Reducing Diagnostic Errors in Medicine

This review considers the feasibility of reducing or eliminating the three major categories of diagnostic errors in medicine: "No-fault errors" occur when the disease is silent, presents atypically, or mimics something more common. These errors will inevitably decline as medical science advances, new syndromes are identified, and diseases can be detected more accurately or at earlier stages. These errors can never be eradicated, unfortunately, because new diseases emerge, tests are never perfect, patients are sometimes noncompliant, and physicians will inevitably, at times, choose the most likely diagnosis over the correct one, illustrating the concept of necessary fallibility and the probabilistic nature of choosing a diagnosis. "System errors" play a role when diagnosis is delayed or missed because of latent imperfections in the health care system. These errors can be reduced by system improvements, but can never be eliminated because these improvements lag behind and degrade over time, and each new fix creates the opportunity for novel errors. Tradeoffs also guarantee system errors will persist, when resources are just shifted. "Cognitive errors" reflect misdiagnosis from faulty data collection or interpretation, flawed reasoning, or incomplete knowledge. The limitations of human processing and the inherent biases in using heuristics guarantee that these errors will persist. Opportunities exist, however, for improving the cognitive aspect of diagnosis by adopting system-level changes (e.g., second opinions, decision-support systems, enhanced access to specialists) and by training designed to improve cognition or cognitive awareness. Diagnostic error can be substantially reduced, but never eradicated.