Methods Using a physician survey, we elicited each physician's estimate of the NCR by three approaches. In the primary approach, we asked: “How many individuals without ACS are you willing to admit to prevent one individual with ACS from being discharged home?” In the second approach, we asked: “At which probability of disease (ACS) does your decision change from discharge to admit?” The third approach employed a visual analogue scale to record valuations of health outcomes associated with admission or discharge of patients with suspected ACS.
For each approach, we estimated the median NCR with quartiles, and we determined which physician characteristics were associated with higher or lower NCRs, using rank tests.
Results Thirty-one physicians were interviewed. The mean age of the physicians interviewed was 40 years; 80% were male; the mean number of years as an attending was seven; 47% were emergency physicians and 53% were cardiologists.
Based on the primary approach, the median NCR was 100 (25th percentile: 20, 75th percentile: 1000, range: 5 to 1,000,000). Only physician specialty had a statistically significant association with NCR. The median NCR was 350 for emergency physicians and 40 for cardiologists (p = 0.014). NCR was not otherwise influenced by physician age, gender, or years as an attending.
NCR estimates differed significantly across approaches (p <0.001). Based on the second approach, the median NCR was 25 (25th percentile: 9, 75th percentile: 99). Based on the third approach, the median NCR was 12 (25th percentile: 5, 75th percentile: 27). A significant between-specialty difference persisted for the second approach (median estimate: 99 for emergency physicians, 9 for cardiologists), but not for the third approach (median estimate: 13 for emergency physicians, 11 for cardiologists).
Conclusions Physicians vary greatly in estimating the net consequence ratio for the decision to admit patients for the diagnostic work-up for ACS. Emergency physicians and cardiologists differ in their estimations, but this difference may depend on the elicitation approach.