Introduction
Maternal mortality, defined as death of a mother due to complications from child birth or pregnancy, is still an important global public health problem [
1], majorly in low income countries such as Sub-Saharan African countries [
2] due to majorly poor maternal health services utilization (MHSU) [
3]. Globally, approximately 0.3 million women and adolescent girls died in 2015 from pregnancy and childbirth-related complications and 2.6 million stillbirth babies occurred [
4,
5] with 60% of stillbirth occurring during the antepartum period due to untreated infections, poor fetal growth and hypertension [
6]. The Sustainable Development Goal (SDG) 3.1 aims at reducing maternal mortality ratio to less than 70 per 100,000 live births globally by 2030. However, based on recent trends, maternal mortality remains a huge challenge [
7].
In 2015, Uganda was ranked among the top ten countries with the highest maternal mortality in the world, with a maternal mortality rate of 343 per 100 000; and number of maternal deaths of 57,000 mothers [
4]. Stillbirth, maternal mortality and morbidity and other poor maternal health outcomes are high in Uganda and are associated with inadequate utilization of maternal health services, including inadequate utilization of antenatal care (none, incomplete and late ANC attendance), failure to deliver in health facilities, untimely postnatal checkups or no checkups at all [
8‐
14].
The promise of early and full attendance of ANC visits is that, it would improve facility-based deliveries, postnatal care utilization and consequently improve maternal and child health [
15,
16]. During pregnancy, ANC attendance plays an important role towards positive pregnancy outcomes because it is through these visits that screening and treatments of pregnancy complications such as preeclampsia, anemia, sexually transmitted infections, and non-communicable disease such as diabetes is done. Other services provided during this time include weight and height measure, tetanus immunization, provision of supplements such as folic acid, provision of information on behavioral modification and prevention and treatment of intermittent malaria [
17‐
19]. Without proper management of pregnancy, adverse pregnancy outcomes such as low birth weight, preterm delivery, spontaneous abortion maternal and perinatal mortality and morbidity may result [
18,
20].
Most studies have investigated factors affecting ANC, facility delivery, skilled birth attendance and postnatal care, and some studies have investigated how ANC affects neonatal and infant mortality, its association to low birth weight, stunting and underweight [
21] and its relationship with facility-based delivery and perinatal survival [
22]. Through the use of conventional logistic regression, positive associations between ANC attendance on facility-based delivery [
23‐
30] and PNC utilization [
31‐
37] have been observed. In addition, facility-based delivery has been associated with PNC utilization [
38‐
42].
However, ANC in and of itself may not directly result in facility delivery and early PNC utilization, rather it may be due to individual differences in unknown factors that enable facility-based delivery and early PNC among mothers who utilize ANC [
43]. For example, these mothers may be from wealthy households, educated and exposed to media. The use of propensity score matched analysis offers a better option compared to conventional logistic regression analyses in controlling for confounding that may exist in analyzing associations between ANC and facility delivery and early PNC utilization. The use of Propensity Score (PS) matches women who attended 4 + ANC visits (exposed) and those who attended less than 4 ANC visits (unexposed) with similar conditional probabilities of attending 4 + ANC visits hence reducing the bias that may persist when conventional logistic regression is used. This study applied PS matched analysis in examining whether a mother having had four or more ANC visits increases probability of facility-based delivery and early PNC utilization and also whether having facility-based delivery leads to increased probability of PNC utilization in Uganda. Four or more ANC visits were considered because, it is believed that having 4 + visits increases the likelihood of a pregnant woman receiving a full range of required maternal health interventions during pregnancy [
44,
45] and by the time of data collection, the Uganda’s Ministry of Health equally recommended at least 4 or more ANCs for pregnant mothers. Studies that have examined the effect of ANC visits on health outcomes, specifically health facility delivery have used logistic regression models. Analyses using PSM to answer the same research question not only checks on consistency of previous results using another method but also reduces the bias in the intervention effect estimate.
Propensity score analysis (PSA) involves statistical methods for estimating treatment effects with observational data [
46]. It offers an alternative approach for program evaluation in cases where randomized controlled trials are either infeasible, unethical or when researchers need to evaluate treatment effects from survey data. Associations between an outcome and given set of exposures may be biased due to unobservable individual characteristics in survey research. The use of propensity score matching (PSM) reduces such bias by matching women who attended 4 or more ANC visits (exposed) and those who attended less than 4 ANC visits (unexposed) with similar conditional probabilities to receive the treatment and is thus more preferred than traditional regression adjustments, such as logistic regression [
43]. The PS is a balancing score that balances baseline characteristics between the exposed and unexposed groups based on survey data, therefore mimicking characteristics of randomized trials. [
47‐
49]. It also helps create comparable balanced groups of respondents with respect to observed covariates and help minimize the influence of confounders such as age, education level, wealth index [
50‐
53]. Propensity score matched analysis is used to estimate the average treatment effects of the treated (ATT) of a given covariate on outcome of interest [
43,
54]. In this study, we assessed the effect of four or more ANC visits on facility-based delivery, the effect of four or more ANC visits on timing of PNC and the effect of facility-based delivery on timing of PNC using data drawn from Uganda Demographic and Health Survey of 2016.
Discussion
We found that ANC attendance of 4 + visits was associated with a 12% higher probability of health facility-based delivery compared to the same women had they not attended 4 + ANC visits. We also found out that ANC attendance of 4 + visits was associated with a 10% higher probability of early PNC check-up among women compared to the same women had that not attended 4 + ANC visits. The study also revealed that having a health facility-based delivery was associated with 52% higher probability of early PNC check-up compared to the same women had they not had had facility-based delivery.
Literature shows that ANC attendance during pregnancy is positively associated with facility-based delivery [
24,
26,
28,
29] and PNC utilization [
31,
32,
37]. It also shows that facility-based delivery is positively associated with PNC utilization [
41,
42], based on conventional regression models. The present study revealed a significant and positive effect of 4 + ANC visits on facility-based delivery and EPNC utilisation, and facility-based delivery on early PNC utilisation after matching exposed and unexposed women on observable and significant characteristics within 2016 UDHS dataset.
The results align with previous studies which highlighted a positive association between appropriate ANC attendance and facility delivery and PNC utilisation [
24,
26,
28,
29,
31,
32,
34,
37] and with those carried out in Uganda [
11,
13,
62,
63]. Regarding the effect of 4 + ANC visits on facility-based delivery, our results agree with similar studies linking ANC with facility-based delivery in Bangladesh and India that used propensity score matched analysis [
43,
58]. This is likely due to the fact that women who attend ANC receive maternal education and are often referred to health facilities for delivery [
43].
The study further observed that ANC attendance affects early PNC utilisation and also that facility-based delivery affected early PNC utilisation. Studies using propensity score matched analysis investigating these effects could not be found in literature. Overall, the findings of this study confirm the belief that ANC attendance improves the likelihood of facility-based delivery and PNC utilisation and also that facility-based delivery improves the probability of early PNC use.
However, the results from this study are based on observational data to infer causality or causal relationship between; 4 + ANC visits and facility-based delivery, 4 + ANC visits and early PNC utilisation and facility-based delivery and early PNC utilisation. Even though propensity score matching removes bias based on observable woman characteristics, bias due to unobservable confounders is not accounted for leading to overestimated effects of ANC visits on facility-based delivery and early PNC utilisation and facility-based delivery on early PNC utilisation [
43]. However, the use of propensity scores provides a better method for assessing interventions where the use of controlled randomized trial is impossible or inappropriate. It matches the treated with controls based on observable confounders which leads to better estimates of treatment effect. It ensures covariate balance across groups leading to unbiased estimates through the use of observational data.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.