Cancer Medicine (Sep 2021)

A simple test‐based frailty index to predict survival among cancer patients with an unplanned hospitalization: An observational cohort study

  • Timothy Hembree,
  • Olga Theou,
  • Sarah Thirlwell,
  • Richard R. Reich,
  • Biwei Cao,
  • Marina Sehovic,
  • Misbahuddin Syed,
  • Neha Verma,
  • Thu‐Cuc Nguyen,
  • Dinesh Keerty,
  • Jaqueline Wesolow,
  • Viktoriya Koverzhenko,
  • Martine Extermann,
  • Jessica Huang,
  • Asha Ramsakal

DOI
https://doi.org/10.1002/cam4.4107
Journal volume & issue
Vol. 10, no. 17
pp. 5765 – 5774

Abstract

Read online

Abstract Background Frailty is a state of increased vulnerability to stressors, and predicts risk of adverse outcomes, such as mortality. Frailty can be defined by a frailty index (FI) using an accumulation of deficits approach. An FI comprised of 20 items derived from our previously studied test‐based frailty index (TBFI) and an additional 33 survey‐based elements sourced from the standard CGA was developed to evaluate if predictive validity of survival was improved. Methods One hundred eighty‐nine cancer patients during acute hospitalization were consented between September 2018 and May 2019. Frailty scores were calculated, and patients were categorized into four groups: non‐frail (0–0.2), mildly frail (0.2–0.3), moderately frail (0.3–0.4), and severely frail (>0.4). Patients were followed for 1‐year to assess FI and TBFI prediction of survival. Area under the curve (AUC) statistics from ROC analyses were compared for the FI versus TBFI. Results Increasing frailty was similarly associated with increased risk of mortality (HR, 4.5 [95% CI, 2.519–8.075] and HR, 4.1 [95%CI, 1.692–9.942]) and the likelihood of death at 6 months was about 11‐fold (odds ratio, 10.9 [95% CI, 3.97–33.24]) and 9.73‐fold (95% CI, 2.85–38.50) higher for severely frail patients compared to non‐frail patients for FI and TBFI, respectively. This association was independent of age and type of cancer. The FI and TBFI were predictive of survival for older and younger cancer patients with no significant differences between models in discriminating survival (FI AUC, 0.747 [95% CI, 0.6772–0.8157] and TBFI AUC, 0.724 [95% CI, 0.6513–0.7957]). Conclusions The TBFI was predictive of survival, and the addition of an in‐person assessment (FI) did not greatly improve predictive validity. Increasing frailty, as measured by a TBFI, resulted in a meaningfully increased risk of mortality and may be well‐suited for screening of hospitalized cancer patients.

Keywords