Background: Opioid medications are one of the most common approaches adopted by primary care providers for chronic pain management but are usually not recommended beyond 3 months. Research studies in chronic pain management at primary care settings remains lacking, and the usual investigations relying majorly on large surveys or questionnaires can lead to bias by missing a large proportion of population.
Aims: Our study aims to provide an overview of chronic pain patients and the primary health care they’ve received using a novel artificial intelligence based machine learning algorithm.
Methods: This study used observational U.S. Medicaid claims data (n>7 million) among 6 geographically different states to identify chronic pain patients in primary care between January 2017 to December 2019. The performance of 8 different machine learning classification algorithms were evaluated to model the risk factors of chronic pain patients prescribed opioids by primary care providers.
Results: Our study demonstrated a prevalence of 25.9% (average age 33.0±19.7 years, 64.0% female) of chronic pain among Medicaid claimants, and 6.5% (average age 37.5±19.4 years, 67.5% female) were considered as high-impact chronic pain. The top 3 performing algorithms for accuracy were XG Boost (0.814), Random Forest (0.810) and Ada Boost (0.806).
Conclusions: With machine learning algorithms, our study yielded a previously under-recognized vulnerable age group of children and adolescents under 18 years old as one of the major risk factors of having an opioid prescription from a primary care provider.