“Uncertainty pervades the diagnostic process”7 in the form of inherently imprecise estimates of probability, limits of scientific knowledge, and incomplete individual clinician knowledge.8 But even with the most accurate probability estimates, high-quality evidence to inform decision making, and optimal clinician knowledge, the diagnostic process will always involve some uncertainty.
Few clinical decisions are made in the context of complete certainty. Beginning with the conceptualization of the diagnostic process created by the National Academies of Sciences, Engineering, and Medicine, we further mapped the process of diagnosis as it relates to probability.1
As illustrated in Figure 1, which depicts the framework we developed, every step of the process of diagnosis incorporates understanding of probabilities that are between 0 and 100 percent, with some uncertainty always present. For example, a patient presenting with acute cough, fever, and shortness of breath who has nonspecific findings on chest x-ray is far from certain to have pneumonia, but treatment with antibiotics is likely appropriate. In the figure, steps in the diagnostic pathway are displayed in the boxes at the top and key points for understanding probability are shown in the boxes below.
Figure 1. Importance of probabilistic understanding in the diagnostic process
Source: Based on Stern SDC, Cifu AS, Altkorn D. Symptom to Diagnosis: An Evidence-Based Guide, Third Edition. New York, NY: McGraw-Hill; 2015. Chapter 1: Diagnostic Process. https://accessmedicine.mhmedical.com/content.aspx?bookid=1088§ionid=61696411.
The concept of uncertainty often provokes anxiety in clinicians, who may misconstrue it as a reflection of individual flaws, but uncertainty must be embraced to optimize diagnostic excellence. Clinicians who believe there is little uncertainty in medicine tend to overtest and overtreat patients.9,10 Engaging with patients to explain and ideally quantify uncertainty can increase understanding, confidence, and accuracy of the diagnostic process.
Probabilistic reasoning that manages the inevitable uncertainty, including adjusting chance of disease and updating degree of uncertainty with new information during clinical care in an explicit fashion, may improve diagnosis. When diagnoses are expressed as possible and uncertainty is recognized, the chance of misdiagnosis and premature closure (failing to consider reasonable alternatives after an initial diagnostic impression) is reduced.11
Understanding the Importance of Language
Clear language reflects organized thinking. Use of nonspecific words can bias and confuse conversations around probability and muddy the diagnostic process. Simple language includes discussing chance or probability of disease or diseases, with numerical quantification.
For example, an 80-year-old man with hypertension and lower extremity edema who is worried about heart failure asks his primary care provider for a brain natriuretic peptide (BNP) test. The clinician might explain the low likelihood of heart failure, the impact of results of a BNP test, and the rationale for not performing the test by stating, “Someone like you has about a 2 percent chance of having heart failure. If we do a BNP test and it’s abnormal, that chance of heart failure would rise to about 8 percent.”
A focus on quantifying probability numerically is important given the inconsistency in interpretation of vague phrases such as “likely,” “probably,” and “maybe,” which have been shown to mean different things to different people.12 Adding to the imprecision, the medical community often further confounds understanding by relying on pseudo-probabilistic aphorisms such as “low threshold” and “diagnosis of exclusion.”13
Biased language can influence clinical thinking and reasoning. For example, the outcome of medical decisions is often incorrectly phrased in terms of “risks” and “benefits” vs. the more neutral, transparent, and quantifiable comparison of “chance of harm” and “chance of benefit.”14