Combining the Power of Artificial Intelligence with the Richness of Healthcare Claims Data: Opportunities and Challenges
PharmacoEconomics, March 8, 2019
Advances in artificial intelligence (AI) and computing power allow researchers to uncover previously undetected patterns in large datasets. One example of such large datasets is patient claims data, which contain information about patients’ treatments, diagnoses, and prescriptions, and can in some instances be linked with other data sources, such as census data and electronic medical records (EMRs). In “Combining the Power of Artificial Intelligence with the Richness of Healthcare Claims Data: Opportunities and Challenges,” published in the journal PharmacoEconomics, six authors – including Analysis Group Managing Principals Lisa Pinheiro and Paul Greenberg, Director of Data Science Razvan Veliche, and Vice President Nick Dadson – introduce readers to recent and potential applications of AI to claims data, and demonstrate some of the ways they can be harnessed to improve various aspects of health care.
For example, claims data can be linked to laboratory information and physician notes to help predict the future onset of diseases, and therefore support prevention. Claims data can also be connected to EMR data to discover patterns of adverse effects from various treatments. In still another example, AI approaches can combine claims data with lab values, specialist visits, and physician notes to help identify underdiagnosed conditions.
While these new tools are powerful, their use presents a number of risks – including confidentiality, transparency, and methodological suitability – that should be addressed forthrightly. And, the authors point out, AI’s predictive abilities can be used to adjust insurance premiums in undesirable and potentially unfair ways. “AI methodologies are therefore not inherently good or bad, biased or clean,” the authors write in the article. “They are what we make of them.”
Authors
Thesmar D, Sraer D, Pinheiro L, Dadson N, Veliche R, Greenberg PE