Estimating Healthcare Costs Associated With Ischemic Heart Disease and Stroke Using Several Different Statistical Models
Author(s)
Koki Idehara, PhD1, Fei Zhao, MSc2, Seok-Won Kim, PhD2.
1Real-World Evidence Solutions/HEOR, IQVIA Solutions Japan G. K, Tokyo, Japan, 2Real-World Evidence Solutions/HEOR, IQVIA Solutions Japan G.K., Tokyo, Japan.
1Real-World Evidence Solutions/HEOR, IQVIA Solutions Japan G. K, Tokyo, Japan, 2Real-World Evidence Solutions/HEOR, IQVIA Solutions Japan G.K., Tokyo, Japan.
OBJECTIVES: As cost-effectiveness analysis gains broader application in healthcare decision-making, healthcare cost estimations have become increasingly important. Especially in Japan, where a fee-for-service payment system is employed as remuneration for medical services provided, healthcare costs associated with diseases are often estimated utilizing real-world data. Nevertheless, challenges remain in robustly and accurately capturing relevant costs while ensuring reliability. This study compared the performance of different regression models for estimating costs for cardio- and cerebrovascular events and their follow-up using a large-scale administrative database covering all ages, called the IQVIA Claims Database.
METHODS: This study investigated model performance by assessing prediction accuracies between observed and predicted costs based on various regression models, including linear, linear with log-transformed costs, gamma, Poisson, negative binomial, and median regression models, for event and the follow-up costs associated with the first hospitalization of acute myocardial infarction (AMI), stroke, and heart failure (HF) occurring between October 2022 to August 2024. The models were developed using two-thirds of each population, and performances were measured using remaining one-third. Demographics and prevalent comorbidities of patients experiencing each event were included as covariates. The analysis was conducted from a Japanese public healthcare perspective.
RESULTS: Among the overall IQVIA Claims population, 6,236 patients experienced AMI, 20,622 experienced ischemic stroke, 9,166 experienced hemorrhagic stroke, and 43,287 experienced HF. The observed event costs were ¥3,403,180 ($23,151) for AMI, ¥3,055,491 ($20,786) for ischemic stroke, ¥5,013,220 ($34,104) for hemorrhagic stroke, and ¥3,316,595 ($22,562) for HF on average. For the event costs, the Poisson model had the lowest mean squared errors (MSE) for ischemic stroke and heart failure, while the negative binomial model presented the lowest MSE for AMI and hemorrhagic stroke.
CONCLUSIONS: According to the prediction accuracies in different cardio- and cerebrovascular events, the optimal regression model for predicting healthcare costs may vary depending on the diseases and scope of costs.
METHODS: This study investigated model performance by assessing prediction accuracies between observed and predicted costs based on various regression models, including linear, linear with log-transformed costs, gamma, Poisson, negative binomial, and median regression models, for event and the follow-up costs associated with the first hospitalization of acute myocardial infarction (AMI), stroke, and heart failure (HF) occurring between October 2022 to August 2024. The models were developed using two-thirds of each population, and performances were measured using remaining one-third. Demographics and prevalent comorbidities of patients experiencing each event were included as covariates. The analysis was conducted from a Japanese public healthcare perspective.
RESULTS: Among the overall IQVIA Claims population, 6,236 patients experienced AMI, 20,622 experienced ischemic stroke, 9,166 experienced hemorrhagic stroke, and 43,287 experienced HF. The observed event costs were ¥3,403,180 ($23,151) for AMI, ¥3,055,491 ($20,786) for ischemic stroke, ¥5,013,220 ($34,104) for hemorrhagic stroke, and ¥3,316,595 ($22,562) for HF on average. For the event costs, the Poisson model had the lowest mean squared errors (MSE) for ischemic stroke and heart failure, while the negative binomial model presented the lowest MSE for AMI and hemorrhagic stroke.
CONCLUSIONS: According to the prediction accuracies in different cardio- and cerebrovascular events, the optimal regression model for predicting healthcare costs may vary depending on the diseases and scope of costs.
Conference/Value in Health Info
2025-09, ISPOR Real-World Evidence Summit 2025, Tokyo, Japan
Value in Health Regional, Volume 49S (September 2025)
Code
RWD96
Topic Subcategory
Health & Insurance Records Systems
Disease
SDC: Cardiovascular Disorders (including MI, Stroke, Circulatory)