Enhancing Heart Failure (HF) Management—Coupling Predictive Model with Clinical Insights Using Administrative Data

Author(s)

Tan H1, Zhou S2, Desai R3, Evers T4, Power TP5, Smith J6, Willey V6
1HealthCore, Inc., San Diego, CA, USA, 2IngenioRx, Morristown, NJ, USA, 3Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA, 4Bayer AG, Wuppertal, Germany, 5AIM Specialty Health, Chicago, IL, USA, 6HealthCore, Inc., Wilmington, DE, USA

Presentation Documents

OBJECTIVES:

Many predictive HF models exist, but most have used only a small number of important clinical predictors to compute risk score. While useful during individual patient-clinician encounters, these models have limited utility to inform and identify population-level interventions. This study aimed to develop and internally validate predictive models of high-risk HF patients to identify a potential target population for intensified interventions or management adjustment in a large United States (U.S.) managed care population.

METHODS:

This observational cohort study used retrospective administrative claims data (January 1, 2016 to May 31, 2019) from the HealthCore Integrated Research Database, representing over 50 million members across the U.S. The cohort was randomly divided (7:3 ratio) into training and test datasets. Three predictive models were developed with the training dataset using least absolute shrinkage and selection operator (LASSO) logistic regression with 10-fold cross-validation to predict 1-year: HF hospitalization risk, top decile total costs, and top quartile total costs. The test dataset was used to assess model performance characteristics using receiver operating characteristics (ROC) and calibration.

RESULTS:

The study population of 70,926 adult HF patients was 53.7% male with a mean (SD) age of 67.5 (14.0) years. All three models demonstrated good performance: area under the ROC curve (AUC) of 0.704, 0.794, and 0.747 for HF hospitalization, top decile, and top quartile total cost models, respectively. The top quartile total cost model was recommended for its balanced predictive performance and clinical relevance.

CONCLUSIONS: While not all individuals in the top quartile cost cohort may be optimal for HF intervention, incorporating clinical insights based on the selected intervention may identify the most likely beneficiaries. A combination of advanced analytics with clinical insights tailored for a specific intervention maximizes the potential value of predictive modelling by better targeting the population most likely to benefit.

Conference/Value in Health Info

2023-05, ISPOR 2023, Boston, MA, USA

Value in Health, Volume 26, Issue 6, S2 (June 2023)

Code

HSD95

Topic

Methodological & Statistical Research, Study Approaches

Topic Subcategory

Artificial Intelligence, Machine Learning, Predictive Analytics

Disease

No Additional Disease & Conditions/Specialized Treatment Areas

Explore Related HEOR by Topic


Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×