Purpose: Advance directives as a mechanism for helping patients exert control over end-of-life medical treatment have been criticized for failing to provide necessary context and guidance for patients to make decisions that are meaningful and consistent with their underlying values and preferences. The purpose of our study was to determine if an interactive computer-based decision aid that uses a multi-attribute utility model to generate an advance directive 1) was acceptable to users, and 2) accurately represented individuals' expressed values and preferences for future medical care. Individuals could self-tailor the program as well as make edits to the advance directive it generated.

Methods: We recruited adult volunteers from an outpatient internal medicine practice to pilot test our program (Making Your Wishes Known: Planning Your Medical Future). We measured 12 aspects of users' satisfaction with the program: how it provided information about medical conditions and treatments; its impact on users' knowledge of advance care planning; and its helpfulness at various aspects of the planning process (e.g. clarification of values/wishes, making important decisions, and putting wishes into words). We measured accuracy by asking individuals whether they made any edits to the advance directive initially generated by the program, and to rate how well the final version expressed their wishes.

Results: A total of 50 individuals used our computer program (mean age 52 years, 68% female, 68% college graduates). Users were very satisfied with the program overall (mean 8.5, where 1=not at all satisfied and 10=extremely satisfied), and also with how it improved their knowledge and helped them make decisions (mean 4.2 for 12-item measure, where 1=very dissatisfied and 5=very satisfied). Also, users were very satisfied with the decisions they made (mean 26.5, where 6=dissatisfied and 30=very satisfied).

Overall, 40% of users made no edits to the advance directive generated by the computer. For the 60% who did make edits, they rated the accuracy of the pre-edited advance directive at 5.7 (1= not accurate and 7=very accurate). For all users, (including those who made no edits) the accuracy rating of the final advance directive was 6.4 on this 7-point scale.

Conclusions: Our computer-based decision aid for advance care planning is well-accepted, and generates an advance directive that users affirm is an accurate reflection of their values and preferences.