The development of process measures for the quality of inpatient care given to children hospitalised with common illnesses in Kenya

Assessing the quality of health services remains a major challenge for the less industrialized nations of the world. Hospitals consume high proportions of national health budgets but approaches to examining their capacities and performance are still rudimentary. Better assessment strategies are essential for understanding what investments in hospital care and health systems are achieving and whether resources are being equitably distributed. Assessing hospital care for children is a particular challenge given the multiplicity of steps to be correctly undertaken in examining patients, diagnosing disease and giving appropriate treatment. This thesis describes the development of a measure of quality of care for children admitted to hospital, which can be routinely deployed to form part of future nation-wide health system assessments. The measure is based on standards of paediatric care described by national clinical practice guidelines for Kenya adapted from WHO guidelines. It is constructed using data from a trial of a new intervention to improve quality of care in first-level referral hospitals in Kenya. The proposed measure is subsequently validated using data from observations of routine hospital care for children admitted to a different set of hospitals over a period not covered by the primary data. A number of statistical techniques are used in this regard: these include factor analysis to explore the dimensions of process of care, logistic regression to study the association between the new measure and mortality, and multilevel modelling to explore the amount of variability in the data lost through modification of the structure of the measure. These analyses show that the items making up the measure are consistent with three conceptual domains of clinical process described in guidelines. They also provide evidence that the measure is associated with other exposures and outcomes in ways that strengthen its validity and suitability of purpose. Specifically, there is strong evidence that adherence to each of six generic recommendations of how care should be provided is associated with a reduction in odds of death by a fifth. This thesis thus demonstrates the usefulness of a generic approach to measuring quality of care, and highlights key issues to be addressed when extending this approach to other settings.