Spatiotemporal Trend and Hotspot Analysis of Stillbirth Prevalence in Ghana

ABSTRACT

Introduction

Stillbirth, although a preventable health outcome, remains a silent contributor to global mortality. Stillbirths have also been deemed to be one of the most significant, albeit least known and reported adverse pregnancy consequences. The regional variations of stillbirth and its risk factors at the district level have not been well studied in Ghana. This study examined the spatiotemporal distribution, hotspots, and determinants of stillbirth in Ghana over 5 years (2017 ‐ 2021).

Methods

This study is an ecological, observational analysis using district‐level aggregated stillbirth data from 2017 to 2021. The main analyses used trend analysis, Moran's I, Gi*, and OLS regression. Global Moran's Index was used to determine spatial autocorrelation. Getis Ord G* was used to determine spatial hotspots, clusters and outliers. An Ordinary Least Squares analysis was performed to determine the association between the stillbirth rate and other risk factors.

Result

The findings showed that the highest stillbirth rates throughout the 5 years were observed in districts in the Greater Accra region (Tema with SBR between 30.0 and 34.5) and Volta region (Akatsi North (SBR = 45.2)). There were persistent hotspot districts throughout the 5 years. These occurred in the Northern and Greater Accra regions. The regression analysis revealed that age of mothers, number of health facilities in the district, and number of beds per facility in the district are significant risk factors for stillbirth in Ghana.

Conclusion

The findings of this study reveal a persistently high stillbirth burden across several districts in Ghana over the 5‐year period, with districts such as Tema, Akatsi North, and La‐Dade‐Kotopon consistently recording the highest stillbirth rates. These high‐burden districts require urgent and targeted quality improvement initiatives, including strengthening emergency obstetric care, improving skilled birth attendance, and enhancing antenatal care coverage.

1. Introduction

Stillbirth continues to be a significant public health problem globally. The World Health Organization (WHO) describes stillbirth as the loss of a viable fetus above 28 weeks of pregnancy and before delivery [1]. There were about 1.9 million stillbirths around the world in 2023, and the stillbirth rate of the world was 14.3 stillbirths per 1000 total births [2], of which 84% occur in low and middle‐income countries [3]. Over 40 percent of all stillbirths occur during labor (intrapartum stillbirth) (WHO, 2019). The highest stillbirth rates are recorded in Sub‐Saharan Africa (SSA), accounting for 35.4 percent of the global burden of stillbirths [4]. The stillbirth rate for SSA in 2019 was estimated at 21.67 per 1000 births, showing over 825,000 stillbirths being recorded in the region. This is alarming when compared to the stillbirth rate of 3.2 per 1000 births in Europe, which indicates approximately 25,000 stillbirths in the same year [2].

With substantial improvements in the healthcare system in Ghana, the initiation of the National Health Insurance Scheme (NHIS) in 2003, and the Free Maternal Healthcare Policy (FMHP) in 2008, there has been increased access to healthcare services among pregnant women and children [5]. This decreased the maternal mortality rate from 2003 to 2017 (398 to 308 per 100,000 live births) [6]. There was a further increase in the proportion of women who had at least four antenatal visits from 70.6 percent in 2003 to 78.7 percent in 2008 and 86.5 percent in 2014 [6]. Moreover, the proportion of delivery by skilled attendants increased from 43.9 percent in 2003 to 67.8 percent in 2008 and 72.8 percent in 2014 [6]. Establishing modern healthcare facilities and recruiting and training health personnel have also contributed to a more efficient healthcare system [7]. However, poor location planning of hospitals has led to huge disparities in the distribution of health facilities between districts [5], which affects stillbirth due to the difference in population distribution and the demand for health care services.

Despite these leaps in the healthcare system, quality of care and inequality in access to healthcare continue to be a problem, especially in certain geographical areas. Saleh [5, 8]. Saleh attributes this problematic situation to the inefficient allocation of health resources. Investments in hospitals are based on regional levels rather than need‐based standards, causing a shortage of lower‐level health facilities and equipment at the subdistrict level [5]. Additionally, poor location planning of hospitals has led to huge disparities in the distribution of health facilities between districts [5]. Although there are still some unknown risk factors associated with stillbirth, the known factors affecting stillbirth can be grouped into 4 main categories. These broad categories are maternal factors, maternal medical conditions, maternal reproductive history, and fetal characteristics. These risk factors include the level of education, socio‐economic status (SES), maternal age (< 20 years and > 35 years), level of antenatal care (visits < = 4), low birth weight, parity (< 1 and > 5), availability of beds in health facilities, obesity, hypertension, malaria, syphilis, anemia during pregnancy, smoking, and drinking [9]. According to UNICEF, Ghana's stillbirth rate of 28.69 per 1000 in 2000 linearly decreased to 21.65 per 1000 in 2019, showing notable improvement (UNICEF, 2020). Across the globe, it is observed that the stillbirth rate is decreasing by 1.1% per year [9]. However, this decreasing trend is significantly attributed to developed countries rather than the low‐middle‐income countries (LMIC). When comparing to developed countries like Australia, with an SBR of 3.21 per 1000 in the year 2000 to a decrease in SBR of 2.22 per 1000 in the year 2019 (UNICEF, 2020), Ghana needs to identify and define well‐planned policies from the current SBR outcome situation based on the prevailing risk factors.

Stillbirths place a significant burden on countries and healthcare systems, particularly in developing countries. Stillbirths outnumber many other global health issues, including HIV/AIDS, and intrapartum stillbirths alone surpass malaria‐related child deaths worldwide [10]. However, Ghana, like many SSA countries, does not actively collect, record, and monitor stillbirth data. Although studies have investigated the prevalence of stillbirths, none of them utilized national data to focus on stillbirth differences by district or region [11, 12]. However, using non‐spatial statistical analysis, a retrospective case‐control study performed in northern Ghana showed that geographical locations affected the stillbirth rates [13]. Furthermore, the application of a Geographical Information System (GIS) to the study of the spatial correlation between stillbirths and its risk factors has not been fully exploited. Despite a growing body of literature on stillbirth in Ghana, there remains a critical gap in the use of national routine health data, specifically the District Health Information Management System (DHIMS‐2) to characterise district‐level spatial clustering, hotspots, and temporal trends in stillbirth rates over time. Most existing studies have relied on facility‐based or survey data with limited geographic scope, restricting the ability to draw nationally representative, district‐level conclusions.

The use of GIS to investigate, assess, and prevent disease is becoming increasingly popular [14, 15]. Disease‐affected regions, risk factors, available services, and preventive and health resource planning can all be studied using GIS. Data visualization, mapping, and geospatial analyses play an essential role in addressing the growing need for better national spatial investigation by identifying hotspot zones of prevalent health issues, identifying significant factors affecting disease in the country, and analyzing the prevalent factors using spatial regression models such as the Geographically Weighted Regression (GWR) model and the Ordinary Least Square (OLS) model [16]. Ordinary Least Square regression is instrumental since it is relatively straightforward to validate model assumptions like linearity, constant variance, and outlier effect using simple graphical approaches [17].

The prevalence of stillbirths in Ghana reveals complex interactions involving maternal health, socio‐economic factors, and geographic disparities. Studies consistently highlight that comprehensive access to healthcare, maternal education, and targeted health interventions are critical to reducing stillbirth rates in the country. Therefore, this study aimed to describe temporal trends, identify spatial clusters and hotspots, and examine district‐level determinants of stillbirth in Ghana from 2017 to 2021, using national DHIMS‐2 data.

2. Methods

2.1. Study Area

The Republic of Ghana (Figure 1), bounded centrally in the south by the Gulf of Guinea, is the second‐most populous country in West Africa [14]. Ghana is only a few degrees north of the Equator; therefore, a warm climate is experienced almost all year round. Ghana spans an area of 238,535 km² (92,099 sq mi) between latitudes 4°45'N and 11° N and longitudes 1°15'E and 3°15'W, with an Atlantic coastline stretch of 560 kilometers (350 miles). The country is divided into 16 administrative regions‐Northern, Ashanti, Western, Volta, Eastern, Upper East, Upper West, Central, Bono East, Greater Accra, Savannah, North‐East, Oti, Western North, Ahafo, and Bono [15] as shown in Figure 1. There are currently 260 districts that are distributed across the 16 administrative regions and ~9260 public and private health facilities, including 6 teaching hospitals in Ghana.

Figure 1.

Map of Ghana showing all the study districts, 2022.

2.2. Data Source

Aggregated secondary data were acquired from District Health Information Management Systems (DHIMS‐2). Monthly midwifery‐reported data on stillbirth were extracted from the DHIMS‐2 database from 1st January 2017 to 31st December 2021 for 260 districts in Ghana. The dataset was extracted, cleaned, arranged, and imported into the district's shape files. The data was imported into the Ghana administrative district shapefile acquired from Ghana Statistical Services. The data on the DHIMS platform was collected by health information officers from various health facilities in the countries monthly and entered in a timely manner. Therefore, software checks, validations, and human checks at various levels were adopted to improve data quality. A study conducted in Ghana on the completeness of antenatal aggregated data from DHIMS‐2 showed completeness of 94.3% (95%, CI = 90.6%−98.0%) which showed the maternal health services data was reliable for use [18].

The United Nations Inter‐Agency Group for Child Mortality Estimation (UN IGME) and its Core Stillbirth Estimation Group (CSEG) used a model‐based approach to obtain a stillbirth rate (SBR) estimate that is nationally representative of all UN countries. The model used data from administrative sources, health management Information Systems, household surveys, and population surveys to estimate the stillbirth rates (UNICEF, 2020). In this study, data were collected from all health facilities through DHIMS, serving as a good representative sample of the population.

2.3. Ethical Review

The data utilizes district‐level aggregated data without a link to individuals collected in routine disease surveillance and control activities. Therefore, it was not considered to constitute human subjects research requiring Institutional Review Board approval. However, approval was sought from the Ghana Health Services to use the data on stillbirth indicators for all 260 districts in the country.

2.4. Data Description

The data used were yearly aggregated data with districts as the observation unit. The dependent variable was the stillbirth count. Variables, including total births and ANC mother registrants, were used as denominators for calculating the proportions of the explanatory variables. The independent variables were the number of health facilities, the aggregate proportion of beds in the health facilities, mothers who are parity 0 or parity 5 or more, mothers who are less than 21 years or 35 years or older, babies with birth weight less than 2500 g, skilled delivery, and low birth weight. The proportions of mothers who were parity 0 or parity 5 or more or mothers who were less than 21 years or 35 years or older, were calculated by dividing by ANC mother registrants and then multiplying by 100 to find the percentages. Low birth weight was divided by total births and multiplied by 100 to find the percentage of low birth weight. The number of stillbirths was divided by total births and then multiplied by 1000 to find the stillbirth rate per 1000, details are shown in Table 1.

Table 1.

The Description of data extracted at the district level for the years 2017 to 2021.

Variable	Role in study	Definition	Unit of measurement	Data source	Formula
Stillbirth Rate	Outcome	The number of stillbirths (foetal deaths at ≥ 28 weeks of gestation) occurring per 1000 total births (live births + stillbirths) in a district within a calendar year.	Rate per 1000 total births	Ghana DHIMS‐2 (Monthly midwifery‐reported data on stillbirth)	(Number of stillbirths ÷ Total births) × 1000
Total Stillbirth	Outcome	Total number of Stillbirth (foetal deaths at ≥ 28 weeks of gestation) occurring per 1,000 total births	Rate per 1000 total births	Ghana DHIMS‐2
Health Facilities	Predictor	The total number of functional health facilities (including hospitals, health centres, clinics, and CHPS compounds) reported within each district in a given year.	Count (number of facilities)	Ghana DHIMS‐2)	Direct count from DHIMS‐2 facility registry per district
Bed Complement	Predictor	The total number of hospital/facility beds available across all health facilities within a district in a given year, used as a proxy for inpatient healthcare capacity.	Count (number of beds)	Ghana DHIMS‐2	Aggregate sum of all reported facility beds per district
4th ANC Visit (PERCENT_4THVISIT)	Predictor	The percentage of pregnant women in a district who attended at least four antenatal care (ANC) visits during pregnancy, in line with the WHO‐recommended focused ANC model. Reflects the coverage and utilisation of antenatal care services.	Percentage (%)	Ghana DHIMS‐2	(Number of women completing ≥ 4 ANC visits ÷ Total expected pregnancies) × 100
Parity 0 — Primipara (PERCENT_P0)	Predictor	The percentage of deliveries in a district attributed to primiparous women (women delivering for the first time, parity = 0). First‐time mothers are at elevated risk of adverse birth outcomes due to inexperience and potential obstetric complications.	Percentage (%)	Ghana DHIMS‐2	(Number of deliveries by primiparous women ÷ Total deliveries) × 100
Parity 5 + — Grand Multipara (PERCENT_P5UP)	Predictor	The percentage of deliveries in a district attributed to grand multiparous women (women with five or more previous deliveries). Grand multiparity is a recognised risk factor for stillbirth due to uterine fatigue, placental insufficiency, and increased likelihood of obstetric complications.	Percentage (%)	Ghana DHIMS‐2	(Number of deliveries by women with parity ≥ 5 ÷ Total deliveries) × 100
Maternal Age < 21 or 35 and above(PERCENT_MAGE_LESS21)	Predictor	The percentage of deliveries in a district occurring among adolescent and young mothers aged below 21 years and above 35 years. Young maternal age is associated with increased risk of stillbirth due to biological immaturity, older mothers also experienced delayed care‐seeking, and limited access to quality obstetric care.	Percentage (%)	Ghana DHIMS‐2	(Number of deliveries by mothers aged < 21 years or 35 and above ÷ Total deliveries) × 100
Low Birth Weight (PERCENT_LBW)	Predictor	The percentage of live births in a district with a recorded birth weight of less than 2500 grams. Low birth weight is a marker of foetal growth restriction and poor maternal nutritional status, both of which are associated with adverse perinatal outcomes, including stillbirth.	Percentage (%)	Ghana DHIMS‐2	(Number of live births weighing < 2500 g ÷ Total live births) × 100
Skilled Delivery (PERCENT_SKILLEDD)	Predictor	The percentage of deliveries in a district attended by a skilled birth attendant (SBA), defined as a trained health professional (doctor, nurse, or midwife) capable of managing normal deliveries and diagnosing or referring obstetric complications. Reflects the quality of and access to intrapartum care.	Percentage (%)	Ghana DHIMS‐2	(Number of deliveries attended by a skilled birth attendant ÷ Total births) × 100

2.5. Data Visualizing

ArcGIS (version 10.7.1) was used to visualize the data on maps. We updated the attribute table in the 260 districts shapefile with variables including health facilities, stillbirths, stillbirth rates, total births, bed complement, mothers who haven't delivered before or delivered 5 or more times (parity 0 and 5 + ), mothers at high risk (< = 20 years, > = 35 years), and babies with low birth weight (< 2.5 kg), ANC registrants, skilled delivery and their respective proportions. Choropleth/density, hotspot, and cluster analysis maps were created to show the stillbirth rates over the 5 years.

2.6. Data Analysis

Trend analysis was performed using Excel to develop a double y‐axis line graph that shows the total stillbirth rate and total birth trend of the country over the 5 years. Global Moran's I was performed to determine whether there was clustering, a random or dispersed distribution with a 95% confidence threshold (Glen, 2022) of cases observed in the population. Moran's I coefficient lies between + 1 (which shows perfect clustering of similar values) and −1 (which shows total spatial dispersion), while a 0 value shows a perfectly random distribution or no autocorrelation [19]. The Moran's index equation is shown below.

$$I=\frac{\sum i\sum j(y_i-\overline{y})\left(y_j-\overline{y}\right)}{s^2\sum i\sum j{w}_{ij}}$$ (1)

Where y is the dependent variable, $\stackrel{\circledR }{y}$ is the mean, $s^2$ is the variance, and w is the binary connectivity matrix. Global Moran's I was used to assess whether the overall spatial distribution of stillbirth rates was clustered, dispersed, or random across the study area. Getis‐Ord Gi* was then applied to identify specific geographic hotspots and cold spots, enabling targeted public health interpretation. LISA complemented these analyses by detecting local spatial clusters and spatial outliers, revealing heterogeneity that global statistics may obscure. Hotspot analysis was performed using the Getis‐Ord‐Gi* statistic, which employs inverse distance square in ArcGIS to identify clusters of low (cold spots) and high (hotspots) to identify where the clustering is observed. For the Getis‐Ord Gi* and LISA analyses, statistically significant clusters were identified at the 95% (z ≥ ± 1.96), 99% (z ≥ ± 2.58), and 99.9% (z ≥ ±3.29) confidence levels. No correction for multiple comparisons was applied, consistent with exploratory spatial analysis practice. The results are therefore interpreted as indicative of spatial patterns rather than confirmatory. Local Indicator of Spatial Association (LISA) analysis was performed to identify the area similarities per values of stillbirth.

The Ordinary Least Square (OLS) Regression statistics integrated into ArcGIS were used to investigate the spatial relationship between the occurrence of stillbirth and its determinants from 2018 to 2021. OLS diagnostics were used to assess and select significant predictors of stillbirth. OLS has an assumption of spherical errors, which is homoscedasticity and no autocorrelation. Hence, when the dependent variable is spatially autocorrelated, the deviation from the true value becomes greater, and the model gives a poor coverage rate [19]. The Ordinary Least Square equation is given as:

$$y={\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+\dots +{\beta }_k{x}_k+\epsilon$$ (2)

where y is the dependent variable, $x_k$ are the independent variables, ${\beta }_k$ are the regression coefficients, and $\epsilon$ is the error term of the residuals.

Model performance was assessed using the coefficient of determination (R²) and adjusted R². Multicollinearity among predictors was examined using the Variance Inflation Factor (VIF) values, with VIF > 10 used as the threshold for concern. Heteroscedasticity was assessed using the Breusch‐Pagan test. Spatial autocorrelation in model residuals was evaluated using Moran's I on OLS residuals to determine whether the OLS assumption of independent errors was violated. Although OLS assumes spatially independent errors, it was retained as the primary regression method to allow straightforward interpretation of covariate relationships and comparability across years. Moran's I on OLS residuals was used to detect spatial autocorrelation; where residual autocorrelation was observed, this is acknowledged as a limitation. Future work could employ spatial lag or spatial error models, or Geographically Weighted Regression (GWR), to more explicitly account for spatial dependency in the data. Separate OLS regression models were built for each year from 2018 to 2021 to capture temporal variation in the spatial relationships between stillbirth and its determinants. The covariates included in the models include health facility, bed complement, maternal age, parity, and ANC visits.

3. Results

3.1. Descriptive Statistics

Table 2 presents the descriptive statistics of the stillbirth rate and district‐level determinants across Ghana's districts from 2017 to 2021. Overall, the mean stillbirth rate declined steadily over the study period, from 12.02 per 1,000 total births (SD = 7.31) in 2017 to 9.72 per 1000 total births (SD = 7.07) in 2021, suggesting a modest but consistent reduction in stillbirth burden across districts. However, considerable variability was observed across districts in all years, as evidenced by the wide ranges recorded; for instance, district‐level stillbirth rates ranged from 0.6 to 33.1 in 2017 and from 0.5 to 41.6 in 2021, indicating persistent geographic inequality in stillbirth outcomes throughout the study period.

Table 2.

Descriptive statistics of stillbirth rate and district‐level determinants across Ghana by year, 2017–2021 (n = 260).

Year	Variable	Mean	SD	Median	IQR	Min	Max
2017
	Stillbirth Rate	12.02	7.31	10.8	11.15	0.6	33.1
	Bed Complement	113.17	122.94	82.2	90.48	7.6	1424.7
	Health Facilities	35.62	15.69	33	18.25	11	134
	4th ANC Visit	73.6	25.76	69.3	25.05	22.2	242.3
	Low Birth Weight	2.24	2.7	1.3	2.58	0.03	20.45
	Maternal age	13.49	4.67	14.3	5.45	0.3	41.4
	Parity 0	27.68	11.45	27.2	4.45	3.7	199.4
	Parity 5 +	9.84	4.54	9.5	5.92	1.2	25
	Skilled Delivery	53.26	33.33	48.05	39.98	0	183.1
2018
	Still Birth Rate	10.96	7.04	10.15	11.45	0.7	34
	Bed Complement	109.29	101.06	81.3	87.7	0.49	683.3
	Health Facilities	35.62	15.69	33	18.25	11	134
	4th ANC Visit	74.86	27.51	69.65	21.95	21	325.1
	Low Birth Weight	1.9	2.14	1.15	2.45	0.01	13.62
	Maternal age	13.2	4.42	13.8	5.35	0.6	37.4
	Parity 0	28.13	10.86	27.65	4.12	4.2	190.7
	Parity 5+	9.9	4.54	9.6	6.23	1.2	25.5
	Skilled Delivery	56.71	35.03	51.85	39.87	0	191.1
2019
	Stillbirth Rate	10.13	6.59	8.95	9.93	0	31.4
	Bed Complement	110.22	100.28	77.7	81	0.08	548.1
	Health Facilities	35.62	15.69	33	18.25	11	134
	4th ANC Visit	75.5	23.75	70.9	21.58	27.3	262.5
	Low Birth Weight	2.09	2.22	1.29	2.71	0.03	10.96
	Maternal age	13.17	4.42	13.7	5.43	0.4	33.5
	Parity 0	28.33	10.39	27.75	4.2	4.4	182
	Parity 5+	9.7	4.48	9.4	5.45	0.8	26.3
	Skilled Delivery	57.95	31.26	53.1	38.42	4.5	195.4
2020
	Still Birth Rate	10.06	6.7	8.85	9.45	0.5	30.3
	Bed Complement	119.68	107.2	88.25	91.45	1	751.3
	Health Facilities	35.62	15.69	33	18.25	11	134
	4th ANC Visit	74.2	20.8	70.5	18.53	27.7	245.1
	Low Birth Weight	2.01	2.26	1.02	2.41	0.03	12.76
	Maternal age	12.45	4.22	13.35	5.75	0.6	24.1
	Parity 0	28.06	8.77	27.55	3.93	5.1	156
	Parity 5+	9.73	4.33	9.15	5.15	1.2	24.2
	Skilled Delivery	58.1	30.48	53.2	36.08	5.1	205.5
2021
	Still Birth Rate	9.72	7.07	9	9.75	0.5	41.6
	Bed Complement	125.48	107.81	90.6	90.5	0.93	779.3
	Health Facilities	35.62	15.69	33	18.25	11	134
	4th ANC Visit	81.14	27.01	77.45	20.62	31.4	356.3
	Low Birth Weight	1.96	2.3	1.08	2.27	0.01	13.05
	Maternal age	12.47	4.34	13.5	5.55	0.3	26.8
	Parity 0	28.61	8.84	28.1	3.92	5	158.4
	Parity 5+	9.49	4.21	9.1	5.4	0.7	21.7
	Skilled Delivery	63.51	31.88	61.2	32.52	10.1	350.9

The mean number of health facilities remained constant across all 5 years (Mean = 35.62, SD = 15.69), reflecting the static nature of health infrastructure data in the dataset. Bed complement showed a general increasing trend, rising from a mean of 113.17 (SD = 122.94) in 2017 to 125.48 (SD = 107.81) in 2021, though substantial variability across districts was noted throughout.

Regarding antenatal care utilisation, the mean percentage of women attending four or more ANC visits increased from 73.6% (SD = 25.76) in 2017% to 81.14% (SD = 27.01) in 2021, suggesting a gradual improvement in antenatal care coverage over the study period. Similarly, skilled delivery coverage showed an upward trend, increasing from a mean of 53.26% (SD = 33.33) in 2017% to 63.51% (SD = 31.88) in 2021. However, the wide standard deviations and observed ranges indicate considerable disparity in skilled delivery uptake across districts.

The mean percentage of low‐birth‐weight births remained relatively low and stable across all years, ranging from 1.90% (SD = 2.14) in 2018% to 2.24% (SD = 2.70) in 2017. The proportion of deliveries by mothers aged below 21 years showed a slight decline from 13.49% (SD = 4.67) in 2017% to 12.47% (SD = 4.34) in 2021. The percentage of primiparous deliveries (parity 0) remained relatively stable across the study period, ranging from 27.68% to 28.61%, while the proportion of grand multiparous deliveries (parity 5 + ) showed a marginal decline from 9.84% (SD = 4.54) in 2017% to 9.49% (SD = 4.21) in 2021.

3.2. Trend Analysis and Choropleth Maps

Figure 2 shows the trend analysis and total birth rate from 2017 to 2021. It was observed that there was an increasing trend of total birth while the trend of stillbirth rate decreases. Stillbirth rate decreased sharply from 2017 to 2019. The decline continued from 2019 to 2021 but at a slow rate. As total birth rates are increasing while stillbirth rates are decreasing, it suggests genuine improvements in maternal‐fetal medicine and obstetric care quality.

Figure 2.

Trend analysis of stillbirth rate and total births from 2017 to 2021 showing the decrease in stillbirth rate over time. The left y‐axis shows total births (orange), and the right y‐axis shows stillbirth rate (blue) with its linear trend (dotted).

Figure 3 denotes the maps showing the stillbirth rate in Ghana in the various districts. In 2017, it was observed that about 14 districts had stillbirth rates less than 2.0 (SBR < 2.0) with Wassa East recording the least (SBR = 0.6). The other 13districts had a stillbirth rate greater than 25.0 (SBR > 25.0). Among the 13 districts, Tema recorded the highest stillbirth rate of 33.1. In 2018, a stillbirth rate below 2.0 (SBR < 2.0) was observed in 19 districts, with Ga South having the lowest stillbirth rate (SBR = 0.7), while 8 districts had stillbirth rates above 25.0 (SBR > 25.0), with Tema having the highest stillbirth rate (SBR = 34.0). Also, in 2019, it was observed that 16 districts had stillbirth rates less than 2.0 (SBR < 2.0), with Akyemansa having the least stillbirth rate (SBR = 0.0), and 6 other districts had stillbirth rates greater than 25.0 (SBR > 25.0), with Tema having the highest (SBR = 31.4).

Figure 3.

(a–e) Maps showing the Stillbirth rate in Ghana in the various districts (2017–2021).

In 2020, it was observed that 21 districts had stillbirth rates less than 2.0 (SBR < 2.0); with Wassa Amenfi Central having the least (SBR = 0.5), while 8 other districts with a stillbirth rate of greater than 25.0 (SBR > 25.0), with Akatsi North having the highest stillbirth rate (SBR = 45.2). In 2021, it was observed that 24 districts had stillbirth rates less than 2.0 (SBR < 2.0); Nzema East had the lowest (SBR = 0.5), while 7 other districts had stillbirth rates greater than 25.0 (SBR > 25.0), with La‐Dade‐Kotopon having the highest (SBR = 37.1).

3.3. Hotspot Analysis

Figure 4 shows hotspot maps in various districts. In 2017, about 4 districts were observed to have hotspots (90% confidence) in Greater Accra and Northern regions, 4 districts with hotspots (95% confidence) in the Northern regions, and 1 district with hotspot (99% confidence), Mion in the Northern region. In 2018, 14 districts had hotspots (90% confidence) in Greater Accra and Northern regions, 1 district with hotspot (95% confidence) in the Northern region, and 1 district with hotspot (99% confidence): Mion in the Northern region.

Figure 4.

(a–e) Hotspot maps of showing where there are significant Stillbirth hotspots in Ghana in the various districts (2017–2021). Spatial clusters were identified using the Getis‐Ord Gi statistic with the following confidence levels: 90% confidence (p < 0.10) indicates a probable spatial cluster; 95% confidence (p < 0.05) indicates a likely spatial cluster; and 99% confidence (p < 0.01) indicates a highly significant spatial cluster with less than 1% probability of occurring by chance.

In 2019, about 16 districts had hotspots (90% confidence) in the Greater Accra region, 7 districts with hotspots (95% confidence) in the Greater Accra and Northern regions, and 2 districts with hotspots (99% confidence), Mion and Sagnerigu, both in the Northern region. In 2020, about 11 districts had hotspots (90% confidence) in the Greater Accra region, 3 districts with hotspots (95% confidence) in the Northern region, and 4 districts with hotspots (99% confidence): Mion, Tolon, Sagnerigu, and Tamale, all in the Northern region. In 2021, about 13 districts had hotspots (90% confidence) in Greater Accra and Northern regions, 2 districts with hotspots (95% confidence) in the northern region, and no district with hotspots (99% confidence). The Northern and Great Accra Regions were the 2 regions with hotspots over the 5 years. Mion was almost consistently a hotspot among satellite districts Tolon, Sagnerigu, Tolon, Savelugu, and Kumbungu.

3.4. Spatial Autocorrelation Report

The analysis in Figure 5 shows that there was no statistically significant spatial clustering in stillbirth rates in Ghana in all 4 years analyzed (2018–2021). The p‐values were all larger than 0.05 which is an indication that the spatial patterns might have been caused by random chance and not because of the actual geographic clustering. Lack of large spatial clustering in all the years indicates several critical things. The risk factors of stillbirth seem to be well spread within Ghana but not representing certain hotspots. This implies that stillbirth is a national health issue as opposed to a community issue. Although the trend indicates that the stillbirth rates in the country are on a downward trend (apparently due to the trend analysis), the pattern of their distribution within Ghana is quite random across the regions of this country. This implies that the Stillbirth prevention campaigns must be national in their approach and be sensitive to any regional differences that may arise or alter with time.

Figure 5.

(a–e) Spatial autocorrelation report of using stillbirth count as dependent variables (2017–2021).

3.5. Cluster and Outlier Analysis

It was observed that there was a high value of stillbirth with similar surrounding neighbors in 7 districts (High ‐High) according to Figure 6. These are Asokwa, Ayawaso North, Ayawaso East, Accra Metro, Korle‐Klottey, La‐Dade‐Kotopon, and Ledzokuku and 3 High‐Low districts and 17 Low‐High districts. However, Anselin Local Moran's Index Analysis was performed on the 2018 ‐ 2021 stillbirth variable with random spatial autocorrelation. In 2018, it was observed that 3 districts had a High‐high correlation: Effia‐Kwesimintsim, Korle‐Klottey, and Asokwa; 1 district had a High‐Low correlation, 16 districts with high values but low values surrounding neighbors and 1 district, Bia West had low value with similar surrounding neighbors (Low‐Low). In 2019, it was observed that 6 districts, Effia‐Kwesimintsim, Asokwa, Accra Metro, Korle‐Klottey, La‐Dade‐Kotopon, Ledzokuku had high values of stillbirth with similar surrounding neighbors (High‐High), 2 districts had High‐Low correlation, 16 districts had Low‐High correlation, and no districts had Low‐Low correlation. In 2020, it was observed that 2 districts, Asokwa and Accra Metro, had high values of stillbirth with similar surrounding neighbors (High‐High), 5 districts had a High‐Low correlation, 12 districts had a Low‐High correlation, and 1 district, Aowin, had low value with similar surrounding neighbors (Low‐Low).

Figure 6.

(a–e) Cluster and Outlier maps showing where there are clusters of Stillbirth in Ghana (2017–2021).

In 2021, it was observed that 4 districts, Old Tafo, Asokwa, Korle‐Klottey, and La‐Dade‐Kotopon had high values of stillbirth with similar surrounding neighbors (High‐High), 3 districts had High‐Low correlation, 14 districts had Low‐High correlation and 3 districts, Bia West, Bole, Kwahu Afram Plains North low value with similar surrounding neighbors (Low‐Low).

3.6. Ordinary Least Square Analysis

Ordinary Least Square has a good coverage area of the model when there is homoscedasticity and no autocorrelation [19]; therefore, no OLS was performed on the 2017 dataset, which showed a spatial autocorrelation with a p‐value of 0.081853. A set of explanatory variables were used to develop the OLS models for 2018 to 2021.

Table 3 presents the results of separate ordinary least squares (OLS) regression models examining the district‐level determinants of stillbirth rates in Ghana for each year from 2018 to 2021. Robust standard errors were applied in all models to account for the heteroscedasticity identified in the diagnostic tests. Across all 4 years, bed complement was a consistent and statistically significant positive predictor of stillbirth rates (p < 0.001), suggesting that districts with higher inpatient bed capacity, likely reflecting larger, busier referral facilities receiving more high‐risk cases, recorded higher stillbirth rates. Similarly, health facility was a significant positive predictor in 2018 (β = 1.26, p = 0.035), 2019 (β = 0.74, p = 0.042), and 2021 (β = 1.22, p = 0.035), indicating that districts with more health facilities tend to record higher stillbirth rates, possibly reflecting referral concentration effects.

Table 3.

Summary of ordinary least squares results using stillbirth count as the dependent variable (2018–2021).

2018	Adjusted R‐Squared: 0.565811		Koenker (BP) Statistic: 100.678344
Variable	Coefficient	Standard Error	95% Confidence Interval	p	VIF
Health facilities	1.258576	0.592706	(0.0969, 2.4203)	0.034685*	1.269367
Bed complement	0.309083	0.059068	(0.1933, 0.4249)	0.000001*	1.558094
4th ANC visit	0.136312	0.081313	(−0.0231, 0.2957)	0.094919	1.125893
Parity 0	0.315948	0.161931	(−0.0014, 0.6333)	0.052151	1.275076
Parity 5+	1.690846	0.801999	(0.1189, 3.2628)	0.035988*	1.532912
Maternal age	−2.055365	0.588516	(−3.2089, −0.9019)	0.000577*	1.598992
Low Birth Weight	−1.189968	0.879838	(−2.9145, 0.5345)	0.177446	1.049842
Skilled Delivery	0.142576	0.123703	(−0.0999, 0.3850)	0.250184	1.386278

2019	Adjusted R‐Squared: 0.627007		Koenker (BP) Statistic: 60.309633
Variable	Coefficient	Standard error	95% Confidence Interval	p	VIF
Health facilities	0.7391	0.361036	(0.0315, 1.4467)	0.041676*	1.305713
Bed complement	0.298059	0.048472	(0.2031, 0.3931)	0.000000*	1.622602
4th ANC Visit	0.211441	0.091565	(0.0320, 0.3909)	0.021733*	1.181685
Parity 0	0.083251	0.145623	(−0.2022, 0.3687)	0.56805	1.27049
Parity 5+	0.713588	0.447868	(−0.1642, 1.5914)	0.112363	1.425485
Maternal age	−1.021464	0.379644	(−1.7656, −0.2774)	0.007608*	1.418201
Low birth weight	−1.347246	0.768663	(−2.8538, 0.1593)	0.080879	1.079965
Skilled delivery	0.382038	0.125924	(0.1352, 0.6288)	0.002676*	1.496364

2020	Adjusted R‐Squared: 0.66694		Koenker (BP) Statistic: 94.470325
Variable	Coefficient	Standard Error	95% Confidence Interval	p	VIF
Health facilities	0.64501	0.337093	(−0.0157, 1.3057)	0.056827	1.275387
Bed complement	0.28819	0.044735	(0.2005, 0.3759)	0.000000*	1.54851
4th ANC visit	0.236311	0.094897	(0.0503, 0.4223)	0.013406*	1.15454
Parity 0	0.083978	0.147303	(−0.2047, 0.3727)	0.569122	1.147013
Parity 5+	0.806411	0.547407	(−0.2665, 1.8793)	0.141976	1.431746
Maternal age	−1.26124	0.385337	(−2.0165, −0.5060)	0.001225*	1.426489
Low Birth Weight	−0.255153	0.677109	(−1.5823, 1.0720)	0.706633	1.060307
Skilled delivery	0.264896	0.088013	(0.0924, 0.4374)	p	1.466799

2021	Adjusted R‐Squared: 0.623016		Koenker (BP) Statistic: 68.910847
Variable	Coefficient	Standard Error	95% Confidence Interval	p	VIF
Health facilities	1.21537	0.574218	(0.0899, 2.3408)	0.035271*	1.248998
Bed complement	0.224231	0.041503	(0.1429, 0.3056)	0.000000*	1.611862
4th ANC visit	0.064318	0.087884	(−0.1079, 0.2366)	0.464939	1.231886
Parity 0	0.086705	0.165695	(−0.2381, 0.4115)	0.601248	1.156073
Parity 5+	1.081658	0.735013	(−0.3590, 2.5223)	0.142388	1.487456
Maternal age	−1.070306	0.45148	(−1.9552, −0.1854)	0.018500*	1.536501
Low birth weight	0.56254	0.820899	(−1.0464, 2.1715)	0.4938	1.06635
Skilled delivery	0.521801	0.105164	(0.3157, 0.7279)	0.000002*	1.621709

^*An asterisk next to a number indicates a statistically significant p‐value (p < 0.05).

Maternal age below 21 years was a statistically significant negative predictor across all 4 years (p < 0.05), indicating that districts with a higher proportion of deliveries by young mothers paradoxically recorded lower stillbirth rates, which may reflect unmeasured confounding or differences in care‐seeking behaviour among younger mothers. Parity 5+ was a significant positive predictor in 2018 (β = 1.69, p = 0.036), consistent with the known obstetric risk associated with grand multiparity.

Skilled delivery was also a significant positive predictor in 2019 (β = 0.38, p = 0.003), 2020 (β = 0.26, p = 0.003), and 2021 (β = 0.52, p < 0.001), which likely reflects a referral bias whereby districts with higher skilled delivery coverage also receive more complicated cases, inflating stillbirth counts. The fourth ANC visit percentage was a significant positive predictor only in 2019 (β = 0.21, p = 0.022) and 2020 (β = 0.24, p = 0.013). Low birth weight, parity 0, and the showed varying levels of significance across years, while the remaining predictors were not statistically significant in most years.

The Ordinary Least Squares models for the respective years returned adjusted R² values of 0.55, 0.62, 0.64, and 0.57. The Joint Wald and Koenker statistics were both found to be statistically significant.

4. Discussion

The current discussion of the prevalence of stillbirth in Ghana, based on the widespread data obtained through the District Health Information Management System (DHIMS) provides the significant tendencies and local differences, which should be closely addressed. Findings of this research indicate that Ghana indeed strives to minimize stillbirths especially towards the realization of the Every Newborn Action Plan (ENAP) goal of below 12 stillbirths per 1000 total births by the year 2030 [20].

The data suggests that there is a significant downward trend in the incidences of stillbirth in different districts in Ghana over the years, which is encouraging since there is an attempt to improve the service delivery of maternal and child health. Research confirms the effectiveness of the policy like Free Maternal Health Care Policy (FMHCP) that was introduced in 2008 and is supposed to mitigate the financial challenges to maternal healthcare [21]. This policy is supposed to have led to the enhanced access to antenatal care (ANC) and qualified deliveries, which are the paramount in mitigating stillbirths as confirmed by the past studies [22, 23, 24]. Interestingly, our analysis has demonstrated a sharp impact of decrease in the rate of stillbirths in the Northern region, which is an indication of efforts to address high rates in the region. It is however, absolutely necessary to mention that the decline rate is not homogenous in all regions and requires subtle health policy and interventions strategies. The occurrence of heterogeneous distribution of stillbirth rates as shown through the use of choropleth maps has created an awareness that there are some districts that record high stillbirth rates than others, especially in the Greater Accra and Northern Regions [25].

The analysis of hotspots revealed that there was a strong spatial concentration of the incidence of stillbirths, and the Mion district of the Northern area was found to be a hotspot throughout the period between 2017 and 2019. Such a unified categorization of this area as a hotspot means that special health interventions are needed in this area [26]. The issues of improper use of Community‐based Health Planning and Services (CHPS), cultural values, lack of healthcare facilities, and obstacles in the way of acquiring needed drugs have been mentioned as some of the contributors to the high rates of stillbirth in these areas [27]. The strong results of the hotspots regions are consistent with the past research which indicated similar clustering patterns of poor birth outcomes in geographical areas [28, 29]. The geographical variation in stillbirth rates implies that the local health policies must be designed according to the needs and problems peculiar to various districts. Policymakers ought to use the information to focus on the high burden areas during interventions and also make sure the resources are distributed efficiently.

Ordinary Least Squares (OLS) regression has given some understanding on the correlation between different predictor variables and stillbirth rates. The count of health facilities, a percentage of women receiving at least 4 ANC visits, and the presence of skilled birth attendants are considered pivotal in the outcomes of stillbirth, showing that good access to healthcare service leads to better pregnancy outcomes [23, 24]. Nevertheless, it is important to note that OLS can be used to explain the associations, but not all the spatial autocorrelation of stillbirth occurrences can be explained. The Index of the spatial autocorrelation analysis was the Moran, which revealed that the distribution of stillbirths in some of the districts in 2017 was very concentrated implying that the districts with close health outcomes are likely to have correlated results. The same trend did not persist in the following years, which suggests a possible underlying shift in healthcare access or quality that is to be explored further [12]. In addition, outlier districts, that is Asokwa and Korle‐Klottey, provide an interesting example of how the access to healthcare facilities does not necessarily correlate with a lower rate of stillbirths. This result is consistent with earlier studies that opine that increasing the number of facilities or the number of bed complements might not be enough to reduce adverse outcomes because other systemic variables associated with healthcare delivery remain a crucial factor [30].

The findings of this extensive research, as well as hotspot locations and spatial clustering, suggest that geospatial approaches should be incorporated into public health. In this research, it is clear that there is a need to reinforce intervention in the health system of the identified hotspots regions, especially in the Northern Region. In addition, thorough continuous studies are essential to uncover the social determinants of stillbirths and model improved predictive tools that would allow tracking community‐specific factors and predicting a positive outcome of maternal and infant health [31, 32].

Finally, the demand to implement interventions and allocate resources locally can be viewed as a sign that health policymakers need to implement evidence‐based strategies based on comprehensive data research. The further development and assessment of spatial data systems will remain the key component of enhancing maternal and newborn health in Ghana.

5. Limitations

This study has several limitations that should be considered when interpreting the findings. First, the ecological nature of the analysis means that associations observed at the district level cannot be directly extrapolated to individual‐level causal relationships. As with all ecological studies, there is a risk of ecological fallacy, whereby district‐level associations may not accurately reflect individual‐level relationships between the exposures and stillbirth outcomes. Second, potential misclassification between stillbirth and early neonatal death in routine DHIMS‐2 data cannot be excluded, as the distinction between these two outcomes may not always be consistently applied across facilities and districts, particularly in settings with varying levels of clinical training and documentation practices. Third, the dataset did not include several important determinants of stillbirth, such as maternal comorbidities (e.g., hypertension, diabetes, anaemia), household socio‐economic indicators, and detailed quality‐of‐care measures. The absence of these variables may have resulted in residual confounding, and their inclusion in future studies would provide a more comprehensive understanding of the district‐level drivers of stillbirth in Ghana. Again, future studies should explore the integration of geocoded health facility location data with spatial hotspot and cold spot analyses, as this would allow for a more granular examination of the relationship between health facility accessibility and stillbirth rates across districts in Ghana.

The spatial findings of this study also have important implications for future research design. Future studies should consider the application of spatial regression models such as geographically weighted regression (GWR) or spatial lag and spatial error models to more rigorously account for spatial autocorrelation and identify locally varying relationships between determinants and stillbirth rates. Additionally, mixed‐methods studies conducted in identified hotspot districts would be particularly valuable in elucidating the local contextual drivers of persistently high stillbirth rates, including health system, community, and socio‐cultural factors that cannot be captured through routine data alone. Longitudinal individual‐level studies and the linkage of DHIMS‐2 data with household survey data such as the Ghana Demographic and Health Survey (GDHS) would further strengthen causal inference and provide a more nuanced understanding of stillbirth determinants in Ghana.

6. Conclusion

The findings of this study reveal a persistently high stillbirth burden across several districts in Ghana over the 5‐year period, with districts such as Tema, Akatsi North, and La‐Dade‐Kotopon consistently recording the highest stillbirth rates. These high‐burden districts require urgent and targeted quality improvement initiatives, including strengthening emergency obstetric care, improving skilled birth attendance, and enhancing antenatal care coverage. Additionally, robust surveillance systems should be established in these districts to monitor trends, identify underlying causes, and evaluate the impact of interventions over time. Lack of adequate number of health facilities also emerged as one of the significant risk factors associated with stillbirth occurrences. Therefore, strategically increasing the number of functional health facilities, particularly in hotspot regions, is essential to improving antenatal care coverage and timely obstetric emergency response, all of which are key to reducing stillbirth rates in Ghana. While education and training interventions for healthcare workers are necessary, they must be embedded within a broader, district‐specific programmatic framework that prioritizes the highest‐burden areas to achieve meaningful reductions in stillbirth rates in Ghana.

Author Contributions

Charllote Boateng: conceptualization, data curation; writing – original draft; writing – review and editing. Michael Arthur Ofori: formal analysis, writing – original draft, writing – review and editing. Shadrach Mintah: conceptualization, formal analysis, data curation, methodology. Emmanuel Abayie Acheampong: conceptualization, methodology, formal analysis, data curation. Brandy Bonnah Swati: writing – original draft, writing – review and editing. Isaac Duah Boateng: writing – original draft, writing – review and editing. Aliyu Mohammed: conceptualization, data curation, supervision, methodology.

Funding

The authors have nothing to report.

Disclosure

The lead (Charllote Boateng) affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Ethics Statement

The data utilizes district‐level aggregated data without a link to individuals collected in routine disease surveillance and control activities. Therefore, it was not considered to constitute human subjects research requiring institutional review board approval. However, approval was sought from the Ghana Health Services to use the data on stillbirth indicators for all 260 districts in the country.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

All data analysed during this study are with the corresponding author and will be made available on reasonable request.