Diagnosis Uncertain Models For Medical Risk Prediction
This addresses a critical failure mode in all-cause risk models for medical triage, offering an incremental improvement in interpretability and accuracy for healthcare practitioners.
The paper tackles the problem of medical risk prediction models that lack access to patient diagnoses, showing they generalize well but systematically underestimate risk for rare, high-risk conditions due to averaging across diagnoses. The authors propose a fix by modeling diagnostic uncertainty, providing interpretable risk assessments beyond a single number.
We consider a patient risk models which has access to patient features such as vital signs, lab values, and prior history but does not have access to a patient's diagnosis. For example, this occurs in a model deployed at intake time for triage purposes. We show that such `all-cause' risk models have good generalization across diagnoses but have a predictable failure mode. When the same lab/vital/history profiles can result from diagnoses with different risk profiles (e.g. E.coli vs. MRSA) the risk estimate is a probability weighted average of these two profiles. This leads to an under-estimation of risk for rare but highly risky diagnoses. We propose a fix for this problem by explicitly modeling the uncertainty in risk prediction coming from uncertainty in patient diagnoses. This gives practitioners an interpretable way to understand patient risk beyond a single risk number.