Frailty measure may improve Medicare cost predictions

Adding a claims-based frailty index to the CMS Hierarchical Condition Category (CMS-HCC) model may improve predictions of patients' annualized Medicare costs in value-based payment programs, according to a recent study.

Researchers performed a retrospective cohort study using data from the Medicare Current Beneficiary Survey linked to Medicare claims from 2006 to 2013. A validated claims-based frailty index was used to classify frailty status and was added to the CMS Hierarchical Condition Category model. The association between the index and annualized Medicare costs was examined, and observed Medicare costs versus predictions based on the standard CMS-HCC model both with and without the frailty index were compared. The results of the study were published April 7 by Annals of Internal Medicine.

The study included data on 16,535 community-dwelling, fee-for-service beneficiaries, encompassing 26,705 patient-years. Each patient had a baseline year, defined as at least one year of continuous enrollment in Medicare Parts A and B from which costs could be predicted, and an evaluation year, defined as at least one year of continuous enrollment or death during the following year when actual costs were measured. Of these patients, 8,910 (46.4% of patient-years) were classified as robust, 8,405 (41.6% of patient-years) were classified as prefrail, 2,215 (9.6% of patient-years) were classified as mildly frail, and 593 (2.5% of patient-years) were classified as moderately to severely frail.

Mean costs for each frailty category were $5,274, $12,462, $26,239, and $44,586, respectively. When the frailty index was added to the CMS-HCC model, it predicted average additional costs of $2,712, $7,915, and $16,449 for prefrail, mildly frail, and moderately to severely frail patients, respectively. Cost predictions were more accurate on average at all levels of frailty when the model that included the frailty measure was used. Observed costs, however, were distributed more widely than predicted with the enhanced model at all frailty levels.

The researchers noted that they looked only at Medicare fee-for-service beneficiaries who participated in the Medicare Current Beneficiary Survey and that the claims-based index is only a proxy for frailty. However, they concluded that average cost predictions with the CMS-HCC model appear to be more accurate when a claims-based measure of frailty is included. “Our findings suggest that the [claims-based index] addition to the CMS-HCC model may improve the accuracy, fairness, and equity of current [value-based payment] programs and thus deserves consideration by CMS and other policymakers,” the authors wrote. “However, because individual-level cost prediction remains poor even under the augmented models, ongoing advancements in methodology are needed.”

The authors of an accompanying editorial said it was not surprising that observed costs were wider than the predicted range across all frailty grades and said that variability in risk among patients with the same level of frailty could be linked to other factors, such as acuity of illness or variations in practice. They also noted that under the current model, hospitals are likely to receive less reimbursement for comprehensive geriatric care than they would under a model that adjusts for frailty.

“Because incorporation of the [claims-based frailty index] improved model performance in predicting cost, value-based payment models could reward providers, nonarbitrarily, for rendering services to frail patients,” the editorialists wrote. “Given that the number of older persons living with frailty is about to greatly increase—those in the leading edge of the baby boom turn 75 in 2021—time is running out to stop the financial disincentive against providers who offer them care and turn toward value-based payments that recognize what otherwise is unmeasured value.”