Early diagnosis and treatment promise deep transformation in healthcare by enabling better treatment options, improving long-term survival and quality of life, and reducing overall cost. However, ED&D research often focuses on novel technologies or risk stratification while neglecting the potential for optimising individual patients’ diagnostic paths by improving medical decision-making. To enable personalised early and timely diagnosis, a foundational framework is needed that delineates the diagnosis process and systematically identifies the time-dependent value of various diagnostic decisions for an individual patient given their unique characteristics and history.
In a recent pre-print, we propose a foundational framework for early and timely diagnosis. The framework is grounded in decision-theoretic approaches and integrates machine learning and statistical methodology to estimate the optimal personalised diagnostic path.
This framework provides:
– formalism and thus clarity for the development of decision support tools
– a holistic view on complementing observed patient information from present and past with estimates of the patient’s future trajectory
– a rationale to estimate the net benefit of counterfactual diagnostic paths and the associated uncertainties
– fundamental definitions for describing this and other frameworks, including a clear differentiation between “early” and “timely” diagnosis
– a systematic mechanism for assessing the value of technologies, including new and existing diagnostic tests in terms of their impact on personalised early diagnosis, resulting health outcomes and incurred costs for individuals, populations and healthcare systems.
This work informs early diagnosis research on all 4 levels of our framework defining the impact of ML on healthcare.