Background:
Little is known about the ability of physicians to predict death in hospitalized patients over the subsequent year. Accurate prediction of death can facilitate discussions for end‐of‐life planning.
Methods:
The study was part of the larger Hospital Medicine Reengineering Network (HOMERuN) Transitions of Care study, a multisite, prospective observational study of randomly selected patients discharged from a medicine service and readmitted within 30 days. This review was conducted in a subset of 100 patients at a single urban academic safety net hospital. Discharging physicians, readmitting physicians (if different) and primary care providers (PCPs) were surveyed about factors related to patient readmission. The surveys included the question “Would you be surprised if the patient died within the next 6‐12 months?” The chart was abstracted for patient characteristics and hospitalization processes including discharge procedures and end‐of‐life discussions. Death within 1 year of the index admission was determined based on chart review (87 patients) or the social security death index database (13 patients).
Results:
There were 100 readmitted patients enrolled between May 2012 and February 2013. Mortality follow‐up data were available through November 26, 2013. Survey response rates were 73/100 (73%) of the admitting physicians, 42/79 (53%) of discharging physicians, and 39/81 (44%) of PCPs. At least one provider response was available for 92% of patients. The mean patient age was 55 years (SD=13 years). Patients had at least one advanced conditions in 36% of readmissions: 18% with stage IV‐V chronic kidney disease, 9% with cancer, 7% with stage III or IV congestive heart failure, 5% with severe COPD, and 1% with a degenerative CNS disorder. Active drug or alcohol abuse was present in 40% of patients. The mortality rate was 28% in the follow‐up period, and among deaths the mean time to death was 127 days (range 13‐336 days). The majority (63%) of responding physicians reported they would not be surprised if their readmitted patient died in the next 6‐12 months, and 66% of patients had at least one physician respond affirmatively. The sensitivity of this prediction was 88%, specificity was 48%, and positive predictive value was 44%. Of patients who died, 50% (14/28) had an early palliative care or goals of care discussion (i.e., prior to the readmission), compared to 14% (10/72) of patients who did not die on follow‐up; overall 82% (23/28) of patients who died had a palliative care discussion before they died. The early palliative care consultation rate was 3.9 times greater for patients who had at least one of their physicians indicating a risk of death in the subsequent year.
Conclusions:
Readmitted medicine patients at this urban safety net hospital had a high mortality over the subsequent year. Patients who died were recognized to be at significant risk of death by their physicians in most cases, and half of these patients had early palliative care involvement.