Purpose: To review clinical interventions designed to change care at t
he end of life. Data Sources: Published results of clinical trials ide
ntified by MEDLINE searches, review of abstracts from meetings, and se
lected bibliographies. Study Selection: Studies were included if data
were presented on a clinical intervention designed to change medical c
are at the end of life. Studies done in nonclinical settings or outsid
e the United States were excluded. Data Extraction: Interventions were
classified as targeting patients, physicians, or both and were analyz
ed by their effect on four common clinical goals: increasing use of pa
tient preferences, decreasing pain and suffering, reducing use of life
-sustaining treatments, and reducing costs. Positive and negative tria
l results were compared for differences in intervention, target group,
and strength of study methods. Data Synthesis: Educational interventi
ons usually increased expression of patient treatment preferences. Suc
cess rates were higher when more severely ill patients were targeted a
nd when written materials were combined with repeated discussions in c
linical encounters. Educational interventions with physicians led to i
ncreased use of patient preferences, but sophisticated educational tec
hniques were needed to motivate physicians to change their behavior. T
hree studies provided limited evidence that physician education reduce
d the use of life-sustaining treatments. No clinical intervention had
an effect on pain, suffering, or cost of medical care. Conclusion: To
change care at the end of life, intensive educational interventions fo
r physicians and broad institutional programs seem more promising than
advance directives. Future innovations should seek to change physicia
n practices, reduce costs, and improve patient-centered out comes, suc
h as pain control and satisfaction.