The Impact of Frailty and Comorbidity on Institutionalization and Mortality in Persons With Dementia: A Prospective Cohort Study

Contains fulltext : 202861.pdf (Publisher’s version ) (Closed access)OBJECTIVES: The predictive value of frailty and comorbidity, in addition to more readily available information, is not widely studied. We determined the incremental predictive value of frailty and comorbidity for mortality and institutionalization across both short and long prediction periods in persons with dementia. DESIGN: Longitudinal clinical cohort study with a follow-up of institutionalization and mortality occurrence across 7 years after baseline. SETTING AND PARTICIPANTS: 331 newly diagnosed dementia patients, originating from 3 Alzheimer centers (Amsterdam, Maastricht, and Nijmegen) in the Netherlands, contributed to the Clinical Course of Cognition and Comorbidity (4C) Study. MEASURES: We measured comorbidity burden using the Cumulative Illness Rating Scale for Geriatrics (CIRS-G) and constructed a Frailty Index (FI) based on 35 items. Time-to-death and time-to-institutionalization from dementia diagnosis onward were verified through linkage to the Dutch population registry. RESULTS: After 7 years, 131 patients were institutionalized and 160 patients had died. Compared with a previously developed prediction model for survival in dementia, our Cox regression model showed a significant improvement in model concordance (U) after the addition of baseline CIRS-G or FI when examining mortality across 3 years (FI: U = 0.178, P = .005, CIRS-G: U = 0.180, P = .012), but not for mortality across 6 years (FI: U = 0.068, P = .176, CIRS-G: U = 0.084, P = .119). In a competing risk regression model for time-to-institutionalization, baseline CIRS-G and FI did not improve the prediction across any of the periods. CONCLUSIONS: Characteristics such as frailty and comorbidity change over time and therefore their predictive value is likely maximized in the short term. These results call for a shift in our approach to prognostic modeling for chronic diseases, focusing on yearly predictions rather than a single prediction across multiple years. Our findings underline the importance of considering possible fluctuations in predictors over time by performing regular longitudinal assessments in future studies as well as in clinical practice