Chronic pain prevalence and opioid prescribing patterns among primary care providers

Abstract

Chronic pain is an increasing global health concern and can seriously affect an individual’s quality of life. Primary care providers are often the initial contact for patients to seek medical help for chronic pain symptoms, and opioid prescriptions remain a common chronic pain treatment option in primary care settings. This is despite the recommendation that opioids should not be prescribed beyond three months, due to potential underlying harms that could result in opioid tolerance, physical dependence, addiction and even death. Electronic health records provide extensive datasets for the investigation of chronic pain and its management. This advent of big data coincides with the rapid development of artificial intelligence and machine learning techniques. This has allowed the reassessment of chronic pain prevalence and opioid prescribing in primary care by leveraging big data. The overall aim of this thesis is to characterise the chronic pain prevalence and opioid prescribing patterns among primary care providers using big data research. Chapter 2 is a scoping review study of big data research that aims to map the chronic non-cancer pain patient population in primary care utilising big data research, investigating the patient characteristics and opioid prescription patterns. The chapter reported an estimated chronic pain prevalence of 3.82% and 10.3% in the primary care setting from two of the eligible studies. The chapter also documented that chronic pain was most prevalent in females compared to males and was often associated with the comorbidities of depression and anxiety. The study results reported that over 30% of chronic pain patients received opioid prescriptions from primary care providers. However, this scoping review study only had three eligible studies, demonstrating a significant gap between previous research and this new area of pain informatics. Chapter 3 presents a clinical epidemiology study and an exploratory computational study. The chapter aims to explore chronic pain prevalence and opioid prescribing patterns for chronic pain and high-impact chronic pain patients among primary care providers within an extensive Medicaid claims dataset of over seven million claimants in six states in the United States of America. The chapter reported an estimated chronic pain prevalence of 24.2% (average age 33.0 ± 19.7 years, 64.0% female), and 7.2% (average age 37.5 ± 19.4 years, 67.5% female) of claimants were identified as high-impact chronic pain patients whose daily activities are significantly impacted. The chapter also evaluated the model performance of eight different machine-learning algorithms. This study demonstrated that the Extreme Gradient Boost (XG Boost) outperformed other algorithms with an accuracy of 0.814. Further, this chapter also identified the age group of children and adolescents under 18 years old, an under-recognised vulnerable group, as one of the major risk factors associated with receiving an opioid prescription in primary care settings. In conclusion, this thesis proposes the utility of big data in advancing pain informatics research. The studies provide a re-assessment of chronic pain prevalence and opioid prescriptions in primary care settings. Future studies should consider clinical pain informatics and methodologies for better pain management and the surveillance of associated risk factors in primary care.