Is "Appearing Chronically Ill" a Sign of Poor Health? A Study of Diagnostic Accuracy

Objective: To determine the sensitivity and specificity of a physician's assessment that a patient "appears chronically ill" for the detection of poor health status. Methods: The health status of 126 adult outpatients was determined using the 12-Item Short Form Health Survey (SF-12). Physician participants (n = 111 residents and faculty) viewed photographs of each patient participant and assessed whether or not the patient appeared chronically ill. For the entire group of physicians, the median sensitivity and specificity of "appearing chronically ill" for the detection of poor health status (defined as SF-12 physical health score below age group norms by at least 1 SD) were calculated. The study took place from February 2009 to January 2011. Results: Forty-two participants (33%) had an SF-12 physical health score >= 1 SD below age group norms, and 22 (18%) had a score >= 2 SD below age group norms. When poor health status was defined as an SF-12 physical score >= 1 SD below age group norms, the median sensitivity was 38.1% (IQR 28.6-47.6%), specificity 78.6% (IQR 69.0-84.0%), positive likelihood ratio 1.64 (IQR 1.42-2.15), and negative likelihood ratio 0.82 (IQR 0.74-0.87). For an SF-12 physical score >= 2 SD below age group norms, the median sensitivity was 45.5% (IQR 36.4-54.5%), specificity 76.9% (IQR 66.3-83.7%), positive likelihood ratio 1.77 (IQR 1.49-2.25), and negative likelihood ratio 0.75 (IQR 0.66-0.86). Conclusions: Our study suggests that a physician's assessment that a patient "appears chronically ill'' has poor sensitivity and modest specificity for the detection of poor health status in adult outpatients. The associated likelihood ratios indicate that this assessment may have limited diagnostic value.