Prediction of 6-month mortality in nursing home residents with advanced dementia: validity of a risk score

OBJECTIVE: Prognostic information is important for guiding palliative care planning for patients with dementia. We aim to validate a risk score that uses Minimum Data Set (MDS) to estimate 6-month mortality for nursing home residents with advanced dementia. DESIGN: Two cohort studies. SETTING: Six nursing homes in The Netherlands, and 35 nursing homes in Missouri. PARTICIPANTS: Long-term stay residents with advanced dementia: 288 Dutch residents and 269 residents from Missouri who also had a lower respiratory tract infection (LRI). MEASUREMENTS: Patient risk factors and 6-month mortality. RESULTS: Six-month mortality rates were 24.3% for Dutch residents, and 36.8% for US residents. The risk score's AUROC was 0.65 (CI 0.58-0.72), and 0.64 (CI 0.58-0.71), respectively. For the large majority of residents, observed mortality in the 2 validation cohorts were comparable to the development cohort. Among the few residents identified as at very high risk according to the risk score, observed mortality was lower than expected. CONCLUSION: The original mortality risk score predicted 6-month mortality with reasonable accuracy in 2 validation cohorts of nursing home residents with advanced dementia. Thus, the performance of the risk score, at least over the range of low to moderate risk (up to around 40% risk of mortality), can be generalized to long-stay (versus recently admitted) residents with advanced dementia, and to those with LRI.