This study is an ecological time-series study across 29 districts in the Greater Accra region of Ghana between 2018 and 2021. Greater Accra is the region where Ghana’s national capital, Accra, is located. The region has health facilities including teaching hospitals, regional hospitals, university hospitals, district hospitals, psychiatric hospitals, polyclinics, clinics, maternity homes, health centers, private health facilities, NGO-run facilities, Christian Health Association of Ghana facilities, and community health posts. In this study, the unit of observation is the population of children and women using health services across the 29 administrative districts. The Greater Accra region has a population of 5.5 million persons according to the 2021 population and housing census [13]. The population of Ghana has not only grown but has also experienced rapid urbanization in the past several decades, particularly in the Greater Accra and Ashanti regions. Today, more than half of the country’s population resides in urban areas [14]. According to a UN-Habitat report, the proportion of the urban population living in slum households in Ghana was 30.4% as of 2018. Using 2010 national census data and UN-Habitat’s definition of slums, the Accra Metropolitan Assembly identified 78 informal settlements and hotspots in Accra [15]. According to the World Bank, approximately 37.4% of people who live in Ghana’s urban regions live in slums [16].

The unit of analysis in this study is the district where the facilities are located, because of the difficulty of identifying catchment populations (unstable denominators) for individual facilities at the sub-district level in the Greater Accra region. In addition, there were no reliable data on population size at the sub-district level. We used routine Health Management Information System (HMIS) data on MNCH health outcomes and service utilization as clients interact with health services at health facilities or during outreach services. Routine HMIS is a comprehensive solution for the reporting and analysis needs of district health administrations and health facilities at every level. Healthcare workers at facilities collect the data and input aggregated reports via the district HMIS database in Ghana. The system receives data from government, NGOs/mission facilities, and private health facilities and has been operational since 2012. We obtained monthly data on the indicators from the district HMIS between 2018 and 2021 from the Office of Policy, Planning, Monitoring, and Evaluation Division of the Ghana Health Service and merged the data with slum information at the district level. The data on service utilization of ANC, skilled birth attendance (SBA), and PNC used were monthly aggregates of women receiving maternal health services at different health facilities. The vaccination coverage indicators used were aggregates of services provided at health facilities and through outreach programs provided by community health nurses.

Slum areas were identified by triangulating field survey, spatial, and census data. The data on slum locations across Greater Accra were derived from the synthesis of data obtained from the field in 2021. Also used were Ghana Statistical Service slum-delineated data for 2021 and a literature review of similar slum identification in Greater Accra using census, survey, and remotely sensed and GPS-located data [17, 18]. The data on the population size and population projections for districts used the national population and housing census 2010 and were obtained from the Ghana Statistical Service.

The primary service coverage measures include ANC attendance, skilled birth attendance, PNC attendance, and vaccination coverage for Bacillus Calmette-Guérin (BCG), oral polio, measles, and pentavalent 1 (Penta1) vaccine at the district level (Table 1).

An essential component of the study was identifying the number of slums in a district. UN-Habitat 2004 defines a slum household as a group of individuals living under the same roof in an urban area who lack one or more of the following: (1) durable housing of a permanent nature that protects against extreme climate conditions; (2) sufficient living space, which means not more than three people sharing a single room; (3) easy access to safe water in sufficient amounts at an affordable price; (4) access to adequate sanitation in the form of a private or public toilet shared by a reasonable number of people; and (5) security of tenure that prevents forced evictions [1].

In this study, we based the determination of slums in Greater Accra on the UN-Habitat definition of a slum. The slum areas were identified by triangulating three data sources: (1) evidence from literature based on the UN-Habitat definition in the last two decades; (2) a listing of slums in Greater Accra from the Ghana Statistical Service (GSS); and (3) a field survey. From the literature, we extracted maps of Accra slums from two published manuscripts [17, 18]. These Accra slum maps were digitized, georeferenced, and compared to establish the location of slums in Accra. The list of slum locations obtained from the GSS was geocoded and mapped, which transformed place names or addresses into spatial data. These two data sources were overlaid to be sure that the borders matched and further validated the slum map based on this overlay through site visits. That is, the research team validated the existence of slums in the locations identified in the literature and the list of slum locations obtained from the GSS through field visits.

The final judgment of what could be defined as slum locations was decided by the team based on the UN-Habitat definition, and they took into consideration that not all slums are homogeneous and not all slum dwellers experience the same degree of deprivation. The degree of deprivation depends on how many of the five conditions that define slums, as per UN-Habitat, are prevalent within a slum household. The final list of slums identified includes households that experience at least two shelter deprivations.

Because districts were used as the unit of analysis in this study, the districts in Greater Accra were categorized into slum and non-slum districts. A district was designated as containing slums if it met one or both of the following conditions: (1) sufficient intersection with the slum areas derived from the literature and UN-Habitat definition (i.e., at least one-quarter or more of the households in the district), and (2) at least one town from the GSS list of towns. A total of 22 out of 29 districts in the Greater Accra region were classified as containing slums, and the remaining seven were considered non-slum districts (Fig. 1).

Districts within the Greater Accra region with slums/no slums and type of health facilities

Descriptive summary measures such as median, 25th and 75th percentile, mean, standard deviation, and range were used to describe the service coverage measures of interest. In addition, time-series tools were used to explore the distribution of the coverage measures and identify the underlying coverage trends between the urban slum and urban non-slum districts, seasonal patterns, and outliers.

Initial data exploration showed the coverage measures were heavily skewed. Therefore, we quantified the impact of living in urban slum districts on MNCH outcomes and service utilization using the quantile regression (least-absolute-value models, median absolute deviation, and minimum L1-norm) models with a robust standard error. Furthermore, we adjusted for seasonal deviation and linear time trends. The quantile regression is a natural extension of the ordinary least square (OLS) regression model that is used when the conditions of OLS regression are not met (i.e., linearity, homoscedasticity, independence, or normality). The quantile regression model equation for the τth quantile is given as follows:

$$_\left(_\right)=_0\left(\tau \right)+_1\left(\tau \right)\textrm+_2\left(\tau \right)\textrm+_3\left(\tau \right)\left(\textrm\right)+_4\left(\tau \right)\left(\textrm\right)+_5\left(\tau \right)\textrm+_6\left(\tau \right)\textrm+_7\textrm+_8\textrm+_\left(\tau \right),i=1,\dots, n$$

where yij is the ith month observation for the jth districts. All the multivariable models adjusted for seasonality in month and year fixed effect, the impact of COVID-19 (a binary indicator indicating observations before and after the onset of COVID-19), total outpatient department (OPD) attendance, number of health facilities in the district (a proxy for access), geographical location (urban or rural), and the monthly population size of the district where appropriate.

WHO defines health equity as the absence of unfair and avoidable or remediable differences in health among population groups defined socially, economically, demographically, or geographically. In this study, inequalities in the coverage of MNCH services were measured and monitored and served as an indirect means of evaluating health inequity. We assessed the inequality of MNCH indicators between urban slum and non-slum districts using the Gini index with bootstrapped standard errors and the generalized Lorenz curve.

All statistical analyses were conducted using Stata MP version 17 (StataCorp LP, College Station, TX, USA) and a p-value less than 0.05 was considered statistically significant.