Background: Living wills are considered clear and convincing evidence
of a person's preferences for end-of-life treatment. Unfortunately, li
ving wills often use vague language that forces physicians and others
to infer specific treatment choices, like the choice to forgo cardiopu
lmonary resuscitation (CPR). To test the validity of such inferences w
e examined the relationship between living will completion and CPR pre
ference. We also examined whether CPR choices were fixed or could be i
nfluenced by detailed information on CPR. Methods: We interviewed 102
retired elderly persons, many of whom had living wills. We obtained CP
R preferences in five hypothetical scenarios before and after providin
g CPR information. We then analyzed differences in desire for CPR betw
een the group of subjects with living wills and the group without. Res
ults: In each scenario there were subjects in both groups who desired
CPR. The group with living wills desired less CPR in scenarios involvi
ng functional impairment and cognitive impairment, but not in scenario
s involving current health, severe illness, and terminal illness. Afte
r receiving CPR information, both groups changed their preferences suc
h that intergroup differences were no longer seen. Conclusions: Prefer
ences for CPR among subjects with living wills are not homogeneous, bu
t distributed across the clinical scenarios. Therefore, one cannot inf
er CPR preference from the mere presence of a living will. Cardiopulmo
nary resuscitation information can influence preferences even among pe
rsons with living wills, implying that preferences are neither fixed n
or always based on adequate information. Physicians should view vaguel
y worded documents as unreliable expressions of treatment preference t
hat should not supplant informed discussion.