Senior Director, Vaccines RWE Pfizer Durham, United States
Background: Frailty algorithms using claims data have been developed and validated in older adults aged ≥65 years. Little is known regarding whether these algorithms can be used to measure frailty in younger adults.
Objectives: To describe the relative prevalence of frailty as measured by a claims-based frailty algorithm in adults aged 18-64 versus 65-74 and to quantify the association between frailty and the 1-year risk of death in each age group.
Methods: Using Optum’s de-identified Clinformatics® Data Mart Database from the US, we identified adults aged 18-74 enrolled on January 1 (index date) in each of the following years: 2019 [pre-COVID period] and 2021 [COVID period]. We required 1 year of continuous enrollment before index and ≥1 claim during that period. We implemented a published frailty algorithm using all claims in the baseline period to create a weighted score. This score was categorized into 5 categories using the following percentiles: ≤10th (Not frail); 11-25th (pre frail); 26-75th (mildly frail); 76-90th (moderately frail); >90th (severely frail). Patients were followed for up to 1 year, starting the day after index until the earliest of death, disenrollment or December 31 of the respective year. Cox proportional hazards models were used to measure adjusted hazard ratios (aHR) between frailty categories and mortality, accounting for demographics, poor health indicators and health seeking behaviors.
Results: There were >5 million and >2 million patients aged 18-64 and 65-74 in each calendar cohort, respectively. Overall, older patients had a higher prevalence of all chronic conditions except for anxiety (15.1% [18-64] vs 13.9% [65-74] in 2019). In 2019, 6% of younger adults and 20% of older adults were classified as severely frail. Severe frailty in younger adults was characterized by high prevalence of immune disorders (79.7%), hypertensive diseases (76.2%), autoimmune syndromes (66.3%), and diabetes (48.5%). Mental health conditions were also common with severe frailty (anxiety: 47.2%; depression: 55.6%). There was a dose response relationship between frailty category and mortality: for younger ages comparing prefrail vs non frail (aHR 1.48 (95% confidence interval [CI]: 1.25-1.77)) compared to severe frail vs non frail (30.58 (95% CI: 26.32-35.53). However, aHRs were lower for the older age group: severe frail vs non frail (9.81 (95% CI: 8.67-11.1). Results were similar in 2021.
Conclusions: After applying a frailty algorithm to younger adults, a dose response relationship was identified with mortality. However, the claims-based algorithm may be measuring a different clinical frailty phenotype than in older adults where the algorithm was validated. Frailty algorithms should be validated in the younger population before applying them more broadly.
