Identification of predictors for discontinuation in patients with diabetes mellitus for policy proposal

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K18957
• 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: The University of Tokyo
• Principal Investigator: Okada Akira 東京大学, 医学部附属病院, 特任講師 (70847574)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 糖尿病 / 重症化予防 / 臨床疫学 / 受診中断

Outline of Final Research Achievements

We conducted studies to help develop health care policy regarding early intervention and continuation of treatment for diabetes mellitus.
We conducted a model building study to predict whether people who met the diagnostic criteria for diabetes at their first medical checkup would receive medical care after a recommendation to see a doctor. A model created by machine learning with four variables had better predictive performance than an existing predictive model using 13 factors (Diabetes Care. 2022 Jun 2;45 (6):1346-1354.) The group that received guideline-recommended treatment after the first visit was less likely to discontinue subsequent visits than the group that did not, even after adjusting for confounders and other factors (J Diabetes Investig. 2021 Sep;12(9):1619-1631.)

Free Research Field

臨床疫学

Academic Significance and Societal Importance of the Research Achievements

これまでは、受診しない人はそもそも解析対象にならないことが多かったが、保険者ベースのレセプトデータベースを用いることにより、研究可能となりかつ機械学習を用いることで効率的な介入の可能性を示すことが出来た。以上のことから、より受診中断に重要な方法を模索した研究は存在したが、どうしても積極的に研究に参加するという選択バイアスが強かったが、レセプトデータベースを用い、疫学的手法を駆使して有効である可能性が高い介入方法を見いだした。シンプルな結果であり、政策立案に取り込まれる可能性もあり、社会的な意義は大きい。