Rwanda is an East African country with an estimated population of 10.5 million in 2012, projected to be around 12.5 million in 2019 (19, 20). The country is subdivided into 5 provinces (North, East, West and South, and Kigali City). Each province is subdivided into districts and there are 30 districts in total, in which malaria is highly endemic in 12 districts. Malaria interventions in Rwanda include core interventions such as 1) mosquito net distributions through annual campaigns and in routine services such as antenatal care and Expanded Program of Immunization, 2) IRS in high burden districts, and 3) expanded access to timely diagnostic and treatment services at health facilities and at community level, known as Home Based Management of Malaria. Supplemental interventions include behavior change communication to improve knowledge and awareness of community members, larviciding in some targeted areas using drones, and use of mosquito repellent products. (21).
The work presented here is part of a larger project to assess and mitigate the effects of the Covid-19 pandemic on Rwanda’s health system. We implemented the project from August–December 2020 in three districts: Rwamagana and Kayonza in Eastern Province, and Gasabo in Kigali city (Figure 1). The Rwanda Biomedical Center selected these districts purposively, as part of its framework for management of implementation partners, which allocates to each partner a specific geographic area of implementation for efficiency and to avoid duplication of activities between partners. Malaria is highly endemic in these three districts, and similar malaria interventions are implemented in each district. However, the two districts of the Eastern province (Rwamagana and Kayonza) receive indoor residual spraying as part of malaria prevention while the Gasabo district in Kigali city receives a new generation of mosquito nets, the effect of which is believed to be comparable to indoor residual spraying (22). The Rwandan health system is structured in a traditional hierarchical cascade as follows: community > health center> district hospital> referral and specialized hospitals. The public health facilities available in these districts are one district hospital and 15 health centers in the Rwamagana district, two district hospitals and 14 health centers in Kayonza district, and one district hospital, one referral hospital, one specialized hospital and 15 health centers in the Gasabo district.
In this mixed-methods study, we quantitatively assessed changes in usage of malaria services, comparing the malaria presentations in the pre-Covid-19 period and the malaria presentations after the first Covid-19 case was diagnosed in Rwanda in March 2020. We used 14 months of data (January 2019 to February 2020) as the pre-Covid-19 period and nine months of data as the Covid-19 period (from March 2020 to November 2020). This is because, during that Covid -19 period, strict public health and social measures to mitigate Covid-19 were implemented across the districts whereas, after November 2021, the government started varying measures according to the epidemiological situation in each district (5–11).
We also qualitatively explored the barriers community members faced in accessing malaria services during the period of lockdown, as well as the perspectives of Health Care Providers (HCPs), Community Health Workers (CHWs), and community members on health system readiness to ensure minimal interruption of the implementation of malaria interventions despite the lockdown.
Data sources and data collection procedures
We used both primary and secondary data in this study. The quantitative secondary data were aggregated data on malaria treatment reported in the Health Management Information System (HMIS) and the Systeme Informatique de Sante Communautaire (SIS-COM). Both systems are electronic nationwide databases used to report service delivery in Rwanda monthly. HMIS is used for reporting health facility data while SIS-COM is used to report CHW-related activities. The quality of data reported in both systems is high (23). The malaria data included the number of malaria tests conducted at health facilities and in the community separately, the number of uncomplicated malaria cases diagnosed in health facilities and the community, and the number of severe malaria cases and malaria deaths.
We collected qualitative data during August– September 2020, conducting focus group discussions with community members and CHWs, as well as in-depth interviews with key informants that included HCPs and staff working in the malaria program at the central level. The key informant interview participants were selected purposively based on their involvement in malaria service provision. We interviewed personnel working in the malaria program at the central level and district level, those working with CHWs at the district level, health facilities managers, and in charge of CHWs at the community level in selected health centers. For the focus group discussions, the CHWs and community members were selected by the health facilities managers according to their convenience to participate in the focus group discussions. We requested that health facility managers diversify participants and include both those living near the health facility and those living far from the health facility, to capture all experiences.
We used semi-structured interviews and focus group discussion guides covering key themes including whether the health system was sufficiently prepared to ensure continuous malaria activities during the period of Covid-19 mitigation measures, and the barriers to access malaria services by the community members during the period of lockdown. We used data collectors who were fluent in both English and Kinyarwanda (the local language). They received a 2-day training on data collection tools before the data collection. Both in-depth interviews and focus groups were conducted in the local language and were audio-recorded.
Data management and analysis
We extracted data from HMIS, and SIS-COM systems and exported them into Stata version 14.2 (StataCorp, College Station, Texas) for cleaning and to calculate malaria indicators. To measure the effect of Covid-19 on malaria service use, we used interrupted time series analysis, which is among quasi-experimental study designs with advantages of using pre-existing data, and account for previous trends in the outcomes, which are very frequent in malaria data (24–26).
For uncomplicated malaria, we calculated the rate per 1000 population by dividing the number of patients reported per month by the estimated total population for the three districts (Table 1). The district population was estimated using the medium population growth rates based on the 2012 census data as reported by the National Institute of Statistics of Rwanda (19). We used the population denominators for each year separately. The combined population for all districts was 1,397,038 and 1,429,640 for the years 2019 and year 2020 respectively. For severe malaria, we calculated the rate per 100,000 malaria cases by dividing the number of patients with severe malaria by the total malaria cases reported per month. Moreover, we also separately calculated the presentation rate for testing (suspected cases) per 1,000 population in the health facilities and the community by dividing the number of patients tested in the health facilities and the community by the projected population. For each indicator, we conducted interrupted time series analysis by fitting generalized least squares models to compare trends and levels between the pre-Covid-19 period and the Covid-19 period, considering autocorrelation lag. Autocorrelation was assessed by plotting autocorrelation and partial autocorrelation functions. The time-series analysis was conducted using R 4.0.2, the “nlme” and “car” packages (27, 28).
The thematic analysis included 12 focus group discussions (six conducted with CHWs, six with community members) and 22 interviews (20 conducted with HCPs and staff working in malaria services at district level, and two with staff from the malaria program at the central level). We used the constant comparative method for thematic analysis of the qualitative data (29, 30). Audio recordings were transcribed in Kinyarwanda. The research team members fluent in both English and Kinyarwanda first read 10 transcripts and inductively developed a codebook. We used the codebook to highlight key excerpts from the transcripts and modified it as new themes emerged from the data during the analysis process. We reported key themes and supplemented them with quotes from participants. The quotes used in reporting were translated into English by a professional translator and then verified by the principal investigators for accuracy. Qualitative analysis was conducted using Dedoose Version 4.12. We obtained approval from the Rwanda National Ethics Committee to conduct the study, and participants consented prior their involvement in the study.