Regional trends, spatial patterns and factors associated with health facility delivery among women of reproductive age in Sierra Leone Authors: Lovel Fornah, Augustus Osborne, Taha Hussein Musa, Abakundana Nsenga Ariston Gabriel, Mulugeta Shegaze Shimbre Published: 2025 ABSTRACT BackgroundHealth facility delivery is a critical intervention for reducing maternal and neonatal mortality. Although notable progress has been made globally, Sierra Leone continues to face high maternal and neonatal mortality rates, with disparities in access to health facility delivery across regions and socio-economic groups. This study examined the regional trends, spatial patterns, and factors associated with health facility delivery among women of reproductive age in Sierra Leone using data from the 2008, 2013, and 2019 Demographic and Health Surveys .MethodsThis cross-sectional study utilized data among 19,809 women aged 15–49 who participated in the 2008, 2013, and 2019 Sierra Leone Demographic and Health Surveys. Descriptive statistics were used to analyze health facility delivery utilization trends across survey years and regions. Spatial analysis, including Moran's I and Getis Ord GI statistics, was performed to identify the spatial clustering of health facility delivery. A pooled multilevel logistic regression model was employed to identify individual and community-level factors associated with health facility delivery.ResultsThe prevalence of health facility delivery increased from 24.6% in 2008 to 83.4% in 2019. Spatial analysis revealed clustering of health facility delivery, with hotspots in the Eastern and Southern regions and cold spots in the Northern and Northwestern regions. Our pooled multilevel analysis revealed that women aged 20–29, those with primary education, those with secondary education, those with higher education, those from middle-wealth households, those from rich households, those who attended at least four antenatal care visits, and those exposed to media had higher odds of health facility delivery. Additionally, women who were Muslims, those practicing traditional/other religions, those who were employed, those with higher parity, those residing in rural areas, those who reported distance to health facilities as a medium problem, those who reported distance to health facilities as a high problem, those from Northern, those from Northwestern, those from the Southern, and those from the Western Area had lower odds of health facility delivery. Health facility delivery odds were higher during the 2013 and 2019 surveys compared to 2008.ConclusionDespite improvements in health facility delivery utilization in Sierra Leone, major regional and socio-economic disparities remain. Targeted interventions to enhance maternal education, reduce geographic barriers, and address community-level inequalities are essential for equitable access to health facility delivery and better maternal and neonatal health outcomes. Introduction Health facility delivery is a critical component of maternal and neonatal healthcare, It’s notably reduces the risks of complications during childbirth and improving survival outcomes for both mothers and newborns [1]. Globally, complications related to pregnancy and childbirth remain a leading cause of death among women of reproductive age, with sub-Saharan Africa (SSA), bears the highest burden of maternal mortality, accounting for approximately 66% of global maternal deaths in 2020, indicating a persistent public health crisis [2]. Skilled birth attendance and access to health facilities during delivery are key strategies for addressing these challenges. They provide access to emergency obstetric care, skilled healthcare workers, and essential medical interventions [3]. Despite global efforts to improve maternal health, many low-resource settings, including Sierra Leone, continue to face major barriers to health facility delivery. As a results maternal and neonatal mortality rates remain persistently high [4]. Sierra Leone, a country in West Africa, has one of the highest maternal mortality ratios globally, with an estimated 717 maternal deaths per 100,000 live births in 2019 [4]. This figure is alarmingly high compared to neighboring countries such as Guinea (724/100,000) and Liberia (557/100,000), highlighting the urgent need for targeted interventions. Neonatal mortality also remains alarmingly high, at 75 deaths per 1,000 live births [4]. The country has made considerable efforts to improve maternal and child health outcomes, including implementing the Free Healthcare Initiative (FHI) in 2010, which aimed to remove financial barriers to accessing maternal and child health services [5]. However, despite these interventions, disparities in health facility delivery persist across regions, socio-economic groups, and urban–rural settings, underscoring the need for a deeper understanding of the factors influencing this critical health behavior. Previous studies have identified a range of factors influencing health facility delivery, including maternal education, household wealth, distance to health facilities, cultural practices, and the availability of skilled birth attendants [6–10]. While these studies have provided valuable insights, they often focus on national-level trends, neglecting the regional and spatial variations that may provide valuable insights into localized barriers and enablers of health facility delivery. For instance, research indicates that in many SSA countries, socio-economic disparities and geographic isolation contribute significantly to limited access to healthcare services [6]. Understanding these regional trends and spatial patterns is particularly important in Sierra Leone, where geographic, socio-economic, and cultural diversity influence healthcare access and utilization [11]. Moreover, a comparative analysis of health facility delivery trends in SSA reveals that while some countries have improved access to skilled birth attendants, others continue to lag behind due to systemic challenges such as inadequate infrastructure, insufficient healthcare workforce, and persistent cultural barriers [3]. These disparities highlight the need for a nuanced approach to maternal health interventions that considers local contexts and addresses specific barriers to health facility delivery. The Ebola outbreak from 2014 to 2016 disrupted maternal health systems in Sierra Leone, leading to a significant decline in health facility deliveries [11]. The crisis strained healthcare resources, diverted attention from maternal health services, and exacerbated existing barriers, resulting in increased maternal and neonatal mortality rates [11]. Understanding the impact of such disruptions is crucial for developing resilient health systems capable of withstanding future shocks. To contextualize this study within the existing literature, we employed a theoretical framework based on Andersen's Framework of Healthcare Services Utilisation Model, which posits that three core factors determine individual healthcare utilization: predisposing, enabling, and need for care factors [12]. This framework will guide our analysis of the determinants affecting health facility delivery in Sierra Leone, providing a structured approach to understanding the complexities of maternal health in the region. This study builds on existing research by examining the regional trends, spatial patterns, and factors associated with health facility delivery among women of reproductive age in Sierra Leone using data from the 2008, 2013, and 2019 Demographic and Health Surveys (DHS). By analyzing three consecutive surveys, this study provides a comprehensive overview of changes in health facility delivery over time and identifies persistent disparities across regions and socio-economic groups. Furthermore, the study employs spatial analysis techniques to map patterns of health facility delivery, offering valuable insights into geographic inequities in access to maternal healthcare. The use of hotspot mapping will inform policymakers by identifying regions with the highest need for intervention, thereby facilitating the allocation of resources to areas that will benefit most. Methods Study setting Sierra Leone, located in West Africa, shares borders with Guinea to the north and Liberia to the southeast. Sierra Leone's population in 2019 was approximately 7.5 million [13], with most residents in the southern and western regions [14]. The country is organized into five provinces, 16 districts, and 190 chiefdoms following a reorganization in 2017 that established the Northwestern Province and added two new districts, Falaba and Karene [14–16]. Data source and study design Our research utilized the Sierra Leone Demographic and Health Survey (SLDHS) datasets from 2008, 2013, and 2019. The SDHS is a population-based cross-sectional survey conducted every five years among women aged 15 to 49 and men aged 15 to 59 in Sierra Leone, aiming to collect data on various health indicators. The sample design for each survey typically employs a two-stage probability sampling approach based on a predetermined sampling frame, which usually originates from the most recent census. For the 2008 and 2013 SLDHS, the sampling frame was based on the population and housing censuses conducted in 2004, while the 2019 SLDHS relied on the 2015 Population and Housing Census as its sampling framework. In the first stage, enumeration areas (EAs) were selected using probability proportional to size, followed by systematic sampling of households within each EA. This study utilized the women's data file (Individual Recode files) from the SLDHS datasets of 2008, 2013, and 2019. The response rates for individual interviews with women were notably high, with 94.0% in 2008, 97% in 2013, and 97% in 2019, reflecting robust participation and data quality [14–16]. The SLDHS targeted women aged 15–49, yielding sample sizes of 7,374 in 2008, 16,665 in 2013, and 15,574 in 2019 (see Fig. 1). This study adhered to the guidelines established by the Strengthening Reporting of Observational Studies in Epidemiology (STROBE) [17].Fig. 1Sampling procedure Sampling procedure Data extraction and management The DHS program website has improved its registration and request processes, enabling users to easily access and download datasets after briefly describing their research objectives. After receiving a letter of authorization and access from (archive@dhsprogram.com), we obtained the individual recode file for the years 2008, 2013, and 2019 in STATA format from the MEASURE DHS program website, accessible at [18], along with sub-national boundary data in shapefile format from [19]. Additionally, we submitted a formal request to acquire the geographical coordinate data from [18] to represent the spatial utilization of health facility deliveries. Data cleaning, recoding, merging, and appending were performed using Stata SE 18 software (STATA Corporation, IC., TX, USA). Initially, each survey was weighted using the sampling weights provided in the respective datasets to account for variations in sampling probabilities and to ensure representativeness. Also, the sample weight technique was used to adjust for non-response bias. The 2008, 2013, and 2019 datasets were appended into a single dataset, which was then re-weighted to reflect the overall population structure across the three survey years. From the total number of women interviewed in each survey year, those who had never given birth were excluded. This approach resulted in 5,876, 12,352, and 11,475 women in 2008, 2013, and 2019, respectively. To align with the objectives of our study, we further restricted the analysis to women aged 15–49 who had live births in the last five years preceding each survey year. This yielded a final pooled weighted sample size of 19,809 women aged 15–49 (See Fig. 1). We addressed the issue of missing data through multiple imputations and conducted a complete case analysis for non-essential missing values when the extent of missing data was minimal. We compared descriptive statistics between the original and imputed datasets to mitigate potential bias, finding no significant differences. In all three SLDHS, the geographic coordinates of participants were adjusted by 5 km to maintain participant confidentiality. We integrated each survey year's dataset with its corresponding Global Positioning System (GPS) coordinates to account for geographic displacement. As a result, all GPS coordinates, including displaced points, were included in the analyses. Sensitivity analyses confirmed that displacement did not significantly affect the spatial patterns of health facility utilization, and no coordinates were excluded from the study. Measures Outcome measurement The primary outcome of interest was health facility delivery of the most recent birth among women of reproductive age. Deliveries that took place in either public or private health facilities were classified as"utilizing health facility delivery"and assigned a code of"1". In contrast, those who delivered at home or in other locations were categorized as"not utilizing health facility delivery"and assigned a code of"0". Conceptual framework for health facility delivery To contextualize this study within previous studies [20–32], we utilize Andersen's Framework of Healthcare Services Utilisation Model to select the factors influencing health facility delivery. Andersen's Framework of Healthcare Services Utilization Model says that individual and contextual factors affect healthcare use. Predisposing characteristics, enabling factors, and perceived or evaluated care for need are the model's main factors [12] (see Fig. 2). Predisposing factors include population characteristics such as age and marital status, social structures like education and religion, health beliefs that influence attitudes toward healthcare, and other factors [12]. Enabling factors include personal or family aspects such as wealth index, employment, media exposure, community-level elements like urban versus rural location, distance to healthcare facilities, and other factors. The model also values perceived (self-reported) or evaluated (clinically diagnosed) need for care factors [12]. The need for care factors are antenatal visits, companionship to health facilities, participation in healthcare decisions, and other factors. Andersen's framework links individual actions to systemic obstacles and policies, unlike psychological models like the Health Belief Model (HBM). Andersen's model places personal beliefs and perceptions within larger enabling or disabling frameworks, unlike HBM. The Andersen framework is handy for policymakers aiming to address disparities through infrastructure, education, or policy reforms.Fig. 2Conceptual Framework Conceptual Framework Exposure variables Considering the SLDHS's hierarchical structure, where participants are nested within a cluster. The exposure variables were also classified into individual and community variables. The individual variables were categorized as follows: Age at last childbirth [15–49], marital status (Women who have never been in a union and those who were previously in a union or living with a man are classified as "not married", while those currently in a union or living with a man are classified as"married"), religion (Christianity, Islam, traditional/others), educational level (no education, primary, secondary, higher), employment status (unemployed, employed), wealth index (poor, middle, rich), media exposure (no, yes), wanted index pregnancy (wanted, not wanted), number of childbirths (1–2, 3–4, ≥ 5), number of antenatal visits (< 4, at least 4), companionship to health facilities (big problem, not a big problem), and participation in healthcare decisions (participates, does not participate). Community variables were similarly categorized as residence (urban, rural) and region (eastern, northern, northwestern, southern, western). Community poverty, women's education, and distance to health facilities were derived through cluster-level aggregation, categorized into tertiles, and classified as low, medium, or high [28]. This methodology is commonly used to extract community variables from DHS datasets [20, 21, 28]. Operational definition Age at last childbirth The age at which a mother gave birth to her last child is calculated by subtracting the mother's date of birth (expressed in century month code, CMC) from the child's date of birth (also in century month code). The resulting value is then divided by 12 to convert it into years. This age is subsequently categorized into groups: 15–19 years, 20–29 years, 30–39 years, and 40–49 years [28]. The selected age groups were designed to represent key biological and demographic stages of reproductive life. The 15–19 age bracket encompasses early reproductive years and adolescence, a critical period for maternal health interventions. The optimal fertility period occurs within the 20–29 age range, while advanced reproductive age, marked by varying risk profiles, spans from 30 to 39. The 40–49 age group includes the late reproductive phase leading up to menopause. This classification facilitates direct comparisons with other national data and aligns with established demographic standards used in major health surveys, such as the Demographic and Health Surveys program. These broader intervals were chosen to ensure adequate sample sizes in each category for reliable statistical analysis while preserving clinical and public health relevance. Collectively, these groups effectively capture the primary variations in reproductive outcomes and can provide sufficient statistical power. Wealth index Was generated as a tertile of the wealth index factor score into poor, middle, and rich using principal component analysis of household assets. Mass media exposure Our interest was in women's frequent media exposure, which we defined as having access to television, radio, newspapers, or the internet at least once a week. To create the media exposure variable, we drew from individual questionnaire items that evaluated regular access to these three traditional media sources across all survey years (SLDHS 2008, 2013, and 2019). Notably, while the 2019 SLDHS included questions about internet use, the 2008 and 2013 surveys did not. We restricted our media exposure variable to television, radio, and newspaper access to ensure consistency across all three survey years in our appended dataset. We defined media exposure as frequent for individuals with at least weekly access to television, radio, or newspapers. In contrast, those with no access or less than weekly exposure were classified as having no/infrequent exposure. This process is consistent with previous research conducted by [33]. This study derived community-level poverty, women's education, and community distance to health facilities from a standard procedure. Community-level poverty was defined as the proportion of women from the poorest communities [28]. Community women's education was defined as the proportion of women from the community with at least secondary education [28, 30]. Community distance to health facilities was defined as the proportion of women for whom distance to health facilities is a big problem, aggregated at the cluster level [28]. All three variables were calculated by aggregating individual characteristics at the cluster level, dividing the measure into tertiles, and classifying them as equal tertiles (low, medium, and high), representing the bottom 33%, middle 33%, and top 33% of the distribution. This aligns with the procedures employed by prior studies to generate community variables in DHS datasets. Data analysis Trend An analysis was conducted using Microsoft 365 Excel (version 2405) to examine the utilization trend of health facility delivery by survey year and region. Spatial analysis We employed ArcGIS to visualize the spatial distribution of health facility deliveries in Sierra Leone. Initially, we used neighborhood averaging buffer distances (1 km for urban areas and 5 km for rural areas) to address GPS displacement and also measured sensitivity at 0–2 km urban and 0–5 km rural distances. We applied Moran's I statistics to evaluate whether the patterns of health facility deliveries were spatially clustered, dispersed, or randomly distributed [31]. Moran's I statistic values span from −1 to + 1, with values near −1 indicating a dispersed delivery pattern, those close to + 1 suggesting a clustered distribution, and values around 0 reflecting a random distribution [25]. Z-scores are calculated alongside p-values, with a significance level set at less than 0.05. These values assess the statistical significance of the distribution patterns of health facility deliveries. We employed an adaptive kernel approach for our hotspot analysis. We then used local Getis Ord GI* statistics to find areas where women of reproductive age either frequently or rarely use health facilities for delivery. We calculated the z-score and p-value for the Getis Ord GI* statistics to assess the significance of spatial clustering. High GI* values indicate"hotspots,"areas characterized by a high proportion of health facility delivery utilization, while low GI* values signify"cold spots,"which represent areas with low utilization [30]. Finally, we utilized spatial interpolation to estimate health facility delivery utilization in regions where data was unavailable. Ordinary kriging served as the interpolation method, yielding a predicted distribution of health facility delivery utilization across the entire study area [26]. Multilevel analysis Given the hierarchical nature of the DHS datasets, a multilevel logistic analysis was conducted to explore the factors that affect health facility delivery utilization at both the individual and community levels [29]. Before conducting the multilevel logistic regression analysis, we performed a bivariate analysis using all the independent variables (Table 1) to pinpoint candidate variables for the multilevel logistic analysis. Age at last childbirth, marital status, religion, educational attainment, employment status, wealth index, media exposure, number of childbirths, Number of antenatal visits, Companionship to health facility, participant healthcare decision-making, community poverty levels, women's education within the community, distance to health facilities, residence, and region were the independent variables that were statistically significant with a p-value of less than 0.05.Table 1Sample characteristics of health facility delivery among women of reproductive age in Sierra LeoneSLDHS 2008N = 3933SLDHS 2013N = 8499SLDHS 2019N = 7377SLDHS 2008–2019 N = 19,809Variablesn (%)n (%)n (%)n (%)Facility-based delivery Utilized health facility delivery1,169 (29.7%)4,982 (58.6%)6,189 (83.9%)12,340 (62.3%) Did not utilize health facility delivery2,764 (70.3%)3,517 (41.4%)1,188 (16.1%)7,469 (37.7%)Age at last childbirth 15–19656 (16.7%)1,647 (19.4%)1,210 (16.4%)3,513 (17.7%) 20–292,004 (51.0%)3,970 (46.7%)3,550 (48.1%)9,524 (48.1%) 30–391,095 (27.8%)2,435 (28.7%)2,230 (30.2%)5,760 (29.1%) 40–49178 (4.5%)447 (5.3%)387 (5.2%)1,012 (5.1%)Marital status Not married544 (13.8%)1,378 (16.2%)1,287 (17.4%)3,209 (16.2%) Married3,389 (86.2%)7,121 (83.8%)6,090 (82.6%)16,600 (83.8%)Religion Christianity927 (23.6%)1,677 (19.7%)1,528 (20.7%)4,132 (20.9%) Islam2,953 (75.1%)6,777 (79.7%)5,846 (79.2%)15,576 (78.6%) Traditional/others53 (1.3%)45 (0.5%)3 (0.0%)101 (0.5%)Educational level No education2,797 (71.1%)5,592 (65.8%)4,085 (55.4%)12,474 (63.0%) Primary526 (13.4%)1,155 (13.6%)1,004 (13.6%)2,685 (13.6%) Secondary546 (13.9%)1,631 (19.2%)2,076 (28.1%)4,253 (21.5%) Higher64 (1.6%)121 (1.4%)212 (2.9%)397 (2.0%)Employment Status Unemployed557 (17.0%)1,287 (17.7%)1,456 (20.2%)3,300 (18.6%) Employed694 (21.2%)1,598 (22.0%)1,514 (21.0%)3,806 (21.5%)Wealth Index Poor1,337 (34.0%)2,429 (28.6%)2,837 (38.5%)6,603 (33.3%) Middle1,203 (30.6%)3,440 (40.5%)1,961 (26.6%)6,604 (33.3%) Rich1,393 (35.4%)2,630 (30.9%)2,579 (35.0%)6,602 (33.3%)Media exposure No3,863 (98.2%)8,354 (98.3%)7,324 (99.3%)19,541 (98.6%) Yes70 (1.8%)145 (1.7%)53 (0.7%)268 (1.4%)Wanted index pregnancy Wanted2,750 (69.9%)7,215 (84.9%)6,054 (82.1%)16,019 (80.9%) Not wanted1,183 (30.1%)1,284 (15.1%)1,323 (17.9%)3,790 (19.1%)Number of childbirths 1–21,627 (41.4%)3,380 (39.8%)3,118 (42.3%)8,125 (41.0%) 3–41,166 (29.6%)2,402 (28.3%)2,268 (30.7%)5,836 (29.5%)≥ 51,140 (29.0%)2,717 (32.0%)1,991 (27.0%)5,848 (29.5%)Number of antenatal visits< 41,643 (41.8%)1,888 (22.2%)1,511 (20.5%)5,042 (25.5%) At least 42,290 (58.2%)6,611 (77.8%)5,866 (79.5%)14,767 (74.5%)Companionship to health facility Big problem748 (19.0%)1,456 (17.1%)1,873 (25.4%)4,077 (20.6%) Not a big problem3,185 (81.0%)7,043 (82.9%)5,504 (74.6%)15,732 (79.4%)Participate in healthcare decisions Participates1,688 (42.9%)3,770 (44.4%)2,582 (35.0%)8,040 (40.6%) Does not participate2,245 (57.1%)4,729 (55.6%)4,795 (65.0%)11,769 (59.4%)Community poverty Low1,089 (27.7%)2,744 (32.3%)2,777 (37.6%)6,610 (33.4%) Medium1,619 (41.2%)2,906 (34.2%)2,109 (28.6%)6,634 (33.5%) High1,225 (31.1%)2,849 (33.5%)2,491 (33.8%)6,565 (33.1%)Community women's education Low1,420 (36.1%)2,921 (34.4%)2,320 (31.4%)6,661 (33.6%) Medium1,389 (35.3%)2,969 (34.9%)2,266 (30.7%)6,624 (33.4%) High1,124 (28.6%)2,609 (30.7%)2,791 (37.8%)6,524 (32.9%)Community distance to health facility Low1,178 (30.0%)2,894 (34.1%)2,572 (34.9%)6,644 (33.5%) Medium1,408 (35.8%)2,897 (34.1%)2,277 (30.9%)6,582 (33.2%) High1,347 (34.2%)2,708 (31.9%)2,528 (34.3%)6,583 (33.2%)Residence Urban1,420 (36.1%)2,813 (33.1%)2,473 (33.5%)6,706 (33.9%) Rural2,513 (63.9%)5,686 (66.9%)4,904 (66.5%)13,103 (66.1%)Region Eastern1,025 (26.1%)1,840 (21.6%)1,475 (20.0%)4,340 (21.9%) Northern1,312 (33.4%)3,319 (39.1%)1,713 (23.2%)6,344 (32.0%) NorthwesternN/AN/A1,016 (13.8%)1,016 (5.1%) Southern898 (22.8%)2,290 (26.9%)1,379 (18.7%)4,567 (23.1%) Western698 (17.7%)1,050 (12.4%)1,794 (24.3%)3,542 (17.9%)N/A: Not applicable Table 1Sample characteristics of health facility delivery among women of reproductive age in Sierra LeoneSLDHS 2008N = 3933SLDHS 2013N = 8499SLDHS 2019N = 7377SLDHS 2008–2019 N = 19,809Variablesn (%)n (%)n (%)n (%)Facility-based delivery Utilized health facility delivery1,169 (29.7%)4,982 (58.6%)6,189 (83.9%)12,340 (62.3%) Did not utilize health facility delivery2,764 (70.3%)3,517 (41.4%)1,188 (16.1%)7,469 (37.7%)Age at last childbirth 15–19656 (16.7%)1,647 (19.4%)1,210 (16.4%)3,513 (17.7%) 20–292,004 (51.0%)3,970 (46.7%)3,550 (48.1%)9,524 (48.1%) 30–391,095 (27.8%)2,435 (28.7%)2,230 (30.2%)5,760 (29.1%) 40–49178 (4.5%)447 (5.3%)387 (5.2%)1,012 (5.1%)Marital status Not married544 (13.8%)1,378 (16.2%)1,287 (17.4%)3,209 (16.2%) Married3,389 (86.2%)7,121 (83.8%)6,090 (82.6%)16,600 (83.8%)Religion Christianity927 (23.6%)1,677 (19.7%)1,528 (20.7%)4,132 (20.9%) Islam2,953 (75.1%)6,777 (79.7%)5,846 (79.2%)15,576 (78.6%) Traditional/others53 (1.3%)45 (0.5%)3 (0.0%)101 (0.5%)Educational level No education2,797 (71.1%)5,592 (65.8%)4,085 (55.4%)12,474 (63.0%) Primary526 (13.4%)1,155 (13.6%)1,004 (13.6%)2,685 (13.6%) Secondary546 (13.9%)1,631 (19.2%)2,076 (28.1%)4,253 (21.5%) Higher64 (1.6%)121 (1.4%)212 (2.9%)397 (2.0%)Employment Status Unemployed557 (17.0%)1,287 (17.7%)1,456 (20.2%)3,300 (18.6%) Employed694 (21.2%)1,598 (22.0%)1,514 (21.0%)3,806 (21.5%)Wealth Index Poor1,337 (34.0%)2,429 (28.6%)2,837 (38.5%)6,603 (33.3%) Middle1,203 (30.6%)3,440 (40.5%)1,961 (26.6%)6,604 (33.3%) Rich1,393 (35.4%)2,630 (30.9%)2,579 (35.0%)6,602 (33.3%)Media exposure No3,863 (98.2%)8,354 (98.3%)7,324 (99.3%)19,541 (98.6%) Yes70 (1.8%)145 (1.7%)53 (0.7%)268 (1.4%)Wanted index pregnancy Wanted2,750 (69.9%)7,215 (84.9%)6,054 (82.1%)16,019 (80.9%) Not wanted1,183 (30.1%)1,284 (15.1%)1,323 (17.9%)3,790 (19.1%)Number of childbirths 1–21,627 (41.4%)3,380 (39.8%)3,118 (42.3%)8,125 (41.0%) 3–41,166 (29.6%)2,402 (28.3%)2,268 (30.7%)5,836 (29.5%)≥ 51,140 (29.0%)2,717 (32.0%)1,991 (27.0%)5,848 (29.5%)Number of antenatal visits< 41,643 (41.8%)1,888 (22.2%)1,511 (20.5%)5,042 (25.5%) At least 42,290 (58.2%)6,611 (77.8%)5,866 (79.5%)14,767 (74.5%)Companionship to health facility Big problem748 (19.0%)1,456 (17.1%)1,873 (25.4%)4,077 (20.6%) Not a big problem3,185 (81.0%)7,043 (82.9%)5,504 (74.6%)15,732 (79.4%)Participate in healthcare decisions Participates1,688 (42.9%)3,770 (44.4%)2,582 (35.0%)8,040 (40.6%) Does not participate2,245 (57.1%)4,729 (55.6%)4,795 (65.0%)11,769 (59.4%)Community poverty Low1,089 (27.7%)2,744 (32.3%)2,777 (37.6%)6,610 (33.4%) Medium1,619 (41.2%)2,906 (34.2%)2,109 (28.6%)6,634 (33.5%) High1,225 (31.1%)2,849 (33.5%)2,491 (33.8%)6,565 (33.1%)Community women's education Low1,420 (36.1%)2,921 (34.4%)2,320 (31.4%)6,661 (33.6%) Medium1,389 (35.3%)2,969 (34.9%)2,266 (30.7%)6,624 (33.4%) High1,124 (28.6%)2,609 (30.7%)2,791 (37.8%)6,524 (32.9%)Community distance to health facility Low1,178 (30.0%)2,894 (34.1%)2,572 (34.9%)6,644 (33.5%) Medium1,408 (35.8%)2,897 (34.1%)2,277 (30.9%)6,582 (33.2%) High1,347 (34.2%)2,708 (31.9%)2,528 (34.3%)6,583 (33.2%)Residence Urban1,420 (36.1%)2,813 (33.1%)2,473 (33.5%)6,706 (33.9%) Rural2,513 (63.9%)5,686 (66.9%)4,904 (66.5%)13,103 (66.1%)Region Eastern1,025 (26.1%)1,840 (21.6%)1,475 (20.0%)4,340 (21.9%) Northern1,312 (33.4%)3,319 (39.1%)1,713 (23.2%)6,344 (32.0%) NorthwesternN/AN/A1,016 (13.8%)1,016 (5.1%) Southern898 (22.8%)2,290 (26.9%)1,379 (18.7%)4,567 (23.1%) Western698 (17.7%)1,050 (12.4%)1,794 (24.3%)3,542 (17.9%)N/A: Not applicable Sample characteristics of health facility delivery among women of reproductive age in Sierra Leone N/A: Not applicable After the bivariate analysis, a variance inflation factor (VIF) analysis was performed to assess multicollinearity among all statistically significant independent variables. The VIF values varied from a minimum of 1.03 (for traditional/other religion, using Christian as a reference) to a maximum of 6.99 (for married marital status, using not married as a reference), with an average VIF of 2.92 across all variables. The results demonstrate an acceptable degree of multicollinearity, with values falling beneath the conservative threshold of 10 (predominantly below 5) [34]. Despite the marital status variable exhibiting the highest VIF, we included it in our final model because its incorporation did not significantly affect the model's overall predictive efficacy, as indicated by consistent AIC values in the reduced models. Consequently, four multilevel logistic models were developed. Model 0 (Null model) had no independent variables and was used as a basic comparison to show how much the health facility delivery changes within the main sampling units (PSUs). Model II incorporated all individual-level variables, Model III included all community-level variables, and Model IV encompassed individual and community variables. The results were presented using odds ratios (OR) and 95% confidence intervals (CIs). To assess each model's fit, we utilized the log-likelihood ratio (LLR), Akaike's Information Criterion (AIC), and Bayesian Information Criterion (BIC). Ultimately, the model with the lowest AIC value was chosen as the best fit for the data. AIC balanced model fit and complexity by penalizing additional parameters, making it suitable for our analysis. This feature prevents overfitting and finds models that best explain data. This makes AIC ideal for analyzing complex, multilevel datasets like ours, where model performance and parsimony are crucial [35]. This study's large sample size, n = 19,809, enhances AIC's asymptotic properties. AIC's convergence properties increase with sample size, boosting our confidence in the selected models [35]. Large sample sizes helped us analyze nationally representative DHS data. AIC's widespread use and endorsement in DHS-based epidemiological research support its selection. The frequent use of this criterion for model selection in complex survey data studies shows its widespread usage in this research context [36]. Results Description of sample characteristics The socio-demographic characteristics patterns among women of reproductive age in Sierra Leone across three survey periods (2008, 2013, and 2019) are described in Table 1. Nearly half of the participants (48.1%) were in the 20–29 age group, with consistent age distribution across survey years. Most women (83.8%) were married, with a slight increase in unmarried women from 13.8% in 2008 to 17.4% in 2019. Islam was the predominant religion (78.6% overall), with Christianity representing 20.9% of the sample and traditional/other religions accounting for less than 1%. Educational attainment has improved over time, with the proportion of women with no education decreasing from 71.1% in 2008 to 55.4% in 2019, while secondary education increased from 13.9% to 28.1% during the same period. Employment status remained relatively stable across survey years, with approximately 21.5% of women employed overall. Wealth distribution was relatively balanced across the three categories (poor, middle, rich) in the combined sample, though there were fluctuations between survey years. Media exposure was consistently low across all survey periods (1.4% overall). Regarding reproductive characteristics, most pregnancies were planned, with 80.9% classified as wanted. There was a marked decline in unwanted pregnancies, dropping from 30.1% in 2008 to 17.9% in 2019. The parity distribution showed relative stability, with 41.0% of women having 1–2 children, 29.5% having 3–4 children, and another 29.5% having five or more children. Additionally, the utilization of antenatal care saw considerable improvement, as the percentage of women attending at least four antenatal visits increased from 58.2% in 2008 to 79.5% in 2019. Moreover, access-related factors showed mixed trends. While most women (79.4%) reported that companionship to health facilities was not a big problem, this proportion decreased from 82.9% in 2013 to 74.6% in 2019. Women's participation in healthcare decisions declined from 42.9% in 2008 to 35.0% in 2019, with 59.4% not participating in healthcare decisions overall. Table 1 further shows that there was a decrease in high-community poverty (from 41.2% to 28.6% for the medium category) and an increase in high-community women's education (from 28.6% to 37.8%) between 2008 and 2019. Community distance to health facilities remained relatively stable across survey years. Most women resided in rural areas (66.1%), with minimal change in urban–rural distribution over time. Regional distribution varied across surveys, with the introduction of the Northwestern region in the 2019 survey and fluctuations in the proportions of women from other regions. Trend analysis of health facility delivery among women of reproductive age in Sierra Leone from 2008 to 2019 Our findings indicate a notable trend in the utilization of health facility delivery services among women of reproductive age in Sierra Leone. The percentage of women using these services increased from approximately 25% in 2008 to 54% in 2013 and then surged to 83% in 2019. Overall, the pooled prevalence of health facility delivery utilization among women of reproductive age in Sierra Leone was 54% (seeFig. 3).Fig. 3Trends of health facility delivery of the most recent birth among women of reproductive age in Sierra Leone 2008, 2013, 2019 and pooled dataset Trends of health facility delivery of the most recent birth among women of reproductive age in Sierra Leone 2008, 2013, 2019 and pooled dataset Figure 4 illustrates the proportion of facility-based deliveries among women of reproductive age in Sierra Leone from 2008 to 2019 by region. Between 2008 and 2019, the percentage of women delivering at health facilities increased across all regions, except for the Northwest region, where data on health facility delivery utilization was only available for 2019. The Northern region recorded the lowest proportion of health facility delivery, starting at 16.9% in 2008, rising to 40.3% in 2013, and reaching 85.1% by 2019. The Western region consistently demonstrated higher rates of health facility delivery compared to other regions. In 2008, the rate of health facility delivery was relatively better than that of most regions, at 37.1%. This upward trend continued in 2013 with an improvement to 64%, and by 2019, the Western region achieved a notable 88.1% in health facility delivery.Fig. 4Regional trends of health facility delivery of the most recent birth among women of reproductive age in Sierra Leone (SLDHS 2008, SLDHS 2013, and SLDHS 2019) Regional trends of health facility delivery of the most recent birth among women of reproductive age in Sierra Leone (SLDHS 2008, SLDHS 2013, and SLDHS 2019) Spatial pattern of health facility delivery of the most recent birth among women of reproductive age in Sierra Leone Spatial autocorrelation The global spatial autocorrelation analysis of FHD among women of reproductive age in Sierra Leone from 2008 to 2019, illustrated in Fig. 5, reveals that the three survey years exhibit statistically significant spatial clustering rather than being randomly distributed. In Figs. 4a-c, Moran's I index values are 0.224831 (p=0.001) for the SLDHS 2008, 0.219134 (p=0.001) for the SLDHS 2013, and 0.102125 (p = 0.001) for the SLDHS 2019, respectively. These results suggest that areas with high FHD utilization were clustered rather than randomly distributed across the country.Fig. 5Spatial autocorrelation analysis of the distribution HFD of the most recent birth among women of reproductive age in Sierra Leone for the survey years 2008 (a), 2013 (b), and 2019 (c) Spatial autocorrelation analysis of the distribution HFD of the most recent birth among women of reproductive age in Sierra Leone for the survey years 2008 (a), 2013 (b), and 2019 (c) Hotspot analysis In the SLDHS 2008, some parts of the eastern and some parts of the southern regions were found to be statistically significant hotspot regions at a 99% confidence level for utilization of health facility delivery. The western area was found to be a hot spot at a 90% confidence level. Conversely, the Northern region was found to have significant cold spots (Fig. 6 upper left panel).Fig. 6Hotspot analysis of health facility delivery of the most recent birth among women of reproductive age in Sierra Leone for the survey years 2008 (upper left panel), 2013 (upper right panel), and 2019 (lower right panel). The red color indicates a hot spot (a region with higher utilization of health facility delivery), and the blue indicates a cold spot (a region with lower utilization of health facility delivery) Hotspot analysis of health facility delivery of the most recent birth among women of reproductive age in Sierra Leone for the survey years 2008 (upper left panel), 2013 (upper right panel), and 2019 (lower right panel). The red color indicates a hot spot (a region with higher utilization of health facility delivery), and the blue indicates a cold spot (a region with lower utilization of health facility delivery) In 2013, hotspot regions (i.e., regions with higher utilization of health facility delivery) were Eastern and some parts of Southern regions at a 99% confidence level. Some parts of the Western area were found to be a hotspot at a 90% confidence level. Cold spot regions (i.e., regions with low utilization of health facility delivery) were the Northern and some parts of the Western regions (Fig. 6 upper right panel). However, in the SLDSH 2019, the Eastern region was found to be a hotspot region at a 99% confidence level. Meanwhile, the Southern and some parts of the Western regions were found to be hotspots at a 90% confidence level. In contrast, Northern and Northwestern regions were identified as cold spot regions at a 99% confidence level (Fig. 6 lower left panel). Ordinary kriging interpolation We use Kriging interpolation analysis to figure out how often health facility delivery happens in places where data wasn't collected. This lets us figure out how far away unknown points are from known points and where those points are in the event occurrence ranges. Yellow indicates regions with low health facility delivery utilization, while dark umber indicates regions with high health facility delivery utilization (Fig. 7). In the SLDHS 2008, the possible regions for health facility delivery to occur were the Southern region(Fig. 7 upper left panel), while in the SLDHS 2013, the Eastern and Southern regions were the possible regions for health facility delivery to occur(Fig. 7 upper right panel). In the SLDHS 2019, the possible regions where health facility delivery could occur were some parts of the Eastern, some parts of the Southern, and some parts of the Northern regions (Fig. 7 lower left panel).Fig. 7Kriging interpolated spatial patterns of health facility delivery of the most recent births among women of reproductive age in Sierra Leone for the survey years 2008 (upper left panel), 2013 (upper right panel), and 2019 (lower right panel). The yellow indicates regions with low health facility delivery utilization, while dark umber indicates regions with high health facility delivery utilization Kriging interpolated spatial patterns of health facility delivery of the most recent births among women of reproductive age in Sierra Leone for the survey years 2008 (upper left panel), 2013 (upper right panel), and 2019 (lower right panel). The yellow indicates regions with low health facility delivery utilization, while dark umber indicates regions with high health facility delivery utilization Multilevel analysis to determine the factors associated with health facility delivery of the most recent births among reproductive-age women in Sierra Leone Random fix effect Our findings indicate significant variation in health facility delivery rates across clusters [σ2 = 1.35, 95% CI = 1.13, 1.61], as shown in Model O (Table 2). The intraclass correlation coefficient (ICC) of 0.29 suggests that 29% of the variability in deliveries is due to differences among clusters. Inter-cluster variation is 26% in Model I, 24% in Model II, and 19% in Model III, indicating that cluster differences primarily drive the variation. The AIC values decline from 24,045.99 in Model O to 19,581.31 in Model III. Thus, we identify Model III as the best model for assessing the determinants of health facility delivery among women of reproductive age in Sierra Leone.Table 2Multi-level modeling of health facility delivery of the most recent birth among women of reproductive age in Sierra LeoneVariableModel OModel IaOR [95% CI]Model IIaOR [95% CI]Model IIIaOR [95% CI]Age at last childbirth 15–1911 20–291.23***[1.12, 1.39]1.18**[1.05, 1.33] 30–391.43***[1.22,1.67]1.17 [0.99, 1.39] 40–491.21[0.97,1.51]0.94[0.74, 1.22]Marital status Not married11 Married1.09[0.95,1.24]1.08[0.94, 1.24]Religion Christianity11 Islam0.83*[0.71,0.98]0.81** [0.70, 0.94] Traditional/others0.26***[0.14,0.47]0.49* [0.27, 0.89]Educational level No education11 Primary1.43***[1.26,1.61]1.26***[1.10, 1.43] Secondary2.33***[2.06,2.65]1.56***[1.37, 1.78] Higher4.75***[3.08,7.33]3.14***[1.95, 5.07]Employment Status Unemployed11 Employed0.91[0.81,1.02]0.86*[0.76, 0.97]Wealth Index Poor11 Middle1.03[0.91,1.17]1.20** [1.07, 1.36] Rich1.47***[1.21,1.78]1.30**[1.06, 1.60]Media exposure No11 Yes1.07[0.69,1.65]1.82** [1.16, 2.87]Number of childbirths 1–211 3–40.93[0.83, 1.05]0.88*[0.78, 1.00] ≥ 50.78*** [0.67, 0.89]0.78***[0.68, 0.90]Number of antenatal visits < 411 At least 42.02* * * [1.80,2.27]1.58** *[1.42, 1.76]Companionship to a health facility Big problem11 Not a big problem1.14 [0.99, 1.32]1.32*** [1.15, 1.52]Participate in healthcare decisions Participates11 Does not participate1.25* * * [1.11, 1.39]1.03[0.93, 1.15]Community poverty Low11 Medium0.70***[0.54, 0.90]0.88[0.70, 1.10] High1.02[0.75, 1.37]1.13[0.86, 1.49]Community women's education Low11 Medium1.06[0.82,1.36]1.00[0.79, 1.30] High1.27 [0.92, 1.75]0.95 [0.72, 1.25]Community distance to a health facility Low11 Medium0.68***[0.53, 0.86]0.62**[0.50, 0.78] High0.56***[0.43, 0.73]0.45***[0.35, 0.57]Residence Urban11 Rural0.72**[0.57, 0.92]0.64** [0.50, 0.84]Region Eastern11 Northern0.26***[0.19, 0.37]0.29***[0.22, 0.38] Northwestern3.04*** [1.87, 4.94]0.35*** [0.20, 0.59] Southern0.64** [0.55, 1.12]0.56*** [0.40, 0.79] Western0.78 [0.54, 1.21]0.05***[0.37, 0.67]Survey years 20081 20134.27*** [3.49,5.22] 201919.90***[15.26,25.95]Random effect model PSU variance (95% CI)1.35 [1.13, 1.61]1.17 [0.96, 1.42]1.01[0.82, 1.25]0.76 [0.62, 0.93] ICC0.29 [0.25, 0.33]0.26 [0.22, 0.30]0.24[0.20, 0.28]0.19 [0.16, 0.22] Wald chi-squareReference531.43***262.12***1130.66***Model fitness Log-likelihood−12,020.997−11,412.183−11,387.646−9757.653 AIC24,045.9922,864.3722,801.2919,581.31 BIC24,061.7823,022.2422,903.9119,841.81 N19,80919,80919,80919,809 Number of clusters578578578578aOR = adjusted odds ratios; CI = Confidence Interval, 1.00 = PSU Primary Sampling Unit, ICC Intra-Class Correlation*p < 0.05, **p < 0.01, ***p < 0.001 Table 2Multi-level modeling of health facility delivery of the most recent birth among women of reproductive age in Sierra LeoneVariableModel OModel IaOR [95% CI]Model IIaOR [95% CI]Model IIIaOR [95% CI]Age at last childbirth 15–1911 20–291.23***[1.12, 1.39]1.18**[1.05, 1.33] 30–391.43***[1.22,1.67]1.17 [0.99, 1.39] 40–491.21[0.97,1.51]0.94[0.74, 1.22]Marital status Not married11 Married1.09[0.95,1.24]1.08[0.94, 1.24]Religion Christianity11 Islam0.83*[0.71,0.98]0.81** [0.70, 0.94] Traditional/others0.26***[0.14,0.47]0.49* [0.27, 0.89]Educational level No education11 Primary1.43***[1.26,1.61]1.26***[1.10, 1.43] Secondary2.33***[2.06,2.65]1.56***[1.37, 1.78] Higher4.75***[3.08,7.33]3.14***[1.95, 5.07]Employment Status Unemployed11 Employed0.91[0.81,1.02]0.86*[0.76, 0.97]Wealth Index Poor11 Middle1.03[0.91,1.17]1.20** [1.07, 1.36] Rich1.47***[1.21,1.78]1.30**[1.06, 1.60]Media exposure No11 Yes1.07[0.69,1.65]1.82** [1.16, 2.87]Number of childbirths 1–211 3–40.93[0.83, 1.05]0.88*[0.78, 1.00] ≥ 50.78*** [0.67, 0.89]0.78***[0.68, 0.90]Number of antenatal visits < 411 At least 42.02* * * [1.80,2.27]1.58** *[1.42, 1.76]Companionship to a health facility Big problem11 Not a big problem1.14 [0.99, 1.32]1.32*** [1.15, 1.52]Participate in healthcare decisions Participates11 Does not participate1.25* * * [1.11, 1.39]1.03[0.93, 1.15]Community poverty Low11 Medium0.70***[0.54, 0.90]0.88[0.70, 1.10] High1.02[0.75, 1.37]1.13[0.86, 1.49]Community women's education Low11 Medium1.06[0.82,1.36]1.00[0.79, 1.30] High1.27 [0.92, 1.75]0.95 [0.72, 1.25]Community distance to a health facility Low11 Medium0.68***[0.53, 0.86]0.62**[0.50, 0.78] High0.56***[0.43, 0.73]0.45***[0.35, 0.57]Residence Urban11 Rural0.72**[0.57, 0.92]0.64** [0.50, 0.84]Region Eastern11 Northern0.26***[0.19, 0.37]0.29***[0.22, 0.38] Northwestern3.04*** [1.87, 4.94]0.35*** [0.20, 0.59] Southern0.64** [0.55, 1.12]0.56*** [0.40, 0.79] Western0.78 [0.54, 1.21]0.05***[0.37, 0.67]Survey years 20081 20134.27*** [3.49,5.22] 201919.90***[15.26,25.95]Random effect model PSU variance (95% CI)1.35 [1.13, 1.61]1.17 [0.96, 1.42]1.01[0.82, 1.25]0.76 [0.62, 0.93] ICC0.29 [0.25, 0.33]0.26 [0.22, 0.30]0.24[0.20, 0.28]0.19 [0.16, 0.22] Wald chi-squareReference531.43***262.12***1130.66***Model fitness Log-likelihood−12,020.997−11,412.183−11,387.646−9757.653 AIC24,045.9922,864.3722,801.2919,581.31 BIC24,061.7823,022.2422,903.9119,841.81 N19,80919,80919,80919,809 Number of clusters578578578578aOR = adjusted odds ratios; CI = Confidence Interval, 1.00 = PSU Primary Sampling Unit, ICC Intra-Class Correlation*p < 0.05, **p < 0.01, ***p < 0.001 Multi-level modeling of health facility delivery of the most recent birth among women of reproductive age in Sierra Leone aOR = adjusted odds ratios; CI = Confidence Interval, 1.00 = PSU Primary Sampling Unit, ICC Intra-Class Correlation *p < 0.05, **p < 0.01, ***p < 0.001 Fixed effect results Table 2 presents the findings regarding factors influencing health facility delivery among reproductive-age women in Sierra Leone. Our study identified several significant determinants of health facility delivery, including age at last childbirth, religious affiliation, level of education, employment, wealth, media exposure, number of childbirths, number of antenatal visits, companionship to health facility, community distance to health facilities, residence, and region. Women aged 20–29 had a 1.18 [aOR = 1.18; 95% CI: 1.05, 1.31] times higher odds of delivering in a health facility than those aged 15–19 years. Muslim Women [aOR = 0.81; 95% CI: 0.70, 0.94] and those practicing traditional/other religions [aOR = 0.33; 95% CI: 0.19, 0.60] were less likely to deliver in a health facility than Christian women. Women with primary [aOR = 1.26; 95% CI: 1.10, 1.43], secondary [aOR = 1.56; 95% CI: 1.37, 1.78], and higher education [aOR = 3.14; 95% CI: 1.95, 5.07] were significantly more likely to deliver in health facilities compared to those with no formal education. Employed women [aOR = 0.86, 95% CI: 0.76–0.97] had lower odds of delivering in health facilities than unemployed women. Household wealth demonstrated a positive association, with women from middle [aOR = 1.20, 95% CI: 1.07–1.36] and rich [aOR = 1.30, 95% CI: 1.06–1.60] households having higher odds of facility delivery compared to those from poor households. Media exposure substantially increased the likelihood of facility delivery as women exposed to media [aOR = 1.82, 95% CI: 1.16–2.87] have higher odds of utilizing health facilities for delivery. Reproductive history was also influential, as women with higher parity (≥ 5 childbirths) had lower odds of facility delivery [aOR = 0.78, 95% CI: 0.68–0.90] compared to those with 1–2 childbirths. Women who attended at least four antenatal care visits [aOR = 1.58, 95% CI: 1.42–1.76] had higher odds of delivering in health facilities, and those who reported companionship to health facilities not being a big problem [aOR = 1.32, 95% CI: 1.15–1.52] had higher odds of facility delivery. At the community level, distance to health facilities emerged as a critical barrier, with women in communities with medium [aOR = 0.62, 95% CI: 0.50–0.78] and high [aOR = 0.45, 95% CI: 0.35–0.57] distances having significantly lower odds of facility delivery compared to those in communities with low distances. Women from rural residences [aOR = 0.64, 95% CI: 0.50–0.84] had lower facility delivery odds than urban residents. Women from Northern [aOR = 0.29, 95% CI: 0.22–0.38], Northwestern [aOR = 0.35, 95% CI: 0.20–0.59], Southern [aOR = 0.56, 95% CI: 0.40–0.79], and Western [aOR = 0.05, 95% CI: 0.37–0.67] regions had lower odds of facility delivery compared to those from the Eastern region. Women surveyed in 2013 [aOR = 4.27, 95% CI: 3.49–5.22] and 2019 [aOR = 19.90, 95% CI: 15.26–25.95] had substantially higher odds of facility delivery than those surveyed in 2008. Discussions This study examined the regional trends, spatial patterns, and factors associated with health facility delivery among women of reproductive age in Sierra Leone using data from the 2008, 2013, and 2019 Demographic and Health Surveys. One of the most interesting findings was the notable increase in HFD utilization over time, rising from 24% in 2008 to 54% in 2013 and 83% in 2019. This finding agrees with a study conducted in Sierra Leone [37]. This upward trend reflects substantial progress in improving access to maternal healthcare services in Sierra Leone. The increasing proportion of health facility delivery utilization aligns with findings from other SSA countries where similar improvements have been observed in delivery rates at healthcare facilities following policy changes, highlighting the importance of continuous investments in healthcare [23, 32, 38]. The introduction of the free healthcare initiative in 2010 and women’s empowerment indices, such as education and wealth, likely contributed to this increase over time [39]. Our study also revealed that the pooled prevalence of health facility delivery among women of reproductive age in Sierra Leone was 54%. The prevalence is higher than that of other studies conducted in SAA [31, 32, 40] and Asia [41]. Our finding is similar to a cross-national study of 28 sub-Saharan African countries that report a 54% prevalence of health facility deliveries, ranging from 23% in Chad to 94% in Gabon [42] and Tanzania [43], indicating a regional trend toward improved facility-based deliveries. Additionally, this shows a collective movement towards skilled attendance at birth and highlights the importance of continued efforts to improve maternal healthcare services across the region. Our study indicates a significant increase in the percentage of women delivering at health facilities across all the regions from 2008 to 2019, reflecting the improvement in access to and utilization of facility-based healthcare services for childbirth. The Northern region recorded the lowest proportion of health facility delivery, but a notable improvement in health facility deliveries was still observed over the study period. This may result from healthcare access and outreach programs to improve maternal health outcomes in the country, as better antenatal care services directly influence the likelihood of health facility delivery [24, 38, 44]. The western region consistently exhibited a higher rate of health facility delivery than other regions. The consistently higher rate of health facility delivery in the Western region, where the capital city is located, suggests there may be disparities in healthcare access and quality between urban and rural areas, potentially impacting maternal health outcomes differently across regions. Additionally, the consistent improvements in health facility delivery rates suggest the effectiveness of community health initiatives that likely enhanced women's perceptions of institutional childbirth [28, 45]. The present study revealed that the spatial distribution of health facility deliveries was spatial clustering rather than randomly distributed. The Moran's I index values for all three survey years were statistically significant, indicating that areas with high health facility delivery utilization were clustered rather than randomly distributed across the country. These observed significant spatial clustering aligns with similar studies conducted in Ethiopia [30, 46] and Bangladesh [47], emphasizing the role of geographic inequities in maternal healthcare access. In our hot and cold spots analysis, our study identifies significant spatial patterns in health facility delivery utilization across the three different survey years, highlighting both progress and disparities. The Eastern and Southern regions consistently emerged as regions with a higher utilization of health facility delivery from 2008 to 2019. This trend indicates better healthcare access and well-established healthcare infrastructures in these regions. The possible reason for this overall improvement in health facility delivery utilization could be the free healthcare initiative implemented in 2010 [48]. The Western region also showed higher utilization of health facility delivery over the survey periods. However, some localized cold spots were identified in 2013, suggesting potential barriers to healthcare access in certain areas within the region, which may stem from inadequate healthcare infrastructure or socio-economic inequalities [49]. The North and Northwest remained as cold spot regions throughout the study period, demonstrating persistently lower utilization of health facility delivery. The unequal distribution of healthcare resources may contribute to these disparities, as access to trained healthcare providers and emergency obstetric care facilities remains inadequate in underdeveloped regions [50]. To ensure equitable access to maternal health services nationwide, policymakers must address these inequalities [49]. However, using Andersen's Healthcare utilization model, our study provides a comprehensive analysis of the factors influencing the utilization of health facility delivery in Sierra Leone from 2008 to 2019. Our study demonstrates that predisposing, enabling, and needing care factors can influence women's decision-making regarding the utilization of health facilities for deliveries. Women aged 20–29 had higher odds of facility-based delivery compared to adolescents (15–19 years), possibly reflecting greater awareness of pregnancy-related risks that come with maturity and experience. This finding is consistent with evidence from Sierra Leone and other low-resource settings [39, 51–53]. However, other studies conducted in Madagascar [32] and in Nigeria [40] were inconsistent with our study. However, this advantage diminished for older women, suggesting that perceived risk may decrease with age and multiple childbirth experiences. This aligns with our finding that women with higher parity (≥ 5 childbirths) had significantly lower odds of facility delivery, potentially indicating a reduced perception of risk after multiple successful deliveries, leading them to opt for home delivery over skilled ones [28]. This finding is consistent with previous studies from various sub-Saharan African countries, including Ethiopia [30], Nigeria [28, 54], Uganda [55]and Senegal [56] and in Bangladesh [57]. Another possible reason could be that traditional beliefs surrounding childbirth can lead women, particularly those with multiple births, to assume their pregnancies are manageable without professional healthcare, thus diminishing the perceived need for health facility deliveries [58]. Education emerged as a strong predictor of health facility delivery utilization, with women with primary, secondary, and higher education being significantly more likely to deliver in health facilities than those without formal education. This finding is supported by numerous studies [28, 30, 32, 39, 56, 59, 60], which have consistently demonstrated that higher levels of education are associated with increased awareness of maternal health risks and greater access to healthcare services. This suggests that education enhances women's understanding of the benefits of skilled birth attendance and medical interventions available at health facilities. Therefore, Programmes to improve health literacy among women can improve the use of health facilities for delivery while promoting and ensuring the education of the girl child and women [28]. Interestingly, employed women had lower odds of facility delivery than those who were unemployed. This finding is consistent with a study conducted in Pakistan [61]. This could be because work commitments take up their time or because their jobs make accessing healthcare harder. Religious affiliation also played a significant role, with women practicing Islam and traditional or other religions being less likely to deliver in health facilities compared to Christians. This aligns with findings from previous studies [26, 62–65], which report that women practicing traditional religions or Islam are less likely to utilize health facilities compared to their Christian counterparts. This perspective is echoed in various studies across multiple countries, including Nigeria and Uganda, where cultural aversion to institutional childbirth has been reported as a significant barrier [66, 67]. This may reflect differences in risk perception influenced by cultural and religious beliefs about childbirth, where traditional practices might be perceived as adequate protection against delivery complications. Additionally, possible reasons were barriers to maternal healthcare, such as lack of privacy, inadequate knowledge of healthcare workers about Muslim women's religious and cultural practices, and absence of cultural or religious-specific maternal healthcare services [62]. This finding calls for health facilities to be more culturally sensitive and inclusive to better serve women from diverse religious backgrounds. It also highlights the importance of addressing specific barriers preventing certain groups from accessing essential healthcare services. Similarly, the influence of enabling factors was pronounced in our study. Household wealth demonstrated a positive association with the utilization of health facilities for deliveries. This finding is consistent with other studies [23, 26, 31, 32, 68, 69]. These factors could be linked to the notion that better economic conditions may encourage individuals to pursue healthcare and empower them to make independent healthcare decisions as they can cover the required medical and transportation costs [31]. Media exposure significantly increased the likelihood of facility delivery by raising awareness about the benefits of institutional delivery through health communications and campaigns. This finding aligns with research conducted in Madagascar [32], Sub-Saharan Africa [42], and Ghana [40]. Furthermore, our study also identified important enabling factors at the community level. Distance to health facilities emerged as a critical barrier, with women in communities with medium and high distances having significantly lower odds of delivering in health facilities. This finding is consistent with previous research that has identified community distance to health facilities as a key determinant of healthcare utilization [28, 40, 54]. One possible reason for this finding is that when health centers are difficult to access, individuals may choose other options. They may circumvent the anticipated challenges of obtaining maternal health services from these facilities [28]. Women living in rural areas face significant barriers when it comes to utilizing health facilities for deliveries. Consistent with previous studies [26, 31, 69], our study found that rural women were less likely to deliver at a health facility than women in urban areas. This difference is often due to travel-related issues, available infrastructure, and other resource limitations. Regional disparities were also evident, with women from the country's Northern, Northwestern, Southern, and Western regions having significantly lower odds of HFD utilization than those in the Eastern region. These findings are consistent with studies documenting regional inequalities in healthcare access due to variations in infrastructure and socio-economic development [40, 70, 71]. Our study identified several factors that can be interpreted as the need for care within Andersen's Framework of Healthcare Service Utilization Model. One key finding was that women who attended at least four ANC visits during pregnancy were more likely to give birth in a healthcare facility. This finding agrees with previous studies conducted in Asia [10, 64], East Africa [31, 72], and SAA [42, 54, 73]. This highlights the crucial role of regular antenatal care, as these visits allow healthcare providers to inform women about the benefits of having their babies in a hospital or clinic and help them prepare a birth plan. Conversely, women with higher parity had lower odds of facility delivery compared to those with few children, suggesting that previous childbirth experiences may reduce the perceived need for professional healthcare during delivery. This is consistent with studies conducted in Madagascar [32], Ethiopia [31], Nigeria [28] and SAA [23, 74]. Women who reported companionship to health facilities was not a big problem had higher odds of delivering in a health facility. This finding is inconsistent with a study conducted in Nigeria [28]. Women in the 2013 and 2019 survey years showed a significantly higher likelihood of using health facilities for deliveries than in 2008. This indicates a positive trend toward seeking professional healthcare during childbirth. This shift may be attributed to increased awareness of the importance of skilled attendance at birth, implementation of the FHCI in the country, and better access to healthcare services for pregnant women. Implications and policy practices The findings of this study provide valuable guidance for policymakers and health professionals striving to improve maternal health outcomes in Sierra Leone. Despite progress in increasing health facility deliveries, persistent disparities highlight the need for targeted interventions addressing the complex factors influencing maternal health behaviors. Strengthening maternal health education is crucial, as education significantly impacts a woman's likelihood of delivering in a health facility. Integrating maternal health topics into school curricula and adult literacy programs can help women better understand the benefits of facility-based delivery and the risks associated with home births. Community-level programs should also focus on addressing cultural and religious beliefs that may hinder the use of health facilities. Regional disparities in health facility delivery rates call for geographically tailored strategies. Women in rural and northern areas face unique challenges, such as limited access to healthcare facilities. Addressing these barriers requires investments in infrastructure, transportation systems, and strategic placement of health facilities. Financial barriers should also be mitigated through sustained free maternal health services and social protection programs targeting vulnerable populations. Antenatal care presents a key opportunity to prepare women for childbirth and complications, emphasizing birth preparedness and readiness through counseling. Expanding culturally sensitive media campaigns can also promote awareness and encourage positive health-seeking behaviors. Empowering women through education, social support, and efforts to address restrictive gender norms is essential, particularly in light of declining trends in women’s participation in healthcare decision-making. To sustain and build upon the progress achieved under the FHCI, health system strengthening is vital. This includes improving supply chain management to ensure the availability of essential medicines, implementing community monitoring systems to enhance accountability, and addressing informal fees. Additionally, quality improvement initiatives focused on creating positive delivery experiences can encourage continued use of health facilities. Strength of the study This study possesses multiple strengths that augment the validity of its findings. Firstly, it employs nationally representative data from three Sierra Leone Demographic and Health Surveys (2008, 2013, and 2019), establishing a robust foundation for examining trends in health facility deliveries across various regions. Secondly, advanced geospatial techniques helped identify important patterns in how health facilities are used, which can direct focused efforts for improvement. The study employed a multilevel modeling approach, examining individual and community factors to understand the determinants influencing health facility deliveries comprehensively. The rigorous data management protocols, including sample weighting and multiple imputation for missing data, enhance the methodology's strength. Limitations and future research direction While this study offers valuable insights into health facility deliveries in Sierra Leone, several limitations should be considered when interpreting the findings. Firstly, the DHS data can only show relationships between variables rather than cause-and-effect. Self-reported data may suffer from recall or social desirability bias, and the displacement of GPS coordinates for confidentiality could affect accuracy in identifying delivery hotspots. The analysis did not account for specific contextual factors, like cultural beliefs and systemic barriers affecting facility usage. While we included community-level factors through multilevel modeling, these were based on aggregated individual data, which may overlook important contextual factors. Our choice of random intercepts over random slopes limits our understanding of how individual predictors vary by cluster, which could obscure geographic differences in factors like education and wealth. Lastly, although multiple strategies were used to address missing data, some bias may remain, particularly for variables with higher rates of missingness. Despite these limitations, our findings provide nationally representative evidence to inform maternal health policies in Sierra Leone. Future research should consider longitudinal designs, comprehensive community-level data collection, and mixed-methods approaches for deeper insights. Conclusion This study shows progress in health facility deliveries in Sierra Leone over the study period. Statistically significant hotspots for health facility delivery utilization were identified in the Eastern and Southern regions, while persistent cold spots were evident in the Northern and Northwestern regions. Factors such as age at last childbirth, religious affiliation, educational attainment, employment, wealth index, media exposure, antenatal care attendance, number of childbirths, companionship to health facility, rural residence, community distance to health facilities, and regional location influence health facility deliveries significantly. Targeted interventions are essential to ensure equitable access to health facility delivery and enhance maternal and neonatal health outcomes. These interventions should improve maternal education, reduce geographic barriers, and address community-level inequalities.