Objectives:Advance directives are used to increase autonomy in decisions regarding care when a person lacks the ability to communicate such wishes. Based on studies showing internal consistency in individuals’ preferences, this study demonstrates a new method for identifying a list of questions that may best predict patient preferences. Method:Participants were 71 residents of a large nursing home and 97 hospital patients. All were administered the Preferences for Life-Sustaining Treatment Questionnaire. Advance directives sequence models were developed using conditional probabilities concerning preferences for utilization of specific treatment options, given prior responses to different treatment options. Results:Models resulted in more than 90% accuracy of treatment preferences for both samples and eliminated two thirds to three quarters of questions asked. Conclusions:Although the specific models need validation using larger samples, they demonstrate a method that facilitates transfer of information concerning treatment preferences if future incapacitation occurs.