Assessing Risk for Adverse Outcomes in Older Adults: The Need to Include Both Physical Frailty and Cognition

Abstract

Background: Physical frailty is a powerful tool for identifying nondisabled individuals at high risk of adverse outcomes. The extent to which cognitive impairment in those without dementia adds value to physical frailty in detecting high-risk individuals remains unclear.

Objectives: To estimate the effects of combining physical frailty and cognitive impairment without dementia (CIND) on the risk of basic activities of daily living (ADL) dependence and death over 8 years.

Design: Prospective cohort study.

Setting: The Health and Retirement Study (HRS).

Participants: A total of 7338 community-dwelling people, 65 years or older, without dementia and ADL dependence at baseline (2006-2008). Follow-up assessments occurred every 2 years until 2014.

Measurements: The five components of the Cardiovascular Health Study defined physical frailty. A well-validated HRS method, including verbal recall, series of subtractions, and backward count task, assessed cognition. Primary outcomes were time to ADL dependence and death. Hazard models, considering death as a competing risk, associated physical frailty and CIND with outcomes after adjusting for sociodemographics, comorbidities, depression, and smoking status.

Results: The prevalence of physical frailty was 15%; CIND, 19%; and both deficits, 5%. In unadjusted and adjusted analyses, combining these factors identified older adults at an escalating risk for ADL dependence (no deficit = 14% [reference group]; only CIND = 26%, sub-hazard ratio [sHR] = 1.5, 95% confidence interval [CI] = 1.3-1.8; only frail = 33%, sHR = 1.7, 95% CI = 1.4-2.0; both deficits = 46%, sHR = 2.0, 95%CI = 1.6-2.6) and death (no deficit = 21%; only CIND = 41%, HR = 1.6, 95% CI = 1.4-1.9; only frail = 56%, HR = 2.2, 95% CI = 1.7-2.7; both deficits = 66%, HR = 2.6, 95% CI = 2.0-3.3) over 8-year follow-up. Adding the cognitive measure to models that already included physical frailty alone increased accuracy in identifying those at higher risk of ADL dependence (Harrell's concordance [C], 0.74 vs 0.71; P < .001) and death (Harrell's C, 0.70 vs 0.67; P < .001).

Conclusion: Physical frailty and CIND are independent predictors of incident disability and death. Because together physical frailty and CIND identify vulnerable older adults better, optimal risk assessment should supplement measures of physical frailty with measures of cognitive function. J Am Geriatr Soc 67:477-483, 2019.

Keywords: cognitive frailty; community-dwelling older people; disability; interaction; mortality.

Figure 1

Cumulative Incidence of (A) Incident ADL Dependence and (B) Death by Combining Physical Frailty and Cognitive Impairment at Baseline (n=7338) For ADL dependence, the curves were computed using the Fine and Gray method that considered the competing risk of death. For mortality, the curves were computed using the Kaplan-Meier estimates. ADL=activities of daily living (i.e., eating, transferring, walking across the room, dressing, toileting, and bathing); CIND=cognitive impairment without dementia.