Adverse pregnancy outcomes in rural Uganda (1996-2013): Trends and associated factors from serial cross sectional surveys

Objective: Community based evidence on pregnancy outcomes in rural Africa is lacking yet it is needed to guide maternal and child health interventions. We estimated and compared adverse pregnancy outcomes and associated factors in rural south-western Uganda using two survey methods. Methods: Within a general population cohort, between 1996 and 2013, women aged 15-49 years were interviewed on their pregnancy outcome in the past 12 months (method 1). During 2012-13, women in the same cohort were interviewed on their lifetime experience of pregnancy outcomes (method 2). Adverse pregnancy outcome was defined as abortions or stillbirths. We used random effects logistic regression for method 1 and negative binomial regression with robust clustered standard errors for method 2 to explore factors associated with adverse outcome. Results: One third of women reported an adverse pregnancy outcome; 10.8 % (abortion = 8.4 %, stillbirth = 2.4 %) by method 1 and 8.5 % (abortion = 7.2 %, stillbirth = 1.3 %) by method 2. Abortion rates were similar (10.8 vs 10.5) per 1000 women and stillbirth rates differed (26.2 vs 13.8) per 1000 births by methods 1 and 2 respectively. Abortion risk increased with age of mother, non-attendance of antenatal care and proximity to the road. Lifetime stillbirth risk increased with age. Abortion and stillbirth risk reduced with increasing parity. Discussion: Both methods had a high level of agreement in estimating abortion rate but were markedly below national estimates. Stillbirth rate estimated by method 1 was double that estimated by method 2 but method 1 estimate was more consistent with the national estimates. Conclusion: Strategies to improve prospective community level data collection to reduce reporting biases are needed to guide maternal health interventions. Data for this analysis are from the General Population Cohort (GPC) in Uganda. The study site is located 120 km west of the capital city, Kampala, in a rural community where demographic surveillance and medical surveys have been conducted since 1989 as described in detail elsewhere [13]. The GPC is a community-based open cohort study with approximately 22,000 residents of 25 neighbouring villages. The cohort was initially established by the UK Medical Research Council in collaboration with the Uganda Virus Research Institute to study the population dynamics of HIV transmission in rural Uganda, and now provides a platform to investigate determinants of other diseases, and health related problems focusing on maternal and child health. Agriculture is the main economic activity with rain-fed, small-holder farms for growing mainly bananas, coffee, beans, groundnuts, vegetables and a few root crops such as cassava and potatoes mainly for subsistence. Levels of education are generally low with about one third of the population attaining secondary education. Five health facilities serve the population with basic medical care, three of which offer family planning, antenatal care and deliveries. One higher level centre within the study area and a hospital 20 km away from the study area offer emergency obstetric services. An annual household survey of GPC residents has been conducted since 1989, with all study village residents eligible for inclusion. Community sensitization activities precede each survey round, including local council briefings and village meetings. All households are visited by, in turn, the mapping, census and survey teams. All consenting adult residents are interviewed at home in the local language by trained survey staff and provide a blood sample for HIV testing. In selected medical surveys between 1996 and 2013, all women aged 15–49 years who had been pregnant in the last 12 months were asked specifically about the outcome of their pregnancy. In 2012–2013, additional data on life time experience of pregnancies (total number, and outcome) were collected to compare with the annual interviews (see questions in Additional file 1). The World Health Organization (WHO) has defined stillbirth as foetal death late in pregnancy deferring the gestational age (GA) when a miscarriage (abortion) becomes a stillbirth to country policy [14]. In Uganda the GA cut-off for abortion and stillbirth is 28 weeks. In this paper we therefore define Abortion as a foetal loss before 28 weeks of gestation and stillbirth as a baby born with no signs of life after 28 completed weeks of gestation. Abortion rate is the number of abortions per 1,000 women of childbearing age and Stillbirth rate is the number of stillbirths per 1000 births. In this paper no distinction is made between spontaneous and induced abortions because induced abortion is illegal in Uganda and is highly stigmatized in rural communities. Adverse pregnancy outcome is defined as a pregnancy that did not result in a livebirth (this included both abortions and stillbirths). Age Specific Fertility Rate (ASFR) is the number of births per 1000 women in a particular age group. It is normally calculated for 5-year age groups over the reproductive ages, which are taken as 15–49 years. We also used Total Fertility Rate (TFR) referring to the number of live births that a woman would have had if she were subject to the current ASFR throughout the reproductive ages (15–49 years). Data were initially collected on paper and double entered in Microsoft Office Access, until 2009 when electronic data capture was introduced. The program contained logic programming skips and verifications that disallowed conflicting data. Stata 13 (Stata Corporation, College Station, USA) and SAS 9.4 (SAS Institute Inc., Cary, NC, USA) were used for analysis. Baseline characteristics were tabulated by study round (roughly corresponding to calendar year). Analysis of pregnancy outcomes in the past 12 months (live birth, stillbirth, abortion) and rates were examined by study round. We explored factors associated with abortion and with stillbirth in all study rounds as separate outcomes, and estimated odds ratios (OR) and 95 % CI for the associations using random effects logistic regression to account for clustering within women who reported more than one pregnancy. Age was included in all models as an a priori confounder. For abortions, factors whose age-adjusted association was significant at p < 0.10 were included in a multivariable model, and retained if they remained associated at p < 0.10. Because the numbers of stillbirths were small, we did not attempt to build a full multivariable model for this outcome. We also analysed pregnancy outcomes based on lifetime experience of pregnancies; computed for those who reported at least one pregnancy, the number and proportion of pregnancies ending as livebirth, abortion and stillbirth and summarised the results by age, marital status, religion education occupation, residence, phone ownership and parity. The proportion of women in the reproductive age reporting live births, stillbirths and abortions was also determined. We examined risk factors for abortions and stillbirths, as separate outcomes; the number of these events was considered as count outcome. Negative binomial regression was used to examine the effect of various risk factors on the number of abortions and stillbirths because the data were over-dispersed (variance greater than the mean); robust clustered standard errors were used to account for correlation of repeated pregnancies among women. The logarithm of the total number of pregnancies for each woman was included in the model as an offset. As with the analysis of outcomes in each round, age was considered an a priori confounder and included in all models. Factors whose age-adjusted association with the outcome was significant at p < 0.10 were included in a multivariable model and retained if they remained associated at p < 0.10. Lastly, we compared the results of two survey approaches; annual surveys between 1996 and 2013, when women were interviewed on their pregnancy experience in the preceding 12 months, versus the single survey in 2012–2013 when women were interviewed on their complete obstetric histories. This was done to evaluate the methodological biases associated with each approach. The study was approved by Uganda Virus Research Institute Research and Ethics Committee and the Uganda National Council for Science and Technology. All participants were given detailed study information before a written informed consent was obtained from them.