Use of unsupervised machine learning to characterise HIV predictors in sub-Saharan Africa.

INTRODUCTION: Significant regional variations in the HIV epidemic hurt effective common interventions in sub-Saharan Africa. It is crucial to analyze HIV positivity distributions within clusters and assess the homogeneity of countries. We aim at identifying clusters of countries based on socio-behavioural predictors of HIV for screening. METHOD: We used an agglomerative hierarchical, unsupervised machine learning, approach for clustering to analyse data for 146,733 male and 155,622 female respondents from 13 sub-Saharan African countries with 20 and 26 features, respectively, using Population-based HIV Impact Assessment (PHIA) data from the survey years 2015-2019. We employed agglomerative hierarchical clustering and optimal silhouette index criterion to identify clusters of countries based on the similarity of socio-behavioural characteristics. We analyse the distribution of HIV positivity with socio-behavioural predictors of HIV within each cluster. RESULTS: Two principal components were obtained, with the first describing 62.3% and 70.1% and the second explaining 18.3% and 20.6% variance of the total socio-behavioural variation in females and males, respectively. Two clusters per sex were identified, and the most predictor features in both sexes were: relationship with family head, enrolled in school, circumcision status for males, delayed pregnancy, work for payment in last 12 months, Urban area indicator, known HIV status and delayed pregnancy. The HIV positivity distribution with these variables was significant within each cluster. CONCLUSIONS /FINDINGS: The findings provide a potential use of unsupervised machine learning approaches for substantially identifying clustered countries based on the underlying socio-behavioural characteristics.

Use of machine learning techniques to identify HIV predictors for screening in sub-Saharan Africa.

AIM: HIV prevention measures in sub-Saharan Africa are still short of attaining the UNAIDS 90-90-90 fast track targets set in 2014. Identifying predictors for HIV status may facilitate targeted screening interventions that improve health care. We aimed at identifying HIV predictors as well as predicting persons at high risk of the infection. METHOD: We applied machine learning approaches for building models using population-based HIV Impact Assessment (PHIA) data for 41,939 male and 45,105 female respondents with 30 and 40 variables respectively from four countries in sub-Saharan countries. We trained and validated the algorithms on 80% of the data and tested on the remaining 20% where we rotated around the left-out country. An algorithm with the best mean f1 score was retained and trained on the most predictive variables. We used the model to identify people living with HIV and individuals with a higher likelihood of contracting the disease. RESULTS: Application of XGBoost algorithm appeared to significantly improve identification of HIV positivity over the other five algorithms by f1 scoring mean of 90% and 92% for males and females respectively. Amongst the eight most predictor features in both sexes were: age, relationship with family head, the highest level of education, highest grade at that school level, work for payment, avoiding pregnancy, age at the first experience of sex, and wealth quintile. Model performance using these variables increased significantly compared to having all the variables included. We identified five males and 19 females individuals that would require testing to find one HIV positive individual. We also predicted that 4·14% of males and 10.81% of females are at high risk of infection. CONCLUSION: Our findings provide a potential use of the XGBoost algorithm with socio-behavioural-driven data at substantially identifying HIV predictors and predicting individuals at high risk of infection for targeted screening.

Comparing Machine Learning Classifiers for Predicting Hospital Readmission of Heart Failure Patients in Rwanda.

High rates of hospital readmission and the cost of treating heart failure (HF) are significant public health issues globally and in Rwanda. Using machine learning (ML) to predict which patients are at high risk for HF hospital readmission 20 days after their discharge has the potential to improve HF management by enabling early interventions and individualized treatment approaches. In this paper, we compared six different ML models for this task, including multi-layer perceptron (MLP), K-nearest neighbors (KNN), logistic regression (LR), decision trees (DT), random forests (RF), and support vector machines (SVM) with both linear and radial basis kernels. The outputs of the classifiers are compared using performance metrics including the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. We found that RF outperforms all the remaining models with an AUC of 94% while SVM, MLP, and KNN all yield 88% AUC. In contrast, DT performs poorly, with an AUC value of 57%. Hence, hospitals in Rwanda can benefit from using the RF classifier to determine which HF patients are at high risk of hospital readmission.

Child stunting prevalence determination at sector level in Rwanda using small area estimation.

BACKGROUND: Stunting among children under 5 years of age remains a worldwide concern, with 148.1 million (22.3%) stunted in 2022. The recent 2019/2020 Rwanda Demographic Health Survey (RDHS) revealed that the prevalence of stunting in Rwanda among under five children was 33.5%. In Rwanda, there is no sufficient evidence on stunting status to guide prioritized interventions at the sector level, the lowest administrative unit for implementing development initiatives. This study aimed to provide reliable estimates of stunting prevalence in Rwanda at the sector level. METHODS: In this article, Small Area Estimation (SAE) techniques were used to provide sector level estimates of stunting prevalence in children under five in Rwanda. By plugging in relevant significant covariates in the generalized linear mixed model, model-based estimates are produced for all sectors with their corresponding Mean Square Error (MSE). RESULTS: The findings showed that, overall, 40 out of 416 sectors had met the national target of having a stunting rate less than or equal to 19%, while 194 sectors were far from meeting this target, having a stunting rate higher than the national prevalence of 33.5% in the year 2020. The majority of the sectors with stunting prevalence that were higher than the national average of 33.5% were found in the Northern Province with 68 sectors out of 89 and in Western Province with 64 sectors out of 96. In contrast, the prevalence of stunting was lower in the City of Kigali where 14 out of 35 sectors had a stunting rate between 0 and 19%, and all sectors were below the national average. This study showed a substantial connection between stunting and factors such as household size, place of residence, the gender of the household head, and access to improved toilet facilities and clean water. CONCLUSION: The results of this study may guide and support informed policy decisions and promote localised and targeted interventions in Rwanda's most severely affected sectors with a high stunting prevalence in Rwanda.

Compartmental mathematical modelling of dynamic transmission of COVID-19 in Rwanda.

Objectives: Mathematical modelling is of interest to study the dynamics of coronavirus disease 2019 (COVID-19), and models such as SEIR (Susceptible-Exposed-Infected-Recovered) have been considered. This article describes the development of a compartmental transmission network model - Susceptible-Exposed-Quarantine-Infectious-Infectious, undetected-Infectious, home-based care-Hospitalized-Vaccinated-Recovered-Dead - to simulate the dynamics of COVID-19 in order to account for specific measures put into place by the Government of Rwanda to prevent further spread of the disease. Methods: The compartments of this model are connected by parameters, some of which are known from the literature, and others are estimated from available data using the least squares method. For the stability of the model, equilibrium points were determined and the basic reproduction number R 0 was studied; R 0 is an indicator for contagiousness. Results: The model showed that secondary infections are generated from the exposed group, the asymptomatic group, the infected (symptomatic) group, the infected (undetected) group, the infected (home-based care) group and the hospitalized group. The formulated model was reliable and fit the data. Furthermore, the estimated R 0 of 2.16 shows that COVID-19 will persist without the application of control measures. Conclusions: This article presents results regarding predicted spread of COVID-19 in Rwanda.

Performance of Machine Learning Classifiers in Classifying Stunting among Under-Five Children in Zambia.

Stunting is a global public health issue. We sought to train and evaluate machine learning (ML) classification algorithms on the Zambia Demographic Health Survey (ZDHS) dataset to predict stunting among children under the age of five in Zambia. We applied Logistic regression (LR), Random Forest (RF), SV classification (SVC), XG Boost (XgB) and Naïve Bayes (NB) algorithms to predict the probability of stunting among children under five years of age, on the 2018 ZDHS dataset. We calibrated predicted probabilities and plotted the calibration curves to compare model performance. We computed accuracy, recall, precision and F1 for each machine learning algorithm. About 2327 (34.2%) children were stunted. Thirteen of fifty-eight features were selected for inclusion in the model using random forest. Calibrating the predicted probabilities improved the performance of machine learning algorithms when evaluated using calibration curves. RF was the most accurate algorithm, with an accuracy score of 79% in the testing and 61.6% in the training data while Naïve Bayesian was the worst performing algorithm for predicting stunting among children under five in Zambia using the 2018 ZDHS dataset. ML models aids quick diagnosis of stunting and the timely development of interventions aimed at preventing stunting.

Comparing growth velocity of HIV exposed and non-exposed infants: An observational study of infants enrolled in a randomized control trial in Zambia.

BACKGROUND: Impaired growth among infants remains one of the leading nutrition problems globally. In this study, we aimed to compare the growth trajectory rate and evaluate growth trajectory characteristics among children, who are HIV exposed uninfected (HEU) and HIV unexposed uninfected (HUU), under two years in Zambia. METHOD: Our study used data from the ROVAS II study (PACTR201804003096919), an open-label randomized control trial of two verses three doses of live, attenuated, oral RotarixTM administered 6 &10 weeks or at 6 &10 weeks plus an additional dose at 9 months of age, conducted at George clinic in Lusaka, Zambia. Anthropometric measurements (height and weight) were collected on all scheduled and unscheduled visits. We defined linear growth velocity as the rate of change in height and estimated linear growth velocity as the first derivative of the mixed effect model with fractional polynomial transformations and, thereafter, used the second derivative test to determine the peak height and age at peak heigh. RESULTS: We included 212 infants in this study with median age 6 (IQR: 6-6) weeks of age. Of these 97 (45.3%) were female, 35 (16.4%) were stunted, and 59 (27.6%) were exposed to HIV at baseline. Growth velocity was consistently below the 3rd percentile of the WHO linear growth standard for HEU and HUU children. The peak height and age at peak height among HEU children were 74.7 cm (95% CI = 73.9-75.5) and 15.5 months (95% CI = 14.7-16.3) respectively and those for HUU were 73 cm (95% CI = 72.1-74.0) and 15.6 months (95% CI = 14.5-16.6) respectively. CONCLUSION: We found no difference in growth trajectories between infants who are HEU and HUU. However, the data suggests that poor linear growth is universal and profound in this cohort and may have already occurred in utero.