Projecting patients' preferences from living wills: an invalid strategy for management of dementia with life-threatening illness

Abstract

Objective: To examine variation in elders' choices of therapies in different clinical scenarios and to assess the validity of extending preferences expressed in scenarios of usual health, terminal illness, and coma to preferences in a scenario of moderately advanced Alzheimer disease.

Design: Questionnaire study of community-dwelling elders.

Setting: Houston metropolitan area.

Participants: 218 community-dwellers age 60 years and older.

Measurements: Responses regarding choices of 10 interventions in 4 scenarios. Interventions were: cardiopulmonary resuscitation (CPR), ventilator, total parenteral nutrition (TPN), i.v. medication and hydration, any medication, enteral feeding, dialysis, ICU admission, hospitalization, and antibiotics. Interventions were selected "never", "always," or a "trial of intervention to assess efficacy." Independent variables were responses in scenarios of usual state of health with a life-threatening illness, irreversible coma, and terminal illness causing pain. Dependent variables were responses in a scenario of moderately advanced Alzheimer disease with a life-threatening illness. Frequencies of responses were calculated using "never," "trial," and "always." Subsequently "trial" and "always" were collapsed into a category of "accepting intervention" for dichotomous analysis with "refusing intervention" (the "never" category). Logistic regression was used to assess validity of predicting responses in one scenario from the others.

Main results: Preferences regarding medical therapies varied across scenarios (P < 0.01). In the Usual Health scenario, all interventions were accepted more frequently than refused. In Terminal Illness and Coma scenarios, CPR, ventilator, TPN, enteral feedings and dialysis were refused more frequently than accepted. In the Alzheimer scenario, medications, ICU admission, hospitalization, and antibiotics were accepted more often than rejected. Trial was preferred to always in 90% of all choices across all scenarios. Preferences expressed in Terminal Illness, Coma, and Usual Health scenarios predicted choices in the Alzheimer disease scenario poorly.

Conclusions: (1) Use of a scenario-based advance directive may be limited to the precise scenario described. (2) The common acceptance of interventions in the Alzheimer disease scenario differs from findings in earlier studies, possibly because of differences in populations surveyed or the stage of the disease described, highlighting the variability of preferences in this scenario. (3) Trial of intervention is attractive to many respondents, perhaps because it allows the advantage of potentially beneficial therapies without commitment to a course of therapy not leading to cure. (4) Results of this study should be interpreted in light of the study population, consisting largely of well educated, healthy Caucasians. Findings are likely not to be generalizable to other populations.