Background. Determining prognosis for nursing home residents is important for care planning, but reliable prediction is difficult. We compared performance of four long-term mortality risk indices for nursing home residents - the Minimum Data Set Mortality Risk Index (MMRI), a recent revision to this index (MMRI-R), and the original and revised Flacker-Kiely models. Methods. We conducted a prospective cohort study in one 92-bed facility in Missouri. Participants were 130 residents who received a Minimum Data Set assessment from May through October, 2007. We collected the Minimum Data Set variables needed to calculate the mortality risk scores. We determined 6- and 12-month mortality for included residents. Using each mortality risk score as the sole independent predictor in logistic models predicting mortality, we determined discrimination (c-statistic) and calibration (Hosmer-Lemeshow goodness-of-fit statistic) for each model. Results. In our sample, discrimination was 0.59 for both the MMRI and the MMRI-R. Discrimination of the original Flacker-Kiely model was 0.69 for both 6 months and 1 year and 0.71 and 0.70, respectively, for the revised model. Model calibration was adequate for all models. Conclusions. Performance of four models that predict long-term mortality of nursing home residents was fair. In our population, the Flacker-Kiely models had similar and markedly better discrimination than either the MMRI or the MMRI-R.
|Number of pages||7|
|Journal||Journals of Gerontology - Series A Biological Sciences and Medical Sciences|
|State||Published - Nov 2010|
- Logistic models
- Risk assessment