A cross sectional comparison of postnatal care quality in facilities participating in a maternal health voucher program versus non-voucher facilities in Kenya

Background: Health service fees constitute substantial barriers for women seeking childbirth and postnatal care. In an effort to reduce health inequities, the government of Kenya in 2006 introduced the output-based approach (OBA), or voucher programme, to increase poor women’s access to quality Safe Motherhood services including postnatal care. To help improve service quality, OBA programmes purchase services on behalf of the poor and marginalised, with provider reimbursements for verified services. Kenya’s programme accredited health facilities in three districts as well as in two informal Nairobi settlements. Methods: Postnatal care quality in voucher health facilities (n = 21) accredited in 2006 and in similar non-voucher health facilities (n = 20) are compared with cross sectional data collected in 2010. Summary scores for quality were calculated as additive sums of specific aspects of each attribute (structure, process, outcome). Measures of effect were assessed in a linear regression model accounting for clustering at facility level. Data were analysed using Stata 11.0. Results: The overall quality of postnatal care is poor in voucher and non-voucher facilities, but many facilities demonstrated ‘readiness’ for postnatal care (structural attributes: infrastructure, equipment, supplies, staffing, training) indicated by high scores (83/111), with public voucher facilities scoring higher than public non-voucher facilities. The two groups of facilities evinced no significant differences in postnatal care mean process scores: 14.2/55 in voucher facilities versus 16.4/55 in non-voucher facilities; coefficient: -1.70 (-4.9, 1.5), p = 0.294. Significantly more newborns were seen within 48 hours (83.5 % versus 72.1 %: p = 0.001) and received Bacillus Calmette-Guerin (BCG) (82.5 % versus 76.5 %: p < 0.001) at voucher facilities than at non-voucher facilities. Conclusions: Four years after facility accreditation in Kenya, scores for postnatal care quality are low in all facilities, even those with Safe Motherhood vouchers. We recommend the Kenya OBA programme review its Safe Motherhood reimbursement package and draw lessons from supply side results-based financing initiatives, to improve postnatal care quality.

This paper uses data from a larger quasi-experimental design that evaluated the impact of the Kenyan OBA voucher scheme on selected reproductive health services’ quality and access. The larger study compared voucher-accredited health facilities with non-voucher facilities in counties with similar characteristics at two points in time (in 2010 and 2012); its study methodology is described in detail elsewhere [31]. For our analysis we used cross sectional data from 2010 to compare quality of postnatal care in health facilities accredited by the voucher programme since its inception in 2006 with similar non-voucher health facilities (a “post-test design”). We compare key outcomes from quality of postnatal care with similar non-voucher facilities. This design was chosen due to the lack of random assignment of health facilities within intervention (voucher) or comparison (non-voucher) sites [31] because voucher sites were identified by MoH and KfW based on service gaps and need for increasing service coverage and availability for low income and remote populations. Facilities in targeted counties were asked to participate and were contracted if they met the accreditation criteria. Population Council, in conjunction with MoH, identified neighbouring counties with similar non-voucher facilities to maximise the likelihood of similar social, cultural, and economic characteristics in addition to similar reproductive health and healthcare behaviours among women 15 to 45 years old to serve as the comparison group [31]. Comparison facilities’ characteristics were similar to voucher facilities in their type of practice, available professional skills, clientele, locations, fees, and services including their levels (hospital, nursing home or health centre) and ownership type (public, private, faith based or non-governmental) [31]. This paper uses quantitative data from health facility assessments in 41 health facilities (20 public and 21 private-for-profit or faith-based). The 21 voucher facilities were randomly selected from 56 accredited health facilities in the three counties (Kiambu, Kisumu, Kitui) and compared with 20 facilities from three non-voucher counties (Nyandarua, Uasin Gichu, Makueni). The health facility assessments used four study instruments: All inventories included a checklist for availability of infrastructure, equipment, commodities, medicines and supplies, as well as number of staff and services provided, along with provider training for staff (n = 41). All maternal and newborn health providers working in facilities during data collection were approached for permission to interview them. Based on normal staffing levels, we expected four to eight providers eligible for interviewing in hospitals, and between two and four at health centres, for approximately 80 providers in each group and a total of 160. A total of 90 providers from voucher facilities and 73 in non-voucher facilities were interviewed. Providers were asked questions on their technical competence and time utilisation during the postnatal period. Client-provider interaction includes both a consultation’s process (how clients are treated and whether they actively participate) and content (what they are told, along with technical competence, accuracy of information and provision of essential information). Trained research assistants with clinical backgrounds observed postnatal consultations, assessing the consultation process with a standardized checklist. Subsequent sessions were observed, after patients’ informed consent, for 18 postnatal care clients in each facility (a total of 360 expected). A total of 499 postnatal care consultations in voucher facilities and 295 in non-voucher facilities were observed. Fewer clients in non-voucher facilities attended during the data collection period. Research assistants interviewed women as they exited their postnatal care consultations, focusing on the services, information, and counselling they received, as well as their fertility desires and postpartum family planning. A total of 484 exit interviews were conducted in voucher facilities, and 244 in non-voucher facilities. Ethical approval for the evaluation was granted by Population Council’s Institutional Review Board (IRB) No. 470 and Kenya Medical Research Institute (KEMRI) SCC 174. All women attending PNC services during the data collection period were asked written permission to observe their consultation and to be interviewed afterwards. Written informed consent was obtained prior to all interviews that were conducted in settings that ensured privacy and confidentiality. Participants were informed they could withdraw from the research at any time. Data collectors were trained on ethical conduct. A study definition and framework for quality of care was adapted from Donabedian and Bruce using three general elements of quality: structure, process, and outcome [32, 33]. The elements of quality of care were assessed using health facility inventories, provider interviews, client–provider interactions, and client exit interviews (Table 1). Total scores were developed for the essential components of quality care. A list of minimum equipment required for postnatal care services, for example, was developed based on MoH guidelines [34, 35] and consultations with service providers: a working blood pressure machine; stethoscope; spotlight, flashlight or examination light; examination couch; baby weight scale; adult weight scale; and autoclave or other sterilizer. Quality of Care Framework To develop an overall quality score, a composite scoring system was generated by combining several indicators into a single score [36]. There are two methods for generating composite quality scores. The “Opportunity Model” is based on the percentage of functions (“quality indicators”) actually performed compared to the total number of targeted functions [37]. If a total of 125 functions, of a targeted total of 250, were performed on 10 clients, then the composite Opportunity Model quality score would equal 0.5 (=125/ 250). Typically, equal weighting is assumed for each function, which helps derive an aggregate composite quality score covering all functions. The second scoring method employs a pre-defined criteria system, the “Grading Model”. For example, in Zambia a national assessment of the quality of antenatal care classified all health facilities into three grades, “optimum”, “adequate”, or “inadequate”, using a set of criteria for access to care, responsiveness and appropriateness, continuity of care, patient safety, and effectiveness and efficiency [38]. We have used the Opportunity Model due to its ease of interpretation, with detailed scores for each function, group of functions, and aggregate. The Opportunity Model allows for easier comparison both within categories (e.g., indicators or groups of indicators) and among groups (e.g., facilities, sub-counties), whereas the Grading Model would make differentiation of two groups with the same grade difficult. To measure the readiness of a facility to provide quality postnatal care, its attributes (infrastructure, equipment, commodities, drugs and supplies required for anyone to provide comprehensive postnatal care) are added, with equal weights, to create a structural score with a maximum of 107 points. The facility inventory study tool recorded these scores as well as each facility’s human resource elements including staffing numbers and availability of appropriate services. An additional scoring section, with a total of 32, assessed provider knowledge and training on maternal and newborn healthcare and postpartum family planning (including technical updates), with data drawn from provider interviews. The provider score was incorporated into the facility’s total score, for a potential composite score of 139 per facility. Process attributes for quality, based on national and international standards [34, 39, 40] were analysed using data from our observations of client–provider interactions during postnatal consultations. Using the methodology described facility scores, summary process attributes for technical competence with equal weights (total score of 47) were derived from the observation checklists of client-provider interactions. These observations included how a provider performed in history taking, physical examinations of mother and baby, and counselling on maternal and newborn danger signs, return to fertility and healthy timing and spacing of pregnancies, infant feeding, and HIV and STI risk assessment and management, as well as consultation documentation. Interpersonal relations (with a score between 0 and 8) including privacy, confidentiality, and rapport between client and provider were also captured (Table 2). Attributes of care: structure, process and outcome Outcome attributes focus on postnatal clients’ experiences and perceptions of their quality of care, including waiting times, perceptions of respect accorded, their understanding and knowledge of their health statuses, whether they asked questions during their consultations, opportunity to ask questions, range of services received for both mother and baby, and mechanisms encouraging follow up appointments (Table 2). Quantitative data were double entered using Epidata and exported to Stata 11 for analysis. For each component of quality of care—structure, process and outcome—summary scores were calculated as the sum of items representing specific aspects of each attribute (as defined above), and these demonstrate the overall quality score. Two-tailed unpaired t-tests with unequal variance evaluated group differences in the average process scores comparing intervention and comparison groups. A p-value of less than or equal to 0.05 was the threshold for significance. Pearson’s chi-square tests were used to evaluate differences in proportions for various patient-reported outcome measures that included waiting times to see providers, time spent with providers, and patients reporting satisfaction. To assess measures of effect, a linear regression model was used for individual and summative process quality score outcomes. We controlled for clustering at facility level, type of provider defined as public or private, and level of care (hospitals and sub-district hospitals versus dispensaries, nursing homes and clinics). In all instances we report coefficients with their 95 % confidence intervals and p values for the coefficients for all three models.

N/A

• Dima Health