Impact of a bottom-up community engagement intervention on maternal and child health services utilization in Ghana: a cluster randomised trial

This was a cluster randomised trial in 64 primary health facilities (32 intervention and 32 control) and their catchment area, as detailed in previous related publications by the lead author [5‐8]. Primary health facilities in this context are clinics and health centres as per the Ghana Health Service (GHS) categorizations and pyramidal levels of healthcare.

Private and public health facilities categorized as clinics or health centres were considered eligible for inclusion in this study. Moreover, health facilities credentialed by the National Health Insurance Authority (NHIA) in Ghana were included in the study. Health facilities with same or similar NHIA credentialed scores were included to ensure homogeneity among the control and intervention facilities. The NHIA is an agency under the Ministry of Health (MoH) established in 2003 by Parliamentary Act (650) and amended Act (852) in 2012. The NHIA is mandated to implement the NHIS, determine scheme membership contributions, registration, issuance of membership cards and regulation of private health insurance schemes in Ghana.

This study setting and population are the same as previous publications by the lead author using the same research design and approach (see Alhassan et al. [5‐8]). The study was thus conducted in the Greater Accra and Western regions of Ghana in 16 administrative districts. Greater Accra region is a coastal region which also hosts the capital of Ghana (Accra); it is predominantly urban and cosmopolitan with a population of about 4 million people [7, 9]. Out of the nearly 3593 credentialed health facilities in Ghana in 2018, a total of 419 were in Greater Accra region [7, 9]. Western region which is also a coastal region has a population of a little over 2 million with 439 NHIS-credentialed health facilities as at 2018 [9]. In both regions over 50% of the credentialed facilities are primary level facilities (i.e. clinics and health centres).

Randomisation and sampling procedure for this study are same as previous publications by the lead author [5‐8] where the sample frame entailed list of primary health facilities credentialed by the NHIA. Primary health facilities were purposively selected for this study because they are relatively less complex in terms of health service delivery and could easily be monitored for impact of the implemented intervention [7]. Moreover, these cadres of health facilities are closer to their communities in terms of service delivery and are often the first port of call in terms of formal healthcare delivery [7].

The cluster randomization involved randomly sampling eight (8) NHIS-district offices, analogous to administrative districts, from each of the two regions same as the approach used in Alhassan et al. [7]. The focus was on districts which had NHIS membership enrolment and NHIS credentialed health facilities at the time of conducting the study. NHIS districts were selected because these districts were analogous to the administrative districts in the two regions at the time of conducting the study. Also, the study focused on all facilities accredited by the NHIA at the time of conducting this study hence, the need to consider districts which had NHIS membership enrolment and NHIS accredited health facilities [5‐8].

Next, four (4) health facilities were randomly picked without replacement as cluster sites from each district; two (2) facilities were then randomly assigned to intervention and control groups, making a total of 32 intervention and 32 control facilities (see Additional file 1). Prior to the randomization, Principal Component Analysis (PCA) was performed on the NHIA credentialed data in all the sixteen (16) districts to select the most homogeneous health facilities. The PCA scores allowed for comparability in the 64 sampled facilities (32 from each region), prior to randomization into control and intervention groups [5‐8]. A profile of health facilities involved in the cluster randomised trial is detailed in Table 1.

Detailed description of the SCE intervention has been published in previous articles by the lead author [5‐8]. Nonetheless, an overview of the SCE intervention is presented in this paper for the purposes of emphasis. The community engagement intervention was implemented in 32 primary health facilities for nearly one year (from June, 2013 to March, 2014) and evaluated over three months [7]. Baseline study was conducted in 2012 and the follow-up conducted in 2014 after the intervention implementation. The bottom-up intervention involved using existing community groups/associations to identify service delivery gaps in healthcare facilities through a systematic community engagement process (see Additional file. 2) [7]. Comprehensive reports on the community engagement intervention have been published by Alhassan et al. [5‐8].

The SCE intervention was implemented using a bottom-up approach to promote community participation in the healthcare quality improvement value chain guided by predetermined healthcare quality proxies. The first step of the engagement process involved recruitment and training of 52 facilitators, and identification of existing community groups/associations. As part of the engagement process, one facilitator was assigned to each of the 52 community groups in the two study regions, thus 26 in each region [7].

The second step of the intervention entailed evaluation of the quality of healthcare services in the intervention health facilities by community members. The assessment was done based on the clients’ most recent encounter with the service providers, in the last six (6) months [7]. Healthcare quality proxies assessed by the community members were non-technical components of service delivery such as staff attitude towards clients, staff punctuality to work, feedback from staff to clients on their health conditions, staff providing right information to clients during service delivery and ability of health facility to dispense all prescribed medications to clients. As part of the assessment process, community members were expected to express their satisfaction or disappointment with these healthcare quality proxies. The community scoring was guided by “cartoon illustrated” five-point Likert Scale score card that ranged from 1 = “Very disappointing” to 5 = “Very Satisfactory”. Assessment scores by community members are published in Alhassan et al. [8]; besides, the focus of this paper is on impact of the community engagement activities on utilization of maternal and child health services in health facilities that benefited from the implemented interventions.

The third step of the intervention implementation involved a validation of the community members assessment scores with the relevant stakeholders such as health managers, NHIS managers, clients and traditional authorities. The validation and feedback sessions were held separately in the regional capitals of the two regions. The aim was to enable service providers address identified service quality gaps and agree with clients and other stakeholders on timelines for addressing these gaps [7].

The fourth step was a follow-up on the intervention health facilities by community facilitators also called “community quality care champions” to ensure action plans were implemented by health managers in the intervention health facilities as agreed during the validation and feedback sessions. The follow-up was done by the “community quality care champions” approximately three (3) months after the third implementation step [7].

The final step involved recognition of health facilities which were perceived by community members to have improved in the quality of their services to clients after the series of engagements with community members. The community members adjudged the best health facilities themselves to promote transparency, ownership and accountability of health providers to clients. Reward for best health facilities in service quality was an inscription of recognition displayed at the outpatient department (OPD) of the facility in addition to a cash amount of approximately US$ 280 given in Ghana Cedis (GHC) equivalence. The aim of this recognition was to stimulate health competition among peer health facilities to render client-centered quality services. See details of all the intervention implementation steps in Additional file 3 and in Alhassan et al. [7].

Data was collected using a tool called Situational Analysis Plus (SA⁺) which has four main sections namely: Facility Information, Access to Care, Clinic Activities/Services, and Personnel/Vacancies. SA⁺ was developed by the PharmAccess Foundation and SafeCare Initiative in the Netherlands and has been tested in a number of African countries including Ghana. Details of the SA⁺ tool can be found in Alhassan et al. [5].

Only data from primary health facilities contacted at baseline and follow-up was used for the final analysis in line with the protocol for evaluating effectiveness of interventions [10]. STATA statistical software version 12.0 (StataCorp, College Station. Texas USA) was used for all analysis. Multivariate multiple regression analysis and paired t-test were used to determine true impact of the community engagement intervention on the key outcome variables of interest after testing for multicollinearity and controlling effects of relevant covariates.

Main outcome variables were: number of spontaneous vaginal deliveries; number of child immunizations conducted; number of female condoms distributed; number of HIV tests for non-pregnant women, number of HIV tests for pregnant women, and number of malarial tests. These outcome variables were recorded on “per month per health facility” basis.

Baseline study was conducted between March and June, 2012 while the follow-up data was collected between July and October, 2014 after the intervention implementation which lasted for nearly a year (i.e. June, 2013 to March, 2014). At baseline, a total of 64 primary health facilities participated in the study. This number included 30 private-for-profit, 26 government owned and 8 faith-based facilities. At follow-up the number of facilities reduced marginally to 63 due to attrition comprising of 30 private-for-profit, 25 government-owned and 8 faith-based facilities (see Table 1).

Results from the baseline data showed that in 2012, the average number of clients accessing outpatient and inpatient services were 1011(SD = 787) compared to an average of 1317 (SD = 157) patients in 2014 at follow-up. The number of clients accessing inpatient and outpatient services increased in the intervention facilities from an average of 952 (SD = 692) at baseline (2012) to 1453 (SD = 1329) at follow-up (see Table 2).

The results further revealed an overall reduction in bed occupancy rate in intervention and control facilities over time. Intervention facilities at baseline recorded an average of 34% (SD = 28) per month per health facility while control facilities recorded 33% (SD = 27) per month per health facility. At follow-up, intervention facilities recorded an average bed occupancy rate of 32% (SD = 27) per month per health facility while control facilities recorded 25% (SD = 22) per month per health facility. Average length of inpatient stay at baseline was approximately 2 days (SD = 2.0) compared to 1 day (SD = 1.7) at follow-up in intervention and control facilities (see Table 2).

There were contextual factors existing in the intervention and control facilities which were independent of the intervention. These were typically material and human resources of the facilities. Overtime, there was an increase in the average number of staff per health facility. During the follow-up survey, some improvement was observed in the number of medical assistants, pharmacists, laboratory technologists, and registered nurses working in both intervention and control facilities. Even though the number of other professional categories also increased, the differences between the baseline and follow-up figures were not statistically significant (see Fig. 1). These were contextual factors which were outside the sphere of control of the study and the intervention might not have influenced their outcome.

In terms of the material resources of intervention and control facilities over time, the average number of wards in a health facility increased marginally from an average of 1.8 at baseline to an average of 2.4 (p < 0.001) at follow-up. Likewise, average number of minor surgical operating theatres per clinic increased marginally from 0.14 at baseline to 1.05 at follow-up (p < 0.001) (see Fig. 1). As stated earlier, these variations might not have been necessarily influenced by the intervention since they were outside the sphere of control of this study.

Other improvements observed in intervention facilities at follow-up were the number of female condoms distribution. Among the 5 intervention facilities that rendered female condom services, no female condom was distributed at baseline but increased to an average of 2 (SD = 2) per facility per month at follow-up. Among control facilities that distributed female condoms, the average number distributed per facility per month increased from 0 at baseline to 6 (SD = 10) at follow-up.

Among the 24 intervention health facilities that run tests for HIV, an average of 55 (SD = 65) tests were conducted for non-pregnant clients per facility per month at baseline compared to 104 (SD = 81) tests per facility per month at follow-up, p = 0.0213. Even though in control facilities there was marginal increase in HIV tests for non-pregnant women, it was not significant. In the case of HIV tests for pregnant women, an average of 40 (SD = 43) tests per facility per month was recorded in intervention facilities at baseline but increased to 119 (SD = 128) per facility per month at follow-up (p = 0.0067). In control facilities, although some increases were recorded, they were marginal and statistically insignificant (see Table 3).

Records of five (5) intervention facilities that run tests for malaria showed that the average number of tests conducted per facility per month was 43 (SD = 76) at baseline but increased to 380 (SD = 236) at follow-up, p = 0.0174. Among the control facilities, the average number of malaria tests conducted per facility per month was 0 (SD = 0) at baseline but increased to 294 (SD = 219) at follow-up, p = 0.0120.

Furthermore, statistical test for predictive power of the intervention on the key outcome variables of interest showed that within the baseline and follow-up sub-samples, the community engagement intervention impacted positively on child immunizations (Coef. = 519.5, 95%CI = [− 599.5 to 1638.5]) and condom distribution (Coef. = 72.1, 95%CI = [− 143.6 to 287.1]). Other significant predictors of maternal and child health service utilization were facility ownership, region of clinic location, distance to nearest referral health facility and income level of clients. For instance, private health facilities were less likely to conduct spontaneous vaginal deliveries (SVDs) (Coef. = − 7.31, 95%CI = [− 15.70 to 1.08], p < 0.05), child immunizations (Coef = − 167.6, 95%CI = [− 313.8 to − 21.39], p < 0.01) and distribute condoms (Coef. = − 182.5, 95%CI = [− 279.2 to − 85.7], p < 0.05) (see Table 4).

Health facilities located in the Greater Accra region were less likely to conduct SVDs relative to health facilities located in the Western region (Coef. = − 10.36, 95%CI = [− 18.80 to − 1.93], p < 0.01), and likewise child immunizations (Coef. = − 231.18, 95%CI = [− 378.2 to − 84.19], p < 0.01) (see Table 4). Health facilities which were nearer to a referral hospital within ≤60 min travel time were more likely to record SVDs than facilities which were farther from a referral hospital of > 60 min of travel time (Coef. = 0.18, 95%CI = [− 0.00 to 0.37], p < 0.05) (see Table 4).

Apart from the core maternal and child health outcome indicators other general health service components that recorded significant improvement in intervention and control facilities over time were: number of general laboratory tests, male child circumcisions and minor surgical procedures per facility per month (see Table 4). Control facilities recorded more improvements than intervention facilities in terms of general laboratory tests, voluntary counselling and testing (VCT) for patients, number of patients treated with sexually transmitted infections (STIs), male child circumcisions, and other minor surgical procedures (see Table 4), suggesting the intervention perhaps had lesser impact on these service components. The researchers however, acknowledge extraneous factors other than the community engagement intervention which might have influenced the outcome of these indicators in the intervention and control facilities.