Understanding the dynamics of patient treatment preferences can be important for end-of life are research, and has particular salience not only to guide a process of advance care planning (ACP) but also as an outcome measure. Ascertaining the reliability and responsiveness of preferences for life-sustaining treatments within and between patients is a necessary foundation for utilizing patient-agent congruence as an outcome for ACP interventions. This study validated a modified version of the Emanuel and Emanuel Medical Directive for use in both research and clinical applications. Seriously ill patients (n = 168) were asked at baseline and 21 days to consider four common end-of-life health state scenarios, to indicate their goals for treatment, and to state their preferences for six specific treatments. We investigated the reliability and validity of this tool. We found that preferences for life-sustaining treatments were highly intercorrelated, and internally consistent across treatments by scenario and across scenarios by treatment. Preferences for pain medications were, however, distinct from preferences for other treatments. Preference scores exhibited stability over follow-up, and demonstrated both concurrent and discriminant validity. We detected a small effect size for change in preferences as a function of health state change, suggesting that re-prioritization response shifts do occur but are small in magnitude in these patient samples over this time frame. We conclude that this measure is reliable and valid for use in clinical settings and for evaluating interventions designed to improve patient-agent congruence about patient preferences for life-sustaining treatments. Clinical applications of the tool are discussed.