We present an analysis of hospitalization rates following opioid and non-pharmacologic treatment among Medicare beneficiaries suffering from chronic back pain. Treatment for chronic back pain has largely relied on opioid therapy, though evidence-based guidelines are increasingly shifting toward non-pharmacologic treatment such as physical therapy and chiropractic care as the health risks of opioid treatment have become well established. An important clinical goal in effective management of chronic pain is reducing risk of hospitalization for the patient, but it is unclear if non-pharmacologic treatment significantly reduces hospitalization rates relative to opioids. A causal approach to determining relative hospitalization rates is motivated by confounding between opioid and non-pharmacologic treatment prescribing and hospitalization by factors such as age, sex, race and ethnicity, condition of health, and opioid use disorder status, as well as the possibility of censoring. Using this clinical setting as a motivating example, we develop a g-computation method for jointly modeling recurrent longitudinal and terminal survival outcomes where hospitalization and survival time are both impacted by treatment prescribed, resulting in a dependent relationship between the two outcomes.
