Publications

BACKGROUND: Frailty is increasingly recognized as a useful measure of vulnerability in older adults. Multiple claims-based frailty indices (CFIs) can readily identify individuals with frailty, but whether 1 CFI improves prediction over another is unknown. We sought to assess the ability of 5 distinct CFIs to predict long-term institutionalization (LTI) and mortality in older Veterans.

METHODS: Retrospective study conducted in U.S. Veterans ≥65 years without prior LTI or hospice use in 2014. Five CFIs were compared: Kim, Orkaby (Veteran Affairs Frailty Index [VAFI]), Segal, Figueroa, and the JEN-FI, grounded in different theories of frailty: Rockwood cumulative deficit (Kim and VAFI), Fried physical phenotype (Segal), or expert opinion (Figueroa and JFI). The prevalence of frailty according to each CFI was compared. CFI performance for the coprimary outcomes of any LTI or mortality from 2015 to 2017 was examined. Because Segal and Kim include age, sex, or prior utilization, these variables were added to regression models to compare all 5 CFIs. Logistic regression was used to calculate model discrimination and calibration for both outcomes.

RESULTS: A total of 3 million Veterans were included (mean age 75, 98% male participants, 80% White, and 9% Black). Frailty was identified for between 6.8% and 25.7% of the cohort with 2.6% identified as frail by all 5 CFIs. There was no meaningful difference between CFIs in the area under the receiver operating characteristic curve for LTI (0.78-0.80) or mortality (0.77-0.79).

CONCLUSIONS: Based on different frailty constructs, and identifying different subsets of the population, all 5 CFIs similarly predicted LTI or death, suggesting each could be used for prediction or analytics.

Background Mortality prediction in critically ill patients with cardiogenic shock can guide triage and selection of potentially high-risk treatment options. Methods and Results We developed and externally validated a checklist risk score to predict in-hospital mortality among adults admitted to the cardiac intensive care unit with Society for Cardiovascular Angiography & Interventions Shock Stage C or greater cardiogenic shock using 2 real-world data sets and Risk-Calibrated Super-sparse Linear Integer Modeling (RiskSLIM). We compared this model to those developed using conventional penalized logistic regression and published cardiogenic shock and intensive care unit mortality prediction models. There were 8815 patients in our training cohort (in-hospital mortality 13.4%) and 2237 patients in our validation cohort (in-hospital mortality 22.8%), and there were 39 candidate predictor variables. The final risk score (termed BOS,MA2) included maximum blood urea nitrogen ≥25 mg/dL, minimum oxygen saturation <88%, minimum systolic blood pressure <80 mm Hg, use of mechanical ventilation, age ≥60 years, and maximum anion gap ≥14 mmol/L, based on values recorded during the first 24 hours of intensive care unit stay. Predicted in-hospital mortality ranged from 0.5% for a score of 0 to 70.2% for a score of 6. The area under the receiver operating curve was 0.83 (0.82-0.84) in training and 0.76 (0.73-0.78) in validation, and the expected calibration error was 0.9% in training and 2.6% in validation. Conclusions Developed using a novel machine learning method and the largest cardiogenic shock cohorts among published models, BOS,MA2 is a simple, clinically interpretable risk score that has improved performance compared with existing cardiogenic-shock risk scores and better calibration than general intensive care unit risk scores.

Estimating the likelihood of urgent mechanical circulatory support (MCS) can facilitate procedural planning and clinical decision-making in chronic total occlusion (CTO) percutaneous coronary intervention (PCI). We analyzed 2,784 CTO PCIs performed between 2012 and 2021 at 12 centers. The variable importance was estimated by a bootstrap applying a random forest algorithm to a propensity-matched sample (a ratio of 1:5 matching cases with controls on center). The identified variables were used to predict the risk of urgent MCS. The performance of the risk model was assessed in-sample and on 2,411 out-of-sample procedures that did not require urgent MCS. Urgent MCS was used in 62 (2.2%) of cases. Patients who required urgent MCS were older (70 [63 to 77] vs 66 [58 to 73] years, p = 0.003) compared with those who did not require urgent MCS. Technical (68% vs 87%, p <0.001) and procedural success (40% vs 85%, p <0.001) was lower in the urgent MCS group compared with cases that did not require urgent MCS. The risk model for urgent MCS use included retrograde crossing strategy, left ventricular ejection fraction, and lesion length. The resulting model demonstrated good calibration and discriminatory capacity with the area under the curve (95% confidence interval) of 0.79 (0.73 to 0.86) and specificity and sensitivity of 86% and 52%, respectively. In the out-of-sample set, the specificity of the model was 87%. The Prospective Global Registry for the Study of Chronic Total Occlusion Intervention CTO MCS score can help estimate the risk of urgent MCS use during CTO PCI.

Clinician payment is transitioning from fee-for-service to value-based payment, with reimbursement tied to health care quality and cost. However, the overarching goals of value-based payment-to improve health care quality, lower costs, or both-have been largely unmet. This policy statement reviews the current state of value-based payment and provides recommended best practices for future design and implementation. The policy statement is divided into sections that detail different aspects of value-based payment: (1) key program design features (patient population, quality measurement, cost measurement, and risk adjustment), (2) the role of equity during design and evaluation, (3) adjustment of payment, and (4) program implementation and evaluation. Each section introduces the topic, describes important considerations, and lists examples from existing programs. Each section includes recommended best practices for future program design. The policy statement highlights 4 key themes for successful value-based payment. First, programs should carefully weigh the incentives between lowering cost and improving quality of care and ensure that there is adequate focus on quality of care. Second, the expansion of value-based payment should be a tool for improving equity, which is central to quality of care and should be a focal point of program design and evaluation. Third, value-based payment should continue to move away from fee for service toward more flexible funding that allows clinicians to focus resources on the interventions that best help patients. Last, successful programs should find ways to channel clinicians' intrinsic motivation to improve their performance and the care for their patients. These principles should guide the future development of clinician value-based payment models.

BACKGROUND: Federal programs measuring hospital quality of care for acute cardiovascular conditions are based solely on Medicare fee-for-service (FFS) beneficiaries, and exclude Medicare Advantage (MA) beneficiaries. In this study we characterize the proportion of Medicare beneficiaries enrolled in MA at the time of acute myocardial infarction (AMI), heart failure (HF), and ischemic stroke hospitalization.

METHODS: Retrospective cross-sectional study of short-term acute care hospitals using Medicare claims in 2009 and 2019.

RESULTS: There were 2,653 hospitals in 2009 and 2,732 hospitals in 2019. Across hospitals, the proportion of Medicare beneficiaries hospitalized for AMI who were enrolled in MA increased between 2009 (hospital-level median 14.4% [IQR 5.1%-26.0%]) and 2019 (33.3% [IQR 20.6%-45.2%]), with substantial variation across hospitals. Similar patterns were observed for HF (13.0% [IQR 5.3%-24.3%] to 31.0% [IQR 20.2%-42.3%]) and ischemic stroke (14.6% [IQR 5.3%-26.7%] to 33.3% [IQR 20.9%-46.0%]). Within each hospital referral region, hospital size (large 36.3% vs small 24.5%; adjusted difference 6.7%, 95% CI, 4.5%-8.8%), teaching status (teaching 34.5% vs nonteaching 28.2%; 2.8%, 1.4%-4.1%), and ownership status (private nonprofit 32.3% vs public 24.5%; 5.2%, 3.5%-6.9%) were each associated with a higher hospital MA proportion.

CONCLUSIONS: The proportion of Medicare beneficiaries hospitalized for AMI, HF, and ischemic stroke enrolled in MA doubled between 2009 and 2019, with substantial variation across hospitals. These findings have implications for federal efforts to measure and improve quality, which currently focus only on FFS beneficiaries.

Background Frailty is rarely assessed in clinical trials of patients who receive dual antiplatelet therapy (DAPT) after percutaneous coronary intervention. This study investigated whether frailty defined using claims data is associated with outcomes following percutaneous coronary intervention, and if there is a differential association in patients receiving standard versus extended duration DAPT. Methods and Results Patients ≥65 years of age in the DAPT (Dual Antiplatelet Therapy) Study, a randomized trial comparing 30 versus 12 months of DAPT following percutaneous coronary intervention, had data linked to Medicare claims (n=1326), and a previously validated claims-based index was used to define frailty. Net adverse clinical events, a composite of all-cause mortality, myocardial infarction, stroke, and major bleeding, were compared between frail and nonfrail patients. Patients defined as frail using claims data (12.0% of the cohort) had higher incidence of net adverse clinical events (23.1%) compared with nonfrail patients (10.7%; P<0.001) at 18-month follow-up and increased risk after multivariable adjustment (adjusted hazard ratio [HR], 2.24 [95% CI, 1.38-3.63]). There were no differences in effects of extended duration DAPT on net adverse clinical events for frail (HR, 1.42 [95% CI, 0.73-2.75]) and nonfrail patients (HR, 1.18 [95% CI, 0.83-1.68]; interaction P=0.61), although analyses were underpowered. Bleeding was highest among frail patients who received extended duration DAPT. Conclusions Among older patients in the DAPT Study, claims-defined frailty was associated with higher net adverse clinical events. Effects of extended duration DAPT were not different for frail patients, although comparisons were underpowered. Further investigation of how frailty influences ischemic and bleeding risks with DAPT are warranted. Registration URL: https://www.clinicaltrials.gov; Unique identifier: NCT00977938.

The rich longitudinal individual level data available from electronic health records (EHRs) can be used to examine treatment effect heterogeneity. However, estimating treatment effects using EHR data poses several challenges, including time-varying confounding, repeated and temporally non-aligned measurements of covariates, treatment assignments and outcomes, and loss-to-follow-up due to dropout. Here, we develop the subgroup discovery for longitudinal data algorithm, a tree-based algorithm for discovering subgroups with heterogeneous treatment effects using longitudinal data by combining the generalized interaction tree algorithm, a general data-driven method for subgroup discovery, with longitudinal targeted maximum likelihood estimation. We apply the algorithm to EHR data to discover subgroups of people living with human immunodeficiency virus who are at higher risk of weight gain when receiving dolutegravir (DTG)-containing antiretroviral therapies (ARTs) versus when receiving non-DTG-containing ARTs.