Ecology and emergence: Understanding factors that drive variation in process quality and clinical outcomes in general practice

Abstract

Clinical practice variation (CPV), where differences in healthcare delivery do not reflect differences in patient preferences or clinical need, is considered a hallmark of poor quality care. 'Unwarranted' variation is the focus of mounting policy attention and a growing body of literature, but remains poorly explained and theorised, with ways of determining when variation is warranted only weakly developed. Many assertions around CPV remain under-explored and untested. Much of the literature operates on the assumption that the legitimacy of variation depends on its source or cause, and that variation in processes of care will lead to related variation in outcomes. This doctoral research focuses on two overarching questions relating to CPV in Australian general practice: (1) what is CPV, and how can it be best conceptualised and understood; and (2) what can routinely-collected clinical data tell us about the phenomenon of CPV in general practice? Accordingly, this thesis explores the operationalisation of CPV as a theoretical construct and also examines variation in a series of clinical performance measures for coronary heart disease (CHD) and diabetes. Together, these lines of inquiry constitute a mixed-methods 'sense-making' exercise that seeks an incremental interplay between literature and data, to shed light on the phenomenon of CPV. Data are drawn from a unique dataset of aggregate reporting metrics, using extracted electronic medical record data, among an affiliated group of 36 general practice clinics serving approximately 189,848 patients over a 5-year period. These data are examined descriptively and ultimately analysed using Qualitative Comparative Analysis (QCA) against an empirically derived explanatory framework. Theory development draws on complexity science, especially complex adaptive systems theory, and the disciplines of social epidemiology and health ecology. Results show that a series of discourses have strongly shaped thinking about CPV, converging around a normative 'bad apples' approach to understanding variation. However, CPV may also contribute to healthcare quality in ways that are not well considered, especially in primary care settings. I demonstrate that there may be unconventional but more illuminating ways to conceptualise variation that enable our collective understanding to progress. These include using an ecological framework to conceive CPV as an emergent property of coupled, complex adaptive systems, and employing an equity lens to distinguish between CPV in processes and outcomes of care. In descriptive analyses, I find that variation frequently behaves differently across different measures, with crucial system information contained in the interstices of the data. Contrary to common assumptions, relationships between processes and outcomes of care are not straightforward. Using a framework of factors associated with CPV in general practice management of diabetes and CHD, I confirm that causality is complex and multifactorial, operating at a number of levels. Employing the case-based configurational method of QCA, I show that there may be no single or primary cause for CPV. Instead, clinics can arrive at a particular outcome via multiple independent causal pathways which are themselves multifactorial. These multi-component causes may be defined as much by the interactions between component elements as by individual elements themselves. The same factor may have differential effects within different combinations, or at different scales. These findings suggest that relying on causal explanations to demarcate unwarranted variation may be insufficient. However, both theory and methods require continued development to ensure an adequate understanding of the role and representation of warranted and unwarranted variation in performance measurement systems. Case-based configurational methods such as QCA may have substantial utility in helping to explain and delineate these phenomena.