Accuracy evaluation of predicted disease state transition by Markov model and its application to cost-effectiveness analysis

2020 Fiscal Year Final Research Report

• Project/Area Number: 18K17381
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 58030:Hygiene and public health-related: excluding laboratory approach
• Research Institution: Hiroshima University
• Principal Investigator: Akita Tomoyuki 広島大学, 医系科学研究科(医), 講師 (80609925)
• Project Period (FY): 2018-04-01 – 2021-03-31
• Keywords: マルコフモデル / 推定精度 / 費用効果分析 / 漸近論

Outline of Final Research Achievements

Markov model is used to estimate the disease progression and cost-effectiveness analysis. This model predicts future pathological progression based on the "transition probability" calculated by data from epidemiological/clinical research data, but so far. few studies examined that the number of data from the original research has affected the prediction accuracy.
In this study, we have developed a formula to evaluate the prediction accuracy by Markov model in the "confidence interval". Next, a numerical simulation was performed to examine the number of data and the accuracy of the confidence interval (covering probability). It was also compared with the existing formula confidence intervals in a special case of the Markov model (two states irreversible). Furthermore, based on this formula, the estimation accuracy of cost and effect in concrete cost-effectiveness analysis was examined.

Free Research Field

数理疫学

Academic Significance and Societal Importance of the Research Achievements

検診・治療等の疾病対策導入を検討するとき、検診を導入した/導入しなかった場合の、生涯にかかる費用とQOLをそれぞれ見積り、導入に要した費用に似合うだけのQOL改善が見込まれるのかが評価されている。方法の一つであるマルコフモデルは、実際の疫学・臨床研究のデータから、1年間の疾患の発症率や進行率を出して、それをもとに仮想的に病態進行をシミュレーションを行う。元のデータの対象者数が少ない場合、予測の精度がよくないと考えられるが、これまでの研究では、ほとんど考慮されていない。そこで本研究では、予測の精度を「信頼区間」として表現するための公式を開発し、その方法の妥当性を理論と実用の両面から検討した。