Purpose: To compare two methods of identifying nursing home residents with a life expectancy of 6 months or less.
Design: Secondary analysis of Minimum Data Set (MDS) assessment data from 111 U.S. nursing homes was used in two approaches to compare the prognostic value of each approach. MDS assessment data were collected during 2003 and 2004, and secondary analyses were completed during 2004 and 2005.
Methods: A bivariate analysis was used to identify variables associated with death within 6 months and a summative index was produced. Second, logistic regression was performed to develop a formula for calculating a Probability of Death score.
Findings: Both methods were reasonably sensitive in identifying dying residents. In the total sample of 21,852 residents, 17.5% died within 6 months. When the bivariate analysis and summative index were used, 51.1% of residents who had a score of 4 or higher died within 6 months, and residents who scored 10 or higher had an 80% mortality rate. With use of the logistic regression and Probability of Death scores, 52% of residents who had scores of 0.4 or higher died within 6 months, and residents who scored 0.9 or higher had an 80% mortality rate.
Conclusions: MDS assessment data, collected in all U.S. nursing homes, can be useful to increase the specificity of identifying terminal residents. These statistical methods have high and similar prognostic values, but the summative index is easier to understand. It also is more user-friendly and could be incorporated into the nursing home assessment process with minimal additional training, and with no additional hardware or changes in software programming.