Home > Research > Publications & Outputs > Identifying fraud in medical insurance based on...


Text available via DOI:

View graph of relations

Identifying fraud in medical insurance based on blockchain and deep learning

Research output: Contribution to Journal/MagazineJournal articlepeer-review

  • Guoming Zhang
  • Xuyun Zhang
  • Muhammad Bilal
  • Wanchun Dou
  • Xiaolong Xu
  • Joel J.P.C. Rodrigues
<mark>Journal publication date</mark>31/05/2022
<mark>Journal</mark>Future Generation Computer Systems
Number of pages15
Pages (from-to)140-154
Publication StatusPublished
Early online date16/12/21
<mark>Original language</mark>English


With the rapid growth of medical costs, the control of medical expenses has been becoming an important task of Health Insurance Department. Traditional medical insurance settlement is paid on a per-service basis, which leads to lots of unreasonable expenses. To cope with this problem, the single-disease payment mechanism has been widely used in recent years. However, the single-disease payment also has a risk of fraud. In this work, we propose a framework to identify fraud of medical insurance based on consortium blockchain and deep learning, which can recognize suspicious medical records automatically to ensure valid implementation on single-disease payment and lighten the work of medical insurance auditors. An explainable model BERT-LE is designed to evaluate the reasonability of ICD disease code for Medicare reimbursement by predicting the probability of a disease according to the chief complaint of a patient. We also put forward a storage and management process of medical records based on consortium blockchain to ensure the security, immutability, traceability, and auditability of the data. The experiments on two real datasets from two 3A hospitals demonstrate that the proposed solution can identify fraud effectively and greatly improve the efficiency in medical insurance reviews.