Tropical Medicine and Infectious Disease

Hospital notification of dengue cases is frequently delayed, contributing to underreporting. At Thai provincial hospitals the bottleneck is administrative, not diagnostic: delayed discharge coding and clerical steps mean notifications are submitted late or omitted. Existing machine learning (ML) studies require clinician-annotated features absent in passive electronic medical records (EMRs). We propose a dual-pathway EMR framework for ML-based pre-population of surveillance reports. We analyzed 1031 dengue-suspected encounters (508 notifiable) from Surin Hospital, Thailand (2019–2022). Pathway A combines EMR data with manually documented clinical features; Pathway B uses only passive EMR data. Using an 80/20 stratified split, we tested 7 ML methods across 11 feature sets (77 combinations) with SHAP-guided feature elimination and 10-fold cross-validation. On the primary split, Pathway A (CatBoost) achieved an F1 score of 96.62%, 95.10% sensitivity, and 98.10% specificity (10-seed mean F1 95.7% ± 1.4%). Pathway B (LightGBM) achieved an F1 score of 84.42%, 94.12% sensitivity, 75.24% specificity, and a projected negative predictive value of 99.1% at 10% prevalence; 10-seed performance was lower (F1 78.5% ± 4.3%; sensitivity 89.1%). An exploratory LSTM-16 stacking model (F1 83.99%; precision 79.49%) performed best in precision and specificity among Pathway B. Test F1 SD across 10 seeds was ≤5.0% in 75 of 77 combinations. Within this cohort, passive EMR modeling achieved 91–94% sensitivity on the primary split (85–89% across seeds) without additional data collection, supporting the feasibility of ML-based notification in resource-constrained hospitals. Whether this reduces dengue underreporting must be established in prospective, multi-center studies.

We conducted a retrospective cohort analysis of routine programmatic data to identify predictors of antiretroviral therapy (ART) initiation, retention in care (RiC), and viral load suppression (VLS) among adolescents and young adults (AYA) aged 10–24 years living with HIV in the Ekurhuleni District, South Africa. Time-to-event outcomes were analysed using Cox proportional hazards regression models. Of the 6667 AYA diagnosed with HIV, 5271 (79.5%) were aged 20–24 years, 5754 (86.8%) were female, and 5430 (81.9%) initiated ART. The median time to ART initiation was 0.0 months (interquartile range [IQR]: 0.00–0.13), indicating that most AYA initiated ART on the same day as HIV diagnosis or within a few days thereafter. Time to ART initiation did not differ significantly across age groups (log-rank χ²(2) = 5.83, p = 0.054). Retention in care was suboptimal, with 56% retained in care at five years. Compared with AYA aged 10–14 years, those aged 15–19 years (adjusted hazard ratio [aHR] = 2.29, 95% confidence interval [CI]: 1.38–3.82; p = 0.001) and 20–24 years (aHR = 2.54, 95% CI: 1.55–4.18; p < 0.001) had a higher hazard of retention in care. AYA who were pregnant at ART initiation also had a higher hazard of retention than those who were not pregnant (aHR = 1.26, 95% CI: 1.13–1.42; p < 0.001). Although 81% of AYA achieved VLS, this proportion remained below the UNAIDS third 95 target of 95% viral suppression among people receiving ART. Compared with AYA aged 10–14 years, those aged 20–24 years had a higher hazard of VLS (aHR = 1.05, 95% CI: 1.03–1.07; p < 0.001). Compared with AYA who initiated ART at World Health Organization (WHO) clinical stage I, those at WHO stage II had a lower hazard of VLS (aHR = 0.76, 95% CI: 0.65–0.89; p = 0.001). These findings highlight the importance of age and clinical and pregnancy-related factors in HIV care outcomes among AYA. Targeted, AYA-centred interventions that support timely ART initiation, sustained engagement in care, and viral suppression are essential to improve treatment outcomes in this population.

Dengue transmission in climate-sensitive and socioeconomically heterogeneous settings arises from nonlinear interactions between environmental suitability, mosquito ecology, and human behaviour. Classical SEIR-type models offer mechanistic interpretability but are often constrained by observational noise and incomplete data, while purely data-driven approaches typically fail to respect underlying transmission dynamics. In this study, we introduce a hybrid epidemiological framework that embeds a temperature–unemployment driven SEIR–SEI model within a Physics-Informed Neural Network (SEIR–PINN). Using weekly dengue incidence data from Medellín, Colombia, we incorporate unemployment as a socioeconomic proxy that modulates transmission alongside climatic suitability. The Model performance was evaluated in two ways; the prediction of observed incidence, the SEIR–PINN achieved a mean absolute error (MAE) of 15.05 and $R^2$ of 0.975, compared with MAE of 43.57 and $R^2$ of 0.870 for a mechanistic SEIR–SEI baseline fitted under identical climatic and socioeconomic assumptions. Shuffled five-fold cross-validation provided held-out prediction of observed incidence within the 2009–2011 period, with mean MAE of 14.87 and mean $R^2$ of 0.977 (out-of-fold MAE 14.89, $R^2$ 0.978). These held-out results interpolate within the observed epidemic and are not an assessment of prospective forecasting performance Importantly, the SEIR–PINN reconstructs latent human and mosquito compartments that remain biologically consistent across epidemic phases. By fusing mechanistic structure with data-adaptive flexibility, the proposed SEIR–PINN provides an interpretable framework for analysing the joint climatic and socioeconomic modulation of dengue transmission in an urban endemic setting. Although using Medellín as a single-city methodological case study, application of this framework to other dengue endemic settings requires external validation using additional locations, longer time periods, and independent datasets. The results therefore support the potential use of physics-informed hybrid models as interpretable tools for dengue surveillance and transmission-dynamics research, while recognising the need for further external validation across longer time periods and additional geographical regions.