Is bigger better? Assessment of self-reported and researcher-collected data on maternal health care quality among high-case-load facilities in Uttar Pradesh: a mixed-methods study

India's most populous state, Uttar Pradesh, has the country's second highest maternal mortality ratios, at 285 compared with the national maternal mortality ratio of 167. Reports of disrespect, abuse, and other types of mistreatment are also commonly reported by both the scientific community and popular media. Across nearly 750 facilities in Uttar Pradesh, the SPARQ Quality-Plus (Q+) study aims to understand which high-case-load facilities provide better maternal health care and why. Our objective was to identify whether “better” quality at facilities varied by the measures used to assess maternal health clinical quality and person-centred care quality in Uttar Pradesh. We compared self-reported government data with delivery patient survey and health provider interview data on maternal health infrastructure, service delivery, and person-centred care outcomes. The study sites were sampled based on self-reported performance data, stratified by facility type and geography. Study materials included a health service readiness checklist completed by 727 health facilities during early 2017 and caesarean section and delivery outcome data by quarter from these facilities for 2015 and 2016. Both sources are reported by the Uttar Pradesh National Health Mission (NHM). These secondary data sources were analysed to create a composite quality score used to select 20 high-performing and 20 low-performing sites from among 246 high-volume facilities (>200 deliveries/month). At these 40 sites, quality was assessed using quantitative and qualitative primary data collection with delivery patients (n=2018) and providers (n=251) and health service readiness checklists (n=40). Across all the study facilities (n=40), little correlation existed between the self-reported and researcher-collected measures of clinical quality. Yet self-reported measures do not necessarily report better levels of quality. For example, our researcher-collected data showed that facility-reported emergency obstetric care was more common than self-reported emergency obstetric care (n=29 [73%] vs n=21 [54%] in the government reported data). We found a strong negative correlation (t=–2·05; p