Menu
Menu
Menu
Menu
Background: As Sierra Leone celebrates the end of the Ebola Virus Disease (EVD) outbreak, we can begin to fully grasp its impact on already weak health systems. The EVD outbreak in West Africa forced many hospitals to close down or reduce their activity, either to prevent nosocomial transmission or because of staff shortages. The aim of this study is to assess the potential impact of EVD on nationwide access to obstetric care in Sierra Leone. Methods and Findings: Community health officers collected weekly data between January 2014 – May 2015 on inhospital deliveries and caesarean sections (C-sections) from all open facilities (public, private for-profit and private non-profit sectors) offering emergency obstetrics in Sierra Leone. This was compared to official data of EVD cases per district. Logistic and Poisson regression analyses were used to compute risk and rate estimates. Nationwide, the number of inhospital deliveries and C-sections decreased by over 20% during the EVD outbreak. The decline occurred early on in the EVD outbreak and was mainly attributable to the closing of private not-for-profit hospitals rather than government facilities. Due to difficulties in collecting data in the midst of an epidemic, limitations of this study include some missing data points. Conclusions: Both the number of in-hospital deliveries and C-sections substantially declined shortly after the onset of the EVD outbreak. Since access to emergency obstetric care, like C-sections, is associated with decreased maternal mortality, many women are likely to have died due to the reduced access to appropriate care during childbirth. Future research on indirect health effects of health system breakdown should ideally be nationwide and continue also into the recovery phase. It is also important to understand the mechanisms behind the deterioration so that important health services can be reestablished.
This study was conducted as part of an on-going collaboration between the Ministry of Health and Sanitation (MoHS) in Sierra Leone, Karolinska Institutet in Sweden, the Norwegian University of Science and Technology, and the non-governmental organization CapaCare, and constitutes part of a new surveillance initiative to monitor the effects of the Ebola epidemic on health services [12]. The director of Research and Non-Communicable Diseases and the Director of Hospitals and Laboratory Services of the MoHS approved the study. Since data was collected retrospectively from operation theatre-, delivery-, and admission logbooks, patient consent was not possible to obtain. Therefore, all information concerning individual patients was anonymized and de-identified prior to analysis. A countrywide study in 2013 systematically mapped all 61 governmental, private, not- and for-profit healthcare facilities that offer in-patient care and major surgery [2]. For our study, all these facilities were surveyed since September 2014, by 21 Community Health Officers (CHOs) (details described previously [12]). The CHOs were on leave from a surgical task sharing training program, due to restrictions on clinical training during the EVD outbreak. To minimize travel and potential exposure to EVD, most data collectors lived nearby, or had recently worked or practised in the facilities they regularly visited. During the first facility visit in the end of September 2014, retrospective data was retrieved for the first 38 weeks of the year, and thereafter on a bi-weekly basis until end of May 2015. Weekly accumulated numbers on deliveries and C-sections was collected from readily available operation theatre-, and admission logbooks. The data collectors were trained for one full day and coordinated locally by a final year medical student, also on leave due to the EVD epidemic. Data was captured on tablets (Huawei Mediapad 7 with data SIM connectivity) using the District Health Information System 2 (DHIS 2) software, designed for collection, validation and analysis of aggregated health data. Data was transferred via a secured Internet connection to a central cloud server and monthly validated. Of visited facilities, 32 performed at least 5 deliveries and/or C-sections during the study period (week 1, 2014 to Week 20, 2015) and were included in our study. When this criterion was met, facilities did not need to be open consistently during all three periods, in fact, many of the included facilities closed after the onset of the EVD outbreak. A list of all hospitals included in this study can be found in S1 Appendix. Data on facility status as well as weekly numbers of deliveries and C-sections can be found in S2 Appendix. Data on weekly number of EVD cases per district was retrieved from the MoHS in Sierra Leone and the WHO [18,19]. Since the first EVD case in Sierra Leone was reported in late May 2014, we define the pre-outbreak period as week 1 to 21, 2014 (period 1) and divide the post-outbreak into two separate periods; week 22 to 52, 2014 (period 2; outbreak peak) and week 1 to 20, 2015 (period 3; outbreak slow-down). To understand expected seasonal variations in C-sections, we compared our 2014 and 2015 numbers with those from 2012. Data on the number of C-sections was collected retrospectively by 12 local medical students in 2013 from operation, anaesthesia, and delivery logbooks and used as a reference for these comparisons [2]. We were not able to obtain 2012 data on the number of in-hospitals deliveries. In order to determine how the number of in-hospital deliveries and C-sections varied over time, we calculated the mean weekly number of deliveries and C-sections over the three time periods. The mean weekly incidence rate ratios were calculated with corresponding 95% confidence intervals (CI), using Poisson regression with the pre-outbreak period as the reference. These numbers are calculated for the whole nation, by province, and by type of facility. EVD incidence rate per 100 000 inhabitants was calculated using population data [8]. Logistic regression was used to compute 95% CI around the C-section proportions. The association between the number of deliveries and C-sections and the number of EVD cases was estimated using Poisson regression, treating the number of EVD cases as the predictor variable. The estimates are incidence rate ratios (IRR) for an increase of 100 EVD cases. To compensate for some missing observations on deliveries and C-sections, we used two different imputation methods: one where a missing value was interpreted as a zero, and one where the last observed value for the specific facility was used. The first method is the one presented in the paper. However, the difference between the two imputation methods was negligible, indicating that missing values did not influence the results in any significant way.
