Detecting medical prescriptions suspected of fraud using an unsupervised data mining algorithm.

Area of Science:

• Health Informatics
• Data Mining
• Medical Fraud Detection

Background:

• Health insurance fraud incurs significant financial losses.
• Traditional fraud detection methods are often manual and time-consuming.
• Identifying fraudulent medical prescriptions is crucial for cost reduction.

Purpose of the Study:

• To develop and implement an unsupervised data mining algorithm for detecting fraudulent medical prescriptions.
• To assist fraud detection experts by automating the screening process.
• To improve the efficiency and accuracy of identifying suspicious prescriptions.

Main Methods:

• Utilized an unsupervised data mining algorithm for outlier detection.
• Implemented a model analyzing medicine codes, patient sex, and patient age.
• Employed a three-step screening process for medical prescription data.

Main Results:

• The model detects 25% to 100% of non-compliant prescription cases for specific medicines.
• Achieved a sensitivity of 62.16%, specificity of 55.11%, and accuracy of 57.2% in fraud detection.
• Demonstrated data mining's potential for faster and more accurate fraud identification compared to manual inspection.

Conclusions:

• Data mining offers a more efficient approach to detecting potential fraud in medical prescriptions.
• The proposed model reduces investigator workload by pre-screening prescriptions.
• Findings can inform policymakers in developing strategies against healthcare fraud.