Application of machine learning methods for predicting infant mortality in Rwanda: analysis of Rwanda demographic health survey 2014-15 dataset.

BACKGROUND: Extensive research on infant mortality (IM) exists in developing countries; however, most of the methods applied thus far relied on conventional regression analyses with limited prediction capability. Advanced of Machine Learning (AML) methods provide accurate prediction of IM; however, there is no study conducted using ML methods in Rwanda. This study, therefore, applied Machine Learning Methods for predicting infant mortality in Rwanda.  METHODS: A cross-sectional study design was conducted using the 2014-15 Rwanda Demographic and Health Survey. Python software version 3.8 was employed to test and apply ML methods through Random Forest (RF), Decision Tree, Support Vector Machine and Logistic regression. STATA version 13 was used for analysing conventional methods. Evaluation metrics methods specifically confusion matrix, accuracy, precision, recall, F1 score, and Area under the Receiver Operating Characteristics (AUROC) were used to evaluate the performance of predictive models. RESULTS: Ability of prediction was between 68.6% and 61.5% for AML. We preferred with the RF model (61.5%) presenting the best performance. The RF model was the best predictive model of IM with accuracy (84.3%), recall (91.3%), precision (80.3%), F1 score (85.5%), and AUROC (84.2%); followed by decision tree model with model accuracy (83%), recall (91%), precision (79%), F1 score (84.67%) and AUROC(82.9%), followed by support vector machine with model accuracy (68.6%), recall (74.9%), precision(67%), F1 score (70.73%) and AUROC (68.6%) and last was a logistic regression with the low accuracy of prediction (61.5%), recall (61.1%), precision (62.2%), F1 score (61.6%) and AUROC (61.5%) compared to other predictive models. Our predictive models showed that marital status, children ever born, birth order and wealth index are the 4 top predictors of IM. CONCLUSIONS: In developing a predictive model, ML methods are used to classify certain hidden information that could not be detected by traditional statistical methods. Random Forest was classified as the best classifier to be used for the predictive models of IM.

Prevalence and characteristics associated with diabetes mellitus and impaired fasting glucose among people aged 15 to 64 years in rural and urban Rwanda: secondary data analysis of World Health Organization surveillance data.

Introduction: diabetes mellitus is an increasing public health burden in developing countries. The magnitude of diabetes association with traditional risk factors for diabetes have been given less attention in rural population. This study aims to determine the prevalence of diabetes and impaired fasting glucose and to assess associated characteristics to hyperglycemia in rural and urban Rwanda. Methods: this is a secondary analysis of data from a population-based cross-sectional study of 7240 people describing risk factors for non-communicable diseases using the WHO stepwise methods (STEPS). Relative frequencies of variables of interest were compared in rural and urban residence using Pearson chi-square tests. Diabetes and impaired fasting glucose were combined in a single hyperglycemia variable and odds ratios with 95% confidence intervals were used to explore associations between hyperglycemia, socio-demographic and health factors in urban and rural populations. Results: the prevalence in rural and urban areas was 7.5% and 9.7% (p.005) for diabetes and 5.0% and 6.2% for impaired fasting glucose (p.079) respectively. Obesity (AOR 2.57: CI: 0.86-7.9), high total cholesterol (AOR 3.83: CI: 2.03-7.208), hypertension (AOR 1.18: CI: 0.69-2.00), increasing age were associated with hyperglycemia in urban participants but only high total cholesterol and low high density lipoproteins (HDL) cholesterol were risk factors for hyperglycemia in rural participants. Conclusion: approximately one in six people in Rwanda have hyperglycemia. The magnitude of the association with traditional risk factors for diabetes differ in rural and urban settings. Different approaches to primary and secondary prevention of diabetes may be needed in rural populations.

Experience of Rwanda on COVID-19 Case Management: From Uncertainties to the Era of Neutralizing Monoclonal Antibodies.

The management of COVID-19 in Rwanda has been dynamic, and the use of COVID-19 therapeutics has gradually been updated based on scientific discoveries. The treatment for COVID-19 remained patient-centered and entirely state-sponsored during the first and second waves. From the time of identification of the index case in March 2020 up to August 2021, three versions of the clinical management guidelines were developed, with the aim of ensuring that the COVID-19 patients treated in Rwanda were receiving care based on the most recent therapeutic discoveries. As the case load increased and imposed imminent heavy burdens on the healthcare system, a smooth transition was made to enable that the asymptomatic and mild COVID-19 cases could continue to be closely observed and managed while they remained in their homes. The care provided to patients requiring facility-based interventions mainly focused on the provision of anti-inflammatory drugs, anticoagulation, broad-spectrum antibiotic therapy, management of hyperglycemia and the provision of therapeutics with a direct antiviral effect such as favipiravir and neutralizing monoclonal antibodies. The time to viral clearance was observed to be shortest among eligible patients treated with neutralizing monoclonal antibodies (bamlanivimab). Moving forward, as we strive to continue detecting COVID-19 cases as early as possible, and promptly initiate supportive interventions, the use of neutralizing monoclonal antibodies constitutes an attractive and cost-effective therapeutic approach. If this approach is used strategically along with other measures in place (i.e., COVID-19 vaccine roll out, etc.), it will enable us to bring this global battle against the COVID-19 pandemic under full control and with a low case fatality rate.

Socio-demographic and clinical characteristics of diabetes mellitus in rural Rwanda: time to contextualize the interventions? A cross-sectional study.

BACKGROUND: Existing prevention and treatment strategies target the classic types of diabetes yet this approach might not always be appropriate in some settings where atypical phenotypes exist. This study aims to assess the socio-demographic and clinical characteristics of people with diabetes in rural Rwanda compared to those of urban dwellers. METHODS: A cross-sectional, clinic-based study was conducted in which individuals with diabetes mellitus were consecutively recruited from April 2015 to April 2016. Demographic and clinical data were collected from patient interviews, medical files and physical examinations. Chi-square tests and T-tests were used to compare proportions and means between rural and urban residents. RESULTS: A total of 472 participants were recruited (mean age 40.2 ± 19.1 years), including 295 women and 315 rural residents. Compared to urban residents, rural residents had lower levels of education, were more likely to be employed in low-income work and to have limited access to running water and electricity. Diabetes was diagnosed at a younger age in rural residents (mean ± SD 32 ± 18 vs 41 ± 17 years; p < 0.001). Physical inactivity, family history of diabetes and obesity were significantly less prevalent in rural than in urban individuals (44% vs 66, 14.9% vs 28.7 and 27.6% vs 54.1%, respectively; p < 0.001). The frequency of fruit and vegetable consumption was lower in rural than in urban participants. High waist circumference was more prevalent in urban than in rural women and men (75.3% vs 45.5 and 30% vs 6%, respectively; p < 0.001). History of childhood under-nutrition was more frequent in rural than in urban individuals (22.5% vs 6.4%; p < 0.001). CONCLUSIONS: Characteristics of people with diabetes in rural Rwanda appear to differ from those of individuals with diabetes in urban settings, suggesting that sub-types of diabetes exist in Rwanda. Generic guidelines for diabetes prevention and management may not be appropriate in different populations.

Atypical forms of diabetes mellitus in Africans and other non-European ethnic populations in low- and middle-income countries: a systematic literature review.

BACKGROUND: Atypical presentations of diabetes mellitus (DM) have been reported in non-European ethnic populations under various names. It is unclear whether those names are used for the same or different clinical phenotypes. Unclear terminology may lead to inappropriate treatment and an underestimation of the burden caused by atypical diabetes phenotypes overlapping with classic types of diabetes. This review aimed to describe the terms used for atypical forms of diabetes and to investigate whether the terms are used for similar or different phenotypes. METHODS: PubMed and Scopus were searched for relevant publications in French or English available before 15 September 2015 using the terms: "Atypical diabetes", "Malnutrition Related Diabetes Mellitus (MRDM)", "Fibro-calculus pancreatic diabetes (FCPD)", Protein deficient Pancreatic Diabetes (PDPD)", "African diabetes", "Ketosis prone-type 2 diabetes", "tropical diabetes", "Flatbush diabetes", "J-type diabetes". Titles, abstracts screening and quality assessment were performed by two independent authors. Observational studies addressing atypical diabetes in humans aged 14 years and above were included. One author extracted data from selected articles. RESULTS: 22 articles among 350 identified articles were retained for data extraction. Two atypical diabetes phenotypes were identified, each of them with a variety of names but similar definitions. One phenotype occurred in very thin people less than 30 years of age, typically from poor socio-economic backgrounds and requires insulin for life. It differs from type 1 diabetes in the tolerance of high blood glucose without ketosis in the absence of exogenous insulin. The second phenotype resembles type1 diabetes as it presents with ketosis at onset but responds well, as type2 diabetes, to oral hypoglycemic drugs after initial stabilization with insulin. It occurs in individuals who are usually over 30 years of age, with normal or overweight and absence of auto antibodies mainly found in type 1 diabetes. CONCLUSION: The scarce existing literature used various terms for similar diabetes phenotypes. Agreement on nomenclature for the various forms of diabetes using the above reported characteristics are needed in populations where atypical forms of diabetes exist as well as better characterization of phenotypes and genotypes to inform evidence based treatment.

Proximal deep vein thrombosis among hospitalised medical and obstetric patients in Rwandan university teaching hospitals: prevalence and associated risk factors: a cross-sectional study.

OBJECTIVES: To determine the prevalence of proximal deep vein thrombosis (DVT) by ultrasound scanning, as well as associated clinical features and known risk factors, among medical and obstetrics-gynaecology inpatients in two Rwandan tertiary hospitals. DESIGN: Cross-sectional study. SETTINGS: Rwanda teaching hospitals: Kigali and Butare University Teaching Hospitals. PARTICIPANTS: 901 adult patients admitted to the Departments of Internal Medicine and Obstetrics-Gynecology (O&G) who were at least 21 years of age and willing to provide a consent. OUTCOMES: Prevalence of proximal DVT, clinical features and known risk factors associated with DVT. METHODS: Between August 2015 and August 2016, participants were screened for DVT by compressive ultrasound of femoral and popliteal veins, conducted as a monthly cross-sectional survey of all consenting eligible inpatients. Patients completed a self-report survey on DVT risk factors. Prevalence of proximal DVT by compression ultrasonography was the primary endpoint, with univariate and multivariate regression analyses performed to assess associated clinical features and risk factors. RESULTS: Proximal DVT was found in 5.5% of the study population, with similar rates in medical and O&G inpatients. The mean age was 41±16 SD (range, 21-91), 70% were female and 7% were pregnant. Univariate analysis showed active malignancy, immobilisation, prolonged recent travel and history of DVT to be significant risk factors for proximal DVT (all p values <0.05); while only active malignancy was an independent risk factor on multivariate regression (OR 5.2; 95% CI 2.0 to 13). Leg pain or tenderness, increased calf circumference, unilateral limb swelling or pitting oedema were predictive clinical features of DVT on both univariate analysis and multivariate regression (all p values <0.05). CONCLUSION: Proximal DVT prevalence is high among hospitalised medical and O&G patients in two tertiary hospitals in Rwanda. For reducing morbidity and mortality, research to develop Africa-specific clinical prediction tools for DVT and interventions to increase thromboprophylaxis use in the region are urgently needed.

Effects of a large-scale distribution of water filters and natural draft rocket-style cookstoves on diarrhea and acute respiratory infection: A cluster-randomized controlled trial in Western Province, Rwanda.

BACKGROUND: Unsafe drinking water and household air pollution (HAP) are major causes of morbidity and mortality among children under 5 in low and middle-income countries. Household water filters and higher-efficiency biomass-burning cookstoves have been widely promoted to improve water quality and reduce fuel use, but there is limited evidence of their health effects when delivered programmatically at scale. METHODS AND FINDINGS: In a large-scale program in Western Province, Rwanda, water filters and portable biomass-burning natural draft rocket-style cookstoves were distributed between September and December 2014 and promoted to over 101,000 households in the poorest economic quartile in 72 (of 96) randomly selected sectors in Western Province. To assess the effects of the intervention, between August and December, 2014, we enrolled 1,582 households that included a child under 4 years from 174 randomly selected village-sized clusters, half from intervention sectors and half from nonintervention sectors. At baseline, 76% of households relied primarily on an improved source for drinking water (piped, borehole, protected spring/well, or rainwater) and over 99% cooked primarily on traditional biomass-burning stoves. We conducted follow-up at 3 time-points between February 2015 and March 2016 to assess reported diarrhea and acute respiratory infections (ARIs) among children <5 years in the preceding 7 days (primary outcomes) and patterns of intervention use, drinking water quality, and air quality. The intervention reduced the prevalence of reported child diarrhea by 29% (prevalence ratio [PR] 0.71, 95% confidence interval [CI] 0.59-0.87, p = 0.001) and reported child ARI by 25% (PR 0.75, 95% CI 0.60-0.93, p = 0.009). Overall, more than 62% of households were observed to have water in their filters at follow-up, while 65% reported using the intervention stove every day, and 55% reported using it primarily outdoors. Use of both the intervention filter and intervention stove decreased throughout follow-up, while reported traditional stove use increased. The intervention reduced the prevalence of households with detectable fecal contamination in drinking water samples by 38% (PR 0.62, 95% CI 0.57-0.68, p < 0.0001) but had no significant impact on 48-hour personal exposure to log-transformed fine particulate matter (PM2.5) concentrations among cooks (ß = -0.089, p = 0.486) or children (ß = -0.228, p = 0.127). The main limitations of this trial include the unblinded nature of the intervention, limited PM2.5 exposure measurement, and a reliance on reported intervention use and reported health outcomes. CONCLUSIONS: Our findings indicate that the intervention improved household drinking water quality and reduced caregiver-reported diarrhea among children <5 years. It also reduced caregiver-reported ARI despite no evidence of improved air quality. Further research is necessary to ascertain longer-term intervention use and benefits and to explore the potential synergistic effects between diarrhea and ARI. TRIAL REGISTRATION: Clinical Trials.gov NCT02239250.

Laparoscopy in Rwanda: A National Assessment of Utilization, Demands, and Perceived Challenges.

BACKGROUND: Laparoscopy has proven to be feasible and effective at reducing surgical morbidity and mortality in low resource settings. In Rwanda, the demand for and perceived challenges to laparoscopy use remain unclear. METHODS: A mixed-methods study was performed at the four Rwandan national referral teaching hospitals. Retrospective logbook reviews (July 2014-June 2015) assessed procedure volume and staff involvement. Web-based surveys and semi-structured interviews investigated barriers to laparoscopy expansion. RESULTS: During the study period, 209 laparoscopic procedures were completed: 57 (27.3%) general surgery cases; 152 (72.7%) ob/gyn cases. The majority (58.9%, 125/209) occurred at the private hospital, which performed 82.6% of cholecystectomies laparoscopically (38/46). The three public hospitals, respectively, performed 25% (7/28), 15% (12/80), and 0% (denominator indeterminate) of cholecystectomies laparoscopically. Notably, the two hospitals with the highest laparoscopy volume relied on a single surgeon for more than 85% of cases. The four ob/gyn departments performed between 4 and 87 laparoscopic cases (mostly diagnostic). Survey respondents at all sites listed a dearth of trainers as the most significant barrier to performing laparoscopy (65.7%; 23/35). Other obstacles included limited access to training equipment and courses. Equipment and material costs, equipment functionality, and material supply were perceived as lesser barriers. Twenty-two interviews revealed widespread interest in laparoscopy, insufficient laparoscopy exposure, and a need for trainers. CONCLUSION: While many studies identify cost as the most prohibitive barrier to laparoscopy utilization in low resource settings, logbook review and workforce perception indicate that a paucity of trainers is currently the greatest obstacle in Rwanda.

Spirometric reference values for an East-African population.

BACKGROUND: Accurate interpretation of lung function testing requires appropriate reference values. Unfortunately, few African countries have produced spirometric reference values for their populations. OBJECTIVES: The present study was carried out in order to establish normal lung function values for subjects living in Rwanda, East Africa. METHODS: The study was conducted in Kigali, capital of Rwanda, and in the rural district of Huye in southern Rwanda. The variables studied were forced expiratory volume in 1 s (FEV1), forced vital capacity (FVC) and peak expiratory flow. Multiple regression analysis was performed using age, height, weight and BMI as independent variables to obtain predicted equations for both sexes. RESULTS: Predicted equations for normal lung functions were obtained from 740 healthy nonsmoking subjects; 394 were females and 346 were males. Minor differences in FEV1 and FVC were observed in comparison with other studies of Africans, African-Americans (difference in FEV1 and FVC of less than 5%), Chinese and Indians. When compared with selected studies from Caucasians and white Americans, our results for FEV1 and FVC were 9-12% and 16-18% lower in men and 12-23% and 17-28% lower in women, respectively. CONCLUSIONS: This study provides reference values for pulmonary function in a healthy, nonsmoking Rwandan population and enables comparisons to be made with other prediction equations from other populations. Spirometric reference values in our study were similar to those obtained in a study of black Americans by Hankinson et al.

Prevalence of atopy, asthma and COPD in an urban and a rural area of an African country.

UNLABELLED: The objectives of this study were to determine the prevalence of asthma, atopy and COPD in Rwanda and to identify risk factors. The survey was conducted in Kigali, the Capital of Rwanda, and in Huye District, a rural area located in southern Rwanda. METHODS: A total of 2138 subjects were invited to participate in the study.1920 individuals (90%) answered to questionnaires on respiratory symptoms and performed spirometry, 1824 had acceptable spirograms and performed skin-prick test. In case of airflow obstruction (defined as pre-bronchodilator ratio FEV(1)/FVC < LLN) a post bronchodilator spirometry was performed. Reversibility was defined as an increase in FEV(1) of 200 ml and 12% above baseline FEV(1) after inhalation of 400 mcg of salbutamol. RESULTS: The mean age was 38.3 years; 48.1% of participants were males and 51.9% females. Airflow obstruction was found in 256 participants (14%); 163(8.9%) subjects were asthmatics and 82 (4.5%) had COPD. COPD was found in 9.6% of participants aged 45 years and above. 484 subjects had positive skin-prick tests (26.5%); house dust mite and grass pollen mix were the main allergens. Risk factors for asthma were allergy, female gender and living in Kigali. COPD was associated with cigarette smoking, age and male sex. CONCLUSION: this is the first study which shows the prevalence of atopy, asthma and COPD in Rwanda. Asthma and COPD were respectively diagnosed in 8.9% and 4.5% of participants. COPD was diagnosed in 9.6% of subjects aged ≥ 45 years.The prevalence of asthma was higher in urban compared to rural area.