Menu
Menu
Menu
Menu
Background: The Community-based Health Planning and Services (CHPS) initiative is a major government policy to improve maternal and child health and accelerate progress in the reduction of maternal mortality in Ghana. However, strategic intelligence on the impact of the initiative is lacking, given the persistant problems of patchy geographical access to care for rural women. This study investigates the impact of proximity to CHPS on facilitating uptake of skilled birth care in rural areas. Methods and Findings: Data from the 2003 and 2008 Demographic and Health Survey, on 4,349 births from 463 rural communities were linked to georeferenced data on health facilities, CHPS and topographic data on national road-networks. Distance to nearest health facility and CHPS was computed using the closest facility functionality in ArcGIS 10.1. Multilevel logistic regression was used to examine the effect of proximity to health facilities and CHPS on use of skilled care at birth, adjusting for relevant predictors and clustering within communities. The results show that a substantial proportion of births continue to occur in communities more than 8 km from both health facilities and CHPS. Increases in uptake of skilled birth care are more pronounced where both health facilities and CHPS compounds are within 8 km, but not in communities within 8 km of CHPS but lack access to health facilities. Where both health facilities and CHPS are within 8 km, the odds of skilled birth care is 16% higher than where there is only a health facility within 8km. Conclusion: Where CHPS compounds are set up near health facilities, there is improved access to care, demonstrating the facilitatory role of CHPS in stimulating access to better care at birth, in areas where health facilities are accessible.
The data for the analysis come from the two most recent rounds (2003 and 2008) of the Ghana Demographic and Health Surveys (GDHS), a geo-referenced database of health facilities and digitised topographic database of national road networks from a national programme of land surveillance. The land surveillance was conducted by the Water Research Institute, the Centre for Remote Sensing and Geographic Information Services (CERSGIS) of the University of Ghana, Department of Feeder Roads, Ghana Survey Department and the Forestry Commission of Ghana between 1995 and 2005. The outcome variable of interest and controls were derived from the GDHS, whilst the main predictor of interest (distance to CHPS and health facilities) were derived from the geo-referenced database of health facilities and national road networks. Description of the datasets, variables and calibration of distances are discussed below. The 2003 and 2008 GDHS are nationally representative cross-sectional surveys of women and men aged 15–49 years and 15–59 years, respectively. The surveys collected demographic and health information on women, men, children and other members of households. Information on where a woman gave birth and who attended the birth was collected for all births five years preceding each survey. The births recorded in the two surveys cover the period July 1999 to October 2008, which coincide with the launch of the CHPS initiative in 1999. Prior to 2005, CHPS was a rural policy [11, 31]. In 2005, the Government of Ghana, the Ghana Ministry of Health and Ghana Health Service adopted CHPS as national policy with the aim of developing urban health systems in marginalised urban communities [14, 31]. However, due to the slow implementation of CHPS in urban areas, only two pilot urban CHPS compounds (in Tema and Glefe, both in the Greater Accra region) supported by the CHPS Technical Assistance Project of the USAID were functional [32] within the study period. The analysis is, therefore, restricted to rural communities. The outcome variable of interest focuses on the proficiency in the skills of the attendant at birth. Skilled attendant is used to refer exclusively to people with midwifery skills (e.g. doctors, midwives, nurses), trained in the skills necessary to manage normal deliveries, diagnosis, management of complications and referrals [33]. The data cover 4349 births. The response variable was binary coded 1 if a birth was attended by a skilled professional and 0 otherwise. To ensure minimal recall bias, we examined the consistency between reported place of birth and type of birth attendant. Non-institutional births reported to have been attended by skilled professionals constitute less than 0.05% and were excluded from the analysis. The choice of control variables was based on literature and data availability. The selected control variables were maternal age and education, ethnicity, religion, parity, number of antenatal visits, partner’s educational status, household wealth status, and region of residence. A national georeferenced database of health facilities providing care at birth and digital topographic data on road networks were used as input to a network analysis algorithm to calculate the distance from each PSU in the GDHS to the nearest health facility and also CHPS referal point, often referred to as a CHPS compound. The calibration of network distance to nearest health facility and CHPS coumpound is discussed in the subsequent section. The list of health facilities was compiled from three sources: a list of 2021 health facilities obtained from the Ghana Ministry of Health, a web resource of health facilities maintained by the Ghana Mininistry of Health and a georeferenced list of 1915 health facilities compiled by CERSGIS, University of Ghana. The lists were cross-checked and reconciled. Facilities without latitude and longitude values were georeferenced manually by matching with facility and town names on Google Earth. In unresolved cases, we contacted district health offices to confirm locations. Facilities such as psychiatric hospitals, supplementary feeding centres, nursing training colleges and administrative offices which do not offer maternity services were excluded. Where there was more than one facility at a single location (either because they shared the same building or because they were georeferenced using a village location) the highest order facility was retained for subsequent analysis, thus, avoiding duplications. In all there were 1688 georeferenced conventional health facilities throughout the country. The EmONC audit was used to classify the EmONC status of the facilities based on the signal-functions they offer [7, 34]. Facilities offering all nine signal functions including availability of blood transfusion and surgical/caesarean section capability (76 in all) were classified as comprehensive-EmONC facilities [7]. Facilities offering between six and eight signal functions (81 in all) were classified as partial-EmONC. Since uptake of maternity care in Ghana is often between partially functioning EmONC which are able to respond appropriately to a range of birth complications and non-EmONC [8], we classfied the facilities into two groups—EmONC facilities (6 or more signal functions) and non-EmONC (less than six signal functions). CHPS covers a set defined catchment areas referred to as CHPS zones. The recommended population covered by a CHPS zone is 3000 to 4500 [14]. A CHPS zone may cover one or more villages or communities. Some CHPS zones have a physical structure purposely built or designated for providing CHPS services, referred to as CHPS compounds and staffed by Community Health Officers. The CHPS compound is located in one of the communities within the zone. In such situations, the CHPS compound serves as the referral point. In CHPS zones where there is no physical structure designated for CHPS activities, the location or residence of the Community Health Officer in charge of the CHPS zone is used as the referral point. The GPS coordinates collated by the Ghana Ministry of Health references the location of the CHPS compounds (where there is a physical structure) and the location of the Community Health Officer (where there is no physical structure). In all, there were 458 CHPS compounds, functional in September 2008. The road network data were modified to incorporate ferry routes connecting settlements on the shores of the Volta lake. The distance between the location of each PSU and the nearest health facility and CHPS referral point was calculated using the ‘closest facility’ functionality in ESRI ArcGIS Network Analyst software. For the routing algorithm to function correctly, the national road network dataset had to be topologically cleaned, ensuring that the segments of the network were connected appropriately prior to undertaking the analysis. The topological checking and cleaning was also carried out in the ESRI ArcGIS software. All distance calculations were carried out using the Ghana Metre Grid projected coordinate system (EPSG code 25000). For the road network analysis, all georeferenced facilities were included irrespective of the rural-urban location, since the closest facility for some rural communities was in an urban location. Since many health facilities and CHPS compounds were not located exactly on the nearest road segment, all community locations and health facilities were assigned to the closest location on the road network measured as a straight line distance, where that distance was less than 5 km, as shown in the sample caption of Fig. 1, labels A and B. The reported distance was a sum of the straight line distances between the community and nearest road segment, the along road distance, and the straight line distance between the road and the facility. Twenty-six of the 463 communities and 36 of the 1688 health facilities were found to be more than 5 km from the nearest road. In this case, the straight line distance to the nearest facility was used (Fig. 1, label C). In this study, topographic obstacles were not taken into account in the calibration of the distances. For each community, the closest distance to health facility and CHPS referral point was calculated. Based on the Ghana Health Service policy framework which defines access as being within 8 km of a facility [35, 36], all communities were classified into those within 8 km of—(a) health facility only, (b) CHPS only, (c) both a health facility and CHPS and (d) outside 8 km of both a health facility and CHPS. Thus, in this study access to a health facility or CHPS is defined as being within 8 km of the referral point as stipulated by the Ghana Health Sevice policy framework. For CHPS, the Ghana Ministry of Health provided the month and year they became functional. Information on when conventional health facilities became functional were not available. Relating the month and year of CHPS becoming functional in the communities to the date of birth of the children in the survey which was available from the GDHS, births in (b) and (c) were futher disaggregated into those that occurred before and after CHPS become functional. Operationalisation of access to health facilities and CHPS is illustrated in Fig. 2. The distribution of uptake of skilled care at birth by access to CHPS compounds and health facilities and wealth status were examined through bivariate analysis. Chi-squared tests were used to investigate significant differences. Multilevel logistic regression techniques were used to examine the adjusted effect of access to CHPS and health facilities on the uptake of skilled care at birth, accounting for potential confounders and clustering of the data. Two-level binary logistic regression models were used with children nested within communities (PSUs) to model uptake of skilled delivery care allowing for between-community homogeneity to be examined. There were 4349 children (level 1) nested in 463 communities (level 2). We conducted additional analysis, separately for communities where CHPS have become functional within 8 km but access to health facilities remain limited (n = 511, PSUs = 51) and also where CHPS has supplemented health facilities within 8 km (n = 502, PSUs = 57) to further examine the before and after impact of CHPS on skilled maternity care uptake in those communities where CHPS has become functional. The longitudes and latitudes of the primary sampling units were mapped using the ArcGIS software to identify duplicated communities. A sequential model-building process was used to examine the extent to which access to CHPS, health facilities and the intermediate factors explained uptake of skilled care at birth between communities. The rationale for adopting a sequential model-building process was to investigate how the association between distance to a facility and uptake of skilled birth care changed when other important intermediate factors were included in the model. Model 1 controlled for the between-community differences to account for the hierarchical structure of the data and year of survey to account for the survey effect. Model 2 added the primary variable (access to health facility and CHPS). Model 3 further added the intermediate factors and Model 4 included the region of residence. At each stage of the model-building process variables that were not significant at p<0.05 were discarded. The significance of these variables was further tested in the final model. The models were fitted using MLwiN 2.26 [37]. The Penalized Quasi-Likelihood (PQL) estimation procedure with a second-order Taylor series approximation was used to estimate the model parameters [38–40].
N/A
