Is quality of care a key predictor of perinatal health care utilization and patient satisfaction in Malawi?

Background: The Malawi government encourages early antenatal care, delivery in health facilities, and timely postnatal care. Efforts to sustain or increase current levels of perinatal service utilization may not achieve desired gains if the quality of care provided is neglected. This study examined predictors of perinatal service utilization and patients’ satisfaction with these services with a focus on quality of care. Methods: We used baseline, two-stage cluster sampling household survey data collected between November and December, 2012 before implementation of CARE’s Community Score Card© intervention in Ntcheu district, Malawi. Women with a birth during the last year (N = 1301) were asked about seeking: 1) family planning, 2) antenatal, 3) delivery, and 4) postnatal care; the quality of care received; and their overall satisfaction with the care received. Specific quality of care items were assessed for each type of service, and up to five such items per type of service were used in analyses. Separate logistic regression models were fitted to examine predictors of family planning, antenatal, delivery, and postnatal service utilization and of complete satisfaction with each of these services; all models were adjusted for women’s socio-demographic characteristics, perceptions of the closest facility to their homes, service use indicators, and quality of care items. Results: We found higher levels of perinatal service use than previously documented in Malawi (baseline antenatal care 99.4%; skilled birth attendance 97.3%; postnatal care 77.5%; current family planning use 52.8%). Almost 73% of quality of perinatal care items assessed were favorably reported by > 90% of women. Women reported high overall satisfaction (≥85%) with all types of services examined, higher for antenatal and postnatal care than for family planning and delivery care. We found significant associations between perceived and actual quality of care and both women’s use and satisfaction with the perinatal health services received. Conclusions: Quality of care is a key predictor of perinatal health service utilization and complete patient satisfaction with such services in Malawi. The current heightened attention toward perinatal health services and outcomes should be coupled with efforts to improve the actual quality of care offered to women in this country. Ntcheu district is centrally located in Malawi along the border with Mozambique. Between January 2012 and December 2015, CARE implemented a Community Score Card© intervention aimed at improving utilization and quality of perinatal health services in this district [22], and used a cluster randomized design to evaluate the intervention. Both the intervention and the evaluation design are described in detail elsewhere [22]. The evaluation involved cross-sectional baseline and endline surveys of women aged 15–49 years who have given birth within the last 12 months and whose babies were alive. This analysis uses baseline data collected from 1301 women (response rate = 98%) before implementation of the intervention between November-December, 2012. The survey took 40–60 min to complete and was conducted in Chichewa in a private area of the house; all data were self-reported. Verbal informed consent was obtained from all study participants (Additional file 1). We assessed the following population (women 15–49 years) characteristics: age, parity (1, 2, 3–4, 5+ children born), religion (Catholic, Presbyterian, other Christian, and other), Ngoni ethnicity (yes/no), marital status (married/in union or unmarried/divorced/widowed), education (completed years), reading level (cannot read simple sentence, reads part of a simple sentence, or reads the entire sentence), and household wealth (quintiles of a household wealth index constructed using principal component analysis of household item possession). Women’s perceptions of the quality of care offered in facilities closest to their residence were also examined given literature-informed expectations that: 1) if interested in obtaining perinatal health services, including family planning, women would most likely seek care at the most easily accessible facility, and 2) if the closest health facility is not perceived as providing high quality services, women will likely not seek services there and will have an overall lower likelihood of seeking services elsewhere. Specifically, we assessed perceptions regarding the cleanliness of the facility, whether the staff ensures patients’ privacy, whether one or more providers are always available at the facility, whether high quality services are offered, and whether unmarried women can access family planning and other reproductive health services there as a proxy for service accessibility. Assesment used 5-point Likert scales (strongly agree, agree, neutral, disagree, strongly disagree) and responses were dichotomized (agree or strongly agree vs. any other response) for use in analyses. In addition, we captured the time needed to reach this closest health facility using readily available means of transportation (<30 min, 30–59 min, 1–2 h, >2 h). The perinatal health services of interest were family planning (ever and current use), antenatal care (use during the last pregnancy, pregnancy trimester of initiation during the last pregnancy, and number of visits obtained during the last pregnancy), delivery care (use at last delivery), and postnatal care (use after last delivery and number of maternal or neonatal checks within 2 months postpartum). Survey questions to assess quality of care were developed separately for each type of health services (family planning, antenatal care, delivery care, and postnatal care) using the Hulton framework on quality of maternity care as guide (Table 3) [4]. We created binary variables for each actual and perceived quality of care items in our assessment – yes, if the quality aspect was reported as met; no, if it was not. We used the distribution of these variables to select up to five quality of care items with the highest response variability (i.e. <95% favorable responses) for each outcome; there was one exception to these criteria for the family planning outcome -- we used one item with 96.7% favorable responses because it represented a very distinct quality of care domain than the other four quality of care items chosen (i.e. whether or not the provider scheduled a follow-up visit). For family planning, of 13 quality of care items assessed, the five items chosen were: patient’s contentment with the specific method chosen (patient got the method she wanted – yes/no) and four elements of actual quality of family planning service provision used to construct an index of family planning quality provision ranging from 0 (none of the 4 quality aspects were reported) to 4 (all 4 quality aspects were reported). For antenatal care, we captured whether women received antenatal care from a skilled provider and used one of five quality of care items assessed (discussion about pregnancy danger signs). For delivery care, we recorded whether the woman had a skilled birth attendant or not, the time (in hours) before the first consult upon arrival at the facility, and used five of 12 related quality of care items to construct an index of L&D management quality; the index ranged from 0 (none of the 5 quality aspects were reported) to 5 (all 5 quality aspects were reported). For postnatal care, we captured whether women received care at the first postnatal check from a skilled provider and used one of the three quality of care items assessed for this outcome (discussion about postpartum danger signs). Quality of care items used in our assessment of maternal and neonatal health services Note: Bolded items were chosen for inclusion in regression analyses. aItems were reverse coded; “no” responses are reported here Patient satisfaction with each type of service when last received was ascertained using a common question with response options measured on a 5-point Likert scale: completely unsatisfied, unsatisfied, neutral, satisfied, and completely satisfied. Given the distribution of responses, we created and used in analyses a dichotomous patient satisfaction variable (completely satisfied vs any other response) for each type of perinatal health services. Faced with the possibility of social desirability bias altering patient satisfaction responses, we attempted to validate these responses by also asking women how likely they would be to recommend the services they received to others. Responses to these questions were also measured on a 5-point Likert scale: very likely, likely, neutral, unlikely, very unlikely. We examined correlation coefficients between reported patient satisfaction and likelihood of recommending the same services to others for each type of service. To examine predictors of perinatal health service utilization, our first study objective, we fitted logistic regression models for both ever and current use of family planning, initiation of antenatal care in 1st trimester, delivery in a health facility, and receipt of at least one postnatal check within 2 months of delivery. All models were adjusted for the socio-demographic characteristics described above with two exceptions: age and parity were highly correlated and only parity was included in all regression models; education attainment (measured in single years) and reading level were also highly correlated, and we included only reading level in all regression models. Parity was chosen over age given our interest in quality of care and women’s satisfaction with the care received -- at the population-level, we consider that parity is a better control variable for the level of maternity care that a woman needs and expects to receive. Women’s proven reading capability was preferred over education level because it is a more objective measure of a woman’s ability to obtain information on her own. All models were also adjusted for women’s perceptions of the quality of care at the closest health facility to their homes and for the time needed to reach this facility. For the second study objective, we restricted analyses to women who have ever used family planning, used antenatal, delivery, and postnatal care during their last pregnancy, respectively. We fitted 4 logistic regression models for complete satisfaction with each of these types of services when last received. Models were adjusted for the same socio-demographic and closest health facility characteristics as well as additional factors that are theoretically expected to predict service satisfaction. Specifically, the regression model of satisfaction with family planning services was also adjusted for the ownership of the facility where services were last received (Government, private/mission-based, other), for whether the woman got the method she wanted, and the index of family planning quality provision. The model fitted for satisfaction with antenatal care services was also adjusted for the trimester when such care was initiated and the total number of visits received; for having received antenatal care from a skilled provider; and for having discussed pregnancy danger signs during antenatal care. Satisfaction with delivery care was additionally modeled on the ownership of the facility where the woman delivered (Government, private/mission-based, other), having had a skilled provider, the time to first consult before delivery, and the index of L&D management quality. The model of satisfaction with postnatal care was also adjusted for having a skilled provider, the timing of the first postnatal check given that the majority of women only had one checkup within 2 months postpartum, and having discussed postnatal danger signs with the provider. All analyses were conducted in Stata version 13 and were adjusted for the complex survey design using Taylor’s linearization method. The research protocol was reviewed and approved by Malawi’s National Health Science Research Committee.