Immediate postnatal care following childbirth in Ugandan health facilities: An analysis of Demographic and Health Surveys between 2001 and 2016

Introduction Progress in reducing maternal and neonatal mortality, particularly in sub-Saharan Africa, is insufficient to achieve the Sustainable Developmental Goals by 2030. The first 24 hours following childbirth (immediate postnatal period), where the majority of morbidity and mortality occurs, is critical for mothers and babies. In Uganda,<50% of women reported receiving such care. This paper describes the coverage, changes over time and determinants of immediate postnatal care in Uganda after facility births between 2001 and 2016. Methods We analysed the 2006, 2011 and 2016 Ugandan Demographic and Health Surveys, including women 15-49 years with most recent live birth in a healthcare facility during the survey 5-year recall period. Immediate postnatal care coverage and changes over time were presented descriptively. Multivariable logistic regression was used to examine determinants of immediate postnatal care. Results Data from 12 872 mothers were analysed. Between 2006 and 2016, births in healthcare facilities increased from 44.6% (95% CI: 41.9% to 47.3%) to 75.2% (95% CI: 73.4% to 77.0%) and coverage of immediate maternal postnatal care from 35.7% (95% CI 33.4% to 38.1%) to 65.0% (95% CI: 63.2% to 66.7%). The majority of first checks occurred between 1 and 4 hours post partum; the median time reduced from 4 hours to 1 hour. The most important factor associated with receipt of immediate postnatal care was women having a caesarean section birth adjusted OR (aOR) 2.93 (95% CI: 2.28 to 3.75). Other significant factors included exposure to mass media aOR 1.38 (95% CI: 1.15 to 1.65), baby being weighed at birth aOR 1.84 (95% CI: 1.58 to 2.14) and receipt of antenatal care with 4+Antenatal visits aOR 2.34 (95% CI: 1.50 to 3.64). Conclusion In Uganda, a large gap in coverage remains and universal immediate postnatal care has not materialised through increasing facility-based births or longer length of stay. To ensure universal coverage of high-quality care during this critical time, we recommend that maternal and newborn services should be integrated and actively involve mothers and their partners. Household surveys are the main source of data used within maternal health to compare coverage trends and inequalities both within and between countries.26 The Demographic and Health Surveys (DHS) are cross-sectional nationally representative household surveys, usually covering 5000–30 000 households. They collect data from women in reproductive age (15–49 years) about births and the use of reproductive and maternal care. We used the DHS collected in Uganda in 2006, 2011 and 2016. The DHS use a multilevel cluster sampling survey design; individual women’s survey weights, and the elements of stratification and clustering are needed in analysis to adjust for this design and for non-response. The most recent live birth within a recall period of 5 years to women aged 15–49 at the time of survey, was included in the analysis, if the birth occurred in a health facility. Data from prior births or from those outside of a facility were excluded (figure 1). This resulted in a total of 12 872 eligible mothers included for analysis. Study population flow diagram. Our main outcome is the women’s report of receiving immediate postnatal health check by a healthcare professional within 24 hours of childbirth while still in the healthcare facility. This was a binary outcome (yes/no). This variable was created with a conceptual link to the WHO Postnatal care recommendations, which state that all women giving birth in healthcare facilities and their babies should remain in the health facility for a minimum of 24 hours following uncomplicated vaginal childbirth, and receive frequent routine postnatal care checks during this period.15 We used four variables to construct this outcome, based on separate questions that women were asked: (1) whether the woman received a postnatal check while still in the facility; (2) length of stay of woman in the facility where the birth took place; (3) timing of the first postnatal check in the facility where birth took place and (4) cadre of professional conducting the first postnatal check on mother. As per the WHO recommendations, we would expect optimal immediate postnatal care to be 100% coverage. There were no differences in the question wording used in the three surveys. Women who reported a stay in the facility of under 24 hours after childbirth needed to have received such a check before discharge. Among women who remained at the facility for 24 hours or more, we used the timing of the postnatal check variable to determine whether the first postnatal health check occurred within 24 hours of childbirth. We categorised health professionals as: doctor, nurse/midwife and medical assistant/clinical officer. To analyse the timing of the postnatal checks within 24 hours among those who received one, we used the women’s response to the question on timing of the first postnatal check in the facility where birth took place and constructed the following categories: (1)<1 hours; (2) 1–4 hours; (3) 4–8 hours; (4) 8–12 hours and (5) 12–24 hours. We were additionally interested in the women’s report of whether babies delivered in health facilities received a postnatal health check by a health care professional within 24 hours of childbirth while still in the healthcare facility or not. This was done to examine missed opportunity where one of the dyad is checked but not the other. To construct this outcome, we used a composite of four variables akin to the process of constructing the maternal receipt of postnatal care outcome. We reviewed existing published literature to form a list of key variables noted to impact coverage of postnatal care to create our conceptual framework (figure 2). These were grouped into four categories (socio-demographic factors, access to healthcare, perceived need, and characteristics of healthcare facility). Some factors fell into multiple conceptual categories which have been highlighted in the shaded intersecting portions of the diagram. Unfortunately for certain factors, there were no matching questions asked in the DHS questionnaire and therefore these factors remained in our framework but were not examined in analysis (these are in italics). Conceptual framework displaying key factors thought to influence coverage of immediate postnatal care. Seven dimensions of perceived need for postnatal care were identified, including wantedness of child (wanted at the time of pregnancy or not), mode of delivery (vaginal or caesarean), and child sex (female, male). Parity was categorised as first birth, 2–3, 4–5 and 6+births. We were not able to include obstetric or neonatal risk factors (eg, maternal comorbidities and fetal abnormalities) because the DHS did not collect data pertaining to these dimensions. For obstetric risk factors, we considered using proxies such as maternal body mass index and anaemia, however, these data related to the woman at time of the survey and not at time of birth. We, therefore, chose not to include these proxies. For neonatal risk factors, we used the proxy perceived size of the baby at birth which was categorised as very small or not. This was reflective of the perceived need for care by mother as babies perceived to be smaller by the mother are more likely to be seen to and checked. For the dimension ‘death of a previous baby’, we constructed a new variable for previous baby death (no child death, child death within 24 hours of birth, child death later than 24 hours of birth). This variable was explored through sensitivity subgroup analysis of women with previous children (parity >1). Antenatal care (ANC) in pregnancy was thought to reflect both perceived need and characteristics of healthcare facility. As those women who received facility based ANC were likely to give birth in that same facility. This dimension was examined by categorising the number of ANC visits during the pregnancy (no ANC, 1–3 visits, 4+visits). Health knowledge and exposure to mass media were thought to reflect the perceived need and sociodemographic factors. There were no questions in the DHS that assessed health knowledge and this dimension could therefore not be analysed further. Exposure to mass media was explored through the variables; any use of television, internet, newspaper and radio (or not) at the time of the survey. We considered eight sociodemographic factors for inclusion into the model. Maternal age group at birth of baby (in 5-year age groups), marital status (married or cohabiting at time of survey or not), highest maternal education level (no education, primary education, secondary and higher education) and ethnicity (Christian, other religions) were assessed. The boundaries of districts and regions changed over the ten-year period covered by the three DHS and were not identical. We, therefore, constructed four larger zones (Eastern, Western, Northern, Central—see online supplemental material 1) which are consistent over time, as done previously.6 bmjgh-2020-004230supp001.pdf Family composition was assessed by number of persons (<4, 4–5, 6+ persons) and number of children under the age of 5 years (0–1, 2–3, 4+) in the woman’s household. Women’s occupation was not examined, as data in DHS pertained to the time of the survey and not at the time of index birth. Household wealth quintile, place of residence (urban vs rural) and the woman’s autonomy were thought to reflect both socio-demographic factors and access to healthcare factors. Household wealth quintiles were provided in the dataset and constructed using principal component analysis of household assets using an established method.6 The dimension of financial autonomy was explored with the binary variable of the woman having a bank account or not. Further exploration of autonomy to healthcare and finances was conducted in sensitivity analysis among married women through the variables who makes decisions about healthcare and finances (respondent alone, respondent and male partner, male partner alone, other). Male partner’s highest education level (no education, primary education, secondary/higher education) was explored further in subgroup analysis among women married at the time of survey. We identified five dimensions related to characteristics of the healthcare facility where the birth occurred. We categorised the sector of the facility as public (government hospital, government health centre, other public sector) or private (private hospital/clinic, other private medical sector). Assistance with the birth was captured by considering the highest cadre listed (doctor/non-physician clinicians, nurse/midwife, other/none). Staff-related factors were conceptually important, but not available on DHS. The dimension patient perceived quality of care was not directly asked within the DHS and no proxies for this dimension could be found. There were no direct variables that asked women to recall the content of their postnatal care. We used the variable of whether the woman reported that the baby was weighed (or not) as a proxy for this dimension as it is reflective of the available staffing, procedures and resources. We were able to assess one dimension—social support and network—for access to healthcare. This dimension was captured by the two created variables: number of persons (<4, 4–5, 6+ persons) and number of children under the age of 5 in the woman’s household (0–1, 2–3, 4+). The distance of the house to the nearest facility or the facility where the birth occurred, or transport/road facilities are not captured on the DHS. Ability to pay for healthcare was thought to reflect both characteristics of healthcare facility and access to healthcare. The variable of whether the woman was covered by health insurance or not was used to reflect this dimension. All analyses were conducted in STATA V.16 SE. Analysis included descriptive statistics of demographic characteristics of women who gave birth in health facilities on all three surveys. Among women who gave birth in health facilities, we computed the percentage who reported receiving immediate maternal postnatal care. Among women with such a check, we described the distribution of the timing of the first check. We calculated the percentage of babies born in health facilities receiving a postnatal check within 24 hours, disaggregated by type of facility. For the 2016 survey, we conducted an analysis of mother–baby dyads and calculated the percentage receiving immediate postnatal care within 24 hours while still in healthcare facilities for mother only, baby only, both and neither. Additionally, we used logistic regression to explore the crude associations between factors outlined in the conceptual framework and the woman’s receipt of immediate postnatal care by a health professional in the facility. A multivariable logistic regression model was created by analysing each individual variable and excluding those that were collinear with existing variables. This enabled the multivariable model to be a reflection of the conceptual model. Two sensitivity analyses using crude and multivariable logistic regression were conducted. First, among the subsample of women married/cohabiting at the time of survey, we additionally included highest level of male partner education, and autonomy with finances and healthcare. Among women with previous children, the model included previous baby death. We used the survey set command to adjust all analyses for survey sampling design and non-response using individual sampling weights, stratification and clustering. There were low levels of missing data in the variables used. We describe how missing values were handled in online supplemental material 2. bmjgh-2020-004230supp002.pdf It was not appropriate or possible to involve patients or the public in the design, or conduct, or reporting, or dissemination plans of our research.