A mixed-methods study of maternal health care utilisation in six referral hospitals in four sub-Saharan African countries before and during the COVID-19 pandemic

Introduction In sub-Saharan Africa, referral hospitals are important sources of key maternal health services, especially during a crisis such as the COVID-19 pandemic. This study prospectively assessed the effect of the COVID-19 pandemic on maternal health service utilisation in six large referral hospitals in Guinea, Nigeria, Tanzania and Uganda during the first year of the pandemic. Methods Mixed-methods design combining three data sources: (1) quantitative data based on routine antenatal, childbirth and postnatal care data collected March 2019-February 2021, (2) qualitative data from recurring rounds of semi-structured interviews conducted July 2020-February 2021 with 22 maternity skilled heath personnel exploring their perceptions of service utilisation and (3) timeline data of COVID-19 epidemiology, global, national and hospital-level events. Qualitative and quantitative data were analysed separately, framed based on the timeline analysis and triangulated when reporting. Results Three periods including a first wave, slow period and second wave were identified. Maternal health service utilisation was lower during the pandemic compared with the prepandemic year in all but one selected referral hospital. During the pandemic, service utilisation was particularly lower during the waves and higher or stable during the slow period. Fear of being infected in hospitals, lack of transportation, and even when available, high cost of transportation and service closures were key reasons affecting utilisation during the waves. However, community perception that the pandemic was over or insinuation by Government of the same appeared to stabilise use of referral hospitals for childbirth. Conclusion Utilisation of maternal health services across the continuum of care varied through the different periods and across countries. In crisis situations such as COVID-19, restrictions and service closures need to be implemented with consideration given to alternative options for women to access and use services. Information on measures put in place for safe hospital use should be communicated to women.

Using a mixed-methods study design, this study employed three data sources which includes (1) an analysis of the timeline of key events that occurred at global, national, subnational and intrafacility levels, (2) routine hospital data before and after the WHO declared the COVID-19 pandemic and (3) semi-structured key-informant interviews (KIIs). Based on guidelines for mixed-method study design by Creswell and Clark,18 all data were collected in parallel, analysed separately and triangulation of the findings at the synthesis stage allowed for an in-depth understanding of the situation in the six hospitals. The six hospitals were purposively selected, with emphasis placed on hospitals with large referral maternity wards in urban areas of different sub-Saharan African countries (two in East Africa, two in West Africa, including a francophone country) were selected. The participating hospitals were Hôpital National Ignace Deen/Ignace Deen National Hospital (HNID) and Hôpital Regional de Mamou/Mamou Regional Hospital (HRM) in Guinea, Lagos University Teaching Hospital (LUTH) in Nigeria, Muhimbili National Hospital (MNH) in Tanzania, Kawempe National Referral Hospital (KNRH) and Mulago Specialised Women’s and Neonatal Hospital (MSWNH) in Uganda; their profiles based on information collected from each hospital’s primary investigator (PI) are shown in table 1. Characteristics of the participating hospitals and maternity wards before the COVID-19 pandemic Initial data regarding key events were collected during the semi-structured KIIs conducted as part of this study. Insights from the interviews helped with establishing the time range of interest (1 January 2020 to 28 February 2021). Subsequently, a pro-forma tool was designed to capture events that could influence service provision and utilisation at the participating hospitals, and on national and global levels. At hospital-level, data collected included any periods of maternity services closure and significant modifications to service delivery that could alter utilisation. These data were completed by the country PIs and hospital co-PIs. National level events included periods of national lockdown(s), curfews and travel ban. These also included other key events that were aimed at or have an established potential to alter behaviour of maternity service users such as introducing subsidies for user fees, tax credits or cash schemes. These data were sourced from the Oxford COVID-19 Government Response Tracker, Blavatnik School of Government and University of Oxford.19 One-off events were distinguished from extended ones. National timelines were verified for accuracy and expanded with contextual detail by the country PIs and hospital co-PIs and through review of weblinks describing the national events. Numbers of cases and deaths due to COVID-19 at the national level were collected from ‘Our World in Data’ (https://ourworldindata.org/covid-deaths) and validated on the WHO COVID-19 dashboard (https://COVID-19.who.int/). Global events were sourced from the WHO’s COVID-19 response timeline (https://www.who.int/emergencies/diseases/novel-coronavirus-2019/interactive-timeline). Using epidemiological week cut-offs defined by Salyer et al,20 and information gathered by country PIs and hospital co-PIs, we divided the timeline for each country into three phases: first wave, slow period and second wave. This was based on weekly incidence of COVID-19 cases in the study countries. For Tanzania, which did not report epidemiological data after June 2020, we used periods observed in countries in proximity to it. National data on COVID-19 cases and deaths were entered into Microsoft Excel and presented as line charts; all events per country were mapped on a timeline visual. Monthly aggregated routine health statistics (from 1 January 2019 to 28 February 2021) were collected from each of the participating hospitals by clinical researchers in collaboration with hospital-based clinicians and data clerks between 1 June 2020 and 28 March 2021. We analysed three routine indicators which represent main aspects of maternal care utilisation: number of outpatient antenatal care (ANC) visits, number of births and number of outpatient postnatal care (PNC) visits. A detailed list of routine statistics indicators and their definitions is included in online supplemental material 1. The aggregate routine data used for calculation of these indicators were extracted from multiple sources within each hospital (eg, labour ward registers, medical records, health management information system, etc) and the number of sources ranged between two and four per hospital (online supplemental material 1). When multiple data sources for the same indicator were available, data were collected from all sources and validated against each other. In case of discrepancy, the researchers included the numbers from the most reliable source according to the hospital PIs and data clerks. bmjgh-2021-008064supp001.pdf Aggregated monthly data were inputted in a standardised Microsoft Excel sheet. In collaboration with the researchers from each country, data review and verification were conducted a minimum of two times for each hospital. This was done to allow for detection of missing values and outliers in each of the selected hospitals. We conducted descriptive analysis of each indicator for a period of 24 months, divided into two 12-month time periods representing a year before the pandemic was declared (from March 2019 to February 2020 labelled as pre-COVID-19) and a year afterwards (from March 2020 to February 2021 labelled as during COVID-19). Frequencies were displayed in line charts. Indicator values for the two time periods were compared, matched with other key events and triangulated against findings from semi-structured interviews. We conducted repeated semi-structured interviews with SHPs who were only included in the study if they principally worked in maternity. We adopted a purposive sampling of key informants to ensure maximum variation in the experiences of SHP of varying seniority levels (junior and senior staff) and cadres (medical doctors, midwives and nurses). Potential participants were first approached by the hospital PIs; if they agreed to be interviewed, they were compensated for their time and use of mobile data. Data were collected over one to four rounds of interviews conducted between July 2020 and February 2021. We interviewed two to six maternity SHP in each participating hospital per round. At LUTH (Nigeria) and MNH (Tanzania), we interviewed four respondents each. At KNRH (Uganda) and MSWNH (Uganda) we interviewed two people each. In HNID (Guinea) we interviewed six people and in HRM (Guinea), we interviewed four key informants. In total 22 SHPs were interviewed, and 50 interviews took place. We used a semi-structured interview guide to comprehensively capture information related to changes in the processes of care utilisation across all hospitals and time-points. The content of this guide was developed to record and understand perceptions of respondents on shifts in maternity case volumes, as well as any observations on influence of COVID-19 measures on service utilisation. Interviews lasting between 20 and 120 min were conducted by two researchers virtually using Zoom (Zoom Video Communications, San Jose, California, USA) for Nigeria, Tanzania and Uganda (LB), and face-to-face in Guinea (ND). All interviews were recorded and transcribed in the language of the interview (English or French), de-identified, and imported into the computer-assisted qualitative data analysis software Dedoose. Analysis was an iterative process which was done concurrently with data collection. This approach allowed the researchers to adapt the interview guide in the repeated interview rounds based on respondents’ answers to the previous rounds and to the country situation. Data analysis was conducted using the framework method.21 Following familiarisation with the data by re-listening to the recordings and reading the transcripts which was done by three researchers (ND, AS and LB), coding of the first six interviews was independently done, from which emerging codes were identified (inductive) keeping the structure of the interview guide in mind (deductive). A coding tree was subsequently developed and systematically applied to the interviews by one researcher and checked by another. The themes and examples emerging from the interviews were mapped on to a matrix by period (first wave, slow period and second wave).20 Emerging themes were further summarised to capture similarities and differences across the three periods and six hospitals, and to identify relationships between the main themes in the data. The triangulation and synthesis of data from the three data sources were conducted in an iterative, prospective and collaborative manner, first by sharing and discussing findings during 21 biweekly research team meetings between May 2020 and March 2021, and on completion of all data collection in April–July 2021. Time trends observed in the routine data indicators were compared against findings from the key event analysis and the qualitative data, including perceptions of SHPs on service utilisation, to identify and discuss intersections between all three data sources. We present the findings for each period using all three data sources. Patients and/or the public were not involved in the design, conduct, reporting, or dissemination of this research.

N/A

• Dima Health