Data science process for rule generation/ Using data science to detect claims leakage
Flagship Program: Enabling Information Discovery and Application
The project aims to apply machine-learning and artificial intelligence techniques to improve the utility of administrative health data. Administrative data makes up a major component of the electronic data collected by the health sector. In spite of this, there is widespread acknowledgement that administrative data is of low quality limiting its utility for generating insights by key industry stakeholders such as healthcare organisations, private health insurers and government bodies. This challenge is particularly evident in insurance claims data, which is a form of administrative data that is exchanged between private health insurers (PHIs) and private hospitals in order for healthcare services to be verified and paid for. Although claims data for the same procedure at different hospitals should be consistent, in actuality there is considerable variation limiting its utility for downstream use.
This project is an important first step in addressing the issue of calms data variation. The project will involve developing a repeatable process to generate data insights for detecting Fraud, Abuse, Waste and Errors (FAWE) in claims data. Currently, FAWE is detected manually using expert-driven outlier detection algorithms. This process is cumbersome and results in a high number of false positive symbols. By utilising machine-learning and artificial intelligence techniques this project will develop algorithms to automatically and repeatedly detect FAWE which can be readily translated into proprietary code.
- To develop a rigorous data science-led process to identify variation in claims data precisely and efficiently.
- To test the validity and accuracy of a process for identifying variation in claims data.
- To explore how fragmentation in healthcare data is impeding our ability to utilise administrative data sources to generate better health outcomes