Menu
Menu
Menu
Menu
Objective To estimate age-specific abortion incidence and unintended pregnancy in Zimbabwe, and to examine differences among adolescents by marital status and residence. Design We used a variant of the Abortion Incidence Complications Methodology, an indirect estimation approach, to estimate age-specific abortion incidence. We used three surveys: the Health Facility Survey, a census of 227 facilities that provide postabortion care (PAC); the Health Professional Survey, a purposive sample of key informants knowledgeable about abortion (n=118) and the Prospective Morbidity Survey of PAC patients (n=1002). Setting PAC-providing health facilities in Zimbabwe. Participants Healthcare providers in PAC-providing facilities and women presenting to facilities with postabortion complications. Primary and secondary outcome measures The primary outcome measure was abortion incidence (in rates and ratios). The secondary outcome measure was the proportion of unintended pregnancies that end in abortion. Results Adolescent women aged 15-19 years had the lowest abortion rate at five abortions per 1000 women aged 15-19 years compared with other age groups. Adolescents living in urban areas had a higher abortion ratio compared with adolescents in rural areas, and unmarried adolescent women had a higher abortion ratio compared with married adolescents. Unintended pregnancy levels were similar across age groups, and adolescent women had the lowest proportion of unintended pregnancies that ended in induced abortion (9%) compared with other age groups. Conclusions This paper provides the first estimates of age-specific abortion and unintended pregnancy in Zimbabwe. Despite similar levels of unintended pregnancy across age groups, these findings suggest that adolescent women have abortions at lower rates and carry a higher proportion of unintended pregnancies to term than older women. Adolescent women are also not a homogeneous group, and youth-focused reproductive health programmes should consider the differences in experiences and barriers to care among young people that affect their ability to decide whether and when to parent.
We used an age-specific variant of the Abortion Incidence Complications Methodology (AICM), an indirect estimation approach, to estimate age-specific abortion rates.5 6 The AICM has been used in over 20 countries with restrictive abortion laws to indirectly measure the incidence of abortion.12 22–35 This indirect method obtains the national number of facility-based PAC cases and estimates the proportion of abortions that would result in women having complications and receiving PAC. An age-specific variant of the AICM was employed in Ethiopia and Uganda, and we followed the approach in those studies.5 6 We calculated abortion incidence by age groups (15–19, 20–24, 25–29, 30–34 and 35–49). We also calculated abortion incidence by marital status and residence subgroups within two age groups: adolescent women (aged 15–19 years) and all women of reproductive age (15–49 years). We defined the marital status subgroup dichotomously based on PAC patients’ self-report of being currently married/in union or not currently married. We categorised the dichotomous residence subgroup based on PAC patients’ self-report of living in urban or rural areas. The AICM approach relies on three key data inputs: the number of PAC cases, the proportion of all abortions that result in treated complications, and the age distribution (as well as marital status and place of residence) of PAC patients. These data inputs come from three surveys that were conducted in Zimbabwe in August to November 2016 as part of a larger study that indirectly estimated the incidence of abortion.12 The data on the estimated annual number of PAC cases came from a Health Facility Survey (HFS), which interviewed PAC providers in a census of 227 facilities that provide PAC in Zimbabwe. This was combined with estimates of the proportion of abortions that resulted in complications that received treatment from a Health Professional Survey (HPS), which was a purposive sample of 118 key informants knowledgeable about abortion provision in Zimbabwe. The data on the characteristics of women receiving PAC came from the Prospective Morbidity Survey (PMS), which was conducted in a nationally representative sample of facilities with PAC capacity, using the HFS universe as the sampling frame. Data were collected on all women presenting to the 127 participating facilities for PAC during the 28-day study period (unweighted n=1002). Sociodemographic characteristics of PAC patients from the PMS can be found in online supplementary file 1. The HFS and PMS collected information on women who received PAC in health facilities for either induced abortions or miscarriages. Further details on the study design, sampling and informed consent processes for the HFS and HPS can be found in Sully et al 12 and in Madziyire et al for the PMS.15 bmjopen-2019-034736supp001.pdf We used age-specific fertility rates from the 2015 Zimbabwe Demographic and Health Survey (ZDHS)13 and age-specific population numbers of women of reproductive age from the Zimbabwe National Statistic Agency’s (ZNSA) Population Projections Report to calculate the age-specific number of births.36 This first step of the age-specific variant of the AICM estimated the number of PAC cases by age group. This was done by multiplying the national number of PAC cases in Zimbabwe12 by the weighted age distribution of PAC patients in the PMS.15 To calculate the number of urban and rural PAC patients within the 15–19 and 15–49 age groups, we multiplied this age-specific number of PAC cases times the proportion of PAC patients within that age group who self-reported living in rural or urban areas. We did the same for PAC patient’s self-reported marital status. Second, we estimated the age-specific number of PAC cases due to induced abortion. To do this, we first estimated the proportion of second trimester miscarriages by age group6 37 and then adjusted this by the proportion of those miscarriages likely to receive treatment. In the absence of data on access to PAC, we assumed the proportion of women receiving care was equal to the age and subgroup specific proportion of women who give birth in a facility from the ZDHS.13 The result was subtracted from the total number of PAC cases to obtain the number that was due to induced abortion. The third step was estimating abortions that do not result in facility-based care. This could be due to two reasons: (1) the person having the abortion did not have complications and therefore did not need facility-based treatment or (2) the person having the abortion had complications but did not receive facility-based treatment. In Zimbabwe in 2016, it was estimated that for every one woman receiving PAC, there were 4.7 women who had abortions that did not result in facility-based care (see online supplementary file 2). This estimate, referred to as the multiplier, was applied to all age groups and marital status subgroups in the absence of age-specific and marital status-specific multipliers. We calculated new multipliers for rural women and urban women separately, using the underlying data from the HPS, which was collected separately by residence. Since the HPS is a purposive sample, we could not estimate 95% CI around the multiplier. We therefore conducted a bootstrapping simulation of 10 000 draws with replacement from the HPS respondents and calculated a multiplier with each draw.12 The upper and lower bounds presented contain 95% of the multiplier values from the bootstrapping. Further details on the multiplier calculations are in online supplementary file 2. bmjopen-2019-034736supp002.pdf Fourth, to estimate the total number of induced abortions by age and subgroup, we multiplied the number of induced abortions that received PAC times the respective multiplier for that age and subgroup. We then adjusted all estimates to account for abortions occurring outside of Zimbabwe, estimated from the HPS as 12% of all abortions, which we assumed was the same across all age and subgroups due to lack of data availability.12 Online supplementary file 2 provides more details on the steps of the AICM, data sources and assumptions, and adjustments made to align all summed age groups and subgroups to national totals. We calculated age-specific abortion rates per 1000 women by dividing the total number of induced abortions by the age-specific population size, which was taken from the ZNSA Population Projections Report.36 We also wanted to account for risk of pregnancy given the lower levels of reported recent sexual activity, defined as sex in the past 12 months, among adolescents (29%) compared with women aged 20–49 years (84%).13 Therefore, we also calculated separate estimates of age-specific abortion rates among women who reported having sex in the previous 12 months. We estimated age and subgroup specific abortion ratios per 100 live births by dividing the number of induced abortions by the age and subgroup specific number of births.13 36 To calculate the number of births by marital status and residence among 15–19 and 15–49 age groups, we multiplied the age-specific number of births by the age-specific proportion of births to married or unmarried women (or the proportion of births among women who lived in urban or rural areas) from the ZDHS.13 Lastly, we estimated unintended pregnancy by age and subgroup using the age and subgroup specific abortion estimates, age and subgroup specific data on births by intention status from the ZDHS,13 and estimated miscarriages, which were estimated to be 20% of births and 10% of abortions and applied uniformly across age groups.38 Using the national multiplier for all age and marital status groups assumes that complication rates and treatment seeking do not differ by age or marital status. We tested the validity of these assumptions using PAC patient data, the only available source of representative data on abortion complications and treatment among women in Zimbabwe. The first assumption we tested was whether experiencing complications differed between adolescent and non-adolescent PAC patients, and if this differed by marital status among adolescents. We operationalised complications using the severity classifications from Madziyire et al 15 and considered moderate and severe complications, maternal near-miss and maternal death as having a complication (coded as 1) and a mild complication (defined as no infection, no organ failure and no blood transfusion needed) as no complication (coded as 0). Using this complication variable as the dependent variable, we ran multivariable logistic regression models comparing adolescents versus non-adolescents, and then included an interaction term between adolescents and marital status. The second assumption we tested was whether adolescent and non-adolescent PAC patients differed in treatment seeking. PAC patients provided information on their delays in seeking care from the (1) time (in hours) it took from realising they had a complication to deciding to seek treatment and (2) time (in hours) it took from them deciding to seek care to the time they arrived at the facility. These two delays are related to user-related health seeking behaviours (delay 1) as well as access to facilities (delay 2), which when combined capture treatment seeking experiences.39–41 We ran a Cox proportional hazards model to estimate differences in hazard ratios between adolescents and non-adolescents, and we included an interaction term between adolescents and marital status. All models controlled for residence, secondary education, wealth, previous pregnancy, whether the pregnancy was reported as unintended and being in the second trimester. This research involved interviews with postabortion care patients. Patients were not invited to comment on the study design or involved in the writing or editing of this document. We sought guidance for study planning and dissemination from our Technical Advisory Committee, which included community representatives and technical experts.
