OBJECTIVES: Methods to quantify overdiagnosis of screen detected cancer have been developed, but methods for quantifying overdiagnosis of non-cancer conditions (whether symptomatic or asymptomatic) have been lacking. We aimed to develop a methodological framework for quantifying overdiagnosis that may be used for asymptomatic or symptomatic conditions, and used Gestational Diabetes Mellitus as an example of how it may be applied.
STUDY DESIGN AND SETTING: We identify two earlier definitions for overdiagnosis, a narrower prognosis-based definition, and a wider utility-based definition. Building on the central importance of the concepts of prognostic information and clinical utility of a diagnosis, we consider the following questions: within a target population, do people found to have a disease using one diagnostic strategy but found not to have the disease using another diagnostic strategy (so called 'additional diagnoses'), have an increased risk of adverse clinical outcomes without treatment (prognosis evidence), and/or a decreased risk of adverse outcomes with treatment (utility evidence)?
RESULTS: Using Causal Directed Acyclic Graphs and Fair Umpires, we illuminate the relationships between diagnostics strategies and the frequency of overdiagnosis. We then use the example of Gestational Diabetes Mellitus to demonstrate how the Fair Umpire framework may be applied to estimate overdiagnosis.
CONCLUSION: Our framework may be used to quantify overdiagnosis in non-cancer conditions (and in cancer conditions), as well as to guide further studies on this topic.