The problem of assessment bias when measuring the hospice effect on nursing home residents' pain.

This study examined the observed differential documentation of pain on nursing home (NH) resident assessments (minimum data sets [MDS]) when dying residents were and were not enrolled in hospice. We studied 9,613 NH residents who died in 6 states in 1999 and 2000. Documented pain was compared among three groups of residents who were categorized by their hospice exposure. At the time of their last MDS completion, residents in hospice were more likely to receive opioids for their moderate to severe pain than were non-hospice residents and residents enrolled in hospice after the last MDS assessments. However, hospice residents were twice as likely as non-hospice residents and 1.3 times as likely as residents who eventually enrolled in hospice to have pain documented. These counterintuitive findings suggest that there is differential documentation of pain on the MDS when hospice is involved in care, perhaps because of superior pain assessment by hospice.