Developing an end-of-life quality measure (QM) for nursing homes based on place of death

This study was designed to develop a quality measure (QM) for end-of-life (EOL) care which can be calculated from administrative data available for all nursing homes. The QM measures the risk adjusted probability of death in the hospital, reflecting the expectation that, when appropriate, residents are better off if they are not transferred to the hospital to die.

The study included 2,526,289 decedents in 15,786 nursing homes nationally, during 2003-2007.

MDS data were linked to hospital, hospice and Medicare enrollment files. Logistic, random effect, risk adjustment models predicting the probability of death in the hospital for each resident, were estimated. The initial risk factors were chosen based on review of the literature and judgment of the geriatricians. These models were estimated on a development dataset and then reevaluated on a test dataset.

The QM was calculated as the difference between the actual and the predicted rate of hospitalization. The distributional characteristics of the QM were investigated.

24.3% of decedents died in the hospital. The models showed close agreement between the development and the test datasets. Risk factors included age, gender, do not resuscitate and do not hospitalize orders, and specific diseases. The average QM for 2006 was 0.020 and the standard deviation was 0.163. The 10th percentile was at -0.146 and the 90th at 0.228.

This study offers a prototype EOL QM that could be used to expand the current set of QMs in the Nursing Home Compare web-based report card to include the EOL domain.