Patient-Provider Racial Concordance and Discharges Against Medical Advice Among Patients with Opioid Use Disorder

• In Event: Increasing Access to Medications for Opioid Use Disorder

Abstract

The ongoing U.S. opioid crisis highlights the critical need for effective treatment for individuals with opioid use disorder (OUD). Approximately 1% of individuals in the United States over age 12 suffer from OUD, and these patients are significantly more likely to discharge from inpatient hospitalization against medical advice (AMA) which is associated with higher costs and adverse health outcomes. A major contributing factor to AMA discharge is untreated opioid withdrawal, which medications for opioid use disorder (MOUD) are crucial for mitigating. Unfortunately, Black patients are less likely than White patients to be prescribed MOUD, to be successfully linked to subsequent substance use disorder care, or to receive opioid prescriptions for pain. Existing evidence suggests that demographic concordance between patients and providers may be associated with better health outcomes. This study aims to explore whether patient-provider racial concordance affects AMA discharges among patients with OUD or those receiving MOUD prior to their hospital admission. I leverage the quasi-random assignment of hospitalized patients to attending hospitalist physicians to address potential selection bias.

I use Medicare fee-for-service and encounter emergency department and inpatient claims data from 2015-2022 and flag unique visits/hospital encounters where a patient was diagnosed with OUD within a year prior to arrival. I use the Medicare master beneficiary summary files to identify patient race, and I link providers from medical claims to physician race data from Texas. I flag whether a patient and provider have the same race, excluding hospital visits where the patient or provider race is unknown or reported as “other”. I then calculate monthly patient-level percentages of AMA discharges for (1) all patients, (2) patients with OUD within a year prior to admission, and (3) patients on MOUD within a year prior to admission. Using high-dimensional linear regression models with patient and month fixed effects, I estimate the impact of patient-provider racial concordance on AMA discharges for each group of patients. Standard errors are clustered at the state level.

I find statistically significant reductions in (1) monthly patient-level AMA discharge rates [-0.00, p-value 0.023, 95% CI: -0.01, -0.00], (2) AMA discharge rates for patients with an OUD diagnosis within a year prior to admission [-0.55, p-value <0.001, 95% CI: -0.77, -0.32], and (3) AMA discharges for patients on MOUD within a year prior to admission [-2.13, p-value, 95% CI: -3.84, -0.44] associated with patient-provider racial concordance.

These findings demonstrate that racial concordance between patients and providers may improve retention in care in the hospital setting. This may significantly benefit patients with OUD diagnoses/on MOUD who are more vulnerable to discharge AMA. This novel evidence suggests that a physician workforce that demographically reflects the population it serves may have beneficial outcomes for vulnerable populations.

Author

• Victoria Elizabeth Bethel, University of Georgia
  Presenting Author