St. Francis Hospital Nsambya, commonly known as Nsambya Hospital, is a hospital in Kampala, the capital of Uganda and the largest city in that country.
Specialist epilepsy expertise is scarce in resource-constrained settings, making LLM-based decision support attractive for frontline clinicians managing longitudinal treatment. Such support systems must do more than apply medical knowledge: they must adapt to local prescribing practice and know when to defer. Public medical AI benchmarks are dominated by high-income clinical settings, leaving prescribing practices, medication availability, and follow-up patterns in low-resource contexts largely unrepresented. We study this problem through a multidisciplinary collaboration in Ugandan pediatric epilepsy care. The task is to predict anti-seizure medication regimens from longitudinal unstructured notes collected by local clinicians across serial visits. Standard prompting achieves non-trivial agreement with physician prescriptions, but neurologists' review of model reasoning traces shows that its errors stem from distribution-miscalibrated prescribing defaults rather than the local care environment. We introduce Manana, a non-parametric prompt-learning framework that learns how to reason about local prescribing decisions from a small patient-level training set. Manana turns observed prescription errors into an auditable prompt memory, instantiated in single-agent and multi-agent variants, and outperforms classical ML models, direct LLM prompting, and prompt-optimization baselines across two independently collected Ugandan cohorts. To make the system uncertainty-aware, we propose Bayesian prompt averaging (BPA), a Bayesian model averaging procedure over the learned prompt trajectory. This converts a sequence of learned prompts into prescription likelihoods and produces a deferral signal. On the independently collected held-out cohort, BPA improves visit-level top-3 prescription accuracy by 4-8 percentage points over the prompt-optimization baselines. More consequentially, it enables clinically meaningful selective prediction: the system can auto-handle the most confident half of cases at 95\% precision, or the most confident quarter at 99\% precision, while deferring lower-confidence cases for specialist review. These results suggest a path toward locally adapted clinical LLM systems that learn from limited site-specific data and reserve scarce specialist attention for the cases where uncertainty is highest.
Background:Dementia diagnosis in sub-Saharan Africa is constrained by limited access to specialist neuroimaging interpretation and reduced specificity of brief cognitive tools in low-literacy populations. We evaluated the agreement, incremental value, and comparative performance of Mini Mental State Exam (MMSE), visual MRI medial temporal atrophy (MTA), and automated brain morphometry in older Ugandan adults with suspected dementia. Methods:In this cross-sectional study, adults aged ≥50 years with suspected dementia were recruited from neurology and psychiatry clinics at two hospitals and from a community cohort. Participants underwent MMSE and standardized 1.5T brain MRI. Visual MRI ratings were performed by radiologists blinded to clinical data, and automated morphometry was generated using NeuroQuant® normative percentiles. Hippocampal occupancy (HOC <5th percentile) was used as a reference MRI biomarker for comparative classification. Agreement between visual and automated measures was assessed using Spearman correlation and intraclass correlation. Incremental value was assessed using regression models, and comparative performance using area under the curve (AUC). Results:Sixty-three participants were included (mean age 75.6 ± 8.7 years; 49 female). Agreement between visual ratings and automated morphometry was poor. MMSE correlated inversely with MTA (ρ = -0.47; p = 0.049) and correlated positively with hippocampal volume percentile (ρ = 0.46; p = 0.056). Adding hippocampal volume to MTA did not improve model fit for MMSE (ΔR2 = 0.028; p = 0.18). For comparative classification, MMSE alone was sensitive but poorly specific, while the combined MMSE-MTA model improved specificity and discrimination (AUC 0.70 vs 0.62 for either measure alone). Conclusion:Visual and automated MRI measures were not interchangeable in this heterogeneous cohort. Automated hippocampal volumetry added limited value beyond visual MTA for global cognition, while combining MMSE with visual MTA showed modest improvement in comparative classification and warrants further validation.
Preterm births are a leading cause of death and illness, with rates rising in Uganda. CPAP helps with respiratory problems, and kangaroo mother care (KMC) stabilises temperature, encourages bonding, and supports health. Yet, KMC for preterm infants on CPAP is rarely used in countries like Uganda. This study explores challenges and enablers for KMC in these cases. The study used a cross-sectional design with qualitative methods at St. Francis Hospital Nsambya. NICU healthcare workers were conveniently sampled and interviewed until saturation. Data was transcribed, coded, and analysed thematically in Atlas.ti Version 9. The study included 8 respondents with an average age of 34 years (range: 26–48 years). Barriers and facilitators were categorised into factors related to parents, babies, health workers, and health systems. A total of 22 themes and 20 subthemes emerged, covering both barriers and facilitators. Key barriers included issues such as mothers being sick, misinformation about KMC and CPAP, lack of maternal knowledge, negative attitudes towards KMC, and maternal fear. Additional barriers involved critically ill babies, poor attitudes and limited knowledge among health workers, absence of protocols, staffing shortages, equipment limitations, and timing of KMC. Facilitator themes highlighted paternal involvement and awareness efforts, such as education and counselling to improve maternal knowledge; fostering positive health worker attitudes; training; incentives; functional systems; access to information; early CPAP application; availability of KMC spaces in neonatal units; and integrating KMC into routine quality improvement projects. Multiple barriers categorised as health worker, baby, maternal, and health system factors still hinder the effective implementation of KMC among preterm babies on CPAP. Understanding these factors and developing category-specific interventions can enhance the utilisation of KMC, thereby improving outcomes for preterm babies. Not applicable.
An emergency caesarean section is a critical intervention to reduce maternal and fetal risks during labour complications. The decision-to-delivery interval, recommended internationally of seventy-five minutes or less, is a key indicator of quality in emergency obstetric care. In rural western Uganda, this interval is often prolonged due to financial and infrastructural barriers. This study evaluated the impact of a sponsorship program’s case management and documentation procedures on compliance with the decision-to-delivery interval in a rural hospital setting. A retrospective cross-sectional study was conducted from January 2020 to December 2023 across four comprehensive emergency obstetric care hospitals in southwestern Uganda: Ibanda, Nyakibale, Rushoroza, and Rwibaale. Data were extracted for all emergency caesarean section cases with indications in categories one and two. Documentation rates and compliance with the decision-to-delivery interval (defined as less than seventy-five minutes) were analyzed using descriptive statistics, trends, and stratified graphs. The study sample compared data from non-sponsored and sponsored patients covered by a financing program offered by African Mission Healthcare, a not-for-profit organization working in Africa. Surgical Access for Everyone–sponsored patients consistently demonstrated higher documentation rates for the decision-to-delivery interval (70–80
Background Chronic inflammation is implicated in coronary atherosclerosis, particularly among people with HIV, who may experience persistent immune activation despite antiretroviral therapy. However, most evidence linking inflammatory biomarkers to coronary artery disease (CAD) comes from populations outside sub-Saharan Africa. We assessed associations between inflammatory biomarkers and CAD among older adults with and without HIV in southwestern Uganda. Methods We analyzed adults aged ≥ 40 years in a community-based cohort in rural southwestern Uganda. CAD was assessed using coronary artery calcium scanning and contrast-enhanced coronary computed tomography angiography. Biomarkers included indicators of macrophage activation (sCD163), hypercoagulability (D-dimer), systemic inflammation (C-reactive protein), gut epithelial integrity (fatty acid binding protein 2), and hematologic inflammatory indices derived from complete blood counts. We fitted Firth penalized logistic regression models to estimate crude, HIV-stratified, and multivariable-adjusted associations. Receiver-operating-characteristic analyses assessed whether select biomarkers improved discrimination beyond the atherosclerotic cardiovascular disease (ASCVD) risk score. Results Among 585 participants, mean age was 57.7 years, 287 (49.1%) had HIV, and 45 (7.7%) had CAD. Most biomarkers, including C-reactive protein, D-dimer, fatty acid–binding protein 2, and complete blood count–derived inflammatory indices, were not associated with CAD after multivariable adjustment and false-discovery-rate correction. Higher sCD163 was inversely associated with CAD in the adjusted overall model (odds ratio per 1-standard-deviation increase, 0.63; 95% confidence interval, 0.47–0.87; q = 0.048), but this finding was not consistent across HIV strata. Adding sCD163 to the ASCVD risk score did not improve discrimination for CAD. Conclusions In this cohort of older adults in rural Uganda, inflammatory biomarkers showed limited associations with CAD and did not improve discrimination beyond conventional risk assessment. These findings support emphasis on established cardiovascular risk assessment among older Ugandans, and further investigation into the pathogenesis of CVD risk in this population.
