Abortion in Zimbabwe: A national study of the incidence of induced abortion, unintended pregnancy and post-abortion care in 2016

Background: Zimbabwe has the highest contraceptive prevalence rate in sub-Saharan Africa, but also one of the highest maternal mortality ratios in the world. Little is known, however, about the incidence of abortion and post-abortion care (PAC) in Zimbabwe. Access to legal abortion is rare, and limited to circumstances of rape, incest, fetal impairment, or to save the woman’s life. Objectives: This paper estimates a) the national provision of PAC, b) the first-ever national incidence of induced abortion in Zimbabwe, and c) the proportion of pregnancies that are unintended. Methods: We use the Abortion Incidence Complications Method (AICM), which indirectly estimates the incidence of induced abortion by obtaining a national estimate of PAC cases, and then estimates what proportion of all induced abortions in the country would result in women receiving PAC. Three national surveys were conducted in 2016: a census of health facilities with PAC capacity (n = 227), a prospective survey of women seeking abortion-related care in a nationally-representative sample of those facilities (n = 127 facilities), and a purposive sample of experts knowledgeable about abortion in Zimbabwe (n = 118). The estimate of induced abortion, along with census and Demographic Health Survey data was used to estimate unintended pregnancy. Results: There were an estimated 25,245 PAC patients treated in Zimbabwe in 2016, but there were critical gaps in their care, including stock-outs of essential PAC medicines at half of facilities. Approximately 66,847 induced abortions (uncertainty interval (UI): 54,000-86,171) occurred in Zimbabwe in 2016, which translates to a national rate of 17.8 (UI: 14.4-22.9) abortions per 1,000 women 15-49. Overall, 40% of pregnancies were unintended in 2016, and onequarter of all unintended pregnancies ended in abortion. Conclusion: Zimbabwe has one of the lowest abortion rates in sub-Saharan Africa, likely due to high rates of contraceptive use. There are gaps in the health care system affecting the provision of quality PAC, potentially due to the prolonged economic crisis. These findings can inform and improve policies and programs addressing unsafe abortion and PAC in Zimbabwe. This paper uses the Abortion Incidence Complications Method (AICM) [13] to estimate the incidence of induced abortion in Zimbabwe. The AICM has been widely applied in over 25 countries worldwide, including 10 in sub-Saharan Africa [14–27]. The methodology indirectly estimates the incidence of abortion by obtaining a national estimate of the number of PAC cases treated in facilities, and then estimating what proportion of all abortions in the country would result in women receiving PAC. The AICM includes a Health Facilities Survey (HFS) consisting of interviews with PAC service providers and a Health Professionals Survey (HPS) with purposively-selected key informants who are knowledgeable about abortion. This study also includes a Prospective Morbidity Survey (PMS), which is a nationally representative, facility-based study that collects prospective data from PAC patients to measure PAC service provision at the facility-level. These three surveys were conducted from August to November 2016. Ethical approval was obtained from the Medical Research Council of Zimbabwe, the Joint Research Ethics Committee for the University of Zimbabwe, College of Health Sciences, and the Parirenyatwa Group of Hospitals and the Guttmacher Institute’s Institutional Review Board. The HFS gathers retrospective data on the number of PAC cases treated annually. All public, private and non-governmental organization (NGO) health facilities that have the capacity to provide PAC were eligible to participate in the survey. Health facilities in Zimbabwe are capable of providing PAC if they have an operating theatre or if staff have been trained to use misoprostol for PAC [12]. We obtained lists of all known health facilities from the MoHCC, the Health Professionals Authority (where all public and private facilities are required to register), the Private Hospitals Association of Zimbabwe, the Association of Health Funders of Zimbabwe, and Population Services-Zimbabwe, a local NGO. We excluded specialized facilities unrelated to PAC, as well as duplicate facilities, individual doctors, and facilities that lacked the capacity to provide PAC. According to the National Guidelines for Comprehensive Abortion Care in Zimbabwe, the lowest level of public health facilities (primary health centers) can only provide PAC using misoprostol if they have trained providers [12]. Only 63 primary health centers meet these criteria, having been included in a 2013 operations research study conducted by MoHCC and Venture Strategies Innovations [11]. Based on the inclusion criteria, we identified 245 facilities in Zimbabwe with the capacity to provide PAC (Table 1). During fielding, 7% of facilities on the list were not eligible for the survey. There was a 99.6% response rate in all eligible facilities. Overall, 227 facilities participated in the HFS (Table 1). a Eligible is defined as a facility having the capacity to provide post-abortion care (PAC). Seven percent of facilities in the HFS were not eligible, this included duplicates (n = 3), closed or reclassified facilities (n = 6), and facilities without PAC capacity or not providing PAC (n = 8). b Facilities were added to the PMS to account for misclassification by province. We included additional facilities to have 100% of facilities in that province at the level where misclassification occurred. This included one private hospital, and six NGO facilities. Seven percent of facilities in the PMS were not eligible. c For the PMS, primary health centers, district and mission hospitals, and provincial hospitals had a 100% response rate. The lowest response rate was among NGO facilities at 82% responding, and private hospitals had a 96% response rate. d Ministry of Health and Child Care (MoHCC) monitoring and evaluation (M&E) data includes public facilities that do not have PAC capacity and therefore were not in the sampled universe. However, they still get PAC patients and report these numbers to the MoHCC, so we included out-of-sample MoHCC facilities to calculate PAC caseloads in Zimbabwe. One respondent who had worked at the facility for at least six months was selected at each participating facility. HFS respondents were selected based on their knowledge of the facility’s PAC services; most often, this was a nurse/midwife (68%), senior nurse (27%), or doctor/hospital administrator (5%). Written informed consent and face-to-face structured questionnaires were administered to HFS respondents by Provincial Reproductive Health Officers (RHOs), who are trained health professionals working for the MoHCC. Information was collected on PAC caseloads, facility infrastructure, functionality and availability of equipment and medicines, and PAC procedures used. (S1 File). The PMS was conducted in a nationally representative sample of facilities with PAC capacity, using the HFS universe as the sampling frame. Facilities were stratified by level, ownership and province. The PMS sample included 56% of facilities in Zimbabwe with PAC capacity: all central hospitals (n = 5), provincial hospitals (n = 8), and higher-level private facilities (n = 19) were included due to their anticipated large caseloads; sampling proportions were smaller for lower-level facilities (Table 1). Overall, 127 facilities took part in the PMS, which corresponds to a facility response rate of 95% (Table 1). At the 127 participating facilities, data was collected on all women presenting at these facilities for PAC during the 28-day study period. Questionnaires administered to both patients and their providers were used to capture a count of the number of women who came in for PAC (S2 File). Verbal informed consent was obtained from the woman for both her interview, as well as for the interview with her provider. Interviewers provided and read to women the consent form, after which interviewers indicated on the form whether consent was obtained, and verified this by providing their own signature, name and date. Parental or guardian consent was not obtained for minors as pregnant (or previously pregnant) minors are considered emancipated in Zimbabwe. The three ethics committees approved this informed consent procedure. Facility-level aggregate information, rather than individual-level responses, were used to calculate PAC caseload totals for this analysis. A number of steps were taken to minimize missed PAC patients from the PMS caseload total. Interviewers were given tracking forms to complete each day where they recorded the number of completed interviews, the number of women with post-abortion complications so severe that they were unable to consent to participate in the survey, the number of maternal deaths, and the number of cases they missed. Interviewers were trained to consult with colleagues across multiple departments within the facilities in order to keep track of any PAC cases presenting in the facility at any time over the 28-day period. The HPS collected information from experts on abortion, abortion-related complications, and access to PAC in Zimbabwe. The study team developed a purposive list of 125 key informants, covering all provinces, which included doctors, nurses, midwives, policy makers, lawyers, pharmacists, traditional health workers, village chiefs, NGO staff, and other individuals well placed to know about the behaviors, outcomes and treatment of women seeking abortions. An effort was made to have sufficient representation of experts with knowledge of rural areas. Three-quarters of the sample were health providers, and the remaining quarter were selected for their knowledge of women’s health issues at the community level. Respondents to the HPS were on average 42 years old, with 14 years of work experience and 60% were female. Six HPS respondents were excluded from the final dataset as they had been incorrectly selected for the survey. These were respondents who struggled to answer the questions, and were assessed as having low knowledge on the topic by both the interviewer and themselves. Only one selected respondent refused to participate in the survey. The final sample size for the HPS was 118. Through face-to-face structured interviews conducted by four highly skilled professionals, respondents were asked about from whom women seek abortions, the likelihood of women experiencing complications, the proportion of women with complications that would receive treatment and about Zimbabwean women seeking abortions outside of the country (S3 File). We designed a visual aid to facilitate data collection for the HPS, due to the complex nature and nesting of questions (S4 File). Given that respondents may not have had enough knowledge to answer all the questions, we also asked respondents how certain they were (on a scale of one to ten) of their answers for abortion in urban and rural areas, and for misoprostol use in urban and rural areas (a more recently introduced abortion method in the country). To calculate abortions rates, we used data on the population of reproductive age women in 2016, nationally and by province, from the Zimbabwe National Statistics Agency Population Projections Report based on the 2012 National Census [28]. We calculated the national and provincial number of births, from age-specific fertility rates from the 2015 Zimbabwe Demographic and Health Survey (ZDHS) [2] applied to the female population by age from the Zimbabwe National Statistics Agency (ZNSA) [28]. To determine the distribution of women based on wealth and place of residence, the Poverty and Poverty Datum Line Analysis Report from 2012 produced by the ZNSA provided the wealth distribution [29] and the ZDHS 2015 provided the urban and rural distribution [2]. The ZDHS was used to estimate unintended and intended births to calculate the total number of unintended pregnancies [2]. The HFS and PMS surveys were designed to measure the number of women arriving at a facility seeking PAC due to complications from both miscarriages and induced abortions. The HPS measures the likelihood that women having complications from induced abortions would receive treatment. Abortion complications were defined in all three surveys as not only the extremely serious cases such as those with sepsis or a perforated uterus, but also incomplete abortions, which are usually identified by heavy bleeding and present a somewhat less severe health risk to the woman but still requires treatment in a health facility. Respondents were instructed that abortion complications do not include the expected bleeding associated with a miscarriage or induced abortion that would resolve itself on its own, and does not require treatment. However, in a clinical setting some of these cases may be indistinguishable; incomplete abortions that would resolve on their own may still receive PAC alongside incomplete abortions that do require treatment to be resolved. There is likely measurement error in respondent’s assessment of abortion complications, and PAC cases enumerated by respondents probably include some induced abortions and miscarriages that did not in fact require treatment but nonetheless received it. We therefore refer to estimates from the HFS and PMS as PAC cases, to acknowledge that not all women receiving PAC have abortion complications. The following details each step of the AICM. Mathematical Appendix A in S1 Appendix provides the detailed equations for the AICM calculations, and Fig 1 diagrams the methodology’s key components and data sources. We calculated abortion incidence nationally and by geographic region. We did not calculate provincial-level estimates as women do not always receive treatment in the provinces in which they reside. Instead, we constructed three geographical regions that included the provinces with higher-level facilities that are likely to treat women from the surrounding provinces. These three regions were Matabeleland (Bulawayo, Matabeleland North, Matabeleland South, Midlands), Mashonaland (Harare, Mashonaland East, Mashonaland West, Mashonaland Central) and South Eastern Region (Manicaland and Masvingo). All analyses were conducted in Stata 14.2 and weighted for non-response. a Abortion methods include surgical methods, misoprostol, and other methods of abortion. b Sub-groups include rural poor women, rural non-poor women, urban poor women, urban non-poor women. The HFS and PMS obtained estimates of PAC patients (inpatient and outpatient) treated for either miscarriage or induced abortion. HFS respondents were asked about PAC cases in the past month and a typical month to account for possible seasonality of obstetric care and minimize reporting errors. Respondents who had difficulty reporting in the month time frame (this usually happened when the caseloads were very small) could provide the information for a typical year and the past year. PAC cases were estimated in the PMS using prospective estimates of PAC patients from both completed interviews and missed cases captured on the tracking sheets. Both PMS and HFS estimates were adjusted for referrals to avoid double-counting of PAC patients who may have been treated at one facility and then referred for additional care to a higher level facility (S2 Appendix). In addition to measuring PAC caseloads at each facility, we also assess the capacity of facilities to provide PAC, including the availability of essential drugs, tests and products, and the most commonly used procedures for PAC. The annual number of women receiving PAC at a facility was calculated as the average of all available data points from our study. For facilities in the PMS, this included the past month estimate, the typical month estimate, and prospectively measured caseloads. For all other facilities, only the past month and typical month estimates were used. Facility-level PAC estimates were then weighted to account for non-response in the HFS to provide national and regional estimates of the number of women who obtained PAC in 2016. Confidence intervals are not applicable for the facility-level PAC caseload estimates as the HFS was conducted as a census, rather than a sample. In the process of validating the assumption to exclude official health statistics due to underreporting [30], we uncovered PAC cases in the MoHCC data being reported from facilities outside of our sample. These were public primary health centers that had not received training to use misoprostol as well as three district hospitals that had been erroneously missed from the HFS facility listing (n = 825, Table 1). The HFS data collectors conducted a brief phone survey with a sample of these primary health centers and all three district hospitals to assess the accuracy of these reported PAC cases (S3 Appendix). While most cases in the out-of-sample public facilities were referred to higher-level facilities due to the lack of PAC capacity, we determined that out-of-sample facilities were providing some, though often insufficient, PAC treatment. We therefore took the reported MoHCC caseloads and subtracted the reported proportion referred. We also applied an underreporting ratio since the referral-adjusted out-of-sample caseloads are likely an underestimate (S3 Appendix). This total out-of-sample caseload was added to the total caseload estimate from the HFS and PMS to estimate the total annual PAC caseload in Zimbabwe. PAC cases include women seeking care for both induced abortions and miscarriages. The second step of the AICM is estimating the proportion of all PAC cases that are due to induced abortions. To do this we make two key assumptions. The first assumption is that women would likely only seek PAC for miscarriages in the second trimester (13–22 weeks gestation), as miscarriages in the first trimester would resolve on their own. The second assumption is that not all women who need treatment would be able to receive it due to access, costs, knowledge, or a variety of other reasons, including death. To quantify the number of women having miscarriages in the second trimester, we estimated the number of miscarriages using clinical data on the risk of miscarriage by gestational age in the US, as these are the only data which exist [31]. We used life table estimates provided by Harlap et al. to calculate a risk of miscarriage between 13–22 weeks, which is 3.41 miscarriages per 100 live births [31]. To quantify the number of women having second trimester miscarriages who would receive care, we used two sources of data. Previous AICM studies have assumed that the treatment-seeking rate for a second trimester miscarriage is equivalent to the proportion of women delivering in a health facility. Delivering in a health facility is seen as the nearest measure of women’s access to obstetric and gynecological care in the absence of specific data on access to PAC services. We also asked HPS respondents what proportion of women with second trimester miscarriages would receive PAC. Those two sources of facility access were within 3 percentage points, so we used the PAC-specific estimate from the HPS. With these two assumptions, we subtracted the second trimester miscarriage cases receiving treatment from the PAC caseload total to arrive at the number of PAC cases due to induced abortion. The HPS was used to calculate a multiplier, or an estimate of the number of women having abortions who either did not experience a complication or did have a complication but did not receive treatment, for every one woman receiving PAC. As complications and treatment access are likely to vary based on where a woman lives and what resources she has, we asked about complications and treatment estimates separately for four sub-groups of women: rural poor, rural non-poor, urban poor and urban non-poor women. HPS respondents provided an estimate of the proportion of women having an abortion by method (surgical, misoprostol and other) and by provider type (doctor, nurses, pharmacists, traditional providers, woman themselves) for each of the four sub-groups of women. The likelihood of women experiencing an abortion complication was asked for each type of abortion (by provider and method) for women in each sub-group. We then multiplied the percent of women having complications for each method and provider by the distribution of women across abortion types within each sub-group. We summed the proportion of women experiencing complications by each provider to get an estimate of the proportion that would result in a complication by method. We then adjusted this estimate of complications by the method-specific sub-group estimated proportion of women that were likely to receive treatment. Taking the sum of the method-specific numbers provides an estimate of the proportion of women who receive treatment for complications from induced abortions within each sub-group. We then weighted these estimates by the population of women in each sub-group from the ZNSA and ZDHS [2,29]. This provided us with regional estimates of the proportion of all women having abortions who receive care for complications. The inverse of this number results is the regional multiplier. A higher multiplier can result when complication rates are low or when women are less likely, or able, to seek PAC treatment. The full equation for calculating the multiplier (and the full AICM equation) is provided in S1 Appendix, and summarized in Fig 1. The multiplier was calculated excluding respondents who rated themselves in the lowest 10% on the ten-point scale of self-reported certainty for groups of questions (e.g., urban, rural, misoprostol urban, misoprostol rural) to account for the fact that not all respondents were equally knowledgeable about all types of abortions across all areas. On the ten-point scale, the bottom 10% cutoff for self-reported certainty was four for question on rural areas, five for questions on urban areas, five for questions on misoprostol abortions in urban areas, and three for questions on misoprostol abortions in rural areas. Censoring the least certain respondents resulted in a 0.1 increase in the multiplier; this adjustment did not have a large impact on the multiplier calculation. The national total of induced abortions is the product of the multiplier and the annual PAC caseload. However, one further adjustment was made after this step to calculate the total number of abortions in 2016. Zimbabwe borders three countries with broader access to safe and legal abortions: South Africa, Mozambique and Zambia [32]. With existing high migratory rates of Zimbabwean women to neighboring countries [33], especially South Africa, it is therefore possible that women are traveling to neighboring countries to access safe abortions. To account for this, we asked HPS respondents to estimate the proportion of women in their province who travel outside of the country to have an abortion. We then took these estimates for each province, weighted by the population of women in each province, to calculate the percent of abortions occurring outside of the country by region. We added the estimated number of abortions occurring outside the country to the national total to produce the overall estimate of the number of abortions in 2016. The regional and national abortion rates for 2016 were calculated per 1,000 women of reproductive age, and the abortion ratio per 100 live births. Upper and lower bounds were calculated for the number of abortions, abortion rate, and abortion ratio using a bootstrapping approach to estimate bounds around the multiplier. Since the HPS is a purposive rather than random sample, bootstrapping does not result in a 95% confidence interval around the mean estimate. However, given the nature of the sample and the potential of sensitivity around which respondents were selected to participate, we wanted to account for how sensitive the results are to the selected respondents. To do so, we took 10,000 draws with replacement from the HPS respondents, and calculated a multiplier after each draw. The number of draws was determined by the convergence of the estimated mean and standard errors. The upper and lower bounds presented around the multiplier contain 95% of the multiplier values estimated in the bootstrapping simulation. Knowing the number of abortions, we were able to estimate the number and rate of pregnancies in Zimbabwe and the proportion of pregnancies that are unintended. Data on intention status of births comes from the 2015 ZDHS, in which women reported on whether births in the last five years were planned [2]. We assumed that all abortions were of unintended pregnancies, which is likely given that despite its legality, it is difficult to obtain an abortion for a fetal anomaly. Unintended pregnancies are comprised of all unplanned births, abortions, and miscarriages, which are estimated to be equal to 20% of the unplanned births and 10% of the abortions [34]. Intended pregnancies include planned births and miscarriages, estimated to be 20% of those births. We used these data to calculate the proportion of all pregnancies that are unintended, and the distribution of all pregnancies by intention and outcome.