Multilevel and spatial analysis of factors influencing non-institutional delivery care utilization among Ugandan women: insights from the 2016 UDHS

Abstract

Background

While Uganda has made notable strides in increasing facility-based deliveries, a significant proportion of women approximately 25% still give birth outside health institutions. This pattern contributes to the country's high maternal mortality rate, estimated at 336 deaths per 100,000 live births. This study explores both individual and community-level determinants of non-institutional delivery, along with its spatial distribution across Uganda.

Methods

The analysis draws on data from the 2016 Uganda Demographic and Health Survey (UDHS), focusing on a weighted sample of 9,630 women aged 15–49 who had given birth within the five years preceding the survey. The data were collected through a multistage stratified sampling approach. A multilevel logistic regression model was used to examine the predictors of non-institutional delivery, analyzed using STATA version 14.2. Spatial patterns were assessed using ArcGIS 10.6.1 and SaTScan 10.1.3.

Results

Key individual-level factors associated with higher odds of non-institutional delivery included being previously married, having two or more children, and reporting an unintended pregnancy. Conversely, women with secondary or higher education, those in the middle or upper wealth quintiles, those who had four or more antenatal care visits, and those who received high-quality antenatal services were less likely to deliver outside a health facility. At the community level, lower odds of non-institutional delivery were observed among women living in the Northern region, in communities with better access to health facilities, and in areas with higher media exposure. Spatial analysis revealed significant clusters of non-institutional deliveries in the Western and Eastern regions, with some spread into the Northern and Central regions.

Conclusions

Non-institutional deliveries in Uganda are shaped by a combination of individual and community-level factors, with clear spatial disparities. Policy interventions should prioritize expanding access to secondary education, improving the quality and reach of maternal healthcare services, enhancing infrastructure, and addressing socioeconomic inequalities to reduce non-institutional deliveries.

Introduction

Maternal and infant mortality continues to be a key public health concern across most of the developing countries [1]. Worldwide, maternal mortality decreased by 34.3% from 2000 to 2020, from 339 to 223 per 100,000 live births. Notwithstanding these advancements, in 2020, around 287,000 women died due to complications related to pregnancy, childbirth, or postnatal care. Low- and middle-income countries constituted over 95% of all maternal deaths [2]. Similarly, Sub-Saharan Africa and South Asia comprised around 87% of global maternal mortalities, with Sub-Saharan Africa representing about 70% and South Asia approximately 16% [2, 3]. According to the World Health Organization (WHO), one of the strategies to reduce maternal and infant mortality is the access to health facility delivery [4]. Despite global initiatives to enhance facility deliveries, low- and middle-income nations continue to exhibit a substantial disparity in the percentage of expectant women utilising health facilities for childbirth [5]. Evidence indicates that home deliveries offer significant health hazards to both mother and child during the postpartum period compared to health facility deliveries [6, 7]. Some of these risks include preterm birth rates [8], blood loss, fatal death, urine rapture, dizziness, and a tender abdomen, which are common among women but critical if a woman gives birth at home [9]. To improve facility delivery and reduce maternal mortality, the Ugandan government has decentralised health services and expanded Emergency Obstetric and Newborn Care (EmONC) services in hospitals and health centres to unreachable areas [10–12]. The revised National Health Policies and Essential Maternal and Newborn Clinical Care Guidelines frequently demonstrate Uganda’s commitment to reducing maternal and infant mortality [13–16]. However, Uganda still remained one of the countries in sub-Saharan Africa with a high maternal mortality rate, as the country recorded a slow decline in maternal mortality from 418 deaths per 100,000 live births in 2006 to 336 deaths per 100,000 live births in 2016 [17, 18]. The current maternal mortality rates are still high, compared to the global Sustainable Development Goal (SDG) 3.1 of reducing the maternal mortality ratio (MMR) to below 70 per 100,000 live births by 2030 [2, 3]. Socio-cultural and service-related factors, such as financial dependence, control over birth processes, dissatisfaction with institutional care, belief in fate, and access to affordable, supportive, and convenient alternatives, influence the slow decline in maternal mortality rates. Additionally, existing gender and traditional norms limit women’s freedom to make decisions about family or health [19–23]. In addition to socio-cultural and service-related factors, research conducted in developing countries has identified individual and community-level determinants, including maternal age, educational attainment of both mother and partner, marital status, parity, wealth index, media exposure, religion, women’s decision-making capacity, unplanned pregnancies, absence of pregnancy complications, place and region of residence, proximity to healthcare facilities, community media exposure, and women’s autonomy as primary predictors of non-institutional delivery [24–29]. Phillips et al. [30], conducted a thorough literature study and discovered that most research focused on environmental variables, such as urban or rural settings, which may not adequately reflect more specialist metrics like service supply. In Uganda, there are many studies on the determinants of health facility delivery at the sub-national level [31–34], with limited studies at national level [35–38]. Moreover, studies on non-institutional delivery in Uganda are primarily qualitative rather than empirical, and they primarily focus on the sub-national level [19–22]. Additionally, Uganda has limited studies on spatial clusters related to non-institutional delivery [39]. Previous research did not adequately address the geographic patterns of non-institutional deliveries in Uganda and often relied on unweighted data, limiting the accuracy of findings [40]. This study seeks to fill that gap by analysing both individual and community-level factors influencing non-institutional deliveries, using weighted data from the 2016 Uganda Demographic and Health Survey (UDHS). Additionally, pinpointing areas with a high concentration of non-institutional births is crucial for designing focused and effective public health strategies aimed at increasing institutional delivery rates. The insights gained from this analysis are intended to support policymakers in crafting interventions that address determinants at multiple levels, ultimately enhancing maternal and child health outcomes across Uganda.

Methods

Study area, design and sample size

Geographically, Uganda lies across the equator and is a landlocked country. The republic is bordered to the east by Kenya, Tanzania to the south, Rwanda to the southeast, the Democratic Republic of the Congo to the west, and South Sudan to the north. The country covers an area of approximately 241,039 km². The Central, Eastern, and Western regions receive two rainy seasons per annum, with the rainfall decreasing as one travels northward, turning to one rainy season per annum. The fertile soil varies according to region; the central and western regions are generally fertile regions, becoming less fertile as you move to the east and northern regions. The topography ranges from tropical rainforest vegetation in the south to savannah woodlands and semi-arid vegetation in the north. The climate determines the agricultural potential and human population settlement. The central and western regions are densely populated, and the population declines towards the northern region [18, 41] (see Fig. 1a, 1b, 1c in the supplementary file).

Fig. 1.

(a) Shows the road networks of Uganda. (b) Shows the map of Uganda and land use type. (c) Shows the Elevation Map of Uganda

This research utilized secondary data from the 2016 Uganda Demographic and Health Survey (UDHS). The survey applied a stratified, two-stage cluster sampling approach, considering both regional and urban-rural distinctions. The sampling frame was derived from the 2014 Uganda National Population and Housing Census (NPHC), provided by the Uganda Bureau of Statistics, and included 696 Enumeration Areas (EAs). From a sample of 20,791 selected households, 19,938 were found to be occupied, and interviews were successfully completed in 19,588 of them resulting in a 98% response rate. Data were accessed through the MEASURE DHS Program website (https://dhsprogram.com). For the purposes of this analysis, a weighted sample of 9,630 women aged 15–49 who had given birth within the five years before the survey was included. Further methodological details can be found in the full 2016 UDHS report [18].

Study variables

Outcome variable

Non-Institutional delivery was the outcome variable in this study. This variable was categorised into non-institutional and institutional delivery. Non-institutional deliveries was defined as any birth that took place in the woman’s home or others’ home moreover institutional deliveries refers to deliveries that occur in a health facility (public and private sector) This variable is recoded “0” if the births took place in health facility “1” if the births took place at home or others’ home [18, 42].

Individual level variables

The individual-level factors considered in this study include educational attainment, employment status, religious affiliation, number of children ever born (parity), age of the mother at her most recent birth, marital status, sex of the household head, household size, wealth status, pregnancy intention at the time of conception, number of antenatal care (ANC) visits, and the quality of ANC received. Maternal age at last birth was categorized into the following groups: 15–19, 20–24, 25–29, 30–34, and 35 years or older. Education was grouped into four categories: no education, primary, secondary, and higher education. Employment status was classified as either employed or unemployed. Religious affiliation was recoded into four categories: Anglican, Catholic, Muslim, and other (including Seventh Day Adventist, Orthodox, Pentecostal/Born Again/Evangelical, Baptist, Presbyterian, Mammon, Jehovah’s Witness, Salvation Army, Traditional, and other groups). Parity was divided into four groups based on birth order: first, second, third, and fourth or more. Marital status was recategorized into single, married, cohabiting, and previously married (including separated, divorced, and widowed). The sex of the household head was categorized as male or female. The wealth index was recoded into three categories: poor (poorest and poorer), middle, and rich (richer and richest). Household size, ranging from 1 to 27 members, was grouped into three categories: 1–4, 5–9, and 10 or more [43]. Pregnancy intention was classified into three groups: wanted then, wanted later, and not wanted at all. The number of ANC visits was treated as a continuous variable but later recoded into two categories: 1–3 visits and 4 or more visits [18]. Antenatal care quality was measured based on women receiving six essential components of antenatal care services: These measurement includes: weight and blood pressure, collection of urine and blood samples, provision of information on potential complications, and receipt of iron supplements. Women who received all six components were considered to have received high-quality ANC, otherwise classified of low-quality ANC [44]. These variables were selected based on their documented association with non-institutional delivery in previous studies conducted in sub-Saharan Africa [42, 45–48].

Community level variables

Community-level variables were derived by aggregating individual characteristics at the Enumeration Area (EA) level. Depending on the data distribution, variables were standardized using either the mean or the median. A total of seven community-level variables were considered: place of residence, region, and perceived distance to a health facility, community socioeconomic status, ethnic diversity, media exposure, and women’s autonomy at the community level. Place of Residence: This refers to the classification of the area where the respondent stayed on the night of the survey, categorized as either urban or rural. Region of Residence: Uganda was divided into four major administrative regions for this study: Northern (including Teso, Karamoja, Lango, Acholi, and West Nile), Western (comprising Tooro, Ankole, Bunyoro, and Kigezi), Eastern (covering Busoga, Bugishu, and Bukedi), and Central (including Kampala, Central 1, and Central 2) [38]. The average distance to the nearest health facility within each Enumeration Area (EA) was calculated to reflect access to healthcare services. The Ministry of Health guidelines in Uganda stipulate that the physical access distance to a health institution must be within a radius of 5 km [49–51]. A household was considered to have adequate access if it was located within 5 km of the closest health facility; otherwise, access was deemed inadequate [52]. Community socioeconomic status: This variable was derived from the individual wealth index of women, which is based on a composite measure of asset ownership and housing quality. The index includes household possessions (e.g., televisions, bicycles, cars) and indicators of living conditions such as the main source of drinking water, type of sanitation facilities, and flooring materials. In the 2016 UDHS, households were classified into five wealth quintiles: poorest, poorer, middle, richer, and richest. For this study, we recoded the quintiles into two categories: poor (poorest and poorer = 0) and rich (middle, richer, and richest = 1). Because the UDHS did not provide direct measures of cluster-level characteristics, community socioeconomic status was computed by aggregating the proportion of women classified as rich within each cluster. Aggregates were calculated as mean proportions of the rich category. Since these continuous aggregate values may lack practical meaning and were not normally distributed, they were dichotomized using the national median. Clusters with values below the 50th percentile were classified as low community socioeconomic status, and those above were classified as high community socioeconomic status [53, 54]. The median was used due to the skewed distribution of the aggregated data. The ethnic diversity index for each community was calculated using Simpson’s Diversity Index, which considers both the number of ethnic groups present and their relative proportions. Scores ranged from 0.00 to 1.00, with values between 0.00 and 0.40 indicating homogeneity and values from 0.41 to 1.00 indicating heterogeneity [52, 55]. Media saturation at the community level was derived by aggregating the proportion of women exposed to at least one form of mass media (radio, newspaper, or television) in each EA. Due to the non-normal distribution of this variable, the median was used as a cut-off point. Communities with media exposure below 50% were classified as having low media saturation, and those above 50% as high [56]. Women’s autonomy: The Uganda Demographic and Health Surveys (UDHS) did not capture data that can directly describe the characteristics of the clusters. We created women autonomy variable by aggregating the individual women’s characteristics within their clusters. Women’s autonomy was computed on four attributes from the UDHS data to describe women as autonomous, autonomous, or not autonomous. These were: (i) person who usually decides on woman’s health care; (ii) person who usually decides on large household purchases; (iii) person who usually decides on visits to family or relatives; and (iv) person who usually decides what to do with money husband earns. The categories of the women’s autonomy variable were based on the number of these decisions, in which women participate, as: involved in at least one of the four decisions was coded 1 otherwise not involved in any of the decisions was coded 0. In the current study, a woman was considered to be autonomous if the decision on any one of the above items was made by the woman alone or together with her partner. Since the aggregate values may not have pragmatic meaning, we categorized the aggregate values of a cluster into groups based on national median values, since all aggregates were not normally distributed. The aggregated value that falls below 50% of the national median value was classified as not autonomous and those above 50% as autonomous [57].

Data management and analysis

Before conducting the descriptive analysis, we applied appropriate weighting to the data, incorporating sampling weights, primary sampling units, and strata to ensure the survey’s results were representative of the target population. We utilized the “svyset” command in STATA version 14.2 to adjust for the complex survey design, allowing for accurate estimation of standard errors and valid statistical inferences [40].

Spatial analysis

We conducted spatial data analysis using ArcGIS version 10.6.1 [58] and SaTScan version 10.1.3 [59]. To assess the presence of spatial autocorrelation in the distribution of non-institutional deliveries in Uganda, we applied the Global Moran’s I statistic [60]. This helped determine whether the pattern was clustered, random, or dispersed. Spatial clustering, including hotspot and coldspot areas, was identified using the Getis-Ord Gi* statistic. To estimate values in locations without sampled data, we used Empirical Bayesian kriging based on observed values [61]. Spatial scan statistics were employed to detect significant clusters, both primary and secondary. For this analysis, we applied the Bernoulli model using Kulldorff’s method, as the outcome variable (place of delivery) is binary. Women who delivered outside health facilities were considered cases, and those who delivered in health institutions were considered controls. The Bernoulli model required information on the number of cases and controls, along with their geographic locations. A default maximum spatial cluster size of less than 50% of the population was set to capture both small and large clusters, with larger ones automatically excluded. Identified clusters were ranked by their log-likelihood ratios and assessed for statistical significance using 999 Monte Carlo simulations [62].

Statistical analysis

Descriptive statistics, expressed as percentages, were used to summarize key variables at the univariate level. At the bivariate level, frequencies and cross-tabulations helped examine how outcome variables varied across individual and community-level factors. The Pearson chi-square test assessed associations between these characteristics. Variables with p-values ≤ 0.25 in the bivariate analysis were further evaluated for multicollinearity using the Variance Inflation Factor (VIF), and those with VIF values exceeding 10 were excluded. The analysis revealed no significant multicollinearity (mean VIF = 1.28, minimum = 1.01, maximum = 1.98), allowing the remaining variables to be included in the final multilevel logistic regression model. Given the binary nature of the outcome variable, a multilevel binary logistic regression approach was employed to explore the influence of individual and contextual factors on non-institutional deliveries. The log of the likelihood of the j^th category of the outcome variable for the i^th observation was modelled as follows [63]:

1

Where β_0j is average intercept γ₀₀ including group-dependent deviation U_0j: that is β_0j = γ₀₀ + U_0j and β_kj = γ_k0 + U_kj. Now, the above two-level model (Eq. (1)) can be restated as more generally, as it contains all levels of variation:

2

In this context, γ₀₀ represents the mean regression intercept, X_kij signifies the k level-one covariates, and Z_qj indicates the q level-two covariates. Additionally, γ_k0 and γ_0q are the respective regression coefficients for level-one and level-two explanatory variables, while U_0j and U_kj denote the random effects of the model parameters at level two. The vector of random effects U_0j, U_1j, U_2j, ., U_kj is anticipated to adhere to a normal distribution with a mean of zero and a variance of σ²u₀. To evaluate the association between individual and contextual factors with non-institutional delivery, four multilevel binary logistic regression models were applied. The first model (Model 1), known as the null or empty model, included no predictors and served to assess the variance between enumeration areas (EAs). The intra-class correlation coefficient (ICC) was computed to determine whether the multilevel structure of the data warranted this analytical approach. Model 2 included only individual-level predictors to assess their impact on the likelihood of non-institutional deliveries. Model 3 incorporated community-level predictors while excluding individual-level ones. Finally, Model 4 combined both individual- and community-level predictors to examine their joint fixed and random effects. The ICC was used again in this model to quantify the proportion of total variance attributable to between-community differences, using the formula ICC = σu² / (σu² + π²/3). Fixed effects were interpreted through odds ratios, and statistical significance was determined at a threshold of p < 0.05. Model fit was assessed using the Akaike Information Criterion (AIC) [64].

Results

Background characteristics of respondents

Table 1 presents the percentage distribution of women by their background characteristics, based on births that occurred within the five years preceding the 2016 Uganda Demographic and Health Survey (UDHS). The weighted sample consisted of 9,630 women aged 15–49 who reported giving birth during that period. Among them, 2,408 women accounting for 25% delivered outside a health facility. The majority (53.3%) were aged 20 to 29, while adolescents (aged 15–19) represented 13.7% of the sample. In terms of marital status, 81.3% were either married or cohabiting, and 73.1% resided in male-headed households. Over half of the women (56.4%) came from households with five to seven members. Educationally, 60% had attained primary-level education, while around 30% had completed secondary education or higher. Regarding economic status, 41.3% fell into the poor wealth category. Religious affiliation was predominantly Anglican or Catholic, comprising 71.2% of the respondents. Employment was also common, with 79% of women reporting having worked in the 12 months prior to the survey. In terms of pregnancy intention, 55.3% reported that their pregnancies were wanted at the time of conception; whereas 11% stated that the pregnancy was not desired at that time. ANC attendance showed that 61.4% of the women had at least four antenatal visits. However, 74% received poor quality ANC. At the community level, 77% of the respondents lived in rural areas, and 74% resided more than five kilometers from a health facility. Additionally, 50.4% were from communities characterized by low socioeconomic status. A large proportion (80%) lived in communities with low ethnic diversity, and half (50.4%) were from areas with limited media exposure. Overall, slightly more than half (54.9%) were part of communities with lower socioeconomic standing (Table 1).

Table 1.

Shows percentage of non-institution delivery by their individual-level characteristics

Variables	Weighted sample N = 9630 (%)	Facility delivery (%)	Home delivery (%)	P-value
Deliveries
Home	7179 (75.3)
Facility	2451(24.7)
Individual variables
Age at last birth				< 0.001
<=19	1310 (13.7)	81.7	18.3
20–29	5095 (53.3)	77.5	22.5
30–39	2707 (17.7)	70.2	29.8
40+	518 (5.3)	63.7	36.3
Marital Status				< 0.001
Single	552 (5.8)	86.5	13.5
Married	4099 (40.5)	74.3	25.7
Living together	3807 (40.8)	75.4	24.6
Previous married	1172 (12.9)	73.4	26.6
Household Head				0.073
Male	7038 (73.1)	74.8	25.2
Female	2592 (26.9)	76.8	23.2
Parity				< 0.001
1	1891 (20.1)	86.9	13.1
2	1734 (18.5)	80.7	19.3
3	1474 (15.5)	79.0	21.0
4+	4531(45.9)	66.8	33.2
Household Size				< 0.001
1–4	3123 (33.4)	79.6	20.4
5–9	5486 (56.4)	72.8	27.2
10+	1021 (10.2)	75.0	25.0
Level of Education				< 0.001
No education	1215 (10.5)	63.5	36.5
Primary	5888 (60.0)	70.0	30.0
Secondary or Higher	2527 (29.5)	90.3	9.7
Wealth Index				< 0.001
Poor	4512 (41.3)	65.6	34.4
Middle	1780 (18.9)	71.9	28.1
Rich	3338 (39.8)	87.0	13.0
Religion				< 0.001
Anglican	3048 (31.2)	74.0	26.0
Catholic	4000 (39.5)	71.0	25.9
Muslim	1099 (13.7)	83.1	16.9
Other’s	1483 (15.6)	74.1	25.9
Employment Status				< 0.001
Unemployed	1911 (21.0)	79.7	20.3
Employed	7719 (79.0)	74.1	25.9
Pregnancy wontedness				< 0.001
Wanted then	5366 (55.3)	77.0	23.0
Wanted later	3202 (33.7)	75.5	24.5
Wanted no more	1059 (11.0)	66.1	33.9
Number of ANC visits				< 0.001
1–3	3665 (38.6)	67.6	32.4
4+	5751 (61.4)	81.2	18.8
Quality of ANC content				< 0.001
Low	7171 (73.8)	71.0	29.0
High	2459 (26.2)	87.3	12.7

Community variables
Place of Residence				< 0.001
Urban	1994 (23.4)	88.8	11.2
Rural	7636 (76.6)	71.2	28.8
Region of residence				< 0.001
Central	1875 (27.3)	81.8	18.2
Eastern	2538 (26.7)	71.7	28.3
Northern	2560 (20.6)	78.1	21.9
Western	2657 (25.4)	69.9	30.1
Community Distance to Health Facility				< 0.001
Inaccessible	6624 (58.6)	67.8	32.2
Accessible	3521 (41.4)	86.1	13.9
Community Ethnicity Diversity Index				0.099
Less Diverse	7565 (79.6)	74.0	26.0
More Diverse	2065 (20.4)	77.0	23.0
Community Socioeconomic Status				< 0.001
Low	5519 (50.4)	67.8	32.2
High	4111 (49.6)	83.0	17.0
Community Media Saturation				< 0.001
Less Saturated	5534 (50.4)	73.6	26.4
Saturated	4096 (49.6)	75.4	22.6
Community level of autonomy				0.046
Low	4928 (54.9)	73.6	26.4
High	4702 (45.1)	77.4	22.6
Total	9630	75.0	25.0

Prevalence of non-institutional delivery across explanatory variables

Table 1 depicts the percentage distribution of women who had their last birth within the five years preceding the 2016 UDHS. A quarter (25%) of the women reported to have had Non-Institutional delivery. Individual-level characteristics such as age at last birth, marital status, sex of the household head, parity, family size, wealth index, religion, employment status, pregnancy intention, number of ANC visits, quality of ANC content, and community level characteristics such as place of residence, region, community distance to health facility, community socioeconomic status, ethnicity diversity index, media saturation and level of autonomy were significantly associated Non-Institutional delivery. Regarding age, women under 19 years of age had a low proportion (18%) of non-institutional delivery than women aged 40 years and above (36.3%). A lower proportion (13.5%) of single women had non-institutional delivery compared to married, living together, and formerly married women (25.7%, 24.6%, and 26.6%, respectively). A quarter (25.2%) of the women from male-headed compared to female headed households (23.2%) had non-institutional delivery. Regarding parity, 33.2% of women who had four or more births had higher proportion of non-institutional delivery compared to those who had one births (19.5%). Above a quarter (27.2%) of the women who belong to households with five to nine members (5–9) had non-institutional delivery compared to those from families of less than four (20.4%). A higher proportion of women with no education (36.5%) had non-institutional delivery. Women in the poor wealth index bracket had higher proportion (34.4%) of non-institutional delivery than those in the rich wealth index brackets (13.0%). A significant low proportion (16.9%) of women from Muslim religion had non-institutional delivery. A quarter (25.9%) of employed women had non-institutional delivery. Women who wanted pregnancy no more had a higher proportion (33.9%) of non-institutional delivery compared to women who wanted pregnancy then (23.0%). Women who had fewer antenatal care visits (1–3) had higher (32.4%) non-institutional delivery compared to those who had at least four or more visits (18.8%). A higher proportion (29.0%) of women who received poor-quality ANC had non-institutional delivery. Furthermore, a significant proportion (28.8%) of women in rural areas reported having had non-institutional delivery compared to women from urban areas (11.2%). A significant proportion of women from the Eastern and Western regions (28.3% and 30.1%) had non-institutional delivery compared to women in the Central (18.2%). A higher proportion (32.2%) of women who lived in communities far from health facilities had non-institutional delivery than their counterparts (13.9%). Less ethnically diverse communities (26.0%) had non-institutional delivery than their counterparts (23.0%). Compared to their counterparts (32.2%), less than 17.0% of women from low socioeconomic status communities use non-institution delivery. Furthermore, in communities with low media saturation, 26.4% of women had non-institutional delivery, compared to 22.6% in communities with higher media saturation. Finally, women from communities with low autonomy (26.4%) had more non-institution than women with high autonomy (22.6%) (Table 1).

Individual and community-level determinants of non-institutional delivery

Random effect and model fit statistics

The community-level variance was found to be 2.28 (95% CI: 1.83–2.84, p < 0.001), indicating significant differences between clusters in relation to non-institutional delivery. The intraclass correlation coefficient (ICC) from the empty model was 41%, reflecting substantial variability attributable to differences between communities. The remaining 59% of the variation could be explained by individual or other independent factors. After incorporating individual-level variables in Model 2, the ICC decreased to 0.34, suggesting that 34% of the variability remained at the community level. When community-level variables were included in Model 3, they explained 31% of the variation (ICC = 0.31). In Model 4, which combined both individual and community-level factors, the ICC was 0.30, indicating that 30% of the variability in non-institutional delivery was still attributable to community-level differences. These findings highlight the role of community characteristics in influencing non-institutional delivery rates. Additionally, a progressive reduction in Akaike Information Criterion (AIC) values from Model 1 through Model 4 indicated improved model fit. Consequently, Model 4, with the lowest AIC, was identified as the best-fitting model for predicting non-institutional delivery among women in Uganda (Table 2).

Table 2.

Multilevel analysis of odd ratios on the effect of Individual, household and community level factors in influencing non-institutional delivery among women in Uganda, UDHS 2016

Variables	Model 1 Null model	Model 2 (AOR 95%CI)	Model 3 (AOR 95%CI)	Model 4 (AOR 95%CI)
Individual variables
Age at last birth
<=19		1.00		1.00
20–29		0.91 (0. 69-1.21.21)		0.93 (0.70–1.23)
30–39		0.83 (0.60–1.14)		0.84 (0.61–1.16)
40+		0.97 (0.66–1.42)		0.99 (0.67–1.44)
Marital Status
Married		1.00		1.00
Never married		0.99(0.68–1.46)		0.95 (0.65–1.41)
Living together		1.02(0.68–1.50)		1.06 (0.88–1.26)
Previous married		1.31 (1.02–1.67)**		1.58 (1.19–2.11)**
Household Head
Male		1.00		1.00
Female		0.95 (0.79–1.13)		0.94 (0.78–1.12)
Family Size
1–4		1.00		1.00
5–9		0.90 (0.76–1.07)		0.90 (0.75–1.07)
10+		0.88 (0.67–1.15)		1.13 (0.67–1.15)
Parity
1		1.00		1.00
2		2.08 (1.58–2.72)***		2.03 (1.55–2.66)***
3		2.07 (1.51–2.83)***		2.03 (1.47–2.78)***
4+		3.44 (2.53–4.68)***		3.34 (2.45–4.56) ***
Level of Education
No education		1.00		1.00
Primary		0.97 (0.79–1.18)		1.01(0.82–1.24)
Secondary or Higher		0.52(0.39–0.68)***		0.58 (0.44–0.78)***
Wealth Index
Poor		1.00		1.00
Middle		0.73 (0.62–0.87)***		0.75 (0.63–0.90)***
Rich		0.56 (0.45–0.69)***		0.67 (0.52–0.85)***
Religion
Anglican		1.00		1.00
Catholic		1.03 (0.87–1.21)		1.04 (0.88–1.23)
Muslim		0.95 (0.72–1.26)		0.97 (0.73–1.29)
Other’s		0.96 (0.75–1.22)		0.95 (0.75–1.21)
Employment Status
Unemployed		1.00		1.00
Employed		1.11 (0.93–1.32)		1.18 (0.99–1.42)*
Pregnancy intention
Wanted then		1.00		1.00
Wanted later		1.20 (1.03–1.39)*		1.21 (1.04–1.41) **
Wanted no more		1.16 (0.92–1.46)		1.15 (0.91–1.45)
Number of ANC visits
1–3		1.00		1.00
4+		0.59 (0.52–0.67)***		0.59 (0.52–0.68) ***
Quality of content of ANC
Low		1.00		1.00
High		0.52 (0.44–0.61)***		0.53 (0.45–0.62)***
Community variables
Place of Residence
Urban			1.00	1.00
Rural			1.41(1.02–1.94)**	1.33 (0.96–1.83) *
Region of residence
Central			1.00	1.00
Eastern			1.01(0.68–1.50)	0.96 (0.65–1.44)
Northern			0.38 (0.23–0.65)***	0.34 (0.21–0.56)***
Western			1.18 (0.75–1.86)	1.09 (0.70–1.71)
Community Distance to Health Facility
Inaccessible			1.00	1.00
Accessible			0.58 (0.48–0.70)***	0.72 (0.67–0.94)**
Community Ethnicity Diversity Index
Less Diverse			1.00	1.00
More Diverse			0.91(0.68–1.21)	0.89 (0.67–1.18)
Community Socioeconomic status
Advantaged			1.00	1.00
Disadvantaged			0.54 (0.39–0.74)***	0.75 (0.55–1.02)*
Community Media Saturation
Less Saturated			1.00	1.00
Saturated			0.47(0.34–0.65)***	0.56 (0.42–076.42)***
Community level of autonomy
Low			1.00	1.00
High			0.75 (0.54–1.03)*	0.82 (0.61–1.11)
Random effect results:
PSU Variance (95% CI)	2.28(1.83–2.84)	1.72(1.38–2.14)	1.46(1.17–1.83)	1.44(1.16–1.78)
ICC	0.41	0.34	0.31	0.30
Wild chi-square and p-value	Ref	χ2 = 332.10, p < 0.003	χ2 = 198.89,p < 0.000	χ2 = 459.00, p < 0.000
Model fitness:
Log-likelihood	−1693935.8	−1518822.8	−1663282.5	−1502350.5
AIC	3,387,890	3,037,698	3,326,587	3,004,771
PSU	649	649	649	649
N	9,630	9,630	9,545	9,333

The Null model contains no predictor variables, but it partitioned the variance into two component parts

AIC Akaike Information Criterion, CI Confidence Interval, ^® Reference category, AOR Adjusted Odd Ratio Significant level

***, **, and * denote 1%, 5% and 10% respectively

The fixed effect analysis results

The results from the multilevel binary logistic regression analysis revealed significant associations between non-institutional delivery and several factors, including marital status, parity, educational attainment, wealth status, pregnancy intention, antenatal care (ANC) attendance and quality, region, proximity to health facilities, and exposure to media. Previously married women had a 67% higher likelihood of delivering outside a health facility compared to single women (AOR = 1.67, 95% CI: 1.08–2.58). Similarly, women with two, three, and four or more previous births had significantly increased odds of non-institutional delivery—103% (AOR = 2.03, 95% CI: 1.55–2.66), 102% (AOR = 2.03, 95% CI: 1.47–2.78), and 234% (AOR = 3.34, 95% CI: 2.45–4.56), respectively compared to first-time mothers. In contrast, having secondary or higher education was associated with a 42% reduction in the odds of non-institutional delivery (AOR = 0.58, 95% CI: 0.44–0.85), relative to women with no formal education. Women from middle and high wealth households were also less likely to deliver outside health facilities, with reduced odds of 25% (AOR = 0.75, 95% CI: 0.63–0.90) and 33% (AOR = 0.67, 95% CI: 0.52–0.85), respectively, compared to women in the poorest category. Women who reported wanting their pregnancies at a later time had a 21% increased likelihood of non-institutional delivery (AOR = 1.21, 95% CI: 1.04–1.41) compared to those whose pregnancies were desired at the time. Attending four or more ANC visits was associated with a 41% decrease in the odds of non-institutional delivery (AOR = 0.59, 95% CI: 0.52–0.68), compared to attending fewer visits. Similarly, women who received high-quality ANC services had 47% lower odds of non-institutional delivery (AOR = 0.53, 95% CI: 0.52–0.85) than those who received lower-quality care. Regional differences were also observed; for instance, women in the northern region were 66% less likely to deliver outside a health facility than those in the central region (AOR = 0.34, 95% CI: 0.21–0.56). Accessibility to health facilities significantly influenced delivery location women living closer to facilities had 33% reduced odds of non-institutional delivery (AOR = 0.67, 95% CI: 0.52–0.85). Finally, living in areas with high media exposure was associated with a 44% reduction in the odds of non-institutional delivery (AOR = 0.56, 95% CI: 0.42–0.76), compared to low-media-exposure areas (Table 2).

Spatial analysis results

Spatial autocorrelation analysis

The global spatial autocorrelation analysis revealed that the spatial distribution of non-institutional delivery was significantly varied across Uganda with Global Moran’s I value of 0.50 (p < 0.000). The clustered pattern (the red box on the right) demonstrates a clustering effect on non-institutional delivery among Ugandan women. The exits have keys generated automatically on each panel’s right and left sides. The z-score of 23.53 indicated that there is less 1% likelihood that this clustered pattern can be the results of random chance. The lighter blue and red colours, up to the tails, have a higher level of significance. The table shows that the observed value is higher than the expected value, and the p < 0.05, indicating that the difference is statistically significant. Thus, there is a spatial dependence on non-institutional delivery among pregnant women in Uganda (Fig. 2).

Fig. 2.

Spatial autocorrelation for non-institutional delivery among women in Uganda, UDHS 2016

Hotspot analysis of non-institutional delivery in Uganda

The Getis-Ord GI* analysis was used to detected hotspot the two extreme areas (hotspots and cold spots). The hotspots depict areas of substantial clusters of statistical significant areas high prevalence non-institutional delivery, while cold spots signify clusters of low prevalence of non-institutional delivery. The hotspots areas include the dots in red colour which were found in the central region; most of the clustering occurred in eastern region, northern region and western Uganda. The cold spot areas include dots in blue colour which was found in Central region around the capital city (Kampala), northern and some few cold spots western and eastern region (Fig. 3).

Fig. 3.

Hotspot analysis results for non-institutional delivery across the regions in Uganda, UDHS 2016

Interpolation of non-institutional delivery in Uganda

Spatial interpolation was conduct using empirical Kriging technique to estimate the spatial distribution of non-institutional delivery for areas of unsampled across Uganda. This measured the distance from the known point to predict unknown points and showed the point in the ranges of non-institutional delivery occurrences. More than 37% of the women who lived in an area with red colour had non-institution delivery. In contrast, less than 16% of women living in green areas had a non-institutional delivery (Fig. 4).

Fig. 4.

Spatial Interpolation of non-institutional delivery among women in Uganda, UDHS 2016

SaTScan spatial analysis on non-institutional delivery in Uganda

In Uganda, the SaTScan spatial analysis identified seven statistically significant clusters with high non-institutional delivery. This means that the prevalence of non-institutional delivery was higher within the SaTScan circular window than outside the SaTScan window. With a relative risk (RR) of 1.63 and a log-likelihood ratio of 42.8 at p < 0.0001, the primary cluster spatial window was in the western region, centered at 0.516256 N, 31.053829 E/75.24 km of geographical location radius. It revealed that women who reside within the spatial window were 40.82 times more likely to have non-institutional delivery compared to women who reside outside this window. The secondary cluster spatial window was located in the eastern region, centered at 1.493320 N, 33.928590 E/158.55 km of geographical location radius, with a relative risk of 1.35 and a log-likelihood ratio of 38.30 at p < 0.000. It revealed that women who reside within the spatial window were 38.30 times more likely to have non-institutional delivery compared to women who reside outside this window. The tertiary cluster spatial window was located in the western region, centered at 1.924309 N, 31.963844 E/63.36 km of geographical location radius, with a relative risk of 1.68 and a log-likelihood ratio of 20.41 at p < 0.000. It revealed that women who reside within the spatial window were 20.41 times more likely to have non-institutional delivery compared to women who reside outside this window (Table 3 and Fig. 5, respectively).

Table 3.

Spatial SaTScan analysis on non-institutional delivery utilization, a bernoulli based model

Cluster	Enumeration Area(Clusters) identified	Coordinate or Radius	RR	LLR	P-value
1	545, 507, 506, 509, 508, 510, 109, 108, 514, 497, 511, 133, 546, 513, 490, 542, 556, 482, 491, 505, 106, 496, 517, 516, 554, 107, 488, 495, 515, 484, 492, 489, 558, 487, 486, 540, 541, 65, 539, 52, 483, 493, 110, 571, 512, 590, 485	(0.516256 N, 31.053829 E)/75.24 km	1.63	42.82	< 0.001
2	297, 268, 267, 270, 266, 271, 264, 265, 296, 260, 269, 211, 262, 263, 282, 209, 207, 261, 293, 276, 206, 290, 210, 205, 258, 275, 220, 298, 281, 208, 202, 299, 212, 280, 242, 214, 229, 249, 274, 253, 228, 294, 248, 203, 278, 287, 254, 279, 245, 255, 288, 213, 243, 256, 226, 222, 225, 221, 259, 247, 204, 244, 241, 246, 200, 170, 199, 219, 273, 201, 223, 277, 250, 195, 171, 240, 238, 227, 173, 224, 291, 198, 295, 257, 234, 239, 237, 309, 197, 286, 168, 188, 231, 300, 232, 302, 172, 285, 292, 283, 251, 233, 289, 272, 192, 235, 167, 189, 230, 310, 305, 252, 236, 194, 187, 218, 330, 193, 190, 155, 303, 196, 284, 217, 151, 377, 153, 169, 191, 164, 370, 307, 160, 183, 335, 180, 369, 340, 162, 175, 301, 368, 154, 176, 148, 163, 149, 150, 216, 161, 215, 342, 304, 184, 343, 144, 166, 371, 152, 159, 186, 311, 329, 313, 373, 336, 145, 312, 338, 120, 668, 306, 671, 165, 328, 181, 174, 122, 146, 669, 121, 158, 157, 349, 372, 683, 341, 337, 670, 156, 345, 185, 147, 182, 378, 177, 308, 363, 123, 673, 315, 374, 331, 672, 143, 179, 346, 334, 327, 375, 178, 141, 664, 316, 333, 677, 339, 675, 347, 379, 657, 680, 682, 676, 361, 352, 678, 681, 674, 142, 317, 127, 679, 684, 662, 362, 667, 351, 314, 100, 117, 656, 129	(1.493320 N, 33.928590 E)/158.55 km	1.35	38.30	< 0.001
3	462, 499, 464, 498, 463, 470, 469, 501, 500, 465, 367, 356, 471, 468, 466, 364, 365, 467, 407	(1.924309 N, 31.963844 E)/63.36 km	1.68	20.41	< 0.001
4	432, 434, 430, 442	(3.139208 N, 30.852192 E)/12.88 km	2.34	15.98	< 0.001
5	587, 568, 564, 586, 563	(0.883288 S, 30.545914 E)/20.41 km	2.15	15.81	< 0.001
6	559, 61	(0.794600 S, 31.139360 E)/9.52 km	2.76	15.02	< 0.001
7	394	(2.940537 N, 31.776773 E)/0 km	3.41	13.60	< 0.001

RR Relative risk, LLR Log likelihood ratio

Fig. 5.

Spatial scan statistical analysis high likely clusters for non-institutional delivery among women in Uganda, UDHS 2016

Discussion

This study investigated both individual and community-level factors influencing non-institutional childbirth in Uganda, along with its spatial distribution, using nationally representative data from the 2016 Uganda Demographic and Health Survey (UDHS). The findings revealed that in 2016, approximately a quarter (25%) of women gave birth outside of health facilities. Moreover, the results indicated a decline in the rate of non-institutional deliveries compared to earlier UDHS rounds, beginning from 2000 to 01 up to 2016 [18]. However, this proportion was lower to that of Ethiopia (52.5%) [42], Nigeria (56.8%) [45], and sub-Saharan Africa (31.2%) [28] region in general. The variations on non-institutional deliveries among these countries can be associated to differences in the level of socio-economic development, health infrastructure, health system policy, cultural differences, and attitude of women towards place of delivery [28]. The study also found that women who were previously married, had given birth multiple times (first through fourth or more), or wished to delay their next pregnancy were more likely to deliver outside of a health facility. In contrast, the likelihood of non-institutional delivery decreased among women with at least a secondary education, those in the middle or upper wealth brackets, and those who attended four or more antenatal care visits. Additional protective factors included receiving high-quality antenatal care at both the individual and community levels, living in the northern region, residing in communities with easy access to health facilities, and being part of communities with widespread media exposure. The odds of non-institutional delivery were increased among the previously married women compared to women who were single. Similarly, Shah and Ali [23] and Regassa et al. [29] corroborates this findings. Contrary to these findings, in Ethiopia studies found no association between marital status and non-institutional delivery [53, 65, 66]. The study revealed that women who had been previously married, had multiple childbirth experiences (ranging from their first to fourth or more), or intended to postpone their next pregnancy were more inclined to give birth outside health facilities. Conversely, the chances of delivering outside a medical institution were lower among women with at least secondary education, those in middle or higher income groups, and those who attended four or more antenatal care (ANC) visits. Other factors associated with increased use of health facilities for delivery included receiving high-quality ANC at both the individual and community levels, living in the northern region, residing in areas with better access to healthcare services, and being part of communities with high media exposure [34, 62]. Similarly, the odd of having non-institutional delivery is higher among women who desired a pregnancy later than women who desired pregnancy. This finding is similar to those of Geta et al. [67], in Ethiopia where they found an association between planned pregnancy and non-institutional delivery. However, other studies found no association between planned pregnancy and non-institutional deliveries [68]. This can be attributed to mothers who have planned their last pregnancy having an interest in seeking medical care, following health recommendations, and cooperating with their partners compared to their counterparts [67]. Furthermore, we hypothesize that women who did not want to get pregnant when they did could be associated with a lack of finances to pay for transportation as well as contraceptives. The odds of non-institutional delivery were higher among women who had two or more deliveries. Similar results were found among other studies [24, 42, 46, 69] which show that women had two or more births were more likely to have non-institutional deliveries compared to those who had one birth. The result of this study contradicts the study conducted in sub-Saharan Africa which established no statistical significance between non-institutional delivery and parity [28]. The plausible explanation is that women who had at least two births have more childbearing experience and confidence compared to those who had their first birth [19]. Besides, Atukunda et al. [19]., argued that pregnant women prefer non-institutional deliveries for future pregnancies because their first hospital birth was unsatisfactory. This study found that having secondary or higher education decreased the odds of women having non-institutional delivery. This has been corroborated by several studies in sub-Saharan Africa [28, 53, 66, 70] and Asia [24, 26] who established similar findings. Educated women have a better understanding, are more easily exposed to the media, live in urban areas, and are more aware of PNC, pregnancy warning signs, childbirth risks, and the quality of services. However, the level of decision-making within their families and the socio-cultural context may influence this variable [28, 53, 70]. Therefore, this calls for promotion of continued universal secondary education policy. Similarly, the odds of non-institutional delivery decreased among wealthier index bracket status [65, 69], this may be due to the limited economic budget available to pregnant women from lower socioeconomic classes when they have to pay various fees, such as transfer to the hospital and the delivery service itself in private for profit health facility [53, 69]. Poor women are more likely to seek assistance from Traditional Birth Attendant (TBA) compared to the rich who can afford the cost of skilled health personnel [71]. Similarly, frequency of ANC attendance reduces non-institutional delivery, and this finding is consistent with other studies from Ethiopia [42, 53], the plausible explanation is that frequency of visits to health facility help a woman to build rapport with the service providers, in return they receive enough advice which encourage them to deliver in health facility [53]. In addition, women who absconded from the ANC class missed an opportunity to receive quality ANC and had a higher likelihood of non-institutional delivery compared to their counterparts. In this study we found that women who received high quality antenatal care content were less likely to have non-institutional delivery [54], quality of antenatal content is often associated with adequate utilization of ANC as well as postnatal care services [52, 54]. At the community level, we found community access to media to be a significant factor. This finding was in line with other study by Regassa et al., [29], however, some studies in sub-Saharan Africa found no association between community media exposure and non-institutional delivery [28, 53, 65], despite the lack of significance between non-institution delivery and media exposure, the plausible explanations is linked to socio-cultural variation and community trust in the media among this countries. However, we cannot underestimate the role of the media such as the radio, newspapers and television in spreading information, hence service providers use these media to persuade mothers to adopt a positive attitude towards the use of health facilities for delivery [29]. Therefore, different forms of media covers wide range of distance across different regions in terms of information coverage. Similarly, the odds of non-institution deliveries decreased with distance, this was corroborated by previous study from sub-Sharan Africa [28, 46, 65, 66], and Bangladesh [71], which found distance as a significant factor. Distance as a geographic factor is an obstacle toward accessing health facilities across sub-Saharan Africa. This is due to the fact that most population are rural based and majority of the advanced health facility infrastructure and services are mainly based in urban areas [28]. Therefore, due to the disparities of uneven development resulted to poor road network, women have to travel long distance to access maternity services; this is compounded with lack of financial resources to pay for transport and back door payments to access services at government health facilities acted as a disincentive for women to use health facilities for deliveries [28, 46]. This study also found that regional variation as an important factor. This finding is consistent with other Ugandan researches [35, 36, 38]. These regions’ histories provide plausible explanations for why women in the northern region were less likely to give birth outside of a hospital. Compared to the central region, which houses the capital city of Kampala, northern Uganda is underdeveloped in terms of development, despite being home to the most advanced health facilities. Since 1986, war has ravaged northern Uganda, including Acholi, Lango, and West Nile, with peace being restored in 2005. As a result of this protracted conflict, the majority of individuals were living in displaced refugee camps in northern Uganda. Despite the impoverished conditions of this region, the government, through the Ministry of Health and various non-governmental organizations (NGOs) such as the United Nations High Commissioner for Refugees (UNHCR) and the Uganda Red Cross Society (URCS), has redirected its focus to the greater northern region by providing health services at no cost. This initiative aims to alleviate poverty among the displaced population, thereby enhancing accessibility to healthcare for women residing in displacement camps within this area [35, 36, 38]. The geographic disparities in non-institutional delivery are attributable to sociocultural factors, such as the preference for home births in certain rural communities in Uganda, as well as the varying levels of infrastructural development, including health facilities, across different regions of Uganda [19–22].

Strengths and limitations of the study

This study adopted an improved approach to determining level weights for multilevel model analysis. The advantage of applying level-weights is that multilevel modelling is difficult to apply in DHS data because DHS databases lack sampling weights that correspond to the levels of the two-stage sampling technique. Therefore, researchers using the Uganda Demographic and Health Surveys (UDHS) dataset rarely apply approximated level weights, which aid in resolving non-response issues and ensuring data representativeness at the cluster and individual levels. Furthermore, to establish the geographical distribution of non-institutional delivery in Uganda, this study used a combination of spatial and multilevel analysis methodologies, which are relatively rare in the Ugandan context. ArcGIS and spatial SaTScan software provide a visual aid in viewing Uganda’s primary windows of non-institutional delivery. Additionally, we used a near function in ArcGIS software to measure the direct distance from the enumerated areas (EAs) to health facilities. We averaged the distances from each EA and assigned them to the corresponding EAs, thereby establishing the community distance to a health facility. The secondary nature of the data in this study hinders the establishment of missing values in the dataset and the confirmation of coordinates outside the study area. The study employed cross-sectional data, hence causality cannot be determined. SaTScan, utilised for identifying circular clusters, may fail to detect irregularly shaped clusters. Another drawback is that the UDHS geographic data utilised displaced coordinates, which may be approximate rather than exact. Furthermore, the secondary character of the sampled coordinates utilised for kriging precludes the possibility of ground truth verification. In particular, the use of spatial and multilevel methods helps to provide robust results for policy-making regarding non-institutional delivery in Uganda.

Conclusion

Uganda has made progress in facility deliveries, but significant proportions (25%) of women continue to have non-institutional delivery. The findings show that being previously married, having at least two births, and desiring a later pregnancy increased women’s odds of having non-institutional delivery, whereas having secondary or higher education, being in the middle or rich wealth index bracket, attending antenatal visits for four or more visits, and receiving high-quality content of antenatal care decreased women’s odds of having non-institutional delivery. At the community level, residing in the northern region, living in communities that have access to a health facility, and living in media-saturated communities decreased the odds of non-institutional delivery. Therefore, the promotion of antenatal programmes through radio campaigns, which are easily accessible to both the rural and urban areas, should be encouraged by the Ministry of Health. The study identified spatial clusters of non-institutional deliveries primarily in the Western and Eastern regions, with a partial extension to the Northern and Central regions of Uganda. These results provide further insights into identifying the true picture of the spatial clusters of non-institutional delivery in the country and enable timely spatial targeting of factors facilitating non-institutional delivery. This study’s findings suggest that policymakers and other stakeholders should promote universal secondary education and poverty eradication programmes to ensure universal access to health facilities.

Abbreviations

AIC

Akaike Information Criteria

AOR

Adjusted Odd Ratio

ANC

Antenatal Care

CI

Confidence Interval

UDHS

Uganda Demographic and Health Survey

WHO

World Health Organization

EmONC

Emergency Obstetric and Newborn Care

SDGs

Sustainable Development Goals

MMR

Maternal Mortality Ratio

NGO

Non-governmental organization

NPHC

Uganda National Population and Housing Census

EAs

Enumerated Areas

LR

Log-likelihood ratio

VIF

Variance Inflation Factor

TBA

Traditional Birth Attendant

ICC

Inter-class Correlation Coefficient

PNC

Postnatal Care

RR

Relative Risk

UNHCR

United Nations High Commissioner for Refugees

URCS

Uganda Red Cross Society

Authors’ contributions

Conceptualization of article: Moses Festo Towongo: Moses Festo Towongo. Data analysis: Moses Festo Towongo. Interpretations of the data: Moses Festo Towongo, Enock Ngome, Kannan Navaneetham. Software used: Moses Festo Towongo. Revision of the article draft: Moses Festo Towongo, Enock Ngome, Kannan Navaneetham, Letamo Gobopamang, approval of submission Moses Festo Towongo, Enock Ngome, Kannan Navaneetham, Letamo Gobopamang. All authors read and approved the final manuscript.

Funding

This study received no external funding.

Data availability

The data used in this study are third-party data from DHS (https://www.dhsprogram.com/data/available-datasets.cfm) and can be accessed following the protocol outlined in the Methods section.

Declarations

Ethics approval and consent to participate

The authors obtained permission to access the Uganda Demographic and Health Survey (UDHS) datasets through the DHS Program’s official website (https://www.dhsprogram.com/data/available-datasets.cfm). All data provided by the Demographic and Health Surveys (DHS) program are anonymized before being released for public use. The 2016 UDHS received ethical clearance from the Institutional Review Board (IRB) of ICF, which, along with ORC Macro and ICF Macro IRBs, adhered to the ethical guidelines set forth by the U.S. Department of Health and Human Services (45 CFR 46) for the protection of human research subjects. Informed verbal consent was obtained from all study participants, with consent from parents or legal guardians for those under 18 years of age. Additionally, geographic data collected by DHS is intentionally imprecise to protect respondent confidentiality. Locations are randomly displaced within defined limits: urban clusters are shifted up to 2 km, rural clusters up to 5 km, and 1% of the data points are displaced up to 10 km. Full details regarding the methodology and ethical considerations of the survey are available in the 2016 UDHS report [18] and at http://goo.gl/ny8T6X. The authors affirm that all procedures followed the ethical principles outlined in the Declaration of Helsinki and complied with relevant regulatory standards.

Consent for publication

Not required.

Competing interests

The authors declare no competing interests.