Measuring management practices in India's district public health bureaucracy.

Weak management is widely recognised as a key impediment to scaling-up coverage of health interventions and ensuring health systems are responsive to population needs. Yet there is scant evidence linking management practices in the public administration to effective health service delivery. We report on the development of a tool to measure management practices in India's district health bureaucracy. We first developed a conceptual framework based on a review of the literature and qualitative interviews with district public health managers. Across 16 management practices, we then drafted and piloted questions to be used with a scoring grid to evaluate process-orientated management practices. We implemented the tool in 34 districts of Maharashtra between April and July 2016, interviewing up to three district public health managers per district (n = 99). Using rigorous psychometric methods, we assessed the acceptability, reliability and validity of the tool. We present three key findings. First, the tool was feasible to implement, response rates were high, and there were no missing data. Second, internal consistency of the tool was high and test-retest reliability was comparable with other management tools used in the literature. Third, there was evidence of validity. The number of staff with a management qualification was positively associated with better management practices. Factor analysis showed that one principal component loaded positively on all the management practices although there was little support for management sub-scales. These findings provide novel evidence on the psychometric properties of a tool designed to measure management practices in the public administration of a developing country. Our framework and tool provide the basis to examine associations between district health management practices and health service delivery, and test the effectiveness of management strengthening interventions in India's public health sector.