Presentation on October 7, 2015, at the 3rd Annual BayesiaLab Conference:
Learning Dynamic Bayesian Networks from fMRI Time Series under Conditions of Chronic Pain and Opioid Addiction
Larry Price, Ph.D.
Director – Interdisciplinary Initiative for Research Design & Analysis
Professor of Psychometrics and Statistics
Texas State University
There is a prevalence of comorbid chronic pain and opioid addiction, (Rosenblum et al, 2003; Clark et al, 2008), presenting a serious healthcare challenge. Independently, chronic pain and opioid addiction are difficult to treat, and the comorbidity only increases the complexity. Patients with a substance use disorder (SUD) and co-occurring physical pain are more likely to misuse opioids than SUD patients without pain (Potter et al., 2008). Chronic pain is positively associated with substance use disorder severity, psychiatric disorders, psychological distress, medical comorbidities (Rosenblum et al, 2003), generally physical health problems, medical care utilization (Rosenblum et al, 2003; Trafton et al, 2004) and psychosocial factors (Jamison et al, 2000; Rosenblum et al, 2003; Potter et al, 2004; Trafton et al, 2004). Data were collected during pain induction in 18 opioid-addicted participants who displayed chronic low back pain and 18 age- and sex-matched healthy controls. Identification of a plausible model included employing augmented naïve Bayes classification within Bayesian Networks. Model performance involved study (target) group sensitivity analysis, mutual information and statistical tests of edge parameter differences based on regional (node) alterations.