Incidence of induced abortion in Malawi, 2015

Background In Malawi, abortion is legal only if performed to save a woman’s life; other attempts to procure an abortion are punishable by 7-14 years imprisonment. Most induced abortions in Malawi are performed under unsafe conditions, contributing to Malawi’s high maternal mortality ratio. Malawians are currently debating whether to provide additional exceptions under which an abortion may be legally obtained. An estimated 67,300 induced abortions occurred in Malawi in 2009 (equivalent to 23 abortions per 1,000 women aged 15-44), but changes since 2009, including dramatic increases in contraceptive prevalence, may have impacted abortion rates. Methods We conducted a nationally representative survey of health facilities to estimate the number of cases of post-abortion care, as well as a survey of knowledgeable informants to estimate the probability of needing and obtaining post-abortion care following induced abortion. These data were combined with national population and fertility data to determine current estimates of induced abortion and unintended pregnancy in Malawi using the Abortion Incidence Complications Methodology. Results We estimate that approximately 141,044 (95%CI: 121,161-160,928) induced abortions occurred in Malawi in 2015, translating to a national rate of 38 abortions per 1,000 women aged 15-49 (95%CI: 32 to 43); which varied by geographical zone (range: 28-61). We estimate that 53%of pregnancies in Malawi are unintended, and that 30%of unintended pregnancies end in abortion. Given the challenges of estimating induced abortion, and the assumptions required for calculation, results should be viewed as approximate estimates, rather than exact measures. Conclusions The estimated abortion rate in 2015 is higher than in 2009 (potentially due to methodological differences), but similar to recent estimates from nearby countries including Tanzania (36), Uganda (39), and regional estimates in Eastern and Southern Africa (34-35). Over half of pregnancies in Malawi are unintended. Our findings should inform ongoing efforts to reduce maternal morbidity and mortality and to improve public health in Malawi.

We used an indirect estimation approach called the Abortion Incidence Complications Method (AICM) [20], which has been used in over 20 countries, including Malawi in 2009 [18]. It requires two surveys: a Health Facilities Survey (HFS) conducted in facilities with the potential to treat abortion complications, and a Knowledgeable Informants Survey (KIS), conducted among individuals knowledgeable about abortion in the country. Women undergoing induced abortion have three potential outcomes: experiencing no complications, experiencing complications but not obtaining treatment in a health facility, or experiencing complications and obtaining treatment in a health facility. Data from the HFS is used to estimate the latter. KIS respondents estimate the distribution of abortions by provider type, the probability of experiencing complications by provider type, and the probability of obtaining PAC in a facility. Given differentials in women’s access to abortion and to PAC, these probabilities are estimated separately for four sub-groups: rural poor and non-poor women, and urban poor and non-poor women. This information is used to calculate a multiplier which represents, for each induced abortion complication treated, how many induced abortions occurred for which treatment was either not required or not obtained. Applying the multiplier (calculated from the KIS) to the estimated number of induced abortions with treated complications (calculated from the HFS) yields an estimate of all induced abortions in the country. Fieldwork occurred during October-December 2015, led by the Centre for Reproductive Health at the College of Medicine, University of Malawi, with technical support from the Guttmacher Institute. Twelve interviewers with previous experience administering surveys related to reproductive health conducted HFS and KIS interviews, supervised by three MPH-level interviewers who also administered several KIS interviews. All of these interviewers were in turn supervised by a Project Coordinator. A week-long training for all interviewers, supervisors, and study staff was held in Blantyre just prior to fieldwork initiation. We pilot tested questionnaires in a small number of interviews with respondents outside of our sample. Ethical approval to conduct the study was obtained from Guttmacher’s Institutional Review Board and the Research and Ethics Committee of the College of Medicine. We also obtained a letter of support from the Malawi Ministry of Health, Department of Planning and Policy Development. Respondents in both surveys provided written informed consent, and interviews were conducted in English or Chichewa (the vernacular of Malawi), according to the respondent’s preference. We did not provide incentives for participation in either survey. We obtained a list of 977 facilities from the 2013–14 Malawi Service Provision Assessment [21], which represented 92% of all known health facilities in the country (Table 1). We excluded 49 facilities too specialized to provide PAC (e.g., dental offices, prison clinics, podiatry clinics), plus 47 dispensaries and 20 health posts, since PAC services are not offered at this level of health care facility in Malawi. Among 861 remaining facilities, we selected a nationally representative sample using stratified random sampling by facility level (central hospital, district hospital, rural/community hospital, other hospital, health centre, clinic, maternity unit), administrative zone of the Ministry of Health (North, Central-East, Central-West, Southeast, and Southwest) and ownership type (government, non-governmental organization [NGO], or private). We included all hospitals and maternity units, and randomly sampled 40% of health centers and 12% of clinics, for a total of 334 sampled facilities. * To determine the number of health centers and clinics to be sampled within each geographical zone, we determined the proportion of all health centers and clinics represented in each geographical zone. We then applied the relevant proportion to the total number of health centers (187) or clinics (34) we planned to sample. To determine the number of health centers and clinics to be sampled within each geographical zone and by ownership, we determined the proportion of health centers and clinics within each geographical zone that were within each ownership category (public, NGO, and private). We then applied the relevant proportion for each type of health facility (health centre or clinic) and zone to the total number of health centers or clinics that we planned to sample in each geographical zone. To minimize refusals, the Malawian Ministry of Health sent letters introducing the study and noting an upcoming interview request to the 334 sampled facilities in advance of data collection. Interviewers used a standardized questionnaire to conduct a face-to-face interview with a senior staff member knowledgeable about the facility’s provision of abortion services; generally a nurse-midwife (68%) or other clinician (27%). In consultation with zonal officers and reproductive health experts in Malawi, we developed a purposive sample of 125 potential knowledgeable informants; 25 per zone, representing 27 of Malawi’s 28 districts (the small Lake Malawi island district of Likoma was excluded due to accessibility issues). Trained interviewers successfully interviewed all 125 invited respondents, who held a range of professions, including formal sector medical providers (45%), community health workers (19%), NGO employees (11%), village health committee members (6%), traditional birth attendants (5%), or other professions (14%). Respondents were balanced by gender (52% male and 48% female). Almost half (46%) worked in the public sector, with 29% in NGOs, 12% in the private sector, 5% affiliated with a Christian Health Association of Malawi (CHAM) facility, and 8% in other sectors (for example, a traditional leader or member of a community health committee). The 2009 Malawi AICM [18] noted concern about potential underrepresentation of informants familiar with the situation of abortion in rural areas, and recommended that future studies identify informants with recent rural experience. Thus, we paid particular attention to this factor. While 16% of our respondents worked in urban areas, 42% worked in rural areas and another 42% worked in both urban and rural areas. Interviewers determined that the respondent knew about abortion in rural areas very well or moderately well in 91% of interviews. In 10 of 125 KIS interviews, interviewers noted concern about respondent uncertainty in responses to key questions. We excluded these 10 respondents from analysis, after confirming that they were similar to the remaining sample of KIS respondents in terms of location, gender, age, or years in profession. Out of 334 HFS facilities sampled, we successfully interviewed 294 facilities (88% of sampled facilities) (Table 2). Nine were either closed, not feasible to access, or had merged with another clinic in our sampling frame. An additional 31 facilities refused to participate or did not have the necessary staff member available to give an interview. Of these, 3 refused interviews for unknown reasons and 6 did not have a staff member available to give the interview. Twenty-two refusals (including 11 clinics, 7 health centers, and 4 “other” hospitals) occurred because that facility did not offer PAC services. Of the 294 interviewed facilities, 202 reported providing PAC, with greater likelihood in higher-tier facilities (i.e., hospitals, 79–100%) than lower-tier facilities (i.e., clinics or health centers, 50–60%) or maternity units (25%) (Table 2). * Out of 334 HFS facilities sampled: 4 were not feasible to access; 1 had merged with another clinic in our sampling frame; 4 were permanently closed; 6 did not have a provider available to give an interview during the fieldworkers’ visit to the area; 22 refused to provide an interview because they didn’t offer PAC; and 3 refused an interview for some other reason. ** Weighted by the inverse of the product of the sampling fraction and the response rate for each facility type to yield results that were nationally representative. From these 202 facilities, we obtained estimates of the number of PAC patients (inpatient or outpatient) treated for either miscarriage or induced abortion (as the two are often clinically indistinguishable, and may not be reported separately for fear of legal sanctions) in an average month and in the past month. We averaged these two figures and multiplied by 12 to estimate the total number of PAC cases in interviewed, PAC-providing facilities in 2015. We then weighted those values by the inverse of the product of the sampling fraction and response rate for each facility type to provide nationally representative estimates of PAC caseloads by facility type. We assumed that the proportion, by facility type and ownership, of PAC-providing facilities in the sample was similar to the proportion of PAC-providing facilities in the country, and adjusted the sampling frame accordingly. To avoid double-counting PAC patients who may have been treated at one facility and then referred for additional care to a higher-level facility, we subtracted cases believed to be referrals. In order to do this, we asked each facility how many PAC patients they treated and then referred elsewhere. In the absence of a reliable referral follow-up rate estimate in Malawi, we assumed that 75% of referred patients followed up, as was done in a recent AICM study in Tanzania [22], based on data from a prospective study of follow-up for obstetric complications (specifically intrauterine fetal death) in a rural district of Tanzania [23]. Typically in Malawi, health centers, clinics, maternity units, and rural/community hospitals refer to district and CHAM hospitals, which in turn refer to central and top-tier private hospitals. In each zone, we subtracted 75% of referrals out of health centers, clinics, maternity units, and rural/community hospitals from the number of PAC patients treated at district hospitals and CHAM hospitals; and also subtracted 75% of referrals from district and CHAM hospitals from the number of PAC patients treated at central hospitals and high-level private hospitals (Table 3). This correction for double-counting reduced the number of PAC cases by 10%. * Adjusted 95% CIs were calculated by subtracting referrals and late miscarriages from the upper and lower bounds of the 95% CIs for the estimated number of all PAC cases. **Refers to the number of PAC cases stemming from induced abortions per 1,000 women ages 15–49 ***Malawi is divided into three administrative regions: Northern, Central, and Southern. The Ministry of Health divides the country into five administrative zones: the Southern region is comprised of the Southwest and Southeast zone, the Central region is comprised of the Central-West and Central-East zone, and the Northern region is comprised of the North zone. Next, we subtracted PAC cases likely due to miscarriage, in order to estimate PAC cases stemming specifically from induced abortion. The number of pregnancies ending in late miscarriage (between 13 and 21 weeks gestation) is estimated to equal 3.41% of all live births [24;25], and we assumed that only later miscarriages likely require PAC. We estimated the total number of live births by multiplying, separately for urban and rural women, age-specific fertility rates with the number of women in the population by age groups [19], and then adding these together for the total number of births among women aged 15–49. Since not all late miscarriages require care in a facility, we asked KIS respondents to estimate, separately for urban and rural women, the proportion of late miscarriages treated in a health facility, and calculated the median of their responses (90% among urban women, 70% among rural women). Applying these proportions to the estimated number of late miscarriages represents the total number of PAC cases likely due to miscarriage (Table 3). To generate the multiplier, we multiplied the estimated probabilities of experiencing complications of induced abortion according to provider type from KIS data against the probabilities of receiving treatment for complications, among urban poor, urban non-poor, rural poor, and rural non-poor women, also from KIS data. This yields the estimated proportion of abortions with complications that receive treatment among these four groups. Separately for each health zone, these estimates are combined into a single weighted proportion based on the population distribution of the four groups [26], defining “poor” as belonging to a household with below median wealth and assets, according to the Malawi 2010 Demographic and Health Survey. We used Taylor linearized variance estimation, using a finite population correction factor, to calculate 95% confidence intervals (CIs) around PAC caseloads, reflective of the complex survey design. We used a scaling factor (the average of the variances from strata with multiple sampling units) for strata containing a single population sampling unit. Since we subtracted likely referrals and late miscarriages from the PAC caseloads estimate, we adjusted the 95% CIs accordingly, by subtracting the same amount from both the upper and lower bounds. We applied multipliers to the estimated number of induced abortions requiring and receiving treatment (and to the adjusted upper and lower bounds), to produce an estimate (and associated 95% CI) for the total number of induced abortions in Malawi in 2015 (Table 3). Mathematical expressions for our calculations are provided in the Appendix (Text A in S1 File). Finally, we used these abortion estimates to calculate the number and rate of unintended pregnancies nationally and by zone, by building upon information on unplanned births from DHS to estimate rates, outcomes, and intention status for all pregnancies (versus only those ending in a live birth). We used information about the planning status of births (collected retrospectively, pertaining to a woman’s reported fertility desires at the time she became pregnant) in the past five years from the 2010 DHS to obtain the percentage of live births that were unplanned. We assume that all abortions were the result of unintended pregnancies. Miscarriages of any gestational age were estimated as 20% of live births plus 10% of abortions [25]. The estimate for unintended pregnancies, then, is the sum of unintended births, abortions, and miscarriages of pregnancies believed to be unintended, i.e. 20% of unintended live births plus 10% of abortions. Likewise, the estimated number of intended pregnancies consists of intended births plus the estimate of miscarriages of intended pregnancies, which is equal to 20% of intended births. From the estimated number of unintended pregnancies, we calculated an unintended pregnancy rate and the percentage of pregnancies in Malawi and in each zone that was unintended.

• polis, chelsea b.
• mhango, chisale
• philbin, jesse
• chimwaza, wanangwa
• chipeta, effie
• msusa, ausbert