additionally, the identification of substantial and critical mortality risk phenotype with distinct clinical profiles could guide prioritisation of care or more aggressive interventions, potentially improving outcomes in these groups. With the global rise in Antimicrobial resistance (AMR) risk, the use of antibiotics in > 45% in the CIN admissions within minimal and low mortality risk clusters might be indicative of antibiotic overuse or misuse in these hospital settings, and suggest possibility of less aggressive treatments. This study identifies groups which are more granular than the WHO categories that may be amenable to different antibiotic use strategies, e.g. shorter intravenous antibiotic use or and perhaps oral antibiotics etc. These findings can either (a) be used to augment the clinical judgement of clinicians in risk stratification of neonatal sepsis patients at admission [ 51 ] and/or (b) inform the design and implementation of emulated trials [ 52 ] to evaluate different antibiotic treatment strategies for neonatal sepsis at admission informed by the identified clusters [ 53 , 54 ]. By relying on clinical signs and symptoms readily available at admission, the model-based clusters have better chances of being incorporated into routine clinical workflows at the point of care across SSA, with the potential to enhance early tailored interventions and treatment to ensure critical care is efficiently administered where it is most needed. Strengths and limitations of findings This study has several notable strengths. First, it is based on a large multi-centre cohort of neonatal admission (n = 56798) across 21 hospitals over three years, ensuring geographical and time coverage thus increasing confidence in the applicability of the modelling approaches across diverse neonatal care contexts. The inclusion criteria ensured the study sample was reflective of the typical multi-morbidity of neonatal admissions, thereby strengthening the real-world relevance of the findings. From a methodological standpoint, this study advances neonatal sepsis phenotyping by moving beyond conventional distance-based clustering techniques to more superior probabilistic, multivariate model-based approaches, which unlike distance-based methods, do not use a forced-choice mechanism to assign patients to clusters. The absence of microbiological validation to confirm the presence of bacteria indicative of sepsis is a limitation to this study. Since the aim was to classify neonates based on clinical pSBI signs in recognition of limited laboratory capacity in LMICs, the clusters should hence be interpreted as clinical phenotypes rather than pathogen confirmed syndromes. Future work would involve validating and extending this study’s findings by integrating microbiological testing with clinical clustering to ascertain whether the data-driven phenotypic clusters truly reflect bacterial infections. Prospective studies could also evaluate whether these groups respond differently to interventions or management strategies, providing insights into their clinical utility and potential to inform targeted risk-based neonatal care. Conclusions In this retrospective cohort study, we used pSBI signs that are readily available at neonatal admission across SSA for model-based clustering approach and identified four clinically distinct clusters differentiated by mortality risk: minimal, moderate, substantial and critical risk. From discrimination and calibration statistics, the model-based clustering approach outperformed clustering patients using WHO expert-based guidelines and typical distance-based clustering approaches. These clusters demonstrate the potential to complement clinical judgement in mortality risk stratification at admission and could inform future randomised clinical trials to test whether triage guided by these clusters could improve clinical outcomes such as mortality and length of hospital stay. Future work could also integrate microbiological validation to confirm whether clusters map onto pathogen-specific syndromes and investigate possible variation in treatment responses across the clusters. These efforts could help strengthen the case for more targeted and risk-based neonatal sepsis management, informed by data-driven clustering. Abbreviations CIN Clinical Information Network EHRs Electronic Health Records HICs High Income Countries HCWs Health Care Workers KPA Kenya Paediatric Association LCA Latent Class Analysis LMICs Low and middle-income countries MoH Ministry of Health NBU Newborn Unit pSBI possible Serious Bacterial Infection (pSBI) SERU KEMRI’s Scientific and Ethics Review Unit SSA Sub-Saharan Africa WHO World Health Organization. Declarations Ethics approval and consent to participate Ethical approval was provided by the KEMRI Scientific and Ethical Review Committee (SERU 3459). Individual consent for retrospective access to the de-identified anonymised patient records was not required and was waived by KEMRI with the authors having no access to the information that could identify the patients. Consent for publication This study is published with the permission of the Director of Kenya Medical Research Institute (KEMRI). Availability of Data and Materials The datasets generated and/or analysed during the current study are not publicly available due to the primary data being owned by the hospitals and their counties with the Ministry of Health; The research staff do have permission to share the data without further written approval from both the KEMRI-Wellcome Trust Data Governance Committee and the Facility, County or Ministry of Health as appropriate to the data request. Requests for access to primary data from quantitative research by people other than the investigators will be submitted to the KEMRI-Wellcome Trust Research Programme data governance committee as a first step through [email protected] , who will advise on the need for additional ethical review by the KEMRI Research Ethics Committee. Competing interests The authors have declared that no competing interests exist. Funding This work was primarily supported by a Wellcome Trust Early Career Research Fellowship (#227562/Z/23/Z) awarded to TT and a Wellcome Trust Senior Fellowship (#097170) awarded to ME. Additional support was provided by a Wellcome Trust core grant awarded to the KEMRI-Wellcome Trust Research Programme (#092654). The funders had no role in the preparation of this report or the decision to submit for publication. Author contributions Authorship eligibility guidelines for the final reports adhere to the Contributor Roles Taxonomy (CRediT) statement guidelines. Conceptualization: TT, JA; Supervision: TT, JA; Methodology: TT, JA; Formal Analysis: TT, TM; Investigation: TT, TM, JA, ME; Resources: TT, ME; Data Curation: TT, TM; Writing – original draft: TT, TM; Writing – review & editing: TT, TM, ME, JA; Funding acquisition: TT, ME; Acknowledgements The Clinical Information Network (CIN) Group : The CIN group hospital teams who are tagged to collaborate in the network’s development, data collection, data management, implementation of audit and feedback interventions and who will participate in this study include the following focal persons: Paediatricians: Juma Vitalis, Nyumbile Bonface, Roselyne Malangachi, Christine Manyasi, Catherine Mutinda, David Kibiwott Kimutai, Rukia Aden, Caren Emadau, Elizabeth Atieno Jowi, Cecilia Muithya, Charles Nzioki, Supa Tunje, Penina Musyoka, Wagura Mwangi, Agnes Mithamo, Magdalene Kuria, Esther Njiru, Mwangi Ngina, Penina Mwangi, Rachel Inginia, Melab Musabi, Emma Namulala, Grace Ochieng, Lydia Thuranira, Felicitas Makokha, Josephine Ojigo, Beth Maina, Catherine Mutinda, Mary Waiyego, Bernadette Lusweti, Angeline Ithondeka, Julie Barasa, Meshack Liru, Elizabeth Kibaru, Alice Nkirote Nyaribari, Joyce Akuka, Joyce Wangari; Nurses: Amilia Ngoda, Aggrey Nzavaye Emenwa, Patricia Nafula Wesakania, George Lipesa, Jane Mbungu, Marystella Mutenyo, Joyce Mbogho, Joan Baswetty, Ann Jambi, Josephine Aritho, Beatrice Njambi, Felisters Mucheke, Zainab Kioni, Jeniffer, Lucy Kinyua, Margaret Kethi, Alice Oguda, Salome Nashimiyu Situma, Nancy Gachaja, Loise N. Mwangi, Ruth Mwai, irginia Wangari Muruga, Nancy Mburu, Celestine Muteshi, Abigael Bwire, Salome Okisa Muyale, Naomi Situma, Faith Mueni, Hellen Mwaura, Rosemary Mututa, Caroline Lavu, Joyce Oketch, Jane Hore Olum, Orina Nyakina, Faith Njeru, Rebecca Chelimo, Margaret Wanjiku Mwaura, Ann Wambugu, Epharus Njeri Mburu, Linda Awino Tindi, Jane Akumu, Ruth Otieno, Slessor Osok; Health Record Information Officers (HRIOs): Seline Kulubi, Susan Wanjala, Pauline Njeru, Rebbecca Mukami Mbogo, John Ollongo, Samuel Soita, Judith Mirenja, Mary Nguri, Margaret Waweru, Mary Akoth Oruko, Jeska Kuya, Caroline Muthuri, Esther Muthiani, Esther Mwangi, Joseph Nganga, Benjamin Tanui, Alfred Wanjau, Judith Onsongo, Peter Muigai, Arnest Namayi, Elizabeth Kosiom, Dorcas Cherop, Faith Marete, Johanness Simiyu, Collince Danga, Arthur Otieno Oyugi, Fredrick Keya Okoth. The Clinical Information Network (CIN) Group ’smonitored email address is [email protected] and the list can change when new paediatrician(s), nurse(s) or HRIO leave or come into the hospital. Open access This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. References United Nations Inter-agency Group for Child Mortality Estimation (IGME) (2024) Levels and Trends in Child Mortality, Report 2023. UNICEF, New York Breiman RF et al (2021) Postmortem investigations and identification of multiple causes of child deaths: an analysis of findings from the Child Health and Mortality Prevention Surveillance (CHAMPS) network. PLoS Med 18(9):e1003814 Seale AC et al (2013) Neonatal severe bacterial infection impairment estimates in South Asia, sub-Saharan Africa, and Latin America for 2010. Pediatr Res 74(1):73–85 Seymour CW et al (2019) Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis. JAMA, 321(20) Celik IH et al (2022) Diagnosis of neonatal sepsis: the past, present and future. Pediatr Res 91(2):337–350 Hornik CP et al (2012) Early and late onset sepsis in very-low-birth-weight infants from a large group of neonatal intensive care units. Early Hum Dev 88:69–74 Opiyo N, English M (2011) What clinical signs best identify severe illness in young infants aged 0–59 days in developing countries? A systematic review. Arch Dis Child 96(11):1052–1059 Otu A et al (2020) How to close the maternal and neonatal sepsis gap in sub-Saharan Africa. BMJ Global Health, 5(4) Raturi A, Chandran S (2024) Neonatal Sepsis: Aetiology, Pathophysiology, Diagnostic Advances and Management Strategies, vol 18. Pediatrics, Clinical Medicine Insights, p 11795565241281337 Russell NJ et al (2023) Patterns of antibiotic use, pathogens, and prediction of mortality in hospitalized neonates and young infants with sepsis: A global neonatal sepsis observational cohort study (NeoOBS). PLoS Med 20(6):e1004179 McGovern M et al (2020) Challenges in developing a consensus definition of neonatal sepsis. Pediatr Res 88(1):14–26 Molloy EJ et al (2020) Neonatal sepsis: need for consensus definition, collaboration and core outcomes. Pediatr Res 88(1):2–4 Rosa-Mangeret F et al (2024) Challenges and opportunities in neonatal sepsis management: Insights from a survey among clinicians in 25 Sub-Saharan African countries. BMJ Paediatrics Open 8(1):e002398 World Health Organization (2015) Guideline: Managing possible serious bacterial infection in young infants when referral is not feasible. Geneva, World Health Organisation Jang JY et al (2022) Identification of the robust predictor for sepsis based on clustering analysis. Sci Rep 12(1):2336 Sinha P et al (2023) Identifying molecular phenotypes in sepsis: an analysis of two prospective observational cohorts and secondary analysis of two randomised controlled trials. Lancet Respiratory Med 11(11):965–974 Knox DB et al (2015) Phenotypic clusters within sepsis-associated multiple organ dysfunction syndrome. Intensive Care Med 41(5):814–822 von Elm E et al (2008) The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 370(9596):1453–1457 Collins GS et al (2015) Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): The TRIPOD Statement. J Br Surg 102(3):148–158 Tuti T et al (2016) Innovating to enhance clinical data management using non-commercial and open source solutions across a multi-center network supporting inpatient pediatric care and research in Kenya. J Am Med Inform Assoc 23(1):184–192 Ogero M et al (2020) Examining which clinicians provide admission hospital care in a high mortality setting and their adherence to guidelines: an observational study in 13 hospitals. Arch Dis Child 105(7):648–654 Harris PA et al (2009) Research electronic data capture (REDCap)—A metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 42(2):377–381 Aluvaala J et al (2025) Situational analysis of antibiotic prescriptions in Kenyan neonatal units for antimicrobial stewardship: a retrospective longitudinal study. EClinicalMedicine, 82 Hu C et al (2022) Interpretable machine learning for early prediction of prognosis in sepsis: a discovery and validation study. Infect Dis therapy 11(3):1117–1132 Jovičić M et al (2023) Predictors of Mortality in Early Neonatal Sepsis: A Single-Center Experience. Medicina 59(3):604 Danny) Liang L et al (2018) Predictors of mortality in neonates and infants hospitalized with sepsis or serious infections in developing countries: A systematic review. Front Pead 6:277 Gan MY et al (2022) Contemporary Trends in Global Mortality of Sepsis Among Young Infants Less Than 90 Days: A Systematic Review and Meta-Analysis. Front Pead 10:890767 Gachau S et al (2020) Handling missing data in modelling quality of clinician-prescribed routine care: Sensitivity analysis of departure from missing at random assumption. Stat Methods Med Res 29(10):3076–3092 Depaoli S, Jia F, Visser M (2025) Addressing Missing Data in Latent Class Analysis When Using a Three-Step Estimation Approach. Struct Equation Modeling: Multidisciplinary J 32(2):287–303 Van Buuren S, Groothuis-Oudshoorn K (2011) mice: Multivariate Imputation by Chained Equations in R. ournal Stat Softw 45:1–67 Ruppert D (2004) The Elements of Statistical Learning: Data Mining, Inference, and Prediction Tuti T et al (2022) Improving in-patient neonatal data quality as a pre-requisite for monitoring and improving quality of care at scale: A multisite retrospective cohort study in Kenya. PLOS Global Public Health 2(10):e0000673 Riley RD et al (2020) Calculating the sample size required for developing a clinical prediction model. BMJ, 368 Harrell F (2024) The Burden of Demonstrating Statistical Validity of Clusters , in Statistical Thinking . F. Harrell, Editor. Harrell F (2017) Statistical Errors in the Medical Literature , in Statistical Thinking . Statistical Thinking, F. Harrell, Editor Alagöz EC, Vermunt JK (2022) Stepwise Latent Class Analysis in the Presence of Missing Values on the Class Indicators. Struct Equation Modeling: Multidisciplinary J 29(5):784–790 Tang D, Tong X (2023) A Comparison of Full Information Maximum Likelihood and Machine Learning Missing Data Analytical Methods in Growth Curve Modeling. The Annual Meeting of the Psychometric Society, : pp. 99–109 Cabezas LM, Izbicki R, Stern RB (2023) Hierarchical clustering: Visualization, feature importance and model selection. Appl Soft Comput 141:110303 Landau S, Leese M, Stahl D, Everitt B (2011) and Cluster Analysis. Moschidis O, Markos A, Chadjipadelis T (2023) Hierarchical clustering of mixed-type data based on barycentric coding. Behaviormetrika 50(1):465–489 Costa E, Papatsouma I, Markos A (2023) Benchmarking distance-based partitioning methods for mixed-type data. Adv Data Anal Classif 17(3):701–724 Liu P et al (2024) A modified and weighted Gower distance-based clustering analysis for mixed type data: a simulation and empirical analyses. BMC Med Res Methodol 24(1):305 Gower JC (1971) A General Coefficient of Similarity and Some of Its Properties. Biometrics 27(4):857–871 Ward JH Jr (1963) Hierarchical Grouping to Optimize an Objective Function. J Am Stat Assoc 58(301):236–244 Rousseeuw PJ (1987) Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. J Comput Appl Math 20:53–65 Yuan K-C et al (2020) The development an artificial intelligence algorithm for early sepsis diagnosis in the intensive care unit. Int J Med Informatics 141:104176 Chen T, Guestrin C (2016) Xgboost: A scalable tree boosting system . in The 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . San Francisco California: Association for Computing Machinery Lundberg S An introduction to explainable AI with Shapley values . Welcome to the SHAP documentation 2018 2018 [cited 2025 17 December]; Available from: https://shap.readthedocs.io/en/latest/example_notebooks/overviews/An%20introduction%20to%20explainable%20AI%20with%20Shapley%20values.html Takada T et al (2021) Internal-external cross-validation helped to evaluate the generalizability of prediction models in large clustered datasets. J Clin Epidemiol 137:83–91 Puri D et al (2021) Prevalence of clinical signs of possible serious bacterial infection and mortality associated with them from population-based surveillance of young infants from birth to 2 months of age. PLoS ONE 16(2):e0247457 van Doorn WP et al (2021) A comparison of machine learning models versus clinical evaluation for mortality prediction in patients with sepsis. PLoS ONE 16(1):e0245157 Matthews AA et al (2022) Target trial emulation: applying principles of randomised trials to observational studies. BMJ, 378 Shimabukuro DW et al (2017) Effect of a machine learning-based severe sepsis prediction algorithm on patient survival and hospital length of stay: a randomised clinical trial. BMJ open respiratory Res, 4(1) Chiew CJ et al (2019) Heart rate variability based machine learning models for risk prediction of suspected sepsis patients in the emergency department. Medicine 98(6):e14197 Lo Y, Nancy R, Mendell, Rubin DB (2001) Testing the number of components in a normal mixture. Biometrika 88(3):767–778 Nylund-Gibson K, Choi AY (2018) Ten frequently asked questions about latent class analysis. Translational issues Psychol Sci 4(4):440–461 Celeux G, Soromenho G (1996) An entropy criterion for assessing the number of clusters in a mixture model. 13:195–212 Wang MC et al (2017) Performance of the entropy as an index of classification accuracy in latent profile analysis: A Monte Carlo simulation study. Acta Physiol Sinica 49(11):1473–1482 Shanahan L et al (2013) Sex-differentiated changes in C-reactive protein from ages 9 to 21: the contributions of BMI and physical/sexual maturation. Psychoneuroendocrinology 38(10):2209–2217 Riley RD et al (2019) External validation of clinical prediction models using big datasets from e-health records or IPD meta-analysis: Opportunities and challenges. BMJ, 365 Additional Declarations The authors declare no competing interests. Supplementary Files SupplementaryFigure1.tiff pSBI signs and symptoms present in the different datasets SupplementaryFigure2.tiff Patterns of pSBI signs and symptoms missingness in the different datasets SupplementaryFigure3.tiff Common admission diagnoses for neonates managed for sepsis (n=8012), excluding sepsis diagnosis SupplementaryFigure4.png Feature importance for determining cluster membership, derived from use of extreme gradient boosted regression trees (xgboost) Appendix.docx Cite Share Download PDF Status: Posted Version 1 posted 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-9480999","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":626861409,"identity":"b7ed10d0-023c-46e2-a500-ada88b5b9533","order_by":0,"name":"Timothy Tuti","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA60lEQVRIiWNgGAWjYLCChIoEIMnYACQsZAiq5mEDKk04A9ciwUOcFsa2BBifCC328j3mDx7OS5PTnX24gZmnQoKHv7078QFDxT27Bpy28Bg2JG7LMTY7lwjUckaCR+LM2c0GDGeKkwloqUjcdoax/XdumwSPgUTuNgmgU5Nx+wWkZU5FPVBLA3PuP7CW7T8Ia2nISTADa2mA2MIA1GKHU8uxtMIZCcfSDMG2/DkG8YsEMNgTcGlhbz684eOPmmR5szPsDxhn1NjI8bf3bvzwoSLBHpcWHABoRWIDiXqAMUayjlEwCkbBKBiuAADDjFHpUncf6QAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0002-7915-3004","institution":"KEMRI-Wellcome Trust Research Programme","correspondingAuthor":true,"prefix":"","firstName":"Timothy","middleName":"","lastName":"Tuti","suffix":""},{"id":626862811,"identity":"ed17dd78-5da1-4cd2-802e-7cb576f22efe","order_by":1,"name":"Tabitha Muema","email":"","orcid":"","institution":"KEMRI-Wellcome Trust Research Programme","correspondingAuthor":false,"prefix":"","firstName":"Tabitha","middleName":"","lastName":"Muema","suffix":""},{"id":626863449,"identity":"d1c2b988-c2a9-4986-9d03-3d6126bee02b","order_by":2,"name":"Mike English","email":"","orcid":"","institution":"Centre for Tropical Medicine \u0026 Global Health (CTM\u0026GH), Nuffield Department of Medicine, Oxford University","correspondingAuthor":false,"prefix":"","firstName":"Mike","middleName":"","lastName":"English","suffix":""},{"id":626863450,"identity":"69982b10-8a69-45c6-89cf-1ccb44cb37cf","order_by":3,"name":"Jalemba Aluvaala","email":"","orcid":"","institution":"4Department of Paediatrics and Child Health, University of Nairobi","correspondingAuthor":false,"prefix":"","firstName":"Jalemba","middleName":"","lastName":"Aluvaala","suffix":""}],"badges":[],"createdAt":"2026-04-21 08:22:39","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9480999/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9480999/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107500829,"identity":"5d3588a5-f938-43d6-b4f2-a701f2101307","added_by":"auto","created_at":"2026-04-22 05:50:58","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":108941,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePatients included in analyses grouped by development and external validation dataset\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-9480999/v1/7936c22bd3523346bb6ada5d.png"},{"id":107705596,"identity":"1718610b-a930-4676-9942-c5f5b401ed0d","added_by":"auto","created_at":"2026-04-24 09:13:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":547703,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePatterns of signs and symptoms across the different clusters from table 1.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-9480999/v1/264b711ddbf9613226dd76e5.png"},{"id":107705745,"identity":"903dfbdd-4b2c-4e9e-b24f-dc1818537b51","added_by":"auto","created_at":"2026-04-24 09:14:58","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":442823,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of clustering approach's mortality risk predictive accuracy.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-9480999/v1/e2b1a1355797036952b2f5bf.png"},{"id":108803474,"identity":"47d20cfd-7ee4-4d78-97a2-67fe8b882043","added_by":"auto","created_at":"2026-05-08 14:55:50","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1716324,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9480999/v1/26fb2a8f-d973-4abe-8a12-43ea21012738.pdf"},{"id":107705654,"identity":"34a84acf-1698-475c-9095-2285dbd7583d","added_by":"auto","created_at":"2026-04-24 09:14:11","extension":"tiff","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":760240,"visible":true,"origin":"","legend":"\u003cp\u003epSBI signs and symptoms present in the different datasets\u003c/p\u003e","description":"","filename":"SupplementaryFigure1.tiff","url":"https://assets-eu.researchsquare.com/files/rs-9480999/v1/2e5c2c22cc0d5bdcd7e094b2.tiff"},{"id":107705903,"identity":"9dd9a7ab-738e-419e-8fc0-e87706b5cc94","added_by":"auto","created_at":"2026-04-24 09:15:41","extension":"tiff","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":631834,"visible":true,"origin":"","legend":"\u003cp\u003ePatterns of pSBI signs and symptoms missingness in the different datasets\u003c/p\u003e","description":"","filename":"SupplementaryFigure2.tiff","url":"https://assets-eu.researchsquare.com/files/rs-9480999/v1/54f1eef1e6204b3b2458995b.tiff"},{"id":107500834,"identity":"fc8d05e1-7aaa-4401-93b9-a3588deeacb3","added_by":"auto","created_at":"2026-04-22 05:50:58","extension":"tiff","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":147986,"visible":true,"origin":"","legend":"\u003cp\u003eCommon admission diagnoses for neonates managed for sepsis (n=8012), excluding sepsis diagnosis\u003c/p\u003e","description":"","filename":"SupplementaryFigure3.tiff","url":"https://assets-eu.researchsquare.com/files/rs-9480999/v1/db59ccb5854de3704ac1fae5.tiff"},{"id":107705983,"identity":"6139576b-10fe-4205-a4d1-8d4ad5a88d7a","added_by":"auto","created_at":"2026-04-24 09:17:00","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":685808,"visible":true,"origin":"","legend":"\u003cp\u003eFeature importance for determining cluster membership, derived from use of extreme gradient boosted regression trees (xgboost)\u003c/p\u003e","description":"","filename":"SupplementaryFigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-9480999/v1/f7a5f5e20953b38a78ebc383.png"},{"id":107500836,"identity":"728434c8-baf5-4a9f-8db8-4fabe6d3fc08","added_by":"auto","created_at":"2026-04-22 05:50:58","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":22616,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-9480999/v1/d6c466fcf4c55c09d583de5f.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eA cluster analysis of neonates using clinical signs of possible serious bacterial infection at hospital admission in Kenya: A retrospective multicentre cohort study\u003c/p\u003e","fulltext":[{"header":"Author summary (Lay summary) ","content":"\u003cp\u003e\u003cstrong\u003eWhy was this study done?\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eNeonatal sepsis, a major cause of mortality in children aged \u0026lt;28 days, is a heterogenous syndrome with variation in clinical presentation and mortality outcomes.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eDefining distinct phenotypes from a limited set of clinical signs easy to use by clinicians (usually busy junior clinicians) may facilitate better emergency triaging, more targeted treatment and improve patient outcomes.\u003c/li\u003e\n \u003cli\u003eThis research sought to address the sepsis variation by grouping neonates based on clinical signs readily available on admission. The focus on non-microbiology variables is due to limited microbiology results in typical SSA public hospital settings at admission.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eWhat did the researchers do and find?\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eWe applied a model-based clustering approach that uses probabilistic framework that accounts for uncertainty to assign each patient a probability of belonging to each cluster given the presence/absence of clinical signs at admission.\u003c/li\u003e\n \u003cli\u003eWe identified five clinically relevant clusters that effectively stratified neonatal sepsis into sub-populations with differing risks of mortality.\u0026nbsp;\u003c/li\u003e\n \u003cli\u003eThese clusters demonstrated superior discrimination and calibration in predicting in-hospital mortality when compared to both the WHO expert-based and typical distance-based clustering approaches.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003eWhat do these findings mean?\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eThe five clusters capture the heterogeneity in neonatal sepsis at admission and reliably stratify sepsis into groups predictive of in-hospital mortality at admission\u003c/li\u003e\n \u003cli\u003eThis work’s findings lay the foundation for future work exploring the association of cluster membership with microbiological results and use of emulated trials to inform antibiotic use strategies.\u003c/li\u003e\n \u003cli\u003eThis work’s findings also lay the foundation for future research into the utility and usability of the identified clusters in risk stratification at hospital admission in typical SSA hospital settings.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"Background","content":"\u003cp\u003eThe most vulnerable time for a child’s survival are the first 28 days of life – the neonatal period, with 2.3\u0026nbsp;million children dying during this period globally, with a child from Sub-Saharan Africa (SSA) being 10 times more likely to die than a child from a high-income country [\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e]. Sepsis, which is a dysregulated immune response to infection that leads to acute organ dysfunction, is a leading contributor to burden of disease in neonates in SSA, both as primary cause of death and as a frequent contributor [\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e], and carries a high risk of death even when care is provided promptly [\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn newborns, sepsis occurs in two forms: early-onset neonatal sepsis, which occurs within the first 72 hours of life, often linked to complications during birth, and late-onset neonatal sepsis, which develops between after 72 hours of life [\u003cspan class=\"CitationRef\"\u003e5\u003c/span\u003e]. These infections are more common in premature or low-birth-weight babies, and their rates are higher in regions with limited or poor quality healthcare [\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e]. One-sixth of neonates treated for sepsis face significant challenges including functional limitation, cognitive impairment, and mental health disorders after recovery with 40% of sepsis patients requiring rehospitalisation within 90 days of discharge [\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eMany of these neonatal sepsis deaths in SSA are preventable with early detection and proper treatment. To achieve this healthcare providers rely on a high index of suspicion of sepsis (in the absence of laboratory confirmation) based on signs of possible Serious Bacterial Infection (pSBI) like poor feeding, lethargy, temperature instability, or respiratory distress [\u003cspan class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e8\u003c/span\u003e]. While laboratory tests such as blood cultures to identify pathogens, complete blood counts to assess the immune response, and urine cultures for accurate sepsis diagnosis are crucial for effective treatment [\u003cspan class=\"CitationRef\"\u003e9\u003c/span\u003e], in many health facilities in SSA, these diagnostic tests are unavailable at the time of admission [\u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThere is no consensus on the clinical definition of neonatal sepsis where microbiological confirmations are unavailable [\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e]. It is a heterogeneous syndrome representing a diverse group of patients, ranging from neonates with minor infections that will resolve quickly to those with severe, life-threatening sepsis. Using broad-spectrum, empiric antibiotic treatment for all suspected cases of neonatal sepsis may not be ideal, as it can lead to unnecessary prolonged antibiotic use in non-infected infants, potentially causing harm [\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e]. Tailoring treatment based on specific diagnoses and risk of mortality could improve outcomes and reduce overuse of antibiotics [\u003cspan class=\"CitationRef\"\u003e13\u003c/span\u003e] .\u003c/p\u003e \u003cp\u003eThe challenge lies in developing alternative effective strategies for early diagnosis and treatment of neonatal sepsis in environments such as SSA where microbiological investigation or laboratory testing is either unavailable or inaccessible [\u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e]. Different combinations of pSBI signs may naturally cluster into previously undescribed subsets or phenotypes that may have different risks for the outcome and may respond differently to treatments. Identification of distinct clinical phenotypes may allow more precise therapy and improve care [\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e]. However, many guidelines and hospital protocols continue to recommend a one-size-fits-all approach; recommending the same initial inpatient treatment and follow up [\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eEfforts to address the variability in neonatal sepsis have included the WHO’s triaging guidelines for pSBI [\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e14\u003c/span\u003e] as illustrated in supplementary table 1. Other approaches have included use of (1) cluster analysis of neonates based on clinical and laboratory data from electronic health records (EHRs) collected within the first six hours of the patients’ hospital stay, combined with serum biomarkers[\u003cspan class=\"CitationRef\"\u003e15\u003c/span\u003e], (2) latent class analysis (LCA) to identify molecular phenotypes in sepsis patients using both clinical and protein biomarker data [\u003cspan class=\"CitationRef\"\u003e16\u003c/span\u003e], and even used self-organizing maps to identify sepsis groups susceptible to multiple organ dysfunction syndrome based on their age and sequential organ failure assessment score [\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e]. Most of these approaches rely on clinical variables that are typically not collected for neonates in SSA at admission, given the limited laboratory capacity and lack of point-of-care diagnostics at scale, making findings from inclusion of such variables and subsequent clusters, largely ungeneralisable to many SSA hospital settings [\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eUsing routinely collected data that is available in typical SSA hospital settings, the aim of this study is to identify clusters of neonates at time of admission that can be candidates for appropriate interventions to maximise patient outcomes given the resource constraints. More specifically, the objectives of this study were:\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eTo use unsupervised model-based cluster analysis to group and assign neonates into clinically meaningful clusters based on key clinical features (i.e. pSBI) from data routinely collected at admission in many SSA health facilities.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eTo explores the external validity of the model-based clustering approach to different patient populations in similar clinical contexts.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eTo assesses the identified clusters ability to predict in-hospital mortality and compares this with the “Gold standard”, the World Health Organization (WHO) classification of neonatal sepsis severity.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e"},{"header":"Methods","content":"\u003cp\u003eEthics and reporting\u003c/p\u003e\n\u003cp\u003eThe reporting of this study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines, which provide a set of recommendations for transparent and comprehensive reporting of observational studies using cohort, case-control, or cross-sectional designs [\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e]. The reporting also follows the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines, which is a set of recommendations for the reporting of studies developing, validating, or updating prediction models for prognostic purposes [\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eThe Scientific and Ethics Review Unit of the Kenya Medical Research Institute (KEMRI) approved the collection of the de-identified data that provides the basis for this study as part of the Clinical Information Network (CIN) and is run in partnership with the Ministry of Health and the participating hospitals. We elaborate on the CIN in the study design and settings section. Individual consent for retrospective access to the de-identified anonymised patient records was not required and was waived by KEMRI with the authors having no access to the information that could identify the patients.\u003c/p\u003e\n\u003cp\u003eStudy design and participants\u003c/p\u003e\n\u003cp\u003eThis is a retrospective cohort analysis utilizing data from the Clinical Information Network (CIN). The CIN collects standardized routine admission and discharge data from newborn units (NBU) across 21 public county hospitals in 14 out of the 47 counties in Kenya with a detailed description of the methods of data collection and management is provided elsewhere [\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e]. In brief, neonatal admission data are recorded using paper forms such as Neonatal Admission Records (NAR), treatment sheets, continuous monitoring charts, supplementary forms etc. These documents provide a structured checklist for typically junior clinicians [\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e], covering nine essential domains: demographics, admission details, maternal history, presenting complaints, cardinal signs, physical examinations, nursing monitoring, discharge status and supportive care. Each hospital has a clerk who then extracts data from these typical hospital forms into a Research Electronic Data Capture (REDCap) database immediately after patient death or discharge [\u003cspan class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eThe study participants included inborn neonates aged 0\u0026ndash;28 days admitted into newborn units between 01st January 2022 and 31st December 2024 with at least one pSBI sign/symptom at admission, regardless of whether they had a neonatal sepsis admission diagnosis, or were prescribed first-line intravenous antibiotics (e.g. Crystalline Penicillin and Gentamicin) at admission [\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e]. Data from 8 hospitals (n\u0026thinsp;=\u0026thinsp;33094) was used as model derivation dataset, with the external geographical validation dataset coming from 13 hospitals (n\u0026thinsp;=\u0026thinsp;23704); Hospitals in the derivation dataset represent large hospitals with \u0026ge;\u0026thinsp;1000 NBU admissions per year, with the validation dataset having hospitals with \u0026lt;\u0026thinsp;1000 NBU admissions per year. Exclusion of out-born neonates was to rule out participants with community-acquired pSBI and/or those that might had outpatient antibiotic treatments [\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eOutcome\u003c/p\u003e\n\u003cp\u003eThe primary objective of this study is to identify clinically meaningful sub-populations of neonates with pSBI signs at admission based on key clinical features, derived using unsupervised clustering techniques fitted on the development dataset (n\u0026thinsp;=\u0026thinsp;33094 admissions from 8 hospitals). The selected pSBI clinical variables included signs of local infection, fever, hypothermia, difficulty feeding, difficulty breathing, convulsions, apnoea, floppy, indrawing, grunting, crackles, jaundice, slow capillary refill, central cyanosis, irritable, hypoxic, and bulging fontanelle [\u003cspan class=\"CitationRef\"\u003e14\u003c/span\u003e]. The second objective\u0026rsquo;s outcome was the predicted cluster membership in the external dataset (n\u0026thinsp;=\u0026thinsp;23704 admissions from 13 hospitals) from a parsimonious classification model compared to the derived cluster membership from using the original model-based clustering approach from objective 1. For the third objective, the outcome of interest was mortality at discharge.\u003c/p\u003e\n\u003cp\u003ePredictors\u003c/p\u003e\n\u003cp\u003eThis study analysed two sets of predictors corresponding to the study objectives. For the first and second objective, clinical variables listed in the section above which are commonly associated with sepsis/pSBI [\u003cspan class=\"CitationRef\"\u003e14\u003c/span\u003e] were used to identify and characterise distinct clusters of neonates. Additionally, objective two adopted feature selection based on feature importance to identify a parsimonious set of pSBI predictors for cluster membership classification [\u003cspan class=\"CitationRef\"\u003e24\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eThe identified clusters from objective one are the primary predictor variables for objective three, when assessing their association with mortality outcome; To ensure robust estimation, adjustments were made for potential confounders, neonate\u0026rsquo;s age and birth weight, given their well-established impact on length of stay and mortality [\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eMissing data\u003c/p\u003e\n\u003cp\u003ePrevious studies using CIN data showed that missing at random (MAR) is a reasonable assumption for our dataset [\u003cspan class=\"CitationRef\"\u003e28\u003c/span\u003e]; For objective one, the model-based clustering approach adopts a Full Information Maximum Likelihood (FIML) approach where the cluster memberships are computed using all available information while taking the missingness into account under the MAR assumption, therefore no imputation was required [\u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eFor objective three, with the MAR assumption, multiple imputation using the chained equation approach was separately applied for both the development and validation datasets [\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e] in predicting in-hospital mortality outcome.\u003c/p\u003e\n\u003cp\u003eSample size\u003c/p\u003e\n\u003cp\u003eFor the first objective, given clustering is an unsupervised statistical learning technique that does not rely on predefined sample sizes but rather extracted patterns from the available data [\u003cspan class=\"CitationRef\"\u003e31\u003c/span\u003e], coupled with a large development dataset (n\u0026thinsp;=\u0026thinsp;33094 admissions), a formal sample size calculation was deemed unnecessary. All eligible neonatal records were thus included to maximize the robustness of the clustering approach.\u003c/p\u003e\n\u003cp\u003eFor the second objective, we based our sample size calculation on previously reported mortality outcome prevalence of between 9% \u0026minus;\u0026thinsp;14% with R-squared values of 0.453 derived from previously-developed mortality risk prediction models in similar neonatal patient populations [\u003cspan class=\"CitationRef\"\u003e32\u003c/span\u003e]. Using the pmsampsize library in R [\u003cspan class=\"CitationRef\"\u003e33\u003c/span\u003e] and assuming a maximum of six clinically-meaningfully clusters identified [\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e35\u003c/span\u003e] after including age (in days) and birth weight, the required sample size for the mortality risk model development and validation would be 260 patients with 24 deaths; making both our development dataset with 33094 admissions (with 9768 deaths) and validation dataset with 23704 admissions (with 4180 deaths) satisfactory for clinical prediction model development and external validation.\u003c/p\u003e\n\u003ch3\u003eStatistical analysis methods\u003c/h3\u003e\n\u003ch3\u003eClustering approaches for neonates with pSBI managed for sepsis\u003c/h3\u003e\n\u003cp\u003eLatent Class Analysis (LCA) which is a Finite Mixture Modelling approach was used since it offered a model-based (i.e. probabilistic) approach to derive clusters [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]; LCA adopts Full Information Maximum Likelihood (FIML) approach making use of observations with missing data in the cluster estimation process [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. LCA uses the distribution of the dataset to assess probabilities that certain patients are members of certain clusters. Best model-fit was determined by evaluating different metrics from LCA approach summarised in supplementary table two; In summary the most probable number of clusters is determined by clear delineation (i.e. entropy thresholds\u0026thinsp;\u0026gt;\u0026thinsp;0.8), the smallest cluster size having considerable number of admissions, with the cluster model able to accurately predicting class membership for individual patients (i.e. Average Latent Class Posterior Probability (ALCPP) thresholds\u0026thinsp;\u0026gt;\u0026thinsp;0.85) (Supplementary table two). The final LCA model used to generate the clusters in the development dataset was also used to also predict the cluster membership of the validation dataset.\u003c/p\u003e\n\u003cp\u003eAs part of sensitivity analysis, we compared the model-based clustering approach to a distance-based unsupervised machine learning technique of agglomerative hierarchical cluster analysis (HCA) that groups patients with similar pSBI into clusters without imposing a specific sequential order on them [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. Agglomerative HCA was selected for sensitivity analysis due to its efficiency in handling mixed data types and its ability to generate clinically meaningful clusters that are stable and easy to interpret [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. Given our dataset consists of both continuous and categorical variables, standard Euclidean distance measures suited for numerical data, were unsuitable [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]; Gower\u0026rsquo;s distance which is well suited for mixed data types was used instead [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e], together with Ward\u0026rsquo;s minimum variance method as the linkage criterion to group neonates into meaningful clusters by minimising the total within-cluster variance [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. The optimal number of clusters from HCA approach was determined using the silhouette score [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. The Silhouette Score evaluated how well individual data points fit within their assigned clusters relative to the nearest alternative cluster, and ranges from \u0026minus;\u0026thinsp;1 to 1 with higher scores indicating better clustering [\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e].\u003c/p\u003e\n\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003eParsimonious classification model for cluster-membership\u003c/h2\u003e\n \u003cp\u003eFor objective two, we used eXtreme Gradient Boosting (xgboost), a machine learning technique with implicit features selection that is robust to missing data patterns, to identify a subset of pSBI features based on feature importance to predict cluster membership [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. Feature selection was based on SHAP (SHapley Additive exPlanations) values where the SHAP values derived from xgboost model which explain how the variables contribute to the clinical model\u0026apos;s mortality prediction, showing if the variable pushed the prediction higher or lower than the average [\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e].\u003c/p\u003e\n \u003cp\u003eThe parsimonious multivariate logistic regression model was developed from the development dataset (n\u0026thinsp;=\u0026thinsp;33094 admissions from 8 hospitals) and externally validated on the validation dataset (n\u0026thinsp;=\u0026thinsp;23704 admissions from 13 hospitals). For the process of model validation, the predicted class-membership from the parsimonious model (which was determined based on which cluster the patient had the highest probability of being a member of) was compared to class-membership from the original model-based clustering model. Typical classification metrics of accuracy, sensitivity, specificity, positive predictive value (PPV) etc. were used to report performance of the parsimonious classification model to correctly cluster patients into the model-based clusters.\u003c/p\u003e\n\u003c/div\u003e\n\u003ch3\u003eLogistic regression for in-hospital mortality\u003c/h3\u003e\n\u003cp\u003eFor objective three, Logistic regression without variable selection was used together with the imputed datasets with parameter estimates being combined using Rubin\u0026rsquo;s rule. The number of imputation datasets to use was determined from the integer value of the percentage of patients in the derivation dataset that had one or more missing values, rounded upwards. To examine heterogeneity in model performance while incorporating geographical external validation, we compared the logistic regression models internal-external cross-validation performance where we omitted one hospital at a time using it as the validation dataset, built the model on the remaining hospitals, and evaluated model\u0026rsquo;s discrimination and calibration performance on the hospital left out. We repeated this process with each iteration using a different hospital as the validation data source [\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. The predictive performance of in-hospital mortality by the model-based and hierarchical clustering approaches was then compared to the WHO\u0026rsquo;s pSBI expert-based clustering approach (i.e. \u0026ldquo;Gold standard\u0026rdquo;)(Supplementary Table\u0026nbsp;1). Model performance was assessed using calibration and discrimination performance metrics detailed in supplementary table 3, with the confidence intervals for both c-statistic and calibration slope and intercept, calculated through bootstrapping (i.e., iterative sampling with replacement).\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eDescription of participants\u003c/p\u003e\n\u003cp\u003eBetween 01st January 2022 and 31st December 2024, 73140 neonates aged 0\u0026ndash;28 days were admitted to newborn units across the 21 CIN hospitals. For analysis, 59302/73140 (81.08%) of the admissions had at least one pSBI sign/symptom and were eligible for inclusion in the study (Supplementary Fig.\u0026nbsp;1) with patterns of pSBI signs/symptoms illustrated in Supplementary Fig.\u0026nbsp;2; Of the 59302 admissions with \u0026ge;\u0026thinsp;1 pSBI sign/symptom, 2504/59302 (4.22%) had signs of systematic missingness (i.e. \u0026ge;7/19 (\u0026ge;\u0026thinsp;35%) pSBI signs not documented; Within the CIN, this level of missingness is indicative of missing documentation due to patients with either severe symptoms, in emergency situations or paper documents removed from patient files; The patients with systematic missingness were excluded due to high likelihood of introducing bias if included in subsequent analysis due to violating the missing at random (MAR) assumption. The neonatal admissions in CIN during this period were divided geographically by hospitals into a development dataset from 8 CIN hospitals (n\u0026thinsp;=\u0026thinsp;33094/56798, 58.27%), and a validation set from the remaining 13 CIN hospitals (n\u0026thinsp;=\u0026thinsp;23704/56798, 41.73%). Excluding sepsis admission diagnoses, respiratory distress syndrome, birth asphyxia, and meconium aspiration were the most common admission diagnoses in the included patients (Supplementary Fig.\u0026nbsp;3).\u003c/p\u003e\n\u003cp\u003eTable \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the model-based clustering fit statistics in the development dataset. The identified clusters are of patients with pSBI managed for sepsis grouped to maximise similarity in pSBI characteristics. From the model fit statistics described in supplementary table 2, the maximum number of clusters with clear delineation (i.e. entropy thresholds\u0026thinsp;\u0026gt;\u0026thinsp;0.8), whose smallest cluster size had arguably considerable number of admissions, and whose cluster model was able to accurately predicting class membership for individual patients (i.e. Average Latent Class Posterior Probability (ALCPP) thresholds\u0026thinsp;\u0026gt;\u0026thinsp;0.85), is five (Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eModel fit indices from Latent Class Analysis Fit Indices1\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eClusters\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eParameters\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLL\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAIC\u003csup\u003e\u003cem\u003e3\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eBIC\u003csup\u003e\u003cem\u003e4\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eaBIC\u003csup\u003e\u003cem\u003e5\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eEntropy\u003csup\u003e\u003cem\u003e6\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSmallest Cluster Size\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eALCPP\u003csup\u003e\u003cem\u003e7\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLMR\u003c/p\u003e\n \u003cp\u003eP-value\u003csup\u003e\u003cem\u003e8\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-367766\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e735639\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e736084\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e735916\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.870\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10792 (32.61%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.957\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e3\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e79\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e-360415\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e720989\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e721653\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e721402\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e0.785\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e7162 (21.64%)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e0.897\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026lt;\u0026thinsp;0.001\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e4\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e105\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e-357058\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e714327\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e715210\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e714876\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e0.752\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e4551 (13.75%)\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e0.857\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026lt;\u0026thinsp;0.001\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e131\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-351102\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e702466\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e703568\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e703151\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.803\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3368 (10.18%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.883\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e157\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-350771\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e701857\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e703177\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e702678\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.799\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1754 (5.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.844\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"10\" align=\"left\"\u003e\n \u003cp\u003e\u003csup\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sup\u003e Clustering mixture model performance metrics from MPLUS 7.1 software\u003c/p\u003e\n \u003cp\u003e\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e Log-Likelihood\u003c/p\u003e\n \u003cp\u003e\u003csup\u003e\u003cem\u003e3\u003c/em\u003e\u003c/sup\u003e Akaike Information Criteria (AIC)\u003c/p\u003e\n \u003cp\u003e\u003csup\u003e\u003cem\u003e4\u003c/em\u003e\u003c/sup\u003e Bayesian Information Criteria (BIC)\u003c/p\u003e\n \u003cp\u003e\u003csup\u003e\u003cem\u003e5\u003c/em\u003e\u003c/sup\u003e Sample-Size adjusted Bayesian Information Criteria (aBIC)\u003c/p\u003e\n \u003cp\u003e\u003csup\u003e\u003cem\u003e6\u003c/em\u003e\u003c/sup\u003e How clear the class delineation is; How well the classes separate the population into distinct subgroups of patients\u003c/p\u003e\n \u003cp\u003e\u003csup\u003e\u003cem\u003e7\u003c/em\u003e\u003c/sup\u003e Average Latent Class Posterior Probability (ALCPP). The average probability of the class model accurately predicting class membership for individuals\u003c/p\u003e\n \u003cp\u003e\u003csup\u003e\u003cem\u003e8\u003c/em\u003e\u003c/sup\u003e Lo-Mendell-Rubin (LMR) p-value from likelihood ratio test for K-1 versus K classes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eFigure 2 illustrates the difference in the probability of pSBI signs and symptoms given cluster membership. The pSBI signs of floppy, tachypnoea, grunting, hypothermia, indrawing, hypoxia, difficulty in feeding and difficulty breathing appear to be useful in distinguishing different clusters (Fig. 2). There is substantive variation with in-hospital mortality based on cluster membership, with the minimal risk cluster having the least mortality cases (0.98%) and critical risk having the highest mortality cases (67.8%). The same pattern is present in the validation dataset, with 3.06% and 61.76% case fatality rate between the minimal risk cluster and critical risk cluster (Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). Patients in the critical risk cluster had low gestational age in weeks and low birth weight while patients in the substantial risk cluster had normal gestational age and birth weight but both clusters had at least 4 pSBI signs/symptoms and substantially high mortality rates of \u0026gt;\u0026thinsp;40% in both the development and validation datasets (Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). Admissions in the minimal and low mortality risk clusters had 2 (IQR: 1\u0026ndash;3) pSBI signs/symptoms with antibiotics prescribed in \u0026ge;\u0026thinsp;45% of the cases within each of these clusters in both the development and validation datasets.\u0026nbsp;\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eCharacteristics of patients included in model development and external validation\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003eIndicator\u003c/p\u003e\n \u003c/th\u003e\n \u003cth colspan=\"5\" align=\"left\"\u003e\n \u003cp\u003eDevelopment (n\u0026thinsp;=\u0026thinsp;33094)\u003csup\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003cth colspan=\"5\" align=\"left\"\u003e\n \u003cp\u003eValidation (n\u0026thinsp;=\u0026thinsp;23704)\u003csup\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMinimal Risk n\u0026thinsp;=\u0026thinsp;3368 (10.18%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLow Risk n\u0026thinsp;=\u0026thinsp;12907 (39%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eModerate Risk n\u0026thinsp;=\u0026thinsp;8879 (26.83%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSubstantial Risk n\u0026thinsp;=\u0026thinsp;4130 (12.48%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCritical Risk n\u0026thinsp;=\u0026thinsp;3810 (11.51%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMinimal Risk n\u0026thinsp;=\u0026thinsp;3763 (15.87%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLow Risk n\u0026thinsp;=\u0026thinsp;8333 (35.15%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eModerate Risk n\u0026thinsp;=\u0026thinsp;5841 (24.64%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSubstantial Risk n\u0026thinsp;=\u0026thinsp;2940 (12.4%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCritical Risk n\u0026thinsp;=\u0026thinsp;2827 (11.93%)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth colspan=\"11\" align=\"left\"\u003e\n \u003cp\u003eNeonatal Essential Signs \u0026amp; Symptoms: Binary\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eJaundice\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1510 (44.83%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e21 (0.16%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e94 (1.06%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99 (2.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e53 (1.39%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1630 (43.32%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e14 (0.17%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e149 (2.55%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e178 (6.05%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e55 (1.95%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCrackles\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e29 (0.86%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e469 (3.63%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e304 (3.42%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e993 (24.04%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e298 (7.82%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e51 (1.36%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e394 (4.73%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e154 (2.64%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e751 (25.54%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e199 (7.04%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBulging Fontanelle\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e57 (1.69%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e156 (1.21%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e101 (1.14%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e88 (2.13%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40 (1.05%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97 (2.58%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e149 (1.79%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e71 (1.22%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e102 (3.47%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e35 (1.24%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLocal Infection\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e141 (4.19%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95 (0.74%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e38 (0.43%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e70 (1.69%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (0.26%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e318 (8.45%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e75 (0.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e39 (0.67%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99 (3.37%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10 (0.35%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eApnoea\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6 (0.18%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e88 (0.68%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e190 (2.14%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e656 (15.88%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e620 (16.27%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e20 (0.53%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e135 (1.62%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e250 (4.28%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e654 (22.24%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e618 (21.86%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDifficulty Eating\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e675 (20.04%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2892 (22.41%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1982 (22.32%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2284 (55.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1667 (43.75%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e925 (24.58%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1577 (18.92%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1295 (22.17%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1837 (62.48%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1335 (47.22%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDifficulty Breathing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e75 (2.23%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5997 (46.46%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5193 (58.49%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3949 (95.62%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3095 (81.23%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e172 (4.57%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3341 (40.09%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2858 (48.93%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2762 (93.95%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2142 (75.77%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eConvulsions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e107 (3.18%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e991 (7.68%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e39 (0.44%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e739 (17.89%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e29 (0.76%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e261 (6.94%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e973 (11.68%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e45 (0.77%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e780 (26.53%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e33 (1.17%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFloppy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e656 (19.48%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5616 (43.51%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5101 (57.45%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3183 (77.07%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3175 (83.33%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e779 (20.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2937 (35.25%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2984 (51.09%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2141 (72.82%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2210 (78.17%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIndrawing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e52 (1.54%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e279 (2.16%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e842 (9.48%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1427 (34.55%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1192 (31.29%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e33 (0.88%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e118 (1.42%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e359 (6.15%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e748 (25.44%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e748 (26.46%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGrunting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15 (0.45%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e675 (5.23%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1722 (19.39%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2378 (57.58%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1514 (39.74%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67 (1.78%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e408 (4.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e715 (12.24%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1323 (45%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e797 (28.19%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSlow Capillary Refill\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e232 (6.89%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e812 (6.29%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e391 (4.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e407 (9.85%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e243 (6.38%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e298 (7.92%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e582 (6.98%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e410 (7.02%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e339 (11.53%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e287 (10.15%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCentral Cyanosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e22 (0.65%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e248 (1.92%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e310 (3.49%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e536 (12.98%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e444 (11.65%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e43 (1.14%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e216 (2.59%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e194 (3.32%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e360 (12.24%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e276 (9.76%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIrritability\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e592 (17.58%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e774 (6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e236 (2.66%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e562 (13.61%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e136 (3.57%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e852 (22.64%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e639 (7.67%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e254 (4.35%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e534 (18.16%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e104 (3.68%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTachypnoea\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e358 (10.63%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4957 (38.41%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3415 (38.46%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2077 (50.29%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1344 (35.28%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e569 (15.12%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3296 (39.55%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1968 (33.69%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1276 (43.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e846 (29.93%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHypoxic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e82 (2.43%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1883 (14.59%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1296 (14.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1421 (34.41%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1033 (27.11%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e149 (3.96%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1136 (13.63%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e874 (14.96%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e935 (31.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e675 (23.88%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFever (Examined)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1267 (37.62%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e406 (3.15%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e123 (1.39%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e188 (4.55%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e25 (0.66%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1605 (42.65%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e452 (5.42%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e215 (3.68%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e370 (12.59%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e45 (1.59%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHypothermia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e17 (0.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e437 (3.39%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e837 (9.43%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e474 (11.48%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e744 (19.53%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e31 (0.82%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e490 (5.88%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e787 (13.47%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e366 (12.45%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e803 (28.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAbnormal Heart Rate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e121 (3.59%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e418 (3.24%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e291 (3.28%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e383 (9.27%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e305 (8.01%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e208 (5.53%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e401 (4.81%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e309 (5.29%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e364 (12.38%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e289 (10.22%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1896 (56.29%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7814 (60.54%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4602 (51.83%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2580 (62.47%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2043 (53.62%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2132 (56.66%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4967 (59.61%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2890 (49.48%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1754 (59.66%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1455 (51.47%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"11\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eNeonatal Essential Signs \u0026amp; Symptoms: Continuous\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003epSBI signs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (IQR: 1\u0026ndash;2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (IQR: 1\u0026ndash;3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (IQR: 1\u0026ndash;3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (IQR: 4\u0026ndash;6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4 (IQR: 3\u0026ndash;5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (IQR: 1\u0026ndash;3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (IQR: 1\u0026ndash;3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 (IQR: 1\u0026ndash;3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (IQR: 4\u0026ndash;6)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4 (IQR: 3\u0026ndash;5)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGestational Age\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e38 (IQR: 38\u0026ndash;40)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e38 (IQR: 38\u0026ndash;40)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e34 (IQR: 32\u0026ndash;35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e38 (IQR: 38\u0026ndash;40)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e28 (IQR: 27\u0026ndash;30)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e38 (IQR: 38\u0026ndash;40)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e39 (IQR: 38\u0026ndash;40)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e33 (IQR: 32\u0026ndash;35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e38 (IQR: 38\u0026ndash;40)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e28 (IQR: 27\u0026ndash;30)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBirth Weight\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.1 (IQR: 2.8\u0026ndash;3.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.1 (IQR: 2.79\u0026ndash;3.435)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.86 (IQR: 1.6\u0026ndash;2.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.99 (IQR: 2.6\u0026ndash;3.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.17 (IQR: 0.94\u0026ndash;1.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.1 (IQR: 2.78\u0026ndash;3.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.1 (IQR: 2.785\u0026ndash;3.46)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.8 (IQR: 1.54\u0026ndash;2.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 (IQR: 2.6-3.3525)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.2 (IQR: 1-1.47)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 (IQR: 2\u0026ndash;4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (IQR: 1\u0026ndash;1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (IQR: 1\u0026ndash;1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (IQR: 1\u0026ndash;1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (IQR: 1\u0026ndash;1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 (IQR: 2\u0026ndash;5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (IQR: 1\u0026ndash;1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (IQR: 1\u0026ndash;1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (IQR: 1\u0026ndash;2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 (IQR: 1\u0026ndash;1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTemperature\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37.4 (IQR: 36.7\u0026ndash;38.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36.7 (IQR: 36.3\u0026ndash;36.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36.5 (IQR: 36.1\u0026ndash;36.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36.5 (IQR: 36.1\u0026ndash;36.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36.3 (IQR: 35.675\u0026ndash;36.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37.8 (IQR: 36.8\u0026ndash;38.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36.7 (IQR: 36.2\u0026ndash;37.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36.5 (IQR: 36-36.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36.7 (IQR: 36.1\u0026ndash;37.3)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e36.2 (IQR: 35.2\u0026ndash;36.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRespiratory Rate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e49 (IQR: 45\u0026ndash;56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e57 (IQR: 48\u0026ndash;66)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e57 (IQR: 49\u0026ndash;65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e62 (IQR: 50\u0026ndash;71)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e56 (IQR: 46\u0026ndash;65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e49 (IQR: 44\u0026ndash;57)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e56 (IQR: 48\u0026ndash;67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e55 (IQR: 46\u0026ndash;64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e60 (IQR: 48\u0026ndash;70)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e53 (IQR: 43\u0026ndash;63)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHeart Rate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e138 (IQR: 127\u0026ndash;146)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e138 (IQR: 127\u0026ndash;148)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e139 (IQR: 128\u0026ndash;149)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e140 (IQR: 125\u0026ndash;152)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e140 (IQR: 124\u0026ndash;152)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e139 (IQR: 126\u0026ndash;150)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e138 (IQR: 125\u0026ndash;149)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e141 (IQR: 127\u0026ndash;152)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e142 (IQR: 124\u0026ndash;156)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e140 (IQR: 124\u0026ndash;153)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOxygen Saturation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95 (IQR: 94\u0026ndash;97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95 (IQR: 92\u0026ndash;97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95 (IQR: 92\u0026ndash;97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92 (IQR: 84\u0026ndash;96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93 (IQR: 87\u0026ndash;96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95 (IQR: 93\u0026ndash;97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95 (IQR: 92\u0026ndash;98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95 (IQR: 92\u0026ndash;98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92 (IQR: 83\u0026ndash;96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e94 (IQR: 87\u0026ndash;97)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"11\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eNeonatal Essential Interventions\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOxygen Prescribed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e928 (27.55%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7035 (54.51%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5776 (65.05%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3018 (73.08%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2907 (76.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e398 (10.58%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3702 (44.43%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3033 (51.93%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2093 (71.19%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1978 (69.97%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFluids Prescribed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e103 (3.06%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5109 (39.58%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4159 (46.84%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1973 (47.77%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2038 (53.49%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e72 (1.91%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1298 (15.58%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1016 (17.39%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e398 (13.54%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e603 (21.33%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFeeds Prescribed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e627 (18.62%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1527 (11.83%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e817 (9.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e232 (5.62%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e178 (4.67%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e201 (5.34%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e312 (3.74%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e242 (4.14%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99 (3.37%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e64 (2.26%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAntibiotics Prescribed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2166 (64.31%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8641 (66.95%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6116 (68.88%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2827 (68.45%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2672 (70.13%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1789 (47.54%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3806 (45.67%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2362 (40.44%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1241 (42.21%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1066 (37.71%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePhenobarbital Prescribed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e105 (3.12%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1423 (11.03%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e166 (1.87%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e976 (23.63%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e51 (1.34%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e273 (7.25%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1355 (16.26%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e115 (1.97%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e787 (26.77%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e46 (1.63%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"11\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eNeonatal Outcomes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLength Of Stay\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (IQR: 3\u0026ndash;7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (IQR: 3\u0026ndash;7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e12 (IQR: 5\u0026ndash;23)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4 (IQR: 1\u0026ndash;8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4 (IQR: 1\u0026ndash;27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (IQR: 3\u0026ndash;7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (IQR: 3\u0026ndash;7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e12 (IQR: 5\u0026ndash;22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 (IQR: 1\u0026ndash;8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 (IQR: 1\u0026ndash;29)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDied\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e33 (0.98%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e369 (2.86%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e849 (9.56%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1754 (42.47%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2583 (67.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e115 (3.06%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e355 (4.26%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e524 (8.97%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1441 (49.01%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1745 (61.73%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"11\"\u003e\u003csup\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sup\u003e Continuous variable summaries given as median (Interquartile range). Discrete variable summaries given as proportions\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eExternal validation of a parsimonious model-based clustering approach\u003c/p\u003e\n\u003cp\u003eThe pSBI signs and symptoms used in model-based clustering were 19, excluding birth weight and gestational age and the mortality outcome. From feature selection using xgboost, using we subjectively judged 11 (52%) of pSBI signs and symptoms to be sufficiently capable of assigning the patients into the mortality risk clusters using a multivariate logistic model with the outcome as the clusters based on values (Table \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, Supplementary Fig. 4). From Table \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, the cluster membership of any neonate with a pSBI sign/symptom can be determined based on which cluster the patient had the highest probability of being a member of, with the cluster-specific probabilities computed from the sum of the logits based on the presence/absence of the sign or symptom. The parsimonious model performance in predicting class membership in the external validation dataset showed relatively high accuracy, specificity, sensitivity, negative- and positive predictive performance across all clusters (Table \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eParsimonious multivariate logistic model for determining neonatal cluster membership based on pSBI signs and symptoms1\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003eIndicator\u003c/p\u003e\n \u003c/th\u003e\n \u003cth colspan=\"3\" align=\"left\"\u003e\n \u003cp\u003eMinimal\u003c/p\u003e\n \u003c/th\u003e\n \u003cth colspan=\"3\" align=\"left\"\u003e\n \u003cp\u003eLow\u003c/p\u003e\n \u003c/th\u003e\n \u003cth colspan=\"3\" align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/th\u003e\n \u003cth colspan=\"3\" align=\"left\"\u003e\n \u003cp\u003eSubstantial\u003c/p\u003e\n \u003c/th\u003e\n \u003cth colspan=\"3\" align=\"left\"\u003e\n \u003cp\u003eCritical\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eBeta\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSE\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep-value\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eBeta\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSE\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep-value\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eBeta\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSE\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep-value\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eBeta\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSE\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep-value\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eBeta\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSE\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep-value\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIntercept\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4.936\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.146\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;6.060\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.181\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;2.202\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.092\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;7.620\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.165\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;22.123\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.783\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBirth Weight\u003csup\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.499\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.058\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.260\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.035\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;1.692\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.034\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.645\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.050\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;4.861\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.198\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGestational Age\u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.206\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.017\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.508\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.010\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.017\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.601\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.015\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;1.807\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.059\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEarly Onset Age\u003csup\u003e3\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;5.876\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.134\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6.592\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.183\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.531\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.094\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;1.861\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.131\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.262\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.370\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.479\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDifficulty Breathing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;4.215\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.166\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.015\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.040\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.698\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.045\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.033\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.177\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.578\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.117\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.804\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.145\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFloppy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;2.246\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.082\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.261\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.040\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.218\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.034\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.728\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.071\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.106\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.160\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGrunting\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;2.633\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.414\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;2.109\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.059\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.048\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.042\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.256\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.627\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.069\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.542\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.135\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHypoxic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;1.885\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.174\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.450\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.050\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.508\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.042\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.087\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.066\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.483\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.144\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDifficulty Feeding\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.314\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.085\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.675\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.044\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.572\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.037\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.295\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.059\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.502\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.130\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eTachypnoea\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;1.998\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.092\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.161\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.041\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.104\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.033\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.508\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.062\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.176\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.118\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.134\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIndrawing\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;1.732\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.287\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;2.145\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.086\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.854\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.053\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.941\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.076\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.956\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.156\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCrackles\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;2.264\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.323\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;1.254\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.078\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.571\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.074\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.901\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.084\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.410\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.233\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHypothermia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;2.817\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.374\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;1.031\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.079\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;0.649\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.054\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.421\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.098\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.404\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.152\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eApnoea\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;2.788\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.642\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;2.653\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.139\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026minus;1.624\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.090\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2.006\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.105\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3.641\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.248\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"16\"\u003e\u003cstrong\u003eNote\u003c/strong\u003e: \u003csup\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sup\u003e Birth weight centered at 2.5 kgs; \u003csup\u003e\u003cem\u003e2\u003c/em\u003e\u003c/sup\u003e Gestational age centered at 37 weeks; \u003csup\u003e3\u003c/sup\u003eNeonate\u0026rsquo;s age in \u0026le;\u0026thinsp;72 hours\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003ctable id=\"Tab4\" border=\"1\"\u003e\n \u003ccaption\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003ePerformance of the parsimonious multivariate logistic models in predicting cluster membership in the external validation dataset\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMetric\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMinimal Risk\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLow Risk\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eModerate Risk\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSubstantial Risk\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCritical Risk\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBalanced Accuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e83.16 (82.72 to 83.62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.87 (91.68 to 92.05)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93.94 (93.71 to 94.19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e85.21 (84.68 to 85.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95.36 (95 to 95.71)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eF1 Score\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e78.01 (77.36 to 78.68)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e88.22 (87.93 to 88.49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.1 (90.78 to 91.42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e78.72 (77.92 to 79.47)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93.14 (92.69 to 93.59)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNegative Predictive Value\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e94.14 (93.96 to 94.32)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97.94 (97.8 to 98.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.96 (96.81 to 97.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.32 (96.17 to 96.47)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98.9 (98.81 to 98.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePositive Predictive Value\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92.72 (92.2 to 93.26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e81.02 (80.58 to 81.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.65 (91.21 to 92.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e87.05 (86.15 to 87.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95.02 (94.5 to 95.59)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRecall\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67.33 (66.43 to 68.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.83 (96.61 to 97.04)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90.55 (90.08 to 91.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e71.86 (70.77 to 72.85)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.32 (90.62 to 92.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e67.33 (66.43 to 68.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e96.83 (96.61 to 97.04)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90.55 (90.08 to 91.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e71.86 (70.77 to 72.85)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.32 (90.62 to 92.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99 (98.93 to 99.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e86.91 (86.61 to 87.23)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e97.33 (97.19 to 97.47)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e98.57 (98.47 to 98.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99.39 (99.33 to 99.46)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\"\u003e\u003cstrong\u003eNote\u003c/strong\u003e: The 95% confidence interval was determined through computing the metrics in each of the 1000 iterations of sampling with replacement of 10% of the external validation dataset\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eMortality risk prediction performance based on clusters\u003c/p\u003e\n\u003cp\u003eFigure 3 compares the performance of the three clustering approaches in predicting mortality risk. Overall, 33004/33094 (99%) admissions of the development dataset and 23601/23704 (99%) of the validation dataset were eligible for this comparison; The omitted admissions were excluded since their clinical signs and symptoms did not align with any of the predefined WHO clusters (Supplementary Table\u0026nbsp;1), coupled with variable missingness, posing a challenge in assigning them to a WHO cluster.\u003c/p\u003e\n\u003cp\u003eThe c-statistic (discrimination) for the WHO expert-based clustering approach was 0.721 (95% CI: 0.715 to 0.727) with a calibration intercept of 0.018 (95% CI: -0.005 to 0.041) and calibration slope of 1.015 (95% CI: 0.986 to 1.043). The c-statistic (discrimination) for the distance-based clustering approach was 0.721 (95% CI: 0.714 to 0.727) while its calibration intercept was \u0026minus;\u0026thinsp;0.002 (95% CI: -0.025 to 0.021) and calibration slope was 0.987 (95% CI: 0.96 to 1.014). The WHO and distance-based clustering approach calibration curves both showed substantive variances across the mortality risk spectrum with over-estimation of mortality at the lowest predicted risks and under-estimation at moderate to high risks of mortality (Fig.\u0026nbsp;3).\u003c/p\u003e\n\u003cp\u003eThe model-based clustering approach had the highest c-statistic (discrimination) of 0.867 (95% CI: 0.863 to 0.871), with a calibration intercept of -0.004 (95% CI: -0.031 to 0.023) and a calibration slope of 0.996 (95% CI: 0.979 to 1.043). The model-based clustering approach\u0026rsquo;s calibration curve showed close and relatively better alignment between the predicted risks and observed outcomes across the probability range compared to the alternative clustering approaches (Fig. 3).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eSummary of findings\u003c/p\u003e \u003cp\u003eThis study used a pragmatic approach with a limited set of signs easy to use by clinicians (usually busy junior clinicians) to identify residual heterogeneity in the neonates admitted with signs indicative of sepsis, that can be used to inform to inform antibiotic use treatment decisions and strategies. From applying model-based clustering approaches on routinely collected neonatal data at hospital admission, we identified four clinically distinct clusters of neonates with pSBI managed for sepsis, labelled based on the mortality risk in each cluster: minimal risk, moderate risk, substantial risk and critical risk (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Fig.\u0026nbsp;2). From external validation, we further demonstrate that the model-based clusters outperform distance-based and WHO expert-based clustering approaches in predicting in-hospital mortality based on discrimination and calibration statistics (Fig.\u0026nbsp;3). Among the 19 pSBI signs/symptoms collected at admission and used in the initial model-based clustering, a subset of 11 pSBI signs and symptoms were sufficient to determine cluster membership of a patient and subsequently the associated mortality risk from pSBI at admission (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, Table\u0026nbsp;5).\u003c/p\u003e \u003cp\u003eHypothermia, indrawing, floppy, tachypnoea, grunting, hypoxia, difficulty in feeding and difficulty breathing were key in distinguishing the clusters. These signs were mostly prevalent in the substantial and critical risk groups with tachypnoea being more common in the substantial risk cluster while hypothermia and indrawing were more common in the critical risk cluster. The critical risk cluster also comprised of neonates with low gestation age and birthweight while those in substantial risk cluster had normal gestation age and birthweight yet still faced substantially high mortality rate. The moderate risk cluster on the hand, included neonates with low birth weight but demonstrated a relatively lower rate of mortality risk, highlighting the heterogeneity in mortality risk beyond birth-related factors.\u003c/p\u003e \u003cp\u003eComparison to other findings\u003c/p\u003e \u003cp\u003eThe findings of this study contribute to the growing evidence of how unsupervised machine learning methods might identify meaningful phenotypes in neonatal sepsis populations. For instance, Seymour et al. and Jang et al. [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] employed k-means clustering and identified up to four clusters. Sinha et al. used model-based clustering and derived two clusters, unlike our findings. These studies used patients with different characteristics and clinical variables whose availability at the time of admission varied considerably. Unlike this current study which used routinely collected clinical data from low-resource neonatal units in an SSA country and relied only on clinical signs and symptoms, some of the other prior studies were conducted in high-income settings and included a set of variables combining both clinical and laboratory parameters that are typically unavailable at neonatal admission such as white blood cell count, platelet count, C-reactive protein, premature neutrophil count and serum levels.\u003c/p\u003e \u003cp\u003eFurthermore, these studies neither directly assessed the external validity of the clustering approach to classify new patients into the identified clusters nor did they provide the probability of patients having a combination of signs and symptoms given their cluster membership. These methodological differences highlight how selection of variables and validation approaches influence findings across studies. The observed differences in mortality risk among the clusters in this study and how the risk differs based on the probability of pSBI in each cluster agree with findings from Puri et al [\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e] although our findings support five distinct neonatal sepsis clusters. Hypothermia, among the signs previously identified as strongly predictive of death, was particularly common in the critical risk cluster of our study which had the highest mortality risk.\u003c/p\u003e \u003cp\u003eImplications of findings\u003c/p\u003e \u003cp\u003eThe clusters identified through model-based unsupervised learning approach reliably stratify neonatal sepsis into patient sub-populations predictive of in-hospital mortality at admission, outperforming the WHO expert-based and distance-based clustering methods as exemplified by the reported discrimination and calibration statistics (Fig.\u0026nbsp;3). This work shows that there is residual heterogeneity in the neonatal sepsis in-hospital admissions linked to mortality as an outcome. The improved mortality risk stratification exemplified by our model-based clustering approach provides a foundation for more tailored clinical decision making and potential development of phenotype-specific treatment guidelines that does not neglect the typical multimorbid nature of neonatal sepsis admissions in routine SSA hospitals. Additionally, the identification of substantial and critical mortality risk phenotype with distinct clinical profiles could guide prioritisation of care or more aggressive interventions, potentially improving outcomes in these groups. With the global rise in Antimicrobial resistance (AMR) risk, the use of antibiotics in \u0026gt;\u0026thinsp;45% in the CIN admissions within minimal and low mortality risk clusters might be indicative of antibiotic overuse or misuse in these hospital settings, and suggest possibility of less aggressive treatments. This study identifies groups which are more granular than the WHO categories that may be amenable to different antibiotic use strategies, e.g. shorter intravenous antibiotic use or and perhaps oral antibiotics etc.\u003c/p\u003e \u003cp\u003eThese findings can either (a) be used to augment the clinical judgement of clinicians in risk stratification of neonatal sepsis patients at admission [\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e] and/or (b) inform the design and implementation of emulated trials [\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e] to evaluate different antibiotic treatment strategies for neonatal sepsis at admission informed by the identified clusters [\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e, \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eBy relying on clinical signs and symptoms readily available at admission, the model-based clusters have better chances of being incorporated into routine clinical workflows at the point of care across SSA, with the potential to enhance early tailored interventions and treatment to ensure critical care is efficiently administered where it is most needed.\u003c/p\u003e \u003cp\u003eStrengths and limitations of findings\u003c/p\u003e \u003cp\u003eThis study has several notable strengths. First, it is based on a large multi-centre cohort of neonatal admission (n\u0026thinsp;=\u0026thinsp;56798) across 21 hospitals over three years, ensuring geographical and time coverage thus increasing confidence in the applicability of the modelling approaches across diverse neonatal care contexts. The inclusion criteria ensured the study sample was reflective of the typical multi-morbidity of neonatal admissions, thereby strengthening the real-world relevance of the findings. From a methodological standpoint, this study advances neonatal sepsis phenotyping by moving beyond conventional distance-based clustering techniques to more superior probabilistic, multivariate model-based approaches, which unlike distance-based methods, do not use a forced-choice mechanism to assign patients to clusters.\u003c/p\u003e \u003cp\u003eThe absence of microbiological validation to confirm the presence of bacteria indicative of sepsis is a limitation to this study. Since the aim was to classify neonates based on clinical pSBI signs in recognition of limited laboratory capacity in LMICs, the clusters should hence be interpreted as clinical phenotypes rather than pathogen confirmed syndromes. Future work would involve validating and extending this study\u0026rsquo;s findings by integrating microbiological testing with clinical clustering to ascertain whether the data-driven phenotypic clusters truly reflect bacterial infections. Prospective studies could also evaluate whether these groups respond differently to interventions or management strategies, providing insights into their clinical utility and potential to inform targeted risk-based neonatal care.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this retrospective cohort study, we used pSBI signs that are readily available at neonatal admission across SSA for model-based clustering approach and identified four clinically distinct clusters differentiated by mortality risk: minimal, moderate, substantial and critical risk. From discrimination and calibration statistics, the model-based clustering approach outperformed clustering patients using WHO expert-based guidelines and typical distance-based clustering approaches. These clusters demonstrate the potential to complement clinical judgement in mortality risk stratification at admission and could inform future randomised clinical trials to test whether triage guided by these clusters could improve clinical outcomes such as mortality and length of hospital stay. Future work could also integrate microbiological validation to confirm whether clusters map onto pathogen-specific syndromes and investigate possible variation in treatment responses across the clusters. These efforts could help strengthen the case for more targeted and risk-based neonatal sepsis management, informed by data-driven clustering.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCIN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eClinical Information Network\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eEHRs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eElectronic Health Records\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHICs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHigh Income Countries\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHCWs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHealth Care Workers\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eKPA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eKenya Paediatric Association\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLCA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLatent Class Analysis\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\"\u003eMoH\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMinistry of Health\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNBU\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNewborn Unit\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003epSBI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003epossible Serious Bacterial Infection (pSBI)\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSERU\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eKEMRI\u0026rsquo;s Scientific and Ethics Review Unit\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSSA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSub-Saharan Africa\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":"\u003ch3\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eEthical approval was provided by the KEMRI Scientific and Ethical Review Committee (SERU 3459).\u0026nbsp;Individual consent for retrospective access to the de-identified anonymised patient records was not required and was waived by KEMRI with the authors having no access to the information that could identify the patients.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eConsent for publication\u0026nbsp;\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThis study is published with the permission of the Director of Kenya Medical Research Institute (KEMRI).\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eAvailability of Data and Materials\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThe datasets generated and/or analysed during the current study are not publicly available due to the primary data being owned by the hospitals and their counties with the Ministry of Health; The research staff do have permission to share the data without further written approval from both the KEMRI-Wellcome Trust Data Governance Committee and the Facility, County or Ministry of Health as appropriate to the data request.\u003c/p\u003e\n\u003cp\u003eRequests for access to primary data from quantitative research by people other than the investigators will be submitted to the KEMRI-Wellcome Trust Research Programme data governance committee as a first step through\u0026nbsp;[email protected]\u003c/a\u003e , who will advise on the need for additional ethical review by the KEMRI Research Ethics Committee.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThe authors have declared that no competing interests exist.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThis work was primarily supported by a Wellcome Trust Early Career Research Fellowship (#227562/Z/23/Z) awarded to TT and a Wellcome Trust Senior Fellowship (#097170) awarded to ME. Additional support was provided by a Wellcome Trust core grant awarded to the KEMRI-Wellcome Trust Research Programme (#092654). The funders had no role in the preparation of this report or the decision to submit for publication.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eAuthor contributions\u0026nbsp;\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eAuthorship eligibility guidelines for the final reports adhere to the Contributor Roles Taxonomy (CRediT) statement guidelines.\u003c/p\u003e\n\u003cp\u003eConceptualization: TT, JA; Supervision: TT, JA; Methodology: TT, JA; Formal Analysis: TT, TM; Investigation: TT, TM, JA, ME; Resources: TT, ME; Data Curation: TT, TM; Writing – original draft: TT, TM; Writing – review \u0026amp; editing: TT, TM, ME, JA; Funding acquisition: TT, ME;\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThe\u0026nbsp;\u003cstrong\u003eClinical Information Network (CIN) Group\u003c/strong\u003e: The CIN group hospital teams who are tagged to collaborate in the network’s development, data collection, data management, implementation of audit and feedback interventions and who will participate in this study include the following focal persons:\u0026nbsp;\u003c/p\u003e\n\u003col\u003e\n \u003cli\u003e\u003cstrong\u003ePaediatricians:\u003c/strong\u003e Juma Vitalis, Nyumbile Bonface, Roselyne Malangachi, Christine Manyasi, Catherine Mutinda, David Kibiwott Kimutai, Rukia Aden, Caren Emadau, Elizabeth Atieno Jowi,\u0026nbsp;Cecilia Muithya, Charles Nzioki, Supa Tunje, Penina Musyoka, Wagura Mwangi, Agnes Mithamo, Magdalene Kuria, Esther Njiru, Mwangi Ngina, Penina Mwangi, Rachel Inginia,\u0026nbsp;Melab Musabi, Emma Namulala, Grace Ochieng, Lydia\u0026nbsp;Thuranira, Felicitas Makokha, Josephine Ojigo, Beth Maina, Catherine Mutinda, Mary Waiyego, Bernadette Lusweti, Angeline Ithondeka, Julie Barasa, Meshack Liru, Elizabeth Kibaru, Alice Nkirote Nyaribari, Joyce Akuka, Joyce Wangari;\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eNurses:\u0026nbsp;\u003c/strong\u003eAmilia Ngoda, Aggrey Nzavaye Emenwa, Patricia Nafula Wesakania, \u0026nbsp;George Lipesa, Jane Mbungu, Marystella Mutenyo,\u0026nbsp;Joyce Mbogho, Joan Baswetty,\u0026nbsp;Ann Jambi, Josephine Aritho, Beatrice Njambi, Felisters Mucheke, Zainab Kioni, Jeniffer, Lucy Kinyua, Margaret Kethi, Alice Oguda, Salome Nashimiyu Situma, Nancy Gachaja, Loise N. Mwangi, Ruth Mwai, irginia Wangari Muruga, Nancy Mburu, Celestine Muteshi, Abigael Bwire, Salome\u0026nbsp;Okisa Muyale, Naomi Situma, Faith Mueni, Hellen Mwaura,\u0026nbsp;Rosemary Mututa,\u0026nbsp;Caroline Lavu, Joyce Oketch, Jane Hore Olum, Orina Nyakina, Faith Njeru, Rebecca Chelimo, Margaret Wanjiku Mwaura, Ann Wambugu, Epharus Njeri Mburu, Linda Awino Tindi, Jane Akumu, Ruth Otieno, Slessor Osok;\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eHealth Record Information Officers (HRIOs):\u0026nbsp;\u003c/strong\u003eSeline Kulubi, Susan Wanjala, Pauline Njeru, Rebbecca Mukami Mbogo, John Ollongo, Samuel Soita, Judith Mirenja, Mary Nguri, Margaret Waweru, Mary Akoth Oruko, Jeska Kuya, Caroline Muthuri, Esther Muthiani, Esther Mwangi, Joseph Nganga, Benjamin Tanui, Alfred Wanjau, Judith Onsongo, Peter Muigai, Arnest Namayi, Elizabeth Kosiom, Dorcas Cherop, Faith Marete, Johanness Simiyu, Collince Danga, Arthur Otieno Oyugi, Fredrick Keya Okoth.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe \u003cstrong\u003eClinical Information Network (CIN) Group\u003c/strong\u003e’smonitored email address is [email protected]\u003c/a\u003e and the list can change when new paediatrician(s), nurse(s) or HRIO leave or come into the hospital.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003eOpen access\u0026nbsp;\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThis is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eUnited Nations Inter-agency Group for Child Mortality Estimation (IGME) (2024) Levels and Trends in Child Mortality, Report 2023. UNICEF, New York\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBreiman RF et al (2021) Postmortem investigations and identification of multiple causes of child deaths: an analysis of findings from the Child Health and Mortality Prevention Surveillance (CHAMPS) network. PLoS Med 18(9):e1003814\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeale AC et al (2013) Neonatal severe bacterial infection impairment estimates in South Asia, sub-Saharan Africa, and Latin America for 2010. Pediatr Res 74(1):73\u0026ndash;85\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeymour CW et al (2019) Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis. JAMA, 321(20)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCelik IH et al (2022) Diagnosis of neonatal sepsis: the past, present and future. Pediatr Res 91(2):337\u0026ndash;350\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHornik CP et al (2012) Early and late onset sepsis in very-low-birth-weight infants from a large group of neonatal intensive care units. Early Hum Dev 88:69\u0026ndash;74\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOpiyo N, English M (2011) What clinical signs best identify severe illness in young infants aged 0\u0026ndash;59 days in developing countries? A systematic review. Arch Dis Child 96(11):1052\u0026ndash;1059\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOtu A et al (2020) How to close the maternal and neonatal sepsis gap in sub-Saharan Africa. BMJ Global Health, 5(4)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaturi A, Chandran S (2024) Neonatal Sepsis: Aetiology, Pathophysiology, Diagnostic Advances and Management Strategies, vol 18. Pediatrics, Clinical Medicine Insights, p 11795565241281337\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRussell NJ et al (2023) Patterns of antibiotic use, pathogens, and prediction of mortality in hospitalized neonates and young infants with sepsis: A global neonatal sepsis observational cohort study (NeoOBS). PLoS Med 20(6):e1004179\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcGovern M et al (2020) Challenges in developing a consensus definition of neonatal sepsis. Pediatr Res 88(1):14\u0026ndash;26\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMolloy EJ et al (2020) Neonatal sepsis: need for consensus definition, collaboration and core outcomes. Pediatr Res 88(1):2\u0026ndash;4\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRosa-Mangeret F et al (2024) Challenges and opportunities in neonatal sepsis management: Insights from a survey among clinicians in 25 Sub-Saharan African countries. BMJ Paediatrics Open 8(1):e002398\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization (2015) Guideline: Managing possible serious bacterial infection in young infants when referral is not feasible. Geneva, World Health Organisation\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJang JY et al (2022) Identification of the robust predictor for sepsis based on clustering analysis. Sci Rep 12(1):2336\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSinha P et al (2023) Identifying molecular phenotypes in sepsis: an analysis of two prospective observational cohorts and secondary analysis of two randomised controlled trials. Lancet Respiratory Med 11(11):965\u0026ndash;974\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKnox DB et al (2015) Phenotypic clusters within sepsis-associated multiple organ dysfunction syndrome. Intensive Care Med 41(5):814\u0026ndash;822\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evon Elm E et al (2008) The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 370(9596):1453\u0026ndash;1457\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCollins GS et al (2015) Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): The TRIPOD Statement. J Br Surg 102(3):148\u0026ndash;158\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTuti T et al (2016) Innovating to enhance clinical data management using non-commercial and open source solutions across a multi-center network supporting inpatient pediatric care and research in Kenya. J Am Med Inform Assoc 23(1):184\u0026ndash;192\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOgero M et al (2020) Examining which clinicians provide admission hospital care in a high mortality setting and their adherence to guidelines: an observational study in 13 hospitals. Arch Dis Child 105(7):648\u0026ndash;654\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarris PA et al (2009) Research electronic data capture (REDCap)\u0026mdash;A metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 42(2):377\u0026ndash;381\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAluvaala J et al (2025) Situational analysis of antibiotic prescriptions in Kenyan neonatal units for antimicrobial stewardship: a retrospective longitudinal study. EClinicalMedicine, 82\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu C et al (2022) Interpretable machine learning for early prediction of prognosis in sepsis: a discovery and validation study. Infect Dis therapy 11(3):1117\u0026ndash;1132\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJovičić M et al (2023) Predictors of Mortality in Early Neonatal Sepsis: A Single-Center Experience. Medicina 59(3):604\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDanny) Liang L et al (2018) Predictors of mortality in neonates and infants hospitalized with sepsis or serious infections in developing countries: A systematic review. Front Pead 6:277\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGan MY et al (2022) Contemporary Trends in Global Mortality of Sepsis Among Young Infants Less Than 90 Days: A Systematic Review and Meta-Analysis. Front Pead 10:890767\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGachau S et al (2020) Handling missing data in modelling quality of clinician-prescribed routine care: Sensitivity analysis of departure from missing at random assumption. Stat Methods Med Res 29(10):3076\u0026ndash;3092\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDepaoli S, Jia F, Visser M (2025) Addressing Missing Data in Latent Class Analysis When Using a Three-Step Estimation Approach. Struct Equation Modeling: Multidisciplinary J 32(2):287\u0026ndash;303\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Buuren S, Groothuis-Oudshoorn K (2011) mice: Multivariate Imputation by Chained Equations in R. ournal Stat Softw 45:1\u0026ndash;67\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRuppert D (2004) \u003cem\u003eThe Elements of Statistical Learning: Data Mining, Inference, and Prediction\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTuti T et al (2022) Improving in-patient neonatal data quality as a pre-requisite for monitoring and improving quality of care at scale: A multisite retrospective cohort study in Kenya. PLOS Global Public Health 2(10):e0000673\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiley RD et al (2020) Calculating the sample size required for developing a clinical prediction model. BMJ, 368\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarrell F (2024) \u003cem\u003eThe Burden of Demonstrating Statistical Validity of Clusters\u003c/em\u003e, in \u003cem\u003eStatistical Thinking\u003c/em\u003e. F. Harrell, Editor.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarrell F (2017) \u003cem\u003eStatistical Errors in the Medical Literature\u003c/em\u003e, in \u003cem\u003eStatistical Thinking\u003c/em\u003e. Statistical Thinking, F. Harrell, Editor\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlag\u0026ouml;z EC, Vermunt JK (2022) Stepwise Latent Class Analysis in the Presence of Missing Values on the Class Indicators. Struct Equation Modeling: Multidisciplinary J 29(5):784\u0026ndash;790\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang D, Tong X (2023) \u003cem\u003eA Comparison of Full Information Maximum Likelihood and Machine Learning Missing Data Analytical Methods in Growth Curve Modeling.\u003c/em\u003e The Annual Meeting of the Psychometric Society, : pp. 99\u0026ndash;109\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCabezas LM, Izbicki R, Stern RB (2023) Hierarchical clustering: Visualization, feature importance and model selection. Appl Soft Comput 141:110303\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLandau S, Leese M, Stahl D, Everitt B (2011) and \u003cem\u003eCluster Analysis.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoschidis O, Markos A, Chadjipadelis T (2023) Hierarchical clustering of mixed-type data based on barycentric coding. Behaviormetrika 50(1):465\u0026ndash;489\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCosta E, Papatsouma I, Markos A (2023) Benchmarking distance-based partitioning methods for mixed-type data. Adv Data Anal Classif 17(3):701\u0026ndash;724\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu P et al (2024) A modified and weighted Gower distance-based clustering analysis for mixed type data: a simulation and empirical analyses. BMC Med Res Methodol 24(1):305\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGower JC (1971) A General Coefficient of Similarity and Some of Its Properties. Biometrics 27(4):857\u0026ndash;871\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWard JH Jr (1963) Hierarchical Grouping to Optimize an Objective Function. J Am Stat Assoc 58(301):236\u0026ndash;244\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRousseeuw PJ (1987) Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. J Comput Appl Math 20:53\u0026ndash;65\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan K-C et al (2020) The development an artificial intelligence algorithm for early sepsis diagnosis in the intensive care unit. Int J Med Informatics 141:104176\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen T, Guestrin C (2016) \u003cem\u003eXgboost: A scalable tree boosting system\u003c/em\u003e. in \u003cem\u003eThe 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining\u003c/em\u003e. San Francisco California: Association for Computing Machinery\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLundberg S \u003cem\u003eAn introduction to explainable AI with Shapley values\u003c/em\u003e. Welcome to the SHAP documentation 2018 2018 [cited 2025 17 December]; Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://shap.readthedocs.io/en/latest/example_notebooks/overviews/An%20introduction%20to%20explainable%20AI%20with%20Shapley%20values.html\u003c/span\u003e\u003cspan address=\"https://shap.readthedocs.io/en/latest/example_notebooks/overviews/An%20introduction%20to%20explainable%20AI%20with%20Shapley%20values.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTakada T et al (2021) Internal-external cross-validation helped to evaluate the generalizability of prediction models in large clustered datasets. J Clin Epidemiol 137:83\u0026ndash;91\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePuri D et al (2021) Prevalence of clinical signs of possible serious bacterial infection and mortality associated with them from population-based surveillance of young infants from birth to 2 months of age. PLoS ONE 16(2):e0247457\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan Doorn WP et al (2021) A comparison of machine learning models versus clinical evaluation for mortality prediction in patients with sepsis. PLoS ONE 16(1):e0245157\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatthews AA et al (2022) Target trial emulation: applying principles of randomised trials to observational studies. BMJ, 378\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShimabukuro DW et al (2017) Effect of a machine learning-based severe sepsis prediction algorithm on patient survival and hospital length of stay: a randomised clinical trial. BMJ open respiratory Res, 4(1)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChiew CJ et al (2019) Heart rate variability based machine learning models for risk prediction of suspected sepsis patients in the emergency department. Medicine 98(6):e14197\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLo Y, Nancy R, Mendell, Rubin DB (2001) Testing the number of components in a normal mixture. Biometrika 88(3):767\u0026ndash;778\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNylund-Gibson K, Choi AY (2018) Ten frequently asked questions about latent class analysis. Translational issues Psychol Sci 4(4):440\u0026ndash;461\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCeleux G, Soromenho G (1996) An entropy criterion for assessing the number of clusters in a mixture model. 13:195\u0026ndash;212\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang MC et al (2017) Performance of the entropy as an index of classification accuracy in latent profile analysis: A Monte Carlo simulation study. Acta Physiol Sinica 49(11):1473\u0026ndash;1482\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShanahan L et al (2013) Sex-differentiated changes in C-reactive protein from ages 9 to 21: the contributions of BMI and physical/sexual maturation. Psychoneuroendocrinology 38(10):2209\u0026ndash;2217\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiley RD et al (2019) External validation of clinical prediction models using big datasets from e-health records or IPD meta-analysis: Opportunities and challenges. BMJ, 365\u003c/span\u003e\u003c/li\u003e \u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[{"identity":"b959fe90-01e9-4dc2-b88d-1cfc61aa7c72","identifier":"10.13039/100010269","name":"Wellcome Trust","awardNumber":"227562/Z/23/Z","order_by":0},{"identity":"4113ce1e-44e1-4570-8ca9-e52b2ba4a855","identifier":"10.13039/100010269","name":"Wellcome Trust","awardNumber":"097170","order_by":1},{"identity":"3c3fb17d-2e87-4a7f-b45e-fc2b5022e152","identifier":"10.13039/100010269","name":"Wellcome Trust","awardNumber":"092654","order_by":2}],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"KEMRI-Wellcome Trust Research Programme","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"neonates, newborn, sepsis, hospital, unsupervised learning, Sub-Sahara Africa","lastPublishedDoi":"10.21203/rs.3.rs-9480999/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9480999/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eNeonatal sepsis remains a major cause of mortality in Sub-Saharan Africa (SSA). Despite presenting with considerable clinical heterogeneity, suspected cases are managed uniformly with broad-spectrum antibiotics. Typical data-driven approaches developed in high-resource settings to identify clinically meaningful phenotypes and support management of neonatal sepsis are largely ungeneralisable to typical SSA public hospital settings, due to inclusion of variables that are largely unavailable at admission. This study\u0026rsquo;s objective was to identify sepsis clusters using signs of possible Serious Bacterial Infection (pSBI) readily available at the time of admission, and to assess the clusters performance in predicting mortality.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eWe conducted unsupervised model-based cluster analysis using Latent Class Analysis based on pSBI data collected at admission. All in-born neonates\u0026thinsp;\u0026lt;\u0026thinsp;28 days old admitted to 21 Kenyan hospitals between January 2022 and December 2024 with \u0026ge;\u0026thinsp;1 pSBI sign/symptom at admission were eligible for inclusion. We further explored the external validity of this clustering approach on new patient populations, and assessed the ability of the identified clusters to accurately predict in-hospital mortality compared to the World Health Organization neonatal sepsis severity classification guidelines.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eFive clusters of minimal, low, moderate, substantial and critical mortality risk were identified from development dataset with 33094 patients from eight hospitals. The models had an accuracy, positive predictive value and specificity of at least 83.16% (82.72% to 83.62%), 81.02% (80.58% to 81.45%) and 86.91% (86.61% to 87.23%) respectively in predicting cluster membership of 23704 patients in the external validation dataset admitted to thirteen different hospitals. From an internal-external cross-validation approach of the in-hospital mortality risk, the model-based clustering approach had discrimination (AUROC) of 0.867 (0.863 to 0.871) and calibration intercept and slope of -0.004 (-0.031 to 0.023) and 0.996 (0.979 to 1.014) respectively, outperforming the WHO sepsis severity classification whose discrimination was 0.721 (0.715 to 0.727) and calibration intercept and slope being 0.018 (-0.005 to 0.041) and 1.015 (0.986 to 1.043) respectively.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe identified clusters can complement clinicians\u0026rsquo; judgement in assessing risk among neonates with sepsis at admission. Future work evaluating the utility of these clusters and potential differences in treatment response across clusters are therefore recommended to help strengthen the case for more targeted, risk-based neonatal sepsis management.\u003c/p\u003e","manuscriptTitle":"A cluster analysis of neonates using clinical signs of possible serious bacterial infection at hospital admission in Kenya: A retrospective multicentre cohort study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-22 05:50:48","doi":"10.21203/rs.3.rs-9480999/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e2feac97-89dc-446f-8d78-89868f2c3c3e","owner":[],"postedDate":"April 22nd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-22T05:50:49+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-22 05:50:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9480999","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9480999","identity":"rs-9480999","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}