This framework illustrates how future clinicians can be equipped to make more accurate diagnoses and reduce error through attention to probabilistic diagnosis. Once internalized, this approach is largely intuitive and uses information provided at the point of care, and it would not add substantial time or cognitive load to clinical encounters. Improving diagnostic accuracy will require better clinician skills and tools, including the following achievable steps:
- Focus on embracing uncertainty as a core educational principle.29,35 Medical school and other health professional courses on pathophysiology and history and physical examination should acknowledge and quantify the high degree of uncertainty embedded in all clinical care. Such acknowledgment could be easily integrated into current medical school curricula. Curricula for other clinicians such as nurse practitioners currently include very little content on probabilistic thinking and uncertainty; integration in these settings will be critical but may require more substantial change.
- Emphasize quantifying probability throughout medical education. During clinical training, students and residents should be asked to quantify the likelihood of different diagnoses, expressing it as a probability or range of probabilities. This exercise will highlight the incidence of common diseases in various populations, train intuition, and trigger discussions of determinants of those probabilities to inform more formal probabilistic reasoning.
- Enable clinician access to better data on disease incidence. This goal could partly be accomplished by incorporating links to available data, such as the CDC COVID Data Tracker, into electronic health records (EHRs). Ideally, the approach would use novel EHR-based tools that could refine estimates using clinical characteristics (such as disease calculators on testingwisely.com). It would also require development of better evidence on pretest probability based on epidemiologic studies that report both disease incidence in the population and the nature and frequency of presentations in clinical settings.
- Provide data about test performance at the point of care, ideally through integration with EHRs. Various methods are available to structure such information delivery; doing so at the time of test ordering would be optimal.
Ultimately, better clinician management of probability will lead to better management of patients and fewer diagnostic errors. Clinicians must be better educated to accurately estimate disease probabilities in the general population and in individual patients and to adjust those probabilities in response to test results. Education must train probabilistic intuition, empower more deliberative probability adjustment, and provide needed tools at the point of care. Through understanding probability, clinicians can improve the diagnostic process and optimize patient safety.
