Predictors and Health Outcomes of Treatment-Resistant Depression among Adults with Chronic Non-Cancer Pain Conditions and Major Depressive Disorder

Understanding major depressive disorder (MDD) as a comorbidity in patients with chronic non-cancer pain conditions (CNPC) is of importance because of the high prevalence and well documented bi-directional relationship between MDD and pain. Furthermore, presence of CNPC among adults with MDD often reduces benefits of antidepressant therapy, thereby increasing the possibility of treatment resistance. Treatment-resistant depression (TRD) commonly defined as insufficient response to multiple antidepressant trials, often worsens depression and pain symptoms and can amplify the clinical and economic burden among adults with CNPC and MDD. Additionally, long-term opioid therapy (LTOT) may be prescribed at a higher rate to adults with TRD to treat an undifferentiated state of physical and mental pain, despite lack of evidence on LTOT use in this patient population. Existing literature on TRD has focused on all adults with MDD. Owing to the lack of research on TRD among adults with CNPC and MDD, we conducted this study to fill a critical knowledge gap. This dissertation pursued three related aims: 1) identify leading predictors of TRD; 2) estimate the direct economic burden associated with TRD and identify factors that contribute to the excess cost burden of TRD; and 3) examine the trajectory of LTOT use in adults with CNPC and newly diagnosed MDD and how TRD can alter this trajectory. The study used a retrospective, longitudinal, cohort design, using multiple years (2007-2017) of the de-identified, geographically diverse data on commercially insured adults obtained from the 10% random sample of Optum® Clinformatics® Data Mart. TRD was identified using a comprehensive claims-based algorithm. Among adults (\u3e18 years) with CNPC and newly diagnosed MDD who were treated with antidepressants (N=23,645), approximately one in nine adults transitioned to TRD within a year of MDD diagnosis. First, we identified the use of mental health specialist visits, polypharmacy, psychotherapy use, anxiety, and age to be the five leading predictors of TRD using random forest, a machine learning method and cross-validated logistic regression. In the second aim, we observed that patients with TRD had significantly higher direct all-cause ($21,015TRD vs. $14,712No TRD) and MDD-related costs ($1,201TRD vs. $471No TRD), and healthcare resource utilization (e.g. IRR for inpatient visits = 1.30. 95% CI = 1.14-1.47;