School of Public Health (Public-Health) Collections

Permanent URI for this collection

http://10.29.1.53:4000/handle/10570/184

Browse

Recent Submissions

Now showing 1 - 5 of 970
  • Item
    A Machine Learning Model for Prediction of Antibiotic Resistance with Escherichia Coli Infections Using Demographic, Clinical and Microbiological Data.
    (Makerere University., 2026-01-13) Kahuma, Clare Allelua.
    IntroductionIn, low- and middle-income countries like Uganda, there is growing reliance on empirical prescription of broad-spectrum antibiotics which, while targeting a wide range of pathogens, contributes to the development of resistance to common pathogens such as E. coli. This challenge is compounded by the poor selection of antibiotic panels in many laboratories, which often fail to reflect local resistance patterns and patient-specific factors, leading to inefficient use of scarce resources and delayed appropriate treatment. Objectives of the study The objectives of this study were to; 1) identify risk factors for drug resistant E. coli infections using machine learning techniques; 2) evaluate the performance of different machine learning models in predicting the likelihood of drug resistance among patients with E. coli infections using demographic, clinical and microbiological data; and 3) develop a web-based interface to support proper antibiotic prescription and targeted antimicrobial decision-making. MethodologyA retrospective analysis was conducted on 1,552 records of patients diagnosed with E. coli infections in 10 tertiary healthcare facilities in Uganda. These records were analyzed using machine learning models including Lightgbm, xgboost, random forest, gradient boosting, and decision trees. Feature selection was guided by a weighted importance score and frequency count framework. The best performing model was deployed in a streamlit-based web interface.Results Key predictors of resistance included antibiotic type, patient age, hospital site, specimen type, prior antibiotic use, and hospitalization history. XGboost emerged as the top-performing models for prediction of drug resistance, with an accuracy of 82.32%, a precision of 82.36%, recall of 85.37%, F1 score of 83.84%, and ROC AUC of 90.17%. The web-based interface was implemented using python streamlit technology, intergrated with the best performing model to enable real-time resistance prediction.ConclusionThis study demonstrates the potential of machine learning to transform antimicrobial resistance surveillance and clinical decision-making in resource-limited settings.
  • Item
    Enhancing outbreak surveillance through integration of natural language processing in Uganda’s electronic integrated disease surveillance and response system.
    (Makerere University., 2026-01-13) Nakitandwe, Rebecca,Melisa.
    Introduction: Early detection of diseases or infections is essential to prevent infectious diseases from escalating into large outbreaks. In Uganda, the Electronic Integrated Disease Surveillance and Response (eIDSR) system enables community-level reporting of suspected cases via SMS. However, manual processing of these unstructured messages often delays outbreak detection and response, particularly during high-volume reporting periods. The manual processing of incoming SMS messages within the eIDSR system creates a bottleneck that hinders timely outbreak detection and response. This delay has the potential to increase morbidity and mortality, especially in resource-limited settings. This study aimed to integrate Natural Language Processing (NLP) to automate the extraction of key information, such as disease type, location, and symptoms, from SMS alerts submitted to the eIDSR system. It also sought to understand the contextual factors that influenced model accuracy and performance. Methods: A retrospective design was employed using historical SMS data submitted to the eIDSR system in 2024. A Bidirectional Encoder Representations from Transformers (BERT)-uncased model was fine-tuned on a manually annotated dataset to support named entity recognition. The model was evaluated using precision, recall, F1-score, and processing speed, and its performance was compared with manual extraction. McNemar’s test was used to assess the statistical significance of differences between the two methods. Results: The model achieved an F1-score of 92.6%, with recall of 94.2% and precision of 91.1%, processing approximately 48 messages per second. It extracted high-value entities such as disease, age, gender, and location, with near-perfect accuracy. Errors were concentrated around symptom span boundaries and ambiguous entries. Interviews confirmed the value of automation for reducing analyst workload and outlined key limitations of the current manual workflow, including handling of ambiguous or duplicate messages. Conclusion: This study demonstrated the feasibility of applying NLP to automate SMS-based disease surveillance within Uganda’s eIDSR system. Although human review remains necessary for edge cases, the model showed strong potential to accelerate processing, eliminate backlog, and support timely response under frameworks like 7-1-7. With targeted improvements especially in symptom handling and multilingual input. The model would be suitable for pilot integration under a human-in-the-loop deployment model.
  • Item
    Factors associated with clinically diagnosed malaria among adolescents attending public health centre IVs in Wakiso District, Uganda
    (Makerere University, 2026) Nansubuga, Rahma
    Introduction: The burden of clinically diagnosed malaria in sub-Saharan Africa is increasingly shifting towards adolescents. In Uganda, adolescents are gradually facing an increased malaria risk, yet the specific factors influencing malaria in this group within public health facilities are not well understood. This study assessed the factors associated with clinically diagnosed malaria among adolescents attending public Health Centre IVs in Wakiso District, Uganda. Materials and methods. A cross-sectional survey using quantitative and qualitative methods was conducted among 434 adolescents presenting at 6 selected public Health Centre IVs in Wakiso district, between August and September 2024, using a digitized semi-structured questionnaire. Additionally, 13 key informant interviews were conducted with selected health workers. Quantitative data was analyzed using STATA v14.0 with descriptive statistics and Modified Poisson regression to identify factors associated with clinically diagnosed malaria. Qualitative data was analyzed inductively using Atlas. Ti version 7. Results: Up to 29.95% (130/434) of adolescents were clinically diagnosed with malaria. Most respondents 87.56% (380/434) demonstrated good knowledge about malaria. Although 93.78% (407/434) sought care for their recent malaria episode, only 18.43% (75/407) did so on the day symptoms began. While 80.41% (349/434) received antimalarials, only 67.05% (291/434) adhered to the prescribed regimen and 23.96% (104/434) supplemented it with herbal remedies. Lower occurrence of clinically diagnosed malaria was observed among adolescents residing in rural areas (APR = 0.67, 95% CI: 0.49–0.92), those with higher socioeconomic status (APR = 0.65, 95% CI: 0.46–0.91, APR = 0.70, 95% CI: 0.50–0.97), and older adolescents (APR = 0.42, 95% CI: 0.29 0.61). Conversely, distance to health facilities (APR = 4.71, 95% CI: 2.64–8.41), irregular bed net use (APR = 4.19, 95% CI: 2.41–7.31), and receiving information on malaria from multiple sources (APR = 1.62, 95% CI: 1.17–2.25) were associated with higher occurrence of clinically diagnosed malaria. Health workers reported increased adolescent facility attendance during school term transitions, untimely care seeking, and a preference for injectable over oral antimalarials as challenges to effective management of clinically diagnosed malaria among this age group. Conclusion and Recommendations: The proportion of adolescents with clinically diagnosed malaria was 29.95%, associated with factors such as urban residence, lower socioeconomic status, younger age (10-14 years), greater distance to health facilities, male gender, irregular use of bed nets, and paradoxically, higher exposure to malaria information and good prevention practices. Tailored health education focusing on information critical to clinically diagnosed malaria including early symptom recognition, prompt care seeking within 24 hours, and adherence to prescribed treatment is urgently needed to reduce adolescent malaria morbidity.
  • Item
    Evaluating the effectiveness of facility-based and community-based PrEP delivery models for HIV prevention among key populations in Uganda: a propensity score matching approach
    (Makerere University, 2026) Asaba, Gloria
    Background: HIV remains a major public health concern in Uganda, with new infections disproportionately affecting key and priority populations such as female sex workers, clients of sex workers, boda-boda riders, and adolescent girls and young women. Although the Ministry of Health has adopted both facility-based and community-based models to expand access to oral preexposure prophylaxis (PrEP), gaps persist in retention, adherence, and continuation. Evidence directly comparing the effectiveness of these two delivery approaches in routine program settings remains limited. The goal of this study was to compare the effectiveness of facility-based versus community-based PrEP delivery models in supporting retention, adherence, and HIV seroconversion among clients at Kasangati Health Centre IV. Methods: This was a comparative observational study that applied propensity score matching, to secondary data extracted from the HIV Prevention Tracker using a structured data abstraction tool aligned with national PrEP indicators. The study included all clients initiated on oral PrEP at Kasangati Health Centre IV between January and December 2024, representing key and priority populations who were HIV-negative at initiation and had follow-up records. Because clients were not randomly assigned to delivery models, the community-based model served as the exposure group and the facility-based model as the comparison group. To minimise baseline differences, Propensity Score Matching using 1:1 nearest-neighbor matching without replacement was performed based on age, sex at birth, and population category, yielding a matched sample of 234 clients (117 per model). Outcomes included retention (≥2 PrEP visits in 12 months), adherence (≥95% pill-taking as documented at visits), and HIV seroconversion (any new HIV-positive result during follow-up). After matching, logistic regression was used to estimate the Average Treatment Effect on the Treated (ATT), comparing outcomes between community and facility delivery models. Results: The two matched groups were similar in age distribution with a mean age of 27 years, sex, and population category. After matching, retention did not significantly differ between models (facility 54.7% vs community 52.1%; aOR = 1.15; 95% CI: 0.58–2.27; p = 0.69). Adherence was uniformly high in both models. Crude HIV seroconversion was higher in the facility model (22.2%) compared to the community model (0.9%); however, this difference was not statistically significant after adjustment (aOR = 1.67; 95% CI: 0.32–8.73; p = 0.53). The facility model served a higher proportion of AGYW, while the community model predominantly served sex workers and clients of sex workers. Conclusion: Both service delivery models achieved high adherence and comparable retention, demonstrating that community-based PrEP delivery is as effective as facility-based delivery in real-world conditions. The higher crude seroconversion in facility settings highlights the need for contextual investigation into risk factors among facility clients. Given the comparable effectiveness, future evaluations should assess the cost-effectiveness and coverage benefits of maintaining both models within Uganda’s differentiated PrEP delivery framework.
  • Item
    Effect of insecticide treated nets or indoor residual spraying on Malaria or Anemia risk among children aged 6-59 months in Uganda : a propensity score matched analysis
    (Makerere University, 2026) Kyoheirwe, Apophia
    Background: Malaria and anemia represent critical health challenges for Ugandan children under five, with national prevalence rates of 53% for anemia and persistently high malaria incidence. While Insecticide-Treated Nets (ITNs/LLINs) and Indoor Residual Spraying (IRS) serve as primary vector control interventions, evidence regarding their comparative effectiveness in real-world settings remains limited due to methodological constraints in observational studies. This study addresses this gap by employing Propensity Score Matching methods to evaluate the causal effects of these interventions on both malaria and anemia outcomes while controlling for confounding factors. Methods: A Propensity Score Matching (PSM) analysis was conducted using nationally representative data from the Uganda Malaria Indicator Survey (UMIS) 2018-2019 involving 5,549 children aged 6-59 months. The Average Treatment Effect on the Treated (ATT) was estimated for ITNs/LLINs and IRS adjusting for key sociodemographic and geographic confounders such as maternal education, household wealth and region. Rosenbaum bounds sensitivity analysis assessed robustness to unobserved bias. Results: Malaria and anemia prevalence among children aged 6-59 months was 19.6% and 45% respectively. ITNs/LLINs use was associated with 5.5% reduction in malaria risk (ATT=-0.055; 95% CI: -0.107 to -0.002) and a 15.4% reduction in anemia risk (ATT= -0.154; 95% CI: -0.221 to -0.087). IRS use was associated with 14% reduction in malaria risk (ATT= -0.140; 95% CI: -0.208 to -0.072) and a 11.6% reduction in anemia risk (ATT= -0.116; 95% CI: -0.186 to -0.046). Rosenbaum bounds Sensitivity analyses indicated the effect of IRS on malaria was highly robust to hidden bias, while the effect of ITNs/LLINs on malaria was more sensitive. Conclusion: Both ITNs/LLINs and IRS significantly reduce the risk of malaria and anemia, but with differential efficacy. IRS demonstrates superior protection against malaria, making it a strategic tool for targeted deployment in high-transmission areas. Conversely, ITNs/LLINs showed a stronger effect on reducing anemia and should be maintained as a universal intervention. These findings advocate for an integrated, data-driven malaria control strategy that leverages the complementary strengths of both interventions to mitigate Uganda's dual disease burden effectively.