Translation of a Medicare Claims-based Frailty Index from International Classification of Diseases 9th Revision to 10th Revision

Background: The Faurot frailty index is a validated algorithm using demographic, enrollment, and International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)-based billing information in Medicare claims data as a proxy for frailty. This algorithm has been shown to predict geriatric-relevant outcomes. In October 2015, the US healthcare system transitioned from ICD-9-CM to ICD-10-CM.

Objectives: To update the Faurot frailty index for the ICD-10-CM era and assess the validity of the updated index after the transition.

Methods: We used a random sample of Medicare fee-for-service claims and enrollment data from 2013-2017. Applying General Equivalence Mappings (GEMs) and a forward-backward approach, we translated the code lists for diagnosis-based frailty indicators (e.g., arthritis, dementia) from ICD-9-CM to ICD-10-CM and subsequently manually reviewed code lists. We plotted monthly period prevalence of each frailty indicator and the distribution of the claims-based predicted probability of frailty before and after the ICD transition. To evaluate the association between the updated Faurot frailty index and 1-year risk of key geriatric outcomes (mortality, hospitalization, skilled nursing facility admission), we identified beneficiaries enrolled on Jan 1, 2017 who were continuously enrolled during an 8-month lookback (May-Dec 2016). We assessed frailty indicators and predicted probability of frailty during the lookback, started outcome follow-up on Jan 1, 2017, and estimated 1-year risks for each outcome.

Results: Updated indicators based on GEMs had similar prevalence compared to pre-transition indicator definitions. The exception was rehabilitation services, for which we added Current Procedural Terminology procedure codes to the algorithm to improve capture. The median and interquartile range (IQR) for predicted probability of frailty were similar before and after the ICD transition (pre-transition median [IQR]: 0.034 [0.02-0.07]; post-transition: 0.038 [0.02-0.09]). Using the updated Faurot frailty index, predicted probability of frailty (categorized as < 0.05, 0.05- < 0.1, 0.1- < 0.2, 0.2- < 0.4, ≥0.4) was positively associated with 1-year mortality risk (1.7%, 5.3%, 10.2%, 17.8%, 31.2%). Higher predicted probability of frailty was also associated with increased 1-year risk of hospitalization and skilled nursing facility admission. These associations were similar to those during the ICD-9-CM era.

Conclusions: We updated the Faurot frailty index for the ICD-10-CM era and confirmed its associations with key geriatric-relevant outcomes. Studies of drug effects in older adults should use validated indices, such as the Faurot index, to account for confounding or to assess effect measure modification by frailty.