Methodology for Estimating Abortion Incidence

The HFS data were weighted to project the results nationally. The weights factored in the probability of selection and of non-response according to region and facility type. Facilities that did not provide PAC were also removed from the facility universe. The weighting factor for a given facility type was the inverse of the sampling proportion multiplied by the proportion of completed interviews among sampled facilities of that type.

HFS respondents were asked whether their facility provided treatment of spontaneous or induced abortion complications; if they responded affirmatively, they were asked to report the following: 1) the number of women who received PAC as outpatients in a typical month; 2) the number of women who received PAC as outpatients in the past month; 3) the number of women who received PAC as inpatients in a typical month; and 4) the number of women who received PAC as inpatients in the past month (S1 Appendix).

Respondents were asked about a typical or average month in addition to the past month to differentiate from seasonal peaks in obstetric care and to minimize errors in reporting. Respondents who had difficulty reporting the number of patients for a month timeframe were given the option to provide information for a typical year or for the past year. Typical and past month patient counts were averaged, and then multiplied by 12 to get an estimate for the calendar year. (Typical and past year patient counts were divided by 12, then averaged together, then multiplied by 12.)

Applying the weight to this data yields an estimate of the total number of abortion complications treated per year at all health facilities. As this number includes both spontaneous and induced abortion complications, we needed to exclude women with spontaneous abortions. It is difficult to differentiate between complications due to voluntary pregnancy interruption and those due to a spontaneous abortion because not only are the symptoms of the two often very similar, but stigma also prevents women from admitting to having sought an abortion. To address this issue, we used an indirect method to separate out spontaneous abortion complications from the estimated total number of complications treated in health facilities.

We assume that women who experience late spontaneous pregnancy loss (13–21 weeks’ gestation) are likely to require care at a health facility due to the higher frequency and severity of complications after the 13^th week of gestation [20]. About 2.89% of pregnancies recognized at six weeks’ gestation end in spontaneous abortion during weeks 13–21. As 84.575% of recognized pregnancies result in live births, the number of spontaneous abortions during weeks 13–21 is roughly equal to 3.417% of live births (2.89/84.575) [20,21]. An estimate for live births in Uganda in 2013 was calculated by applying the age-specific fertility rates from the 2011 UDHS (assumed to have remained fairly constant over the 2 year period) to the 2013 regional population estimates for women aged 15–49, by five-year age groups.

The AICM requires a further adjustment because not all women needing treatment for complications of late miscarriages will obtain care in a facility. We assumed that the proportion of women obtaining care in a facility for a late spontaneous pregnancy loss is the same as the proportion of women giving birth in a health facility. According to the 2011 UDHS, this represented 63% of women giving birth; we projected this value to 2013, based on the assumption that births assisted by a trained health worker would continue to increase, as was the case between the 2006 and 2011 UDHSs [15,22]. These cases were removed from the total number of abortion complications treated in a health facility.

Not all women who have induced abortions experience complications, and those that do, do not necessarily obtain treatment in a health facility. To account for these cases, a multiplier or inflation factor must be calculated and applied to the estimated number of induced abortion complications treated in health facilities in order to estimate the total number of women who had an induced abortion. In general, the safer abortion services are, the higher the multiplier will be, because a lower number of women will experience complications that require medical treatment. Likewise, the less safe abortion services are, the lower the multiplier because a larger number of women will have serious complications that will require medical treatment. The multiplier is also affected by women’s access to health facilities. Where facilities are easily accessible, the proportion of women with complications who receive treatment will be higher, implying that the multiplier will be lower. The reverse will be the case in poor or underserved areas.

Data needed to estimate the multiplier is drawn from the HPS. Respondents were asked to make a series of estimations:

the percent distribution of women obtaining abortions according to type of provider;

the proportion of women likely to experience complications requiring medical care;

the proportion of women with complications who were likely to obtain care from a health facility (S2 Appendix).

Provider types included doctors, clinical officers, nurses, trained/certified midwives, pharmacists, traditional birth attendants, and the women themselves. Because the conditions under which women obtain abortions vary by socioeconomic status and their place of residence, respondents were asked to answer the above questions for each of four subgroups of women: urban poor, urban nonpoor, rural poor and rural nonpoor. Poor and nonpoor were defined using women’s level of education, as data on income is not available in Uganda. Poor women were defined as those who had had seven or fewer years of schooling, and nonpoor women as those with eight or more years of schooling [15].

Based on these data, we estimated, for each of the four subgroups, the proportion of women treated for abortion complications among those who had an induced abortion. Percentages were weighted by the relative size of the subgroups to yield the proportion of all women who had an abortion that were treated for complications nationally and regionally. The multipliers are the inverse of these proportions.

The estimated proportion of women getting treatment for abortion complications was 29.7%, therefore the national estimated multiplier was 3.37 (100/29.7). Regional multipliers were calculated for the same four major regions the country was divided into during the 2003 study to facilitate comparison: Central, Eastern, Northern and Western. The regional multipliers were then applied to the regions that they represent currently. Thus, the multiplier estimated for the old Central region (3.46) served as the multipliers for Central 1, Central 2 and Kampala; the multiplier estimated for the Northern region (3.31) was applied for Karamoja, North and West Nile; the multiplier estimated for the Eastern region (3.87) served as the multipliers for East Central and Eastern; and the multiplier estimated for the Western region (3.27) was used for South West and Western. These multipliers were applied to the regional estimates of complications from induced abortions to calculate estimates for the total number of abortions in each region. There is a small difference in the total number of abortions when the national multiplier is applied to the national number of complications compared to the sum of the regional complications. We adjusted the regional totals of abortions slightly, such that they sum up to the total number of abortions obtained using the multiplier. National and regional confidence intervals were constructed around the estimates of induced abortion, based on the 95% confidence intervals around the estimate of abortion complications treated in a year at all health facilities.

Using the estimates of abortion from this study, recent births data from the UDHS, and population data, we estimated the rates of pregnancy and pregnancy outcomes; specifically, the proportion of pregnancies ending as planned and unplanned births, abortions, and miscarriages.

To estimate numbers and rates of unintended pregnancies, we first calculated the number of unplanned births, using data from the 2011 UDHS. This was obtained by estimating the proportion of births during the five years prior to the survey that had been unplanned (i.e. mistimed or unwanted at the time of conception); this proportion was applied to the number of live births in 2013. Summing the numbers of unplanned births, induced abortions (estimated in this study), and miscarriages resulting from unintended pregnancies (calculated as 20% of unplanned births and 10% of abortions) [21,23] yielded estimates of the number of unintended pregnancies. With this number, we estimated the proportion of all pregnancies that were unintended and applied this number to the population of women aged 15–49 to obtain the rate of unintended pregnancies. In the 2003 study, unintended pregnancies ending as miscarriages were not included in the calculations of the unintended pregnancy rates. We have recalculated the unintended pregnancy rate for Uganda for 2003 to facilitate comparison with the 2013 data and allow for analysis of trends over the past decade.