Charles R. Newton’s research while affiliated with KEMRI-Wellcome Trust Research Programme and other places

Background Attrition is a challenge in parameter estimation in both longitudinal and multi-stage cross-sectional studies. Here, we examine utility of machine learning to predict attrition and identify associated factors in a two-stage population-based epilepsy prevalence study in Nairobi. Methods All individuals in the Nairobi Urban Health and Demographic Surveillance System (NUHDSS) (Korogocho and Viwandani) were screened for epilepsy in two stages. Attrition was defined as probable epilepsy cases identified at stage-I but who did not attend stage-II (neurologist assessment). Categorical variables were one-hot encoded, class imbalance was addressed using synthetic minority over-sampling technique (SMOTE) and numeric variables were scaled and centered. The dataset was split into training and testing sets (7:3 ratio), and seven machine learning models, including the ensemble Super Learner, were trained. Hyperparameters were tuned using 10-fold cross-validation, and model performance evaluated using metrics like Area under the curve (AUC), accuracy, Brier score and F1 score over 500 bootstrap samples of the test data. Results Random forest (AUC = 0.98, accuracy = 0.95, Brier score = 0.06, and F1 = 0.94), extreme gradient boost (XGB) (AUC = 0.96, accuracy = 0.91, Brier score = 0.08, F1 = 0.90) and support vector machine (SVM) (AUC = 0.93, accuracy = 0.93, Brier score = 0.07, F1 = 0.92) were the best performing models (base learners). Ensemble Super Learner had similarly high performance. Important predictors of attrition included proximity to industrial areas, male gender, employment, education, smaller households, and a history of complex partial seizures. Conclusion These findings can aid researchers plan targeted mobilization for scheduled clinical appointments to improve follow-up rates. These findings will inform development of a web-based algorithm to predict attrition risk and aid in targeted follow-up efforts in similar studies.

Background: Brain infections pose substantial challenges in diagnosis and management and carry high mortality and morbidity, especially in low-income and middle-income countries. We aimed to improve the diagnosis and early management of patients admitted to hospital (adults aged 16 years and older and children aged >28 days) with suspected acute brain infections at 13 hospitals in Brazil, India, and Malawi. Methods: With hospital stakeholders, policy makers, and patient and public representatives, we co-designed a multifaceted clinical and laboratory intervention, informed by an evaluation of routine practice. The intervention, tailored for each setting, included a diagnostic and management algorithm, a lumbar puncture pack, a testing panel, and staff training. We used multivariable logistic regression and interrupted time series analysis to compare the coprimary outcomes-the percentage of patients achieving a syndromic diagnosis and the percentage achieving a microbiological diagnosis before and after the intervention. The study was registered at ClinicalTrials.gov (NCT04190303) and is complete. Findings: Between Jan 5, 2021, and Nov 30, 2022, we screened 10 462 patients and enrolled a total of 2233 patients at 13 hospital sites connected to the four study centres in Brazil, India, and Malawi. 1376 (62%) were recruited before the intervention and 857 (38%) were recruited after the intervention. 2154 patients (96%) had assessment of the primary outcome (1330 [62%] patients recruited pre-intervention and 824 [38%] recruited post-intervention). The median age across centres was 23 years (IQR 6-44), with 1276 (59%) being adults aged 16 years or older and 888 (41%) children aged between 29 days and 15 years; 1264 (59%) patients were male and 890 (41%) were female. Data on race and ethnicity were not recorded. 1020 (77%) of 1320 patients received a syndromic diagnosis before the intervention, rising to 701 (86%) of 813 after the intervention (adjusted odds ratio [aOR] 1·81 [95% CI 1·40-2·34]; p<0·0001). A microbiological diagnosis was made in 294 (22%) of 1330 patients pre-intervention, increasing to 250 (30%) of 824 patients post-intervention (aOR 1·46 [95% CI 1·18-1·79]; p=0·00040). Interrupted time series analysis confirmed that these increases exceeded a modest underlying trend of improvement over time. The percentage receiving a lumbar puncture, time to appropriate therapy, and functional outcome also improved. Interpretation: Diagnosis and management of patients with suspected acute brain infections improved following introduction of a simple intervention package across a diverse range of hospitals on three continents. The intervention is now being implemented in other settings as part of the WHO Meningitis Roadmap and encephalitis control initiatives. Funding: UK National Institute for Health and Care Research.

The objective of this study was to investigate the impact of common mental disorder (CMD; depression/anxiety) symptoms and risky substance use in people with epilepsy in Ethiopia (four districts) on quality of life (QoL) and functioning over 6 months. A prospective cohort study was carried out. Multivariable linear regression followed by structural equation modelling (SEM) was employed. In the multivariable regression model, neither CMD symptoms (β coef. = −0.37, 95% confidence interval [CI] −1.30, +0.55) nor moderate to high risk of alcohol use (β coef. = −0.70, 95% CI −9.20, +7.81) were significantly associated with a change in QoL. In SEM, the summative effect of CMD on QoL was significant (B = −0.27, 95% CI −0.48, −0.056). Change in functional disability was not significantly associated with common mental disorder (CMD) symptoms (β coef. = −0.03, 95% CI −0.48, +0.54) or with moderate to high risk of alcohol use (β coef. = −1.31, 95% CI −5.89, 3.26). In the SEM model, functional disability was predicted by both CMD symptoms (B = 0.24, 95% CI 0.06, 0.41) and seizure frequency (B = 0.67, 95% CI 0.46, 0.87). In this rural Ethiopian setting, co-morbid CMD symptoms and seizure frequency independently predicted functional disability in people with epilepsy.

Objectives Approximately 80% of people with epilepsy live in low- and middle-income countries (LMICs), where limited resources and stigma hinder accurate diagnosis and treatment. Clinical machine learning models have demonstrated substantial promise in supporting the diagnostic process in LMICs by aiding in preliminary screening and detection of possible epilepsy cases without relying on specialised or trained personnel. How well these models generalise to naïve regions is, however, underexplored. Here, we use a novel approach to assess the suitability and applicability of such clinical tools to aid screening and diagnosis of active convulsive epilepsy in settings beyond their original training contexts. Methods We sourced data from the Study of Epidemiology of Epilepsy in Demographic Sites dataset, which includes demographic information and clinical variables related to diagnosing epilepsy across five sub-Saharan African sites. For each site, we developed a region-specific (single-site) predictive model for epilepsy and assessed its performance at other sites. We then iteratively added sites to a multi-site model and evaluated model performance on the omitted regions. Model performances and parameters were then compared across every permutation of sites. We used a leave-one-site-out cross-validation analysis to assess the impact of incorporating individual site data in the model. Results Single-site clinical models performed well within their own regions, but generally worse when evaluated in other regions (p<0.05). Model weights and optimal thresholds varied markedly across sites. When the models were trained using data from an increasing number of sites, mean internal performance decreased while external performance improved. Conclusions Clinical models for epilepsy diagnosis in LMICs demonstrate characteristic traits of ML models, such as limited generalisability and a trade-off between internal and external performance. The relationship between predictors and model outcomes also varies across sites, suggesting the need to update specific model aspects with local data before broader implementation. Variations are likely to be particular to the cultural context of diagnosis. We recommend developing models adapted to the cultures and contexts of their intended deployment and caution against deploying region- and culture-naïve models without thorough prior evaluation.

Purpose of review Epilepsy disproportionately affects those in low- and middle-income countries (LMICs) where diagnostic and treatment gaps persist. We highlight key recent developments and showcase practical opportunities to improve epilepsy care in resource limited settings. Recent findings In LMICs, cultural, socioeconomic and infrastructural factors drive the epilepsy treatment gap. Robust implementation of the WHO Intersectoral Global Action Plan (WHO IGAP) and Mental Health Gap Action Program (mhGAP), for example, will reduce the epilepsy education gap. Engaging traditional healers and other key community leaders should lessen stigma. The Epilepsy Diagnostic Companion, a culture specific tool that helps identify convulsive seizures, can expedite epilepsy diagnosis at primary care level. Novel, robust 3-D printable EEG headsets prototypes that can be deployed in remote rural communities have been piloted with encouraging results. Levetiracetam has been added to the WHO Essential Medicines List (EML), paving way to safer, less teratogenic antiseizure medications (ASMs). Epilepsy surgery programs in carefully selected patients potentially offer cheap, effective and potentially curative treatments, including in LMICs. Summary Apps, EEG prototypes, better access to ASMs and implementation of WHO iGAP offer current, tangible opportunities to improve epilepsy care in LMICs. Bidirectional learning must be facilitated to also help hard to reach communities in high-income settings.

Clinical practice guidelines (CPGs) and consensus‐based recommendations (CBRs) require considerable effort, collaboration, and time—all within the constraints of finite resources. Professional societies, such as the International League Against Epilepsy (ILAE), must prioritize what topics and questions to address. Implementing evidence‐based care remains a crucial challenge in clinical practice. Using rigorous processes to ensure that the best available research evidence informs health care recommendations is of the utmost importance. We aimed to develop a structured and transparent process for prioritizing future CPGs and CBRs supported by the ILAE. A multidisciplinary group of researchers and experts from the ILAE Prioritization Task Force conducted a scoping review to identify prioritization approaches for CPG and CBR development. This scoping review was reported according to the Preferred Reporting Items for Systematic reviews and Meta‐Analyses extension for Scoping Reviews (PRISMA‐ScR) and Cochrane recommendations. A Problem/population, Concept, and Context (PCC) strategy was applied to the literature search and selection of the studies. We searched Medline/PubMed, Embase, Web of Science, and Scopus without time or language limits. The findings were synthesized qualitatively. A consensus‐based process was followed to develop a prioritization scoring tool for CPGs and another for CBRs. Thirty‐nine participants, including clinicians, experts in the field, methodologists, and other relevant stakeholders, contributed to developing the final instrument (based on a 5‐point Likert scale). Of 721 unique citations, 8 papers reporting prioritization approaches for guideline development were included. Based on these, we developed an initial tool with 10 criteria. It was iteratively optimized and revised by the ILAE Standards and Best Practice Council, which unanimously approved the instrument. The ILAE Executive Committee subsequently approved its final version. The ILAE Prioritization Tool is intended to standardize the prioritization processes and optimize the ILAE's use of resources to select CPGs and CBRs for endorsement.