Effects of combined decision-support and performance-based incentives on reported client satisfaction with maternal health services in primary facilities: A quasi-experimental study in the Upper East Region of Ghana

Background Computerized decision-support systems (CDSS) and performance-based incentives (PBIs) have potential to contribute to client satisfaction with health services. However, rigorous evidence is lacking on the effectiveness of these strategies in lower-income countries such as Ghana. This study aimed to determine the effect of a combined CDSS-PBI intervention on client satisfaction with maternal health services in primary facilities in the Upper East Region of Ghana. Methods We employed a quasi-experimental controlled baseline and endline design to assess the effect of the combined interventions on client satisfaction with maternal health services, measured by quantitative pre/post-test client satisfaction survey. Our analysis used difference-in-difference logistic regression, controlling for potential covariates, to compare variables across intervention and comparison facilities at baseline and endline. Results The combined CDSS-PBI intervention was associated with increased or unchanged client satisfaction with all maternal health services compared at endline. Antenatal client difference-in-difference of mean satisfaction scores were significant at endline for intervention (n = 378) and comparison (n = 362) healthcare facilities for overall satisfaction (DiD 0.058, p = 0.014), perception of providers’ technical performance (DiD = 0.142; p = 0.006), client-provider interaction (DiD = 0.152; p = 0.001), and provider availability (DiD = 0.173; p = 0.001). Delivery client difference-in-difference of satisfaction scores were significant at endline for intervention (n = 318) and comparison (n = 240) healthcare facilities for overall satisfaction with delivery services (DiD = 0.072; p = 0.02) and client-provider interaction (DiD = 0.146; p = 0.02). However, mean overall satisfaction actually reduced slightly in intervention facilities, while DiD for technical performance and provider availability were not significant. Conclusion This combined CDSS-PBI intervention was associated with greater antenatal and delivery client satisfaction with some aspects of maternity services within two years of implementation. It could be expanded elsewhere if funds allow, though further research is still required to assess cost-effectiveness and long-term effects on client satisfaction and maternal health outcomes. The study was conducted in twelve health facilities in Kassena-Nankana (KND) and Builsa districts in the Upper East Region of Ghana. Maternal and child health services are considered particularly poor in the Upper East Region of Ghana, due to insufficient and poorly motivated staff, inadequate use of computers, poor access to reproductive health guidance, staff non-adherence to protocols, poor diagnosis, and inadequate referrals [4, 63–65]. Institutional MMR in the Upper East Region in 2010 was 352 per 100,000 live births. In Kassena-Nankana and Builsa districts in 2010, MMR was estimated at 367 and 259 per 100,000 live births respectively [66]. A cross-sectional study of 400 new mothers indicated 93% had delivered in a health facility and 97% had received antenatal care with 75% having four or more ANC visits [67]. An estimated 207 and 100 health-workers serve health facilities in KND and Builsa Districts, respectively. Nurses and midwives are responsible for the provision of maternal and neonatal health services in health facilities. In these districts in the Upper East Region of Ghana, inadequate health personnel, heavy workload and poor motivation (36, 52) challenge health facilities in the provision of care. KND, the intervention district, covers a total approximate area of 1,675 square kilometres with an estimated population of 152,000 [68]. It is served by one hospital, located in the district capital Navrongo, six health centres, one private clinic, and twenty-seven Community-based Health Planning and Services (CHPS) compounds. Builsa, the comparison district, covers a 2,220 square-kilometre area southwest of KND with an estimated population of 95,800. It is served by the district hospital, located in the district capital Sandema and serving as referral centre for all district health facilities. Additionally, Builsa has six main health centres and thirteen CHPS compounds. A quasi-experimental intervention study was conducted, with data collection at baseline and endline in 6 intervention and 6 comparison primary healthcare facilities in KND and Builsa districts respectively. This design was chosen because it can help establish causal impact of an intervention on a target population without random assignment [69]. As part of the European Union funded Quality of maternal and neonatal health (QUALMAT) study in KND, CDSS and PBIs were implemented in six intervention facilities for two years from April 2012 to April 2014 [32, 36, 40]. The CDSS component consisted of computerized guidance and clinical support for antenatal and delivery care up to 24 hours after delivery. To improve maternal health services [64, 70], 35 purposively-selected maternal healthcare providers (i.e. midwives, community health nurses, health facility managers) completed six trainings on use of computer software to guide implementation of WHO ‘Pregnancy, Childbirth, Postpartum and Newborn Care: a guide for essential practice’ guidelines [71–73]. One laptop, with the CDSS installed and IT support included, was given per intervention facility for three trained midwives and nurses to share during the two-year study period [32, 36, 71]. The rest of the trained health staff were mangers tasked to supervise the use of the CDSS at the facility level. The PBI component, rewarding best-performing midwives and facilities with domestic appliances (e.g. fans, stoves, blenders, freezers, fridges, television sets, blenders, saucepans, cloths, tea kettles, microwaves) and certificates of recognition [40], was implemented from July 2012 to March 2014 to improve CDSS users’ morale. Key performance indicators were defined through meetings between healthcare providers and regional and district health directorates. Performance indicators, focusing on process and outcomes, included proportion of ANC visits recorded in CDSS; proportion of pregnant women who received tetanus vaccination; proportion of pregnant women who received iron supplementation; proportion of pregnant women who received counselling for safe sex; proportion of pregnant women who received counselling for HIV; proportion of births attended by skilled personnel; proportion of partographs completed in CDSS; partograph usage rate; proportion of women referred based on CDSS recommendation; and newborn immunization coverage. Three awards ceremonies (held December 2012, September 2013, and February 2014) additionally acknowledged best-performing midwives and facilities. additional ongoing incentives included regular supervision, verbal appreciation, and provision of furniture and small monthly allowances [40, 64, 74]. Binary outcomes were client satisfaction (yes/no) with overall care and specifically with provider technical performance, interaction, and availability. Sample sizes were calculated to detect a difference of 10% in satisfaction between intervention and comparison arms and between pre and post intervention time-periods, with a two-sided test at a significance level of 5% and a power of 80%, for quality-of-care assuming a baseline of 60% and independence between interviews. This gave us a minimum total sample of 752 women, 376 in intervention facilities and 376 in comparison facilities for both antenatal and delivery objectives. Thus, interviewers aimed to include approximately 63 women per facility who attended an antenatal consultation and 63 per facility who recently gave birth. As we did not have the resources to follow client cohorts prospectively, all ANC and delivery clients visiting study facilities during baseline and endline data collection periods were eligible for inclusion. Quantitative data were collected by client exit-survey interviews lasting approximately 15 minutes. Questions included socio-demographic characteristics, technical performance (i.e. reception of client, compassion, respect shown to client, adequacy of drugs and diagnosis), client-provider interaction, and provider availability–common components of client satisfaction used to assess health staff performance. The questionnaires for antenatal and delivery clients (S1 File) were developed based on similar tools in the literature [75–78] and expert advice, and pretested with 30 ANC clients at the War Memorial Hospital and 30 delivery clients at Sandema Hospital. After tool finalisation, baseline satisfaction survey data were collected in 2010 and endline in 2014. Twelve research assistants were trained at the Navrongo Health Research Center for three days to conduct exit interviews with pregnant and postpartum women leaving health facilities after receiving maternal healthcare from providers. Data collectors attended facilities daily to conduct as many interviews as possible with antenatal and delivery clients. Most interviews in KND were conducted in Kassem and Nankam languages and in Builsa language in Builsa District. Written informed consent was obtained from all women before participation. Data were entered in Microsoft Excel and transferred to Stata version 13 for analysis. Descriptive statistics were summarised, including age, weeks of gestation, parity, number of antenatal visits, educational level, and type of delivery. Antenatal and delivery care satisfaction response categories were recoded as binary (1 for satisfied, 0 for unsatisfied) from the original five categories, to reduce irrelevant detail and increase cell sizes. Exploratory Factor Analysis was used for data reduction. Groups of variables were defined and mean scores calculated for all selected facilities in intervention and comparison areas, per variable and group (i.e. intervention or comparison). Principal Component Analysis was used to extract key factors, followed by a Varimax rotation with Kaiser normalisation. The number of factors retained was based on eigenvalues ≥1. Statistical analysis using difference in difference logistic regression has the potential to account for some initial differences that might have existed between the intervention and comparison health facilities. Controlling for some covariates was also appropriate. Difference-in-Difference (DiD) analysis was performed, using a logit model and controlling for a priori confounders, to compare changes in outcomes between intervention and comparison groups over time (i.e. DiD results are based on differences in means between intervention and comparison groups at baseline compared to endline for each of the four outcome scores measured). Linear regression was used to estimate means, with p<0.05 used as the threshold for statistical significance. Potential confounders adjusted for in analyses as continuous variables were age, gravidity, gestational weeks, and number of ANC visits, while education level was categorical (i.e. never attended, attended primary, attended secondary, attended tertiary). Results were reported using significant digits (82). Ethical approval was obtained from the Institutional Review Board of the Navrongo Health Research Centre, Ghana (reference NHRCIRB116). Written informed consent was obtained from all women before enrolment in the study. To guarantee confidentiality, study tools and outcomes did not include patient identifiers.