Can goals of care be used to predict intervention preferences in an advance directive?

Background: Some have suggested that advance directives elicit goals of care from patients, instead of or in addition to specific intervention preferences, but little is known about whether goals of care can be used in a meaningful way on documents or whether they can predict preferences for specific interventions. Methods: Attending physicians (n=716) at the Massachusetts General Hospital in Boston were surveyed to elicit general goals of care (eg, treat everything or comfort measures only) along with specific preferences for 11 medical interventions in 6 scenarios. In each scenario, each goal was classified as an adequate predictor of acceptance or rejection of an intervention if its predictive value of the preference for that intervention was at least 80%. Results: Goals varied with scenarios (P<.001) in a predictable manner. The goal treat everything was an adequate predictor of acceptance of each intervention, and comfort was an adequate predictor of rejection for nearly every intervention. Attempt cure adequately predicted acceptance of almost every nonaggressive intervention, but did not predict acceptance of aggressive interventions. Quality of life predicted rejection of aggressive interventions in 3 scenarios, but was not useful in other cases. When goals were predictors of preferences, the mean range of 95% confidence intervals for their predictive values was generally 20% or less. Conclusions: Goals have a valid role in advance directives, since the goal choices had a logical relationship to scenarios and intervention choices. However, the 2 goals attempt cure and choose quality of life were not predictive in many instances. If these findings hold true for more general populations of patients, then advance directive documents will need to rely on more than these general goal statements if they are to adequately represent patient preferences.