Leveraging artificial intelligence and data science techniques in harmonizing, sharing, accessing and analyzing SARS-COV-2/COVID-19 data in Rwanda (LAISDAR Project): study design and rationale.

BACKGROUND: Since the outbreak of COVID-19 pandemic in Rwanda, a vast amount of SARS-COV-2/COVID-19-related data have been collected including COVID-19 testing and hospital routine care data. Unfortunately, those data are fragmented in silos with different data structures or formats and cannot be used to improve understanding of the disease, monitor its progress, and generate evidence to guide prevention measures. The objective of this project is to leverage the artificial intelligence (AI) and data science techniques in harmonizing datasets to support Rwandan government needs in monitoring and predicting the COVID-19 burden, including the hospital admissions and overall infection rates. METHODS: The project will gather the existing data including hospital electronic health records (EHRs), the COVID-19 testing data and will link with longitudinal data from community surveys. The open-source tools from Observational Health Data Sciences and Informatics (OHDSI) will be used to harmonize hospital EHRs through the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM). The project will also leverage other OHDSI tools for data analytics and network integration, as well as R Studio and Python. The network will include up to 15 health facilities in Rwanda, whose EHR data will be harmonized to OMOP CDM. EXPECTED RESULTS: This study will yield a technical infrastructure where the 15 participating hospitals and health centres will have EHR data in OMOP CDM format on a local Mac Mini ("data node"), together with a set of OHDSI open-source tools. A central server, or portal, will contain a data catalogue of participating sites, as well as the OHDSI tools that are used to define and manage distributed studies. The central server will also integrate the information from the national Covid-19 registry, as well as the results of the community surveys. The ultimate project outcome is the dynamic prediction modelling for COVID-19 pandemic in Rwanda. DISCUSSION: The project is the first on the African continent leveraging AI and implementation of an OMOP CDM based federated data network for data harmonization. Such infrastructure is scalable for other pandemics monitoring, outcomes predictions, and tailored response planning.

THE EFFECT OF PREGNANCY SPACING ON FETAL SURVIVAL AND NEONATAL MORTALITY IN RWANDA: A HECKMAN SELECTION ANALYSIS.

Most studies on birth intervals and infant mortality ignore pregnancies that do not result in live births. Yet, fetal deaths are important in infant mortality analyses for three reasons: ignoring fetal deaths between two live births lengthens the measured interval between births, implying that short intervals are underestimated; the recommended inter-pregnancy interval (IPI) after a fetal loss is shorter (6 months) than after a live birth (24 months), as the effect of IPI on outcomes might differ according to the previous type of pregnancy outcome; fetal death will selectively reduce the population at risk of neonatal mortality, leading to biased results. This study uses the Heckman selection model to simultaneously estimate the combined effect of IPI duration and the type of pregnancy outcome at the start of the interval on pregnancy survival and neonatal mortality. The analysis is based on retrospective data from the Rwanda Demographic Health Surveys of 2000, 2005 and 2010. The results show a significant selection effect. After controlling for the selection bias, short (60 months) intervals after a fetal death reduce the chances of pregnancy survival, but no longer have an effect on neonatal mortality. For intervals starting with a live birth, the reverse is true. Short intervals (<24 months) do not affect pregnancy survival but increase the odds of neonatal mortality. If the previous child died in infancy, the highest odds are found for neonatal death regardless of the IPI duration.

Inter-Pregnancy Intervals and Maternal Morbidity: New Evidence from Rwanda.

The effects of short and long pregnancy intervals on maternal morbidity have hardly been investigated. This research analyses these effects using logistic regression in two steps. First, data from the Rwanda Demographic and Health Survey 2010 are used to study delivery referrals to District hospitals. Second, Kibagabaga District Hospital's maternity records are used to study the effect of inter-pregnancy intervals on maternal morbidity. The results show that both short and long intervals lead to higher odds of being referred because of pregnancy or delivery complications. Once admitted, short intervals were not associated with higher levels of maternal morbidity. Long intervals are associated with higher risks of third trimester bleeding, premature rupture of membrane and lower limb edema, while a higher age at conception is associated with lower risks. Poor women from rural areas and with limited health insurance are less often admitted to a hospital, which might bias the results.

The role of national nutrition programs on stunting reduction in Rwanda using machine learning classifiers: a retrospective study.

BACKGROUND: In Rwanda, the prevalence of childhood stunting has slightly decreased over the past five years, from 38% in 2015 to about 33% in 2020. It is evident whether Rwanda's multi-sectorial approach to reducing child stunting is consistent with the available scientific knowledge. The study was to examine the benefits of national nutrition programs on stunting reduction under two years in Rwanda using machine learning classifiers. METHODS: Data from the Rwanda DHS 2015-2020, MEIS and LODA household survey were used. By evaluating the best method for predicting the stunting reduction status of children under two years old, the five machine learning algorithms were modelled: Support Vector Machine, Logistic Regression, K-Near Neighbor, Random Forest, and Decision Tree. The study estimated the hazard ratio for the Cox Proportional Hazard Model and drew the Kaplan-Meier curve to compare the survivor risk of being stunted between program beneficiaries and non-beneficiaries. Logistic regression was used to identify the nutrition programs related to stunting reduction. Precision, recall, F1 score, accuracy, and Area under the Curve (AUC) are the metrics that were used to evaluate each classifier's performance to find the best one. RESULTS: Based on the provided data, the study revealed that the early childhood development (ECD) program (p-value = 0.041), nutrition sensitive direct support (NSDS) program (p-value = 0.03), ubudehe category (p-value = 0.000), toilet facility (p-value = 0.000), antenatal care (ANC) 4 visits (p-value = 0.002), fortified blended food (FBF) program (p-value = 0.038) and vaccination (p-value = 0.04) were found to be significant predictors of stunting reduction among under two children in Rwanda. Additionally, beneficiaries of early childhood development (p < .0001), nutrition sensitive direct support (p = 0.0055), antenatal care (p = 0.0343), Fortified Blended Food (p = 0.0136) and vaccination (p = 0.0355) had a lower risk of stunting than non-beneficiaries. Finally, Random Forest performed better than other classifiers, with precision scores of 83.7%, recall scores of 90.7%, F1 scores of 87.1%, accuracy scores of 83.9%, and AUC scores of 82.4%. CONCLUSION: The early childhood development (ECD) program, receiving the nutrition sensitive direct support (NSDS) program, focusing on households with the lowest wealth quintile (ubudehe category), sanitation facilities, visiting health care providers four times, receiving fortified blended food (FBF), and receiving all necessary vaccines are what determine the stunting reduction under two among the 17 districts of Rwanda. Finally, when compared to other models, Random Forest was shown to be the best machine learning (ML) classifier. Random forest is the best classifier for predicting the stunting reduction status of children under two years old.

Estimation of the Population Size of Street- and Venue-Based Female Sex Workers and Sexually Exploited Minors in Rwanda in 2022: 3-Source Capture-Recapture.

BACKGROUND: HIV surveillance among key populations is a priority in all epidemic settings. Female sex workers (FSWs) globally as well as in Rwanda are disproportionately affected by the HIV epidemic; hence, the Rwanda HIV and AIDS National Strategic Plan (2018-2024) has adopted regular surveillance of population size estimation (PSE) of FSWs every 2-3 years. OBJECTIVE: We aimed at estimating, for the fourth time, the population size of street- and venue-based FSWs and sexually exploited minors aged ≥15 years in Rwanda. METHODS: In August 2022, the 3-source capture-recapture method was used to estimate the population size of FSWs and sexually exploited minors in Rwanda. The field work took 3 weeks to complete, with each capture occasion lasting for a week. The sample size for each capture was calculated using shinyrecap with inputs drawn from previously conducted estimation exercises. In each capture round, a stratified multistage sampling process was used, with administrative provinces as strata and FSW hotspots as the primary sampling unit. Different unique objects were distributed to FSWs in each capture round; acceptance of the unique object was marked as successful capture. Sampled FSWs for the subsequent capture occasions were asked if they had received the previously distributed unique object in order to determine recaptures. Statistical analysis was performed in R (version 4.0.5), and Bayesian Model Averaging was performed to produce the final PSE with a 95% credibility set (CS). RESULTS: We sampled 1766, 1848, and 1865 FSWs and sexually exploited minors in each capture round. There were 169 recaptures strictly between captures 1 and 2, 210 recaptures exclusively between captures 2 and 3, and 65 recaptures between captures 1 and 3 only. In all 3 captures, 61 FSWs were captured. The median PSE of street- and venue-based FSWs and sexually exploited minors in Rwanda was 37,647 (95% CS 31,873-43,354), corresponding to 1.1% (95% CI 0.9%-1.3%) of the total adult females in the general population. Relative to the adult females in the general population, the western and northern provinces ranked first and second with a higher concentration of FSWs, respectively. The cities of Kigali and eastern province ranked third and fourth, respectively. The southern province was identified as having a low concentration of FSWs. CONCLUSIONS: We provide, for the first time, both the national and provincial level population size estimate of street- and venue-based FSWs in Rwanda. Compared with the previous 2 rounds of FSW PSEs at the national level, we observed differences in the street- and venue-based FSW population size in Rwanda. Our study might not have considered FSWs who do not want anyone to know they are FSWs due to several reasons, leading to a possible underestimation of the true PSE.

A cross-sectional analysis of the impact of health insurance on the use of health care in Rwanda.

Although the policy in Rwanda aims at ensuring quality healthcare, a portion of the Rwandan population still does not have access to it due to the lack of health insurance. This study investigates the impact of health insurance on healthcare utilization in all 30 administrative districts of Rwanda, using secondary data from the 5th Integrated Household Living Conditions Survey (EICV 5) in Rwanda, with a total of 14,580 households. A logistic regression model was used to evaluate the effects of health insurance on healthcare utilization, and a decision tree model was adopted to categorize districts based on the use of health services. This study has made a novel contribution to the existing research by classifying districts based on similarities in the use of health care services, regarding households with or without health insurance. The results showed a significant age effect on the use of health care services for household heads with an age range of 56-65, a significant increase was observed with an adjusted odds ratio of AO = 1.308, (95% CI: 1.044-1.639). It was the same for the household heads whose age range is 66-75 with an adjusted odds ratio of A0 = 1.589 with (95% CI: 1.244-2.028) and those aged 76 and older with an adjusted odds ratio of AO = 1.524, with (95% CI: 1.170-1.985). Households with health insurance interacted with districts (A0 = 2.76) increased health service use threefold compared to households without health insurance, female-headed households increased health service use (AO = 1.423, 95% CI:1.293-1.566) 1.4-fold compared to male-headed households, while households in the third quintile (AO = 1.198, 95% CI: 1.035-1.385) used health services 1.2 times compared to those in the first quintile; households in the fourth quintile (AO = 1.307, 95% CI: 1.134-1.506) and in the fifth quintile (AO = 1.307, 95% CI: 1.136 1.504) used health services 1.3 times compared to those in the first quintile. Similarly, for the households located in the main district group 4 variable had an odds ratio of 1.386 with (95% CI: 1.242-1.547), indicating that the households located in the main district group 4 use the health care services 1.4 times higher compared to those located in Ruhango district. Households in Rwanda who lack health insurance do not utilize health services to their full capacity, which has a negative influence on the wellbeing of the country's population. The researchers recommend that future policies target households in rural areas with an elderly head of household and those without health insurance that have a low usage of health care services in Rwanda. They also recommend that health insurance fees are reduced in order to increase health coverage rate as recommended by the World Health Organization.

Prediction of Stunting Among Under-5 Children in Rwanda Using Machine Learning Techniques.

OBJECTIVES: Rwanda reported a stunting rate of 33% in 2020, decreasing from 38% in 2015; however, stunting remains an issue. Globally, child deaths from malnutrition stand at 45%. The best options for the early detection and treatment of stunting should be made a community policy priority, and health services remain an issue. Hence, this research aimed to develop a model for predicting stunting in Rwandan children. METHODS: The Rwanda Demographic and Health Survey 2019-2020 was used as secondary data. Stratified 10-fold cross-validation was used, and different machine learning classifiers were trained to predict stunting status. The prediction models were compared using different metrics, and the best model was chosen. RESULTS: The best model was developed with the gradient boosting classifier algorithm, with a training accuracy of 80.49% based on the performance indicators of several models. Based on a confusion matrix, the test accuracy, sensitivity, specificity, and F1 were calculated, yielding the model's ability to classify stunting cases correctly at 79.33%, identify stunted children accurately at 72.51%, and categorize non-stunted children correctly at 94.49%, with an area under the curve of 0.89. The model found that the mother's height, television, the child's age, province, mother's education, birth weight, and childbirth size were the most important predictors of stunting status. CONCLUSIONS: Therefore, machine-learning techniques may be used in Rwanda to construct an accurate model that can detect the early stages of stunting and offer the best predictive attributes to help prevent and control stunting in under five Rwandan children.

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.

Population Size Estimation of Men Who Have Sex With Men in Rwanda: Three-Source Capture-Recapture Method.

BACKGROUND: Globally, men who have sex with men (MSM) continue to bear a disproportionately high burden of HIV infection. Rwanda experiences a mixed HIV epidemic, which is generalized in the adult population, with aspects of a concentrated epidemic among certain key populations at higher risk of HIV infection, including MSM. Limited data exist to estimate the population size of MSM at a national scale; hence, an important piece is missing in determining the denominators to use in estimates for policy makers, program managers, and planners to effectively monitor HIV epidemic control. OBJECTIVE: The aims of this study were to provide the first national population size estimate (PSE) and geographic distribution of MSM in Rwanda. METHODS: Between October and December 2021, a three-source capture-recapture method was used to estimate the MSM population size in Rwanda. Unique objects were distributed to MSM through their networks (first capture), who were then tagged according to MSM-friendly service provision (second capture), and a respondent-driven sampling survey was used as the third capture. Capture histories were aggregated in a 2k-1 contingency table, where k indicates the number of capture occasions and "1" and "0" indicate captured and not captured, respectively. Statistical analysis was performed in R (version 4.0.5) and the Bayesian nonparametric latent-class capture-recapture package was used to produce the final PSE with 95% credibility sets (CS). RESULTS: We sampled 2465, 1314, and 2211 MSM in capture one, two, and three, respectively. There were 721 recaptures between captures one and two, 415 recaptures between captures two and three, and 422 recaptures between captures one and three. There were 210 MSM captured in all three captures. The total estimated population size of MSM above 18 years old in Rwanda was 18,100 (95% CS 11,300-29,700), corresponding to 0.70% (95% CI 0.4%-1.1%) of total adult males. Most MSM reside in the city of Kigali (7842, 95% CS 4587-13,153), followed by the Western province (2469, 95% CS 1994-3518), Northern province (2375, 95% CS 842-4239), Eastern province (2287, 95% CS 1927-3014), and Southern province (2109, 95% CS 1681-3418). CONCLUSIONS: Our study provides, for the first time, a PSE of MSM aged 18 years or older in Rwanda. MSM are concentrated in the city of Kigali and are almost evenly distributed across the other 4 provinces. The national proportion estimate bounds of MSM out of the total adult males includes the World Health Organization's minimum recommended proportion (at least 1.0%) based on 2012 census population projections for 2021. These results will inform denominators to be used for estimating service coverage and fill existing information gaps to enable policy makers and planners to monitor the HIV epidemic among MSM nationally. There is an opportunity for conducting small-area MSM PSEs for subnational-level HIV treatment and prevention interventions.

Prediction of out-of-pocket health expenditures in Rwanda using machine learning techniques.

INTRODUCTION: in Rwanda, the estimated out-of-pocket health expenditure has been increased from 24.46% in 2000 to 26% in 2015. Despite the existence of guideline in estimation of out-of-pocket health expenditures provided by WHO (2018), the estimation of out-of-pocket health expenditure still have difficulties in many countries including Rwanda. METHODS: the purpose of this paper was to figure out the best model which predicts the out-of-pocket health expenditures in Rwanda during the process of considering various techniques of machine learning by using the Rwanda Integrated Living Conditions Surveys (EICV5) of 14580 households (2018). RESULTS: our findings presented the model which predict the out-of-pocket health expenditures with higher accuracy and was found as treenet model. Furthermore, machine learning techniques were used to judge which predictor variable was important in our prediction process and comparison of the performance of the algorithms through train accuracy and test accuracy metric measures. Finally, the findings show that the tests of accuracy of the models were 50.16% for multivariate adaptive regression splines (MARS) model, 74% decision tree model, 87% for treenet model, 83% for random forest model, gradient boosting 81%, predictor total consumption played a significant role in the model for all tested models. CONCLUSION: finally, we conclude that the total consumption of the household came out to be the most important variable which is consistently true to all the algorithms tested. The findings from our study have policy implications for policy makers in Rwanda and in the world generally. We recommend the government to significantly increase public spending on health. Domestic financial resources are key to moving closer to universal health coverage (UHC) and should be increased on a long-term basis. In addition, these results will be useful for the future to assess the out-of-pocket health expenditures dataset.

The Effects of Interpregnancy Intervals and Previous Pregnancy Outcome on Fetal Loss in Rwanda (1996-2010).

In 2005, a WHO consultation meeting on pregnancy intervals recommended a minimum interval of 6 months after a pregnancy disruption and an interval of two years after a live birth before attempting another pregnancy. Since then, studies have found contradictory evidence on the effect of shorter intervals after a pregnancy disruption. A binary regression analysis on 21532 last pregnancy outcomes from the 2000, 2005, and 2010 Rwanda Demographic and Health Surveys was done to assess the combined effects of the preceding pregnancy outcome and the interpregnancy intervals (IPIs) on fetal mortality in Rwanda. Risks of pregnancy loss are higher for primigravida and for mothers who lost the previous pregnancy and conceived again within 24 months. After a live birth, interpregnancy intervals less than two years do not increase the risk of a pregnancy loss. This study also confirms higher risks of fetal death when IPIs are beyond 5 years. An IPI of longer than 12 months after a fetal death is recommended in Rwanda. Particular attention needs to be directed to postpregnancy abortion care and family planning programs geared to spacing pregnancies should also include spacing after a fetal death.