Building a model of Primary Health Care to investigate how PHC systems operate in Brazil and to discuss their impact on child mortality: a Bayesian Confirmatory Factor Analysis approach

Background Prioritizing Primary Health Care (PHC) health systems can potentially improve population health markers, such as maternal and child health. However, in Brazil, the quality of services is substantially variable. We developed and validated a model that could numerically measure PHC quality, and its impact on under-5 mortality. Methods The proposed theoretical model conceives PHC as a complex intervention organized into: 1) Building Blocks - structures and processes needed to deliver any PHC service, and 2) Clusters of Services for Maternal and Child Health – related to our outcome of interest (under-5 mortality). The model adopted was constructed from data of the Programme for the Improvement of Access to and Quality of Primary Health Care. To statistically assess the validity of the components as latent, unobserved variables, we applied Bayesian Confirmatory Factor Analysis to our model. Results The components with the highest (0.692 and 0.670) and almost lowest (0.269 and 0.270) associations with their respective indicators were ‘Facility Infrastructure’ and ‘Planning and organization’, respectively. The other three Building Blocks presented less variable coefficients (between 0.430 and 0.649), except for the indicator ‘Regulation of hospital beds’, which presented the lowest association (0.152) with its component. The Clusters of Services presented slightly higher associations with their respective indicators, ranging between 0.315 and 0.969. The components that were most interrelated were Antenatal care and Childcare (0.937). The next strongest correlations were between these two Clusters of Services and ‘Planning and Organization’ (0.868 and 0.835, respectively). ‘Antenatal care’ scored high associations also with ‘Referral and regulation’ (0.811) and ‘Facility Infrastructure’ (0.708). The lowest correlations were between ‘Immunization infrastructure’ and: ‘Workforce’ (0.198), ‘Planning and organization’ (0.281), ‘Referral and regulation’ (0.304) and ‘Childcare’ (0.388). Conclusions The PHC model and the measures (indicators and components) developed in this study may guide worldwide initiatives to investigate how PHC systems operate and evaluate their impacts on relevant population health outcomes, creating solid evidence to contribute to their improvement. Thus, we reinforce the importance of governments and other international organizations investing in robust data systems and providing enough resources to promote research in the health system, particularly in LMICs.