Prevalence and Alternative Explanations Influence Cancer Diagnosis: An Experimental Study With Physicians
Objective: Cancer causes death to millions of people worldwide. Early detection of cancer in primary care may enhance patients’ chances of survival. However, physicians often miss early cancers, which tend to present with undifferentiated symptoms. Within a theoretical framework of the hypothesis generation (HyGene) model, together with psychological literature, we studied how 2 factors—cancer prevalence and an alternative explanation for the patient’s symptoms—impede early cancer detection, as well as prompt patient management. Method: Three hundred family physicians diagnosed and managed 2 patient cases, where cancer was a possible diagnosis (one colorectal cancer, the other lung cancer). We employed a 2 (cancer prevalence: low vs. high) × 2 (alternative explanation: present vs. absent) between-subjects design. Cancer prevalence was manipulated by changing either patient age or sex; the alternative explanation for the symptoms was manipulated by adding or removing a relevant clinical history. Each patient consulted twice. Results: In a series of random-intercept logistic models, both higher prevalence (OR = 1.92, 95% confidence interval [CI 1.27, 2.92]) and absence of an alternative explanation (OR = 1.70, 95% CI [1.11, 2.59]) increased the likelihood of a cancer diagnosis, which, in turn, increased the likelihood of prompt referral (OR = 22.84, 95% CI [16.14, 32.32]). Conclusions: These findings confirm the probabilistic nature of the diagnosis generation process and validate the application of the HyGene model to early cancer detection. Increasing the salience of cancer—such as listing cancer as a diagnostic possibility—during the initial hypothesis generation phase may improve early cancer detection.