Prediction of five-year survival for patients admitted to a department of internal medicine

Abstract

Objective. The effect of many common forms of therapy, as medication for mild hypertension or hypercholesterolaemia, only reaches clinical significance after years of treatment. The meaningful application of such therapy presupposes that physicians can, at least to some extent, predict the remaining lifetime of patients. We investigated whether clinicians from different disciplines were able to predict the 5-year survival of patients admitted to a department of internal medicine. Design. The members of two groups, each consisting of an internist, a surgeon and a general practitioner, made individual predictions of the expected remaining lifetime of discharged patients from written summaries of clinical information. Each patient was randomized to be assessed by the members of either of the two groups. The predictions were compared with actual 5-year survival. Setting. Department of internal medicine at a university hospital. Subjects. Patients admitted consecutively during a 6-week period. Main outcome measures. Sensitivity, specificity, positive and negative predictive values and areas under the receiver operating characteristic (ROC) curves for predictions of 5-year survival for each of the six experts. Results. A total of 402 patients were included. Five-year survival was 0.63. The sensitivity of the predictions ranged from 0.81 to 0.95, the specificity from 0.61 to 0.77, the positive predictive value from 0.78 to 0.87 and the negative predictive value from 0.68 to 0.87. The areas under the ROC curves ranged from 0.84 to 0.91. Conclusion. The quality of predictions of 5-year survival made by experienced clinicians should permit the rational use of treatments with long-term effects.