Prognoses of seriously ill hospitalized patients on the days before death: implications for patient care and public policy.

Troubling aspects of the experiences of patients at the ends of their lives have fueled interest in special benefits or privileges for this group. There is a presumption that being "at the end of life" is discernible. This study examines this presumption using data from two previously collected databases: the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT) and the Acute Physiology and Chronic Health Evaluation III (APACHE III). Both studies generated multivariable estimates of survival prognosis for hospitalized patients. SUPPORT included 9,105 patients with one of nine serious illnesses in five hospitals over 4 yrs, of whom 2,360 died in the hospital and 4,537 died within 180 days of entry into the study. The APACHE III database describes 2,750 deaths in 16,622 ICU patients in 40 hospitals. The relationship of median estimates with time to death were examined for each source of data, for different diseases, and for ICU settings of care. In SUPPORT, the median predicted chance of survival for 2 months on the day before actual death was .17 (interquartile range, .02-.40) and was .51 (.31-.66) 1 wk before death. Median prognoses varied substantially among diseases: the median for congestive heart failure patients was a .62 chance of living 2 months on the day before death, while lung cancer had only a .17 chance and coma patients only an .11 chance. Median prognostication estimates were not much different when given by physicians and were only a little more pessimistic in APACHE (median estimate for hospital survival on the day before death was .14 and 7 days before was .45). To make plans about care and to optimally support most dying persons and families, conversations must occur while the patient still has a considerable chance of surviving the current episode of illness. Using statistical estimates of prognosis to designate a category of "terminally ill" patients for public policy purposes is unavoidably arbitrary, will often be contested, and will have differential effects upon those dying with differing diseases.