Operationalizing a frailty index from a standardized comprehensive geriatric assessment

J Am Geriatr Soc. 2004 Nov;52(11):1929-33. doi: 10.1111/j.1532-5415.2004.52521.x.

Abstract

Objectives: To construct and validate a frailty index (FI) that is clinically sensible and practical for geriatricians by basing it on a routinely used comprehensive geriatric assessment (CGA) instrument.

Design: Secondary analysis of a 3-month randomized, controlled trial of a specialized mobile geriatric assessment team.

Setting: Rural Nova Scotia. Participants were seen in their homes.

Participants: Frail older adults, of whom 92 were in the intervention group and 77 in the control group.

Measurements: A standard CGA form that accounts for impairment, disability, and comorbidity burden was scored and summed as a frailty index (FI-CGA). The FI-GCA was stratified to describe three levels of frailty. Patients were followed for up to 12 months to determine how well the index predicted adverse outcomes (institutionalization or mortality, whichever came first).

Results: The three levels of frailty were mild (FI-CGA 0-7), moderate (FI-CGA 7-13), and severe (FI-CGA>13). Demographic and social traits were similar across groups, but greater frailty was associated with worse function (r=0.55) and mental status (r=0.33). Those with moderate and severe frailty had a greater risk of adverse outcomes than those with mild frailty (unadjusted hazard ratio=1.9 and 5.5, respectively). There was no difference between frailty groups in mean 3-month goal-attainment scaling scores. Intrarater reliability was 0.95.

Conclusion: The FI-CGA is a valid, reliable, and sensible clinical measure of frailty that permits risk stratification of future adverse outcomes.

Publication types

  • Clinical Trial
  • Randomized Controlled Trial
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Aged
  • Aged, 80 and over
  • Female
  • Frail Elderly*
  • Geriatric Assessment / methods*
  • Humans
  • Male
  • Nova Scotia
  • Proportional Hazards Models
  • ROC Curve