Background: Treatment preferences established before life-threatening illness occurs may differ from actual decisions because of changes in preferences or poor understanding of the link between prospective preferences and outcomes.
Objective: To evaluate the validity of prospective treatment preferences by examining their concordance with ratings of health states.
Design: Survey of seven cohorts of persons with diverse health status. Home- and hospital-based interviews were conducted at baseline and at 6, 18, and 30 months.
Setting: The greater Seattle area.
Participants: Younger and older well adults; persons with chronic conditions, terminal cancer, or AIDS; stroke survivors; and nursing home residents.
Measurements: Concordance between six treatment preferences and five health state ratings (on a seven-point scale) was assessed by using logistic regression to measure the increase in odds of treatment refusal for each one-point change in health state ratings. Preferences were considered concordant if treatments were refused in health states rated as worse than death and were accepted in health states rated as better than death. Reasons for discordance were elicited at the final interview.
Results: The probability of refusal of prospective treatment was strongly related to health state ratings. Odds ratios ranged from 1.7 to 1.9 (P < 0.001) for every treatment. When patients were shown their discordant preferences, they had a coherent explanation or changed their health state rating or treatment preference to make the two concordant.
Conclusions: Prospective life-sustaining treatment preferences show high convergent validity. For most persons, treatment preferences are grounded in a consistent belief system. Concordance and discordance between treatment preferences and health state ratings offer clinicians the opportunity to explore patients' values and reasoning.