Maintenance Notice

Due to necessary scheduled maintenance, the JMIR Publications website will be unavailable from Wednesday, July 01, 2020 at 8:00 PM to 10:00 PM EST. We apologize in advance for any inconvenience this may cause you.

Who will be affected?

Accepted for/Published in: JMIR Medical Informatics

Date Submitted: Jul 5, 2020
Date Accepted: Dec 12, 2020

The final, peer-reviewed published version of this preprint can be found here:

Machine Learning Approach to Decision Making for Insulin Initiation in Japanese Patients With Type 2 Diabetes (JDDM 58): Model Development and Validation Study

Fujihara K, Matsubayashi Y, Yamada Harada M, Yamamoto M, Iizuka T, Miyamura K, Hasegawa Y, Maegawa H, Kodama S, Yamazaki T, Sone H

Machine Learning Approach to Decision Making for Insulin Initiation in Japanese Patients With Type 2 Diabetes (JDDM 58): Model Development and Validation Study

JMIR Med Inform 2021;9(1):e22148

DOI: 10.2196/22148

PMID: 33502325

PMCID: 7875702

Machine learning approach to decision-making for insulin initiation in Japanese patients with type 2 diabetes: an observational study

  • Kazuya Fujihara; 
  • Yasuhiro Matsubayashi; 
  • Mayuko Yamada Harada; 
  • Masahiko Yamamoto; 
  • Toshihiro Iizuka; 
  • Kosuke Miyamura; 
  • Yoshinori Hasegawa; 
  • Hiroshi Maegawa; 
  • Satoru Kodama; 
  • Tatsuya Yamazaki; 
  • Hirohito Sone

ABSTRACT

Background:

Although some studies have evaluated the applications of machine learning for the early detection of diseases that provided clear-cut diagnostic gold standard, little is known about the usefulness of machine learning for decisions such as specialists’ insulin initiation, which has areas of ambiguity in clinical settings.

Objective:

We examined whether the machine learning approach could support decision-making by general physicians for insulin initiation for type 2 diabetes.

Methods:

Prescription data on antidiabetic agents from 2009 to 2015 from diabetes specialists’ patient registries were used to compare machine-learning decisions on insulin initiation to decisions made by general physicians. Gold standard was defined as 80% agreement among nine diabetes specialists based on patients’ information. Twenty-two general physicians chose the most suitable medication according to patients’ information. Analyzed were data on 4,860 patients who received initial monotherapy with either insulin (293 patients) or non-insulin (4,567 patients) and had laboratory data. Neural network output was insulin initiation ranging from 0 to 1 with a cutoff >0.5 for the dichotomous classification. Accuracy was calculated by 5-fold cross validation.

Results:

Average predictive values for insulin initiation by machine learning were 0.40, 0.50 and 0.82, respectively, for down sampling ratios of 1:2, 1:4 and 1:8. General physicians achieved a mean accuracy for insulin initiation of 51%. Overall accuracy was 43%, 57% and 86% for down sampling ratios of 1:2, 1:4 and 1:8, respectively. Difference in accuracy between general physicians and machine learning with the down sampling ratio of 1:8 was significant (P<0.05).

Conclusions:

In this preliminary study, accuracy of machine learning with the down sampling ratio of 1:8 was higher than that of general physicians for insulin initiation defined by diabetes specialists’ choice as the gold standard. Assistance by machine learning may benefit general physicians providing primary care in deciding upon insulin initiation. Clinical Trial: N/A


 Citation

Please cite as:

Fujihara K, Matsubayashi Y, Yamada Harada M, Yamamoto M, Iizuka T, Miyamura K, Hasegawa Y, Maegawa H, Kodama S, Yamazaki T, Sone H

Machine Learning Approach to Decision Making for Insulin Initiation in Japanese Patients With Type 2 Diabetes (JDDM 58): Model Development and Validation Study

JMIR Med Inform 2021;9(1):e22148

DOI: 10.2196/22148

PMID: 33502325

PMCID: 7875702

Download PDF


Request queued. Please wait while the file is being generated. It may take some time.

© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.