More talk, less paper: predicting the accuracy of substituted judgments

To study the accuracy of substituted judgments regarding life-sustaining therapies and other therapies made by surrogate decision makers and to investigate factors associated with more accurate predictions.

A total of 50 pairs of ambulatory patients and surrogates, chosen according to a legal hierarchy, underwent separate interviews in which surrogates were asked to predict the preferences of patients for eight modes of medical therapy in three clinical scenarios, given only yes or no as response options. Patient preferences, their surrogates' predictions, and the extent of agreement between the two were measured. The total number of correct predictions constituted the Surrogate Accuracy in Matching Patient Preferences Scale (SAMPPS). Sociodemographic factors associated with agreement were also assessed.

Agreement between patients and surrogates ranged from 57% to 81%. The mean SAMPPS score was 17 of 24 correct. Kappa (k) coefficients, which measure inter-rater concordance, were positive for 23 of 24 items and were 0.3 or greater (P < 0.05) for 14 of 24 items. Rates of agreement were not related to whether the surrogate interviewed (surrogate determined by state law) was the person the patient would have chosen as a surrogate or whether the patient had an advance directive. In multiple linear regression analysis, both prior discussions of preferences and nonchurchgoing behavior were significantly associated with patient-surrogate agreement, independent of religious denomination and race.

When pressed to choose, surrogates can predict the preferences of patients for life-sustaining therapies with an imperfect accuracy that nonetheless significantly exceeds the agreement expected due to chance alone. Exhorting surrogates to give their "best estimate" and encouraging prior discussions may improve accuracy. Houses of worship might be important target sites for campaigns to improve public awareness about advance directives.