Are bereaved family members a valid proxy for a patient's assessment of dying?

OBJECTIVE: To compare assessments made retrospectively by bereaved family members (or the nearest carer to the patient) with assessments made before death by palliative staff and, where available, by patients themselves or the family member.
METHODS: SETTING--two palliative care support teams. ASSESSMENTS--were recorded prospectively by team staff, patients and their family members for consecutive patients referred, and then were recorded retrospectively by family members during interview seven months after bereavement. MEASURES--seven items each rated 0 (best) to 4 (worst) using standard definitions. The rater was asked to average the severity over one week. ANALYSIS--ratings were tested for percentage agreement, for Cohen's Kappa (which controls for chance agreement) and for Spearman correlations.
RESULTS: Staff ratings and family members' retrospective ratings, which described the last week of life, were available for 35 patients. Six patients and seven family members had also been interviewed shortly before the patient's death. The main problems identified by all raters were similar: family anxiety, symptom control, patient anxiety and pain control. For three items, practical aid, wasted time and communication, agreement was good--all cases except one were equal or within one score. However, problems were rarely identified for these items. For the other four items: pain control, other symptom control, family anxiety and patient anxiety, there was little agreement, Cohen's Kappa ranged 0.05-0.22. Agreement for one item (patient anxiety) was significantly improved if a patient had died at home. Comparison of ratings made by the family members before the death and seven months after bereavement suggests that family members alter their assessments during bereavement.
CONCLUSION: Retrospective assessments by bereaved family members may be valid for some items related to service provision, but not as the sole assessment of a patient's pain, symptoms or anxiety. We suggest that studies which rely on these retrospective ratings should assess the validity of their responses and record more information about the mood and grief of the family member.