INTRODUCTION

Effective and equitable pain management for patients with substance use disorders from minoritized groups presents unique challenges. Patients with substance use disorders, particularly opioid use disorder, often have higher opioid tolerance and hyperalgesia and may concurrently be withdrawing during hospitalization. Rates of certain substance use disorders for minoritized patients vary, and there is demonstrated literature showing inequities in pain management for patients from minoritized groups. Prior research has demonstrated that there is mutual distrust between physicians and patients from both of these groups, including a belief that clinicians do not take the pain reported by a patient seriously. Clinicians may hesitate to prescribe necessary doses of opioids to treat pain in these patient groups because of concerns about drug-seeking behavior, diversion, elopement risk, and respiratory depression. There is a major gap in understanding whether and to what extent inequities in pain management exist for inpatients with substance use disorders who identify as a racial or ethnic minority. We hypothesize that patients from minoritized racial/ethnic groups experience disparate pain management during hospitalization, after accounting for the presence of a substance use disorder. To test this hypothesis, we studied the association between racial/ethnic category and inpatient opioids received, after controlling for demographic and clinical factors, the presence of a substance use disorder, and pain characteristics. This study helps to characterize opioid prescribing for a challenging clinical population where equitable care is an increasing focus.

METHODS

Study Population

This retrospective cohort study included adult hospitalizations (≥ 18 years old) from January 1, 2021, to December 31, 2022, discharged from the general medicine service at the University of California, San Francisco Medical Center at Parnassus Heights, a 785-bed academic medical center that serves a diverse patient population. All data was collected from Epic, our medical system’s electronic health record (EHR), and Clarity, the relational database that extracts and stores inpatient Epic data. We only included patient hospitalizations with complete pain assessment data using self-reported pain assessments. Patient hospitalizations were additionally excluded if the patient spent time in the intensive care unit or if they received inpatient intensive comfort-focused or hospice care because these patients have different pain requirements and have their care often managed by specialists. The UCSF Institutional Review Board for Human Subjects Research approved this study with a waiver of informed consent.

Predictor/Exposure

The primary predictor was the patient’s self-reported race/ethnicity. Consistent with updated US Census, NIH reporting, and institutional standards, we included the following minority race/ethnicity categories: Asian, Black or African American, Latino, Multi-Race/Ethnicity, Native American or Alaska Native, Native Hawaiian or Pacific Islander, Southwest Asian or North African, Other, and Unknown/Declined, with White as a comparison group. These racial/ethnic group identities are socially, not genetically defined. These categorizations are used as a proxy for how race and ethnicity intersect with equity in healthcare and may help guide future research on mechanisms that create inequity, including racism.

Outcome

The primary outcome was the average daily inpatient opioids received during the patient’s hospitalization, measured by morphine milligram equivalents (MMEs). This is calculated by our Division’s Data Core using standardized conversions of opioid medications from the EHR.

Covariates

All analyses were adjusted with demographic, clinical, substance-use, and pain-related variables. Demographic variables included patient age, self-reported sex (male, female, or non-binary/other), limited English proficiency status (defined as having a preferred language for healthcare other than English and requiring a medical interpreter), and insurance status (Medicare, Medi-cal, or Private/Other). Clinical variables included the Elixhauser comorbidity index as a marker of clinical complexity and length of stay. Substance-use-related variables included having a billing diagnosis of any substance-use disorder using International Classification of Diseases (ICD)− 10 codes. These ICD- 10 codes were manually selected by the authors to reflect whether a patient had a billing diagnosis that would best reflect a clinical diagnosis of a substance use disorder (Supplemental Table 2). We also included whether a patient had an existing prescription for Medication for Opioid Use Disorder (MOUD) on admission (methadone or buprenorphine), based on admission medication reconciliation, which includes confirmation with methadone clinic. Pain-related variables included the patient’s average self-reported pain score during their entire hospitalization, whether the patient was admitted with moderate/severe pain (defined as their first pain assessment being ≥ 6 on a 0–10 self-reported pain scale), whether a consult was placed for the pain or palliative medicine service, prior to admission opioid prescription (i.e. oxycodone, morphine, etc.), and the average daily milligrams of acetaminophen and ibuprofen administered to the patient. At our institution, the pain service is staffed by anesthesia, the palliative medicine is staffed by palliative medicine physicians, and either service may be consulted for pain-related issues. The average self-reported pain score was calculated as the mean of all Numeric Rating Scale, Verbal Descriptor Scale, and FACES Pain Scale-Revised Scores standardized to a 0–10 scale, with higher numbers reflective of worse pain. Nursing pain assessments are performed throughout a patient’s hospitalization: on admission, on unit transfers, before, during, and after procedures or analgesic administration, and with routine vital sign checks. These data are inputted by nurses into EHR flowsheets.

Statistical Analysis

All analyses were done using Stata software v.18. To assess for differences in inpatient pain management, we first calculated the unadjusted daily MMEs across race/ethnicity and all the other covariates. For the adjusted analysis, we used multivariable negative binomial regression to account for the heavily dispersed distribution of daily MMEs. The models were adjusted for all demographic, clinical, substance use, and pain-related variables, with clustering by patient medical record number to account for multiple hospitalizations for a given patient during the study period. All hypothesis tests were evaluated at α = 0.05. We prespecified the interaction between race/ethnicity and substance use disorder and utilized omnibus testing. If the interaction was not significant, we refit the model with main effects only. White race was used as the reference category. Results from the negative binomial regression were reported using average marginal effects (AMEs), which describes the average difference in average daily inpatient MMEs between the comparison and reference race/ethnicity categories.

Subgroup Analyses

We performed two subgroup analyses. First, we repeated the above analysis on only patients with an ICD- 10-defined substance use disorder. In this subgroup, we reported the frequency of each specific substance use disorder and performed an adjusted analysis using negative binomial regression. The purpose of this subgroup analysis was to further isolate the effect of race/ethnicity on inpatient opioid administration by minimizing confounding and focusing only on those with a defined substance use disorder. Second, we performed the above analysis excluding all patients who were on either methadone or buprenorphine prior to admission and all the patients who received methadone or buprenorphine during admission. The purpose of this subgroup was to minimize confounding by medications that can be used to treat either OUD or pain and potential variation in MOUD prescribing rates among different racial/ethnic groups. This subgroup did not include any methadone or buprenorphine calculated in the outcome variable of MMEs.

Results

The study included 9102 patients across 13,058 unique hospitalizations from January 1, 2021, to December 31, 2022, discharged from the general medicine service. The cohort had a mean age of 62.7 years (standard deviation (SD) 19.0) and was 51.2% male. The racial/ethnic distribution of the overall cohort was 43.3% White, 23.2% Asian, 13.0% Latino, 12.3% Black or African American, and the remaining groups made up 8.4% (Fig. 1). At the individual patient level, 31.3% of patients had a substance use disorder diagnosis, 4.2% were prescribed MOUD prior to admission, 26.3% were prescribed opioids prior to admission, and 31.8% were admitted with moderate or severe pain. Rates of MOUD prior to admission varied across racial and ethnic groups (White, 5.6%; Black, 6.5%; Asian, 1.4%; and Latino, 2.6%). The mean hospital length of stay was 6.8 days (SD 13.7) (Table 1, see Supplemental Table 1 for data across all race/ethnicity categories).

Primary Analysis

Unadjusted and adjusted analyses examining the relationship between race/ethnicity and average daily inpatient opioid administration in morphine milligram equivalents (MME) are presented in Table 2. In the unadjusted analysis, significant differences were found across racial/ethnic groups (p < 0.001), with Black or African American patients receiving the highest mean MME/day of 81.3 (SD 192.7) and Asian patients the lowest at 12.9/day (SD 65.3) (Table 2).

For the adjusted analysis, after considering demographic factors, clinical variables, substance use, pain characteristics, and accounting for clustering of hospitalizations by patient, significant differences in opioid administration were found across racial/ethnic groups. Every defined racial/ethnic minority group, except for “Other” and “Unknown/Declined” received significantly fewer opioids compared to White patients. The largest racial/ethnic minority groups all received fewer adjusted opioids, including Asian patients (− 61.3 MME/day, 95% confidence interval (CI) − 79.4 to − 43.2, p < 0.001), Black patients (− 44.9 MME/day, 95% CI − 68.9 to − 21.0, p < 0.001), and Latino patients (− 48.8 MME/day, 95% CI − 68.6 to − 29.0, p < 0.001). The largest effect sizes were found for Native American/Alaska Native patients (− 80.4 MME/day, 95% CI − 121.6 to − 39.2, p < 0.001) and Native Hawaiian/Pacific Islander patients (− 72.9 MME/day, 95% CI − 99.5 to − 26.3, p < 0.001). Several other factors were associated with higher opioid administration, including the presence of a substance use disorder diagnosis, being prescribed MOUD or opioids prior to administration, and receiving an inpatient pain or palliative medicine consultation. Patients with higher levels of pain on admission did not receive higher levels of opioids compared to those without higher levels of pain (Table 2).

Subgroup Analysis 1: Patients with SUD

Compared to the overall cohort, this specific subgroup of patients with SUD (n = 2846 patients across 4446 hospitalizations) had a higher proportion of Black or African-American and male patients and fewer Asian patients (Table 3). The most common substance-related disorders were nicotine/tobacco, alcohol, and opioids. For the adjusted analysis of the subgroup, the overall findings were similar, with most racial/ethnic minority groups receiving fewer adjusted opioids compared to White patients. For the largest racial/ethnic minority groups, Asian (− 124.4 MME/day, 95% CI − 168.3 to − 80.6, p < 0.001), Black (− 68.7 MME/day, 95% CI − 120.9 to − 16.6, p < 0.001), and Latino patients (− 110.8 MME/day, 95% CI − 155.8 to − 65.9, p < 0.001) all received fewer opioids than White patients. These effect sizes were all larger than in the overall analysis (Table 4).

Subgroup Analysis 2: Excluding All Patients on Methadone and/or Buprenorphine

This subgroup included a total of 8612 patients across 12,153 hospitalizations. The cohort was very similar in composition to the overall cohort. For the adjusted analysis of this subgroup, our findings were similar to the overall analysis, where most racial/ethnic minority groups received fewer adjusted opioids compared to White patients. For the largest racial/ethnic minority groups, Asian (− 19.9 MME/day, 95% CI − 27.8 to − 11.9, p < 0.001), Black (− 19.3 MME/day, 95% CI − 27.9 to − 10.8, p < 0.001), and Latino (− 13.4 MME/day, 95% CI − 22.3 to − 4.62, p = 0.003) patients all received fewer opioids than White patients (Supplemental Table 3). Notably, average MMEs were lower in this subgroup compared to the overall cohort.

Discussion

In this retrospective study of over 13,000 hospitalizations at an academic medical center, we found clinically meaningful racial/ethnic inequities in inpatient opioid administration for pain management after adjusting for substance use disorders, demographic factors, clinical variables, and pain characteristics. We found compounded inequity for patients with multiple marginalized identities (i.e., race/ethnicity minority, substance use disorder).

Consistent with our initial hypothesis, in our overall cohort, racial/ethnic minority groups received significantly fewer opioids compared to White patients. Large disparities were observed for Asian, Black, Latino, Native American/Alaska Native, Native Hawaiian/Pacific Islander, Southwest Asian/North African, and Multiracial patients. These findings persisted and were even more pronounced in the subgroup analysis restricted to patients with a substance use disorder diagnosis. We again found similar and substantial decreases in opioids received in the subgroup analysis that excluded all patients who received methadone or buprenorphine prior to or during admission. These are major findings to emphasize—even after controlling for key variables, including the presence of a substance use disorder, prior opioid and MOUD prescriptions, the average self-reported pain score, the presence of significant pain on admission, demographic and clinical characteristics, minoritized patients received significantly fewer opioids while admitted.

The effect sizes were not just statistically significant, but also clinically substantial. For context, a standard opioid dose prescribed by a physician is a 5 mg tablet of oxycodone, which is equivalent to 7.5 MMEs. In the subgroup, Asian patients received 124 fewer average daily MMEs than White patients, equivalent to 16 fewer tablets of oxycodone per day.

Our results are consistent with prior research demonstrating inequities for vulnerable groups in pain management across various healthcare settings, including those from racial/ethnic minority groups and those with a substance use disorder. However, this study extends those findings to the unique inpatient population of patients from racial/ethnic minority groups with a concurrent substance use disorder, a challenging clinical scenario where appropriate pain management is particularly complex. This study is also novel in that we studied and found inequities for often underreported racial/ethnic minority groups, including patients who identify as Native American/Alaska Native, Native Hawaiian/Pacific Islander, Southwest Asian/North African, and Multiracial.

There are several potential reasons for our findings, best understood by examination at the clinician, patient, and larger structural levels. At the clinician level, there is likely to be bias in providing care for patients with substance use disorders and those from minority racial/ethnic groups. These biases have been previously identified in the literature for both populations, and this patient population faces intersecting biases and stigma due to race/ethnicity and substance use. Clinicians may be particularly hesitant to prescribe opioids due to concerns about misuse, diversion, or exacerbating substance use disorders, despite evidence that undertreated pain can worsen outcomes.

At the patient level, it is notable that Asian patients received the fewest opioids in both the overall and subgroup analyses. It is possible that there is variation across racial/ethnic groups in (1) the experience and expression of pain, (2) the ability or willingness to communicate a given pain level to a provider, and (3) the willingness to accept an opioid in general, or a higher dose of opioid pain medication for a given pain level. Despite these potential reasons, research in this area has still found that patient-related attitudinal concerns about opioids are more likely to be shaped by undertreatment, not as a cause of it.

Between clinicians and patients, the role of communication is essential. Patients from minoritized racial groups are more likely to have limited English proficiency. While we did not find language status to be a significant predictor of opioid administration, there may still be cultural factors in terms of clinician-patient communication that impact overall pain assessment and management. At the systems level, there has been an increased national focus on opioid deprescribing. The study site was in San Francisco, a city particularly hit hard by the opioid epidemic with record overdose levels during the study period. This background is a likely factor in physician decisions on how aggressively to treat pain with opioids but does not fully explain the racial/ethnic variations in pain management.

There were several additional notable findings. In our models, the most significant predictors of receiving higher inpatient opioid doses, unsurprisingly, were having a prescription for MOUD and/or opioids prior to admission. These patients likely had higher opioid tolerance and the potential for hyperalgesia and concurrent withdrawal. Our subgroup analysis, which removed these patients, demonstrated lower doses of inpatient opioids but persistent racial/ethnic disparities. In the overall model, the average pain score was not associated with opioid administration, but this had a positive association in the subgroup models. As supported by the literature, reported pain is just one of many factors that influence the decision to prescribe opioids, particularly in those with concurrent substance use disorders.

There are limitations to consider. First, this was a single-center study where physicians at the study site likely systematically practice in different ways than other places, which limits our generalizability. Second, we used administrative billing codes to identify patients with substance use disorders, which could either underestimate or overestimate the true clinical prevalence, depending on the circumstance. Third, our capture of inpatient MMEs included methadone and buprenorphine for the overall model, which can be used for both pain control and for treatment for opioid use disorder. We are unable to parse out the indication for these medications in this dataset (for pain vs for OUD). Therefore, the second subgroup model, which eliminated all patients who received these medications prior to and during admission, was performed to minimize confounding from methadone and buprenorphine prescriptions.

Nonetheless, our findings are novel for the fields of general medicine, health equity, and substance use. Patients from minoritized groups who also have a substance use disorder are uniquely vulnerable to inequitable inpatient pain management. Future prospective studies including the specific indication for each opioid medication and clinician-based diagnoses of substance use disorders, and more granular analyses comparing different SUDs are needed to fully elucidate the mechanisms underlying these disparities. We also plan to examine the granular relationship between individual pain assessments and subsequent medication administration in various clinical scenarios across different patient cohorts, which would require multilevel time-series analysis to account for multiple prescribers, varying medication durations, and temporal relationships between pain scores and prescribing decisions. The consistency of all our findings across the overall cohort, the SUD subgroup, and the subgroup without buprenorphine/methadone highlights the importance of developing in-hospital interventions to promote equitable, culturally competent pain care for marginalized populations. Potential strategies include provider education on biases, enhanced patient-provider communication tools, standardized pain assessment and management protocols, and institutional policies that track and promote equitable pain management practices. Ultimately, addressing these disparities is crucial to improving care quality and outcomes for all patients, regardless of their race, ethnicity, or substance use history.

INTRODUCTION

Effective and equitable pain management for patients with substance use disorders from minoritized groups presents unique challenges. Patients with substance use disorders, particularly opioid use disorder, often exhibit higher opioid tolerance and hyperalgesia, and may concurrently be experiencing withdrawal symptoms during hospitalization. Rates of certain substance use disorders for minoritized patients vary, and existing research demonstrates inequities in pain management for patients from minoritized groups. Prior studies have shown mutual distrust between physicians and patients from both these groups, including a belief that clinicians do not adequately address patient-reported pain. Clinicians may hesitate to prescribe necessary opioid doses for pain treatment in these patient populations due to concerns about drug-seeking behavior, diversion, elopement risk, and respiratory depression. A significant gap exists in understanding whether and to what extent inequities in pain management occur for inpatients with substance use disorders who identify as a racial or ethnic minority. The hypothesis is that patients from minoritized racial/ethnic groups experience disparate pain management during hospitalization, even after accounting for the presence of a substance use disorder. To test this hypothesis, the study examined the association between racial/ethnic category and inpatient opioids received, controlling for demographic and clinical factors, the presence of a substance use disorder, and pain characteristics. This study aims to characterize opioid prescribing practices within a complex patient population where equitable care is a growing concern.

METHODS

Study Population

This retrospective cohort study included adult hospitalizations (18 years or older) from January 1, 2021, to December 31, 2022. All included patients were discharged from the general medical service at the University of California, San Francisco Medical Center at Parnassus Heights, a 785-bed academic medical center that serves a diverse patient population. All data were collected from the medical system's electronic health record (EHR) and Clarity, a relational database storing inpatient Epic data. Only hospitalizations with complete self-reported pain assessment data were included. Patient hospitalizations were excluded if the patient spent time in the intensive care unit or received inpatient intensive comfort care or hospice care, as these patients have different pain requirements and their care is often managed by specialists. The study received approval from the UCSF Institutional Review Board for Human Subjects Research, which included a waiver of informed consent.

Predictor/Exposure

The main factor examined was the patient's self-reported race and ethnicity. Following updated US Census, NIH, and institutional guidelines, the minority race/ethnicity categories included Asian, Black or African American, Latino, Multi-Race/Ethnicity, Native American or Alaska Native, Native Hawaiian or Pacific Islander, Southwest Asian or North African, Other, and Unknown/Declined, with White as the comparison group. These racial/ethnic group identities are defined by social constructs, not genetic factors. These categorizations serve as an indicator of how race and ethnicity relate to healthcare equity and may inform future research into the mechanisms that create inequity, such as racism.

Outcome

The primary outcome was the average daily inpatient opioids received during the patient's hospitalization, measured in morphine milligram equivalents (MMEs). This measure is calculated by the Division's Data Core using standard conversion methods for opioid medications from the EHR.

Covariates

All analyses were adjusted for demographic, clinical, substance use, and pain-related factors. Demographic variables included patient age, self-reported sex (male, female, or non-binary/other), limited English proficiency status (defined as having a preferred language for healthcare other than English and requiring a medical interpreter), and insurance status (Medicare, Medi-cal, or Private/Other). Clinical variables included the Elixhauser comorbidity index, which indicates overall clinical complexity, and length of stay. Substance-use-related variables included a diagnosis of any substance use disorder based on International Classification of Diseases (ICD)-10 codes recorded for billing purposes. These ICD-10 codes were carefully chosen by the authors to represent a clinical diagnosis of a substance use disorder. Whether a patient had an existing prescription for medication for opioid use disorder (MOUD) upon admission (methadone or buprenorphine) was also included, based on admission medication reconciliation, which involved confirmation with methadone clinics. Pain-related variables included the patient's average self-reported pain score during their entire hospitalization, whether the patient was admitted with moderate/severe pain (defined as their first pain assessment rating 6 or higher on a 0-10 self-reported pain scale), whether a consultation was requested for the pain or palliative medicine service, prior opioid prescriptions before admission (e.g., oxycodone, morphine), and the average daily doses of acetaminophen and ibuprofen administered. At this institution, the pain service is staffed by anesthesiologists, and the palliative medicine service is staffed by palliative medicine physicians; either service may be consulted for pain-related issues. The average self-reported pain score was calculated as the mean of all Numeric Rating Scale, Verbal Descriptor Scale, and FACES Pain Scale-Revised Scores, standardized to a 0–10 scale, where higher numbers indicated worse pain. Nurses perform pain assessments throughout a patient's hospitalization: upon admission, during unit transfers, before, during, and after procedures or pain medication administration, and with routine vital sign checks. This data is recorded by nurses in the EHR.

Statistical Analysis

All analyses were conducted using Stata software version 18. To assess differences in inpatient pain management, the unadjusted daily MMEs were first calculated across race/ethnicity and all other covariates. For the adjusted analysis, multivariable negative binomial regression was used to manage the wide distribution of daily MMEs. The models were adjusted for all demographic, clinical, substance use, and pain-related variables, with grouping by patient medical record number to account for multiple hospitalizations for the same patient during the study period. All statistical tests used a significance level of 0.05. The interaction between race/ethnicity and substance use disorder was specified in advance, and omnibus testing was utilized. If the interaction was not significant, the model was re-ran using only main effects. White patients served as the comparison group. Results from the negative binomial regression were reported using average marginal effects (AMEs), which represent the average difference in daily inpatient MMEs between the comparison and reference racial/ethnic groups.

Subgroup Analyses

Two subgroup analyses were performed. First, the analysis was repeated exclusively for patients with a substance use disorder diagnosis defined by ICD-10 codes. In this subgroup, the frequency of each specific substance use disorder was reported, and an adjusted analysis using negative binomial regression was performed. The purpose of this subgroup analysis was to more clearly isolate the effect of race/ethnicity on inpatient opioid administration by reducing confounding factors and concentrating on patients with a confirmed substance use disorder. Second, the analysis was performed excluding all patients who received methadone or buprenorphine either before or during admission. The purpose of this subgroup was to reduce the influence of confounding by medications used for both opioid use disorder (OUD) and pain, as well as potential variations in MOUD prescribing rates among different racial/ethnic groups. This subgroup analysis did not include methadone or buprenorphine in the calculation of the outcome variable (MMEs).

RESULTS

The study included 9102 patients across 13,058 unique hospitalizations from January 1, 2021, to December 31, 2022, all discharged from the general medical service. The cohort's average age was 62.7 years (standard deviation 19.0), and 51.2% of patients were male. The overall cohort's racial and ethnic composition was 43.3% White, 23.2% Asian, 13.0% Latino, 12.3% Black or African American, with the remaining groups accounting for 8.4% (Figure 1). Among individual patients, 31.3% had a substance use disorder diagnosis, 4.2% were prescribed MOUD before admission, 26.3% had prior opioid prescriptions, and 31.8% were admitted with moderate or severe pain. MOUD prescription rates before admission differed among racial and ethnic groups (White, 5.6%; Black, 6.5%; Asian, 1.4%; and Latino, 2.6%). The average hospital stay was 6.8 days (standard deviation 13.7) (Table 1; see Supplemental Table 1 for full racial/ethnic data).

Primary Analysis

Table 2 presents the unadjusted and adjusted analyses of the relationship between race/ethnicity and average daily inpatient opioid administration, measured in morphine milligram equivalents (MME). In the unadjusted analysis, significant differences were observed across racial and ethnic groups (p < 0.001). Black or African American patients received the highest mean MME per day (81.3, SD 192.7), while Asian patients received the lowest (12.9 MME per day, SD 65.3) (Table 2).

After adjusting for demographic factors, clinical variables, substance use, and pain characteristics, and accounting for multiple hospitalizations per patient, significant differences in opioid administration were identified across racial and ethnic groups. All defined racial/ethnic minority groups, except "Other" and "Unknown/Declined," received significantly fewer opioids compared to White patients. The largest racial/ethnic minority groups consistently received fewer adjusted opioids. These included Asian patients (−61.3 MME/day, 95% confidence interval (CI) −79.4 to −43.2, p < 0.001), Black patients (−44.9 MME/day, 95% CI −68.9 to −21.0, p < 0.001), and Latino patients (−48.8 MME/day, 95% CI −68.6 to −29.0, p < 0.001). The most substantial differences were observed for Native American/Alaska Native patients (−80.4 MME/day, 95% CI −121.6 to −39.2, p < 0.001) and Native Hawaiian/Pacific Islander patients (−72.9 MME/day, 95% CI −99.5 to −26.3, p < 0.001). Other factors associated with higher opioid administration included a substance use disorder diagnosis, prior prescriptions for MOUD or other opioids, and an inpatient consultation with the pain or palliative medicine service. Patients admitted with higher pain levels did not receive more opioids than those without high pain levels (Table 2).

Subgroup Analysis 1: Patients with SUD

This subgroup of patients with a substance use disorder (n = 2846 patients across 4446 hospitalizations) had a higher proportion of Black or African-American and male patients and fewer Asian patients, compared to the overall cohort (Table 3). The most prevalent substance-related disorders were nicotine/tobacco, alcohol, and opioids. In the adjusted analysis of this subgroup, the overall findings were similar, with most racial/ethnic minority groups receiving fewer adjusted opioids compared to White patients. Among the largest racial/ethnic minority groups, Asian patients (−124.4 MME/day, 95% CI −168.3 to −80.6, p < 0.001), Black patients (−68.7 MME/day, 95% CI −120.9 to −16.6, p < 0.001), and Latino patients (−110.8 MME/day, 95% CI −155.8 to −65.9, p < 0.001) all received fewer opioids than White patients. The magnitudes of these differences were greater than those observed in the overall analysis (Table 4).

Subgroup Analysis 2: Excluding All Patients on Methadone and/or Buprenorphine

This subgroup comprised a total of 8612 patients across 12,153 hospitalizations. The composition of this cohort closely resembled that of the overall cohort. In the adjusted analysis for this subgroup, the findings mirrored the overall analysis: most racial/ethnic minority groups received fewer adjusted opioids than White patients. Among the largest racial/ethnic minority groups, Asian patients (−19.9 MME/day, 95% CI −27.8 to −11.9, p < 0.001), Black patients (−19.3 MME/day, 95% CI −27.9 to −10.8, p < 0.001), and Latino patients (−13.4 MME/day, 95% CI −22.3 to −4.62, p = 0.003) all received fewer opioids than White patients (Supplemental Table 3). It was notable that average MMEs were lower in this subgroup compared to the overall cohort.

DISCUSSION

This retrospective study of over 13,000 hospitalizations at an academic medical center revealed significant racial and ethnic inequities in inpatient opioid administration for pain management. These disparities persisted even after accounting for substance use disorders, demographic factors, clinical variables, and pain characteristics. The findings indicate that patients with multiple marginalized identities, specifically racial/ethnic minorities with a substance use disorder, experience compounded inequities. Consistent with the initial hypothesis, racial/ethnic minority groups in the overall cohort received significantly fewer opioids than White patients. This trend, with large disparities noted across multiple minority groups, was even more pronounced in the subgroup analysis of patients with a substance use disorder diagnosis. Similar substantial decreases in opioid receipt were observed when excluding patients who received methadone or buprenorphine, further emphasizing that minoritized patients received significantly fewer opioids despite comprehensive controls for relevant clinical and demographic factors.

The observed differences were not only statistically significant but also clinically substantial. For perspective, a standard physician-prescribed opioid dose, such as a 5 mg oxycodone tablet, is equivalent to 7.5 MMEs. In the subgroup analysis, Asian patients received 124 fewer average daily MMEs than White patients, which translates to the equivalent of 16 fewer oxycodone tablets per day.

These findings align with previous research indicating inequities in pain management for vulnerable populations across various healthcare settings, including racial/ethnic minority groups and individuals with substance use disorders. This study expands upon existing knowledge by focusing on the unique inpatient population of racial/ethnic minority patients with co-occurring substance use disorders, a clinical situation where appropriate pain management is especially complex. Furthermore, the study is notable for identifying inequities among frequently underreported racial/ethnic minority groups, such as patients identifying as Native American/Alaska Native, Native Hawaiian/Pacific Islander, Southwest Asian/North African, and Multiracial.

The findings may be attributed to several factors operating at the clinician, patient, and systemic levels. At the clinician level, biases are likely present when providing care for patients with substance use disorders and those from racial/ethnic minority groups. These populations experience intersecting biases and stigma. Clinicians may hesitate to prescribe opioids due to concerns about misuse or exacerbating substance use, despite evidence that undertreated pain can worsen outcomes. At the patient level, it is notable that Asian patients consistently received the fewest opioids. Variations across racial/ethnic groups in pain experience, expression, and willingness to accept opioids are possible, though research suggests patient concerns about opioids often stem from undertreatment. Effective communication between clinicians and patients is also crucial; patients from minoritized racial groups may have limited English proficiency, and cultural factors in communication can affect pain assessment and management. At the systemic level, an increased national focus on opioid deprescribing and the severe impact of the opioid epidemic in the study's location (San Francisco) likely influenced physician prescribing decisions, although these factors do not fully explain the observed racial/ethnic variations.

Other significant findings emerged from the analyses. The most notable predictors of higher inpatient opioid doses were prior prescriptions for MOUD and/or other opioids, which suggests higher opioid tolerance or potential for hyperalgesia and withdrawal in these patients. The subgroup analysis excluding these patients demonstrated lower overall opioid doses but still revealed persistent racial/ethnic disparities. In the main model, average pain scores were not associated with opioid administration, but a positive association was observed in the subgroup models. This aligns with literature suggesting that reported pain is one of many factors influencing opioid prescribing, especially for patients with co-occurring substance use disorders.

Several limitations warrant consideration. First, this was a single-center study, and physician practices at this site may differ from other locations, limiting generalizability. Second, administrative billing codes were used to identify patients with substance use disorders, which could either underestimate or overestimate the actual clinical prevalence. Third, the calculation of inpatient MMEs in the overall model included methadone and buprenorphine, which serve both pain control and opioid use disorder treatment. The specific indication for these medications could not be determined from the dataset. Consequently, the second subgroup analysis excluded all patients who received these medications before or during admission to minimize confounding from methadone and buprenorphine prescriptions.

Despite these limitations, the findings offer novel insights for general medicine, health equity, and substance use fields. Patients from minoritized groups with co-occurring substance use disorders are particularly susceptible to unequal inpatient pain management. Future prospective studies are needed to fully clarify the mechanisms behind these disparities, which should include specific indications for opioid medications, clinician-diagnosed substance use disorders, and more detailed analyses comparing different types of substance use disorders. Planned future work will also explore the precise relationship between individual pain assessments and medication administration across various clinical scenarios and patient cohorts, requiring advanced statistical analysis to account for multiple prescribers, varying medication durations, and the timing of pain scores relative to prescribing decisions. The consistency of these findings across the overall cohort, the substance use disorder subgroup, and the buprenorphine/methadone-excluded subgroup underscores the necessity of developing in-hospital interventions to ensure equitable, culturally competent pain care for marginalized populations. Potential strategies include educating providers on biases, improving patient-provider communication, standardizing pain assessment and management protocols, and implementing institutional policies to monitor and promote fair pain management practices. Ultimately, addressing these disparities is essential for enhancing care quality and improving outcomes for all patients, regardless of their race, ethnicity, or substance use history.

Introduction

Managing pain for patients from minority groups who also have substance use disorders is very challenging. Patients with substance use disorders, especially opioid use disorder, often have a higher tolerance to opioids and may experience increased pain sensitivity. They may also be going through withdrawal during a hospital stay. The rates of certain substance use disorders vary among minority patients, and research has shown that these patients often receive unequal pain management. Previous studies indicate a lack of trust between doctors and these patients, with patients sometimes feeling their pain is not taken seriously. Doctors may also hesitate to prescribe enough opioids due to concerns about drug-seeking behavior, drug misuse, patients leaving without permission, or breathing problems. There is a significant gap in understanding how much and what kind of unfairness exists in pain management for hospitalized minority patients with substance use disorders. Researchers expected that patients from minority racial/ethnic groups would receive different pain management during hospitalization, even after accounting for whether they had a substance use disorder. To test this idea, the study examined the connection between a patient's racial/ethnic group and the amount of opioids they received in the hospital, while also considering other personal and medical factors, the presence of a substance use disorder, and details about their pain. This study helps explain how opioids are prescribed for a complex patient group where fair care is becoming increasingly important.

Methods

Study Population

This study looked back at hospitalizations of adult patients (18 years or older) between January 1, 2021, and December 31, 2022. All patients were discharged from the general medicine service at the University of California, San Francisco Medical Center at Parnassus Heights, a large teaching hospital serving many different patient groups. All information was collected from the hospital’s electronic health record system, Epic, and its related database, Clarity. Only hospital stays with complete patient-reported pain assessments were included. Hospitalizations were excluded if the patient spent time in the intensive care unit or received comfort-focused or hospice care because these patients have different pain needs often managed by specialists. The study was approved by the UCSF Institutional Review Board for Human Subjects Research, and patient consent was not required.

Predictor/Exposure

The main factor examined was the patient’s self-reported race/ethnicity. Following current US Census, NIH, and institutional guidelines, the study included these minority race/ethnicity categories: Asian, Black or African American, Latino, Multi-Race/Ethnicity, Native American or Alaska Native, Native Hawaiian or Pacific Islander, Southwest Asian or North African, Other, and Unknown/Declined. White patients were used as the comparison group. These categories are based on social identity, not genetics. These groupings help researchers understand how race and ethnicity are linked to fairness in healthcare and can guide future studies on the causes of unfairness, including racism.

Outcome

The main result measured was the average daily amount of opioids received by the patient during their hospital stay, expressed in morphine milligram equivalents (MMEs). This was calculated by the Division’s Data Core using standard conversions of opioid medications from the electronic health record.

Covariates

All analyses considered other patient details, including demographic factors, medical conditions, substance use history, and pain-related information. Demographic factors included patient age, self-reported sex (male, female, or non-binary/other), if they had limited English proficiency (meaning they preferred a language other than English for healthcare and needed a medical interpreter), and their insurance type (Medicare, Medi-Cal, or Private/Other). Medical factors included a score for overall health complexity (the Elixhauser comorbidity index) and length of hospital stay. Substance use factors included having a diagnosis for any substance use disorder based on specific medical codes (ICD-10 codes). These codes were manually chosen by the researchers to best reflect a clinical diagnosis of a substance use disorder. The study also noted if a patient was already prescribed medication for opioid use disorder (MOUD) like methadone or buprenorphine when admitted, based on a review of their medications confirmed with methadone clinics. Pain-related factors included the patient's average self-reported pain score during their entire hospital stay, if they were admitted with moderate to severe pain (first pain score of 6 or higher on a 0–10 scale), if a pain or palliative medicine specialist was consulted, if they had an opioid prescription before admission, and the average daily amount of acetaminophen and ibuprofen given to the patient. At the study hospital, pain specialists are from anesthesiology, and palliative medicine specialists are also available; both can be consulted for pain issues. The average self-reported pain score was calculated by averaging all pain scores from different scales, standardized to a 0–10 scale, where higher numbers meant worse pain. Nurses recorded pain assessments regularly throughout a patient’s stay.

Statistical Analysis

All data analysis was performed using Stata software v.18. To see differences in hospital pain management, researchers first calculated the daily MMEs for each race/ethnicity group and for all other factors without adjusting for anything else. For the adjusted analysis, a specific statistical method called multivariable negative binomial regression was used to handle the wide range of daily MME amounts. The models were adjusted for all demographic, medical, substance use, and pain-related factors. The analysis also accounted for patients who had multiple hospitalizations during the study period. All tests for a hypothesis used a significance level of 0.05. Researchers decided ahead of time to look at the interaction between race/ethnicity and substance use disorder and used an overall test for this. If no significant interaction was found, the model was rerun with only the main effects of each factor. White patients were used as the reference group for comparison. The results from the negative binomial regression were presented as average marginal effects (AMEs), which show the average difference in daily hospital MMEs between the comparison and reference race/ethnicity groups.

Subgroup Analyses

Two additional analyses were performed on specific patient groups. First, the main analysis was repeated only for patients with a substance use disorder diagnosis. In this group, the frequency of each specific substance use disorder was reported, and an adjusted analysis using negative binomial regression was performed. The goal of this analysis was to better understand how race/ethnicity affected opioid administration specifically among those with a defined substance use disorder, by reducing other influencing factors. Second, the main analysis was repeated excluding all patients who were taking methadone or buprenorphine before admission or who received these medications during their hospital stay. The purpose of this analysis was to reduce the influence of medications that can treat either opioid use disorder or pain, and to account for possible differences in MOUD prescribing rates among different racial/ethnic groups. This second subgroup analysis did not include any methadone or buprenorphine in the MME calculation.

Results

The study included 9,102 patients across 13,058 hospitalizations from January 1, 2021, to December 31, 2022, all discharged from the general medicine service. The average age of the patients was 62.7 years (standard deviation (SD) 19.0), and 51.2% were male. The racial/ethnic breakdown of the entire group was 43.3% White, 23.2% Asian, 13.0% Latino, 12.3% Black or African American, with the remaining groups making up 8.4%. At the individual patient level, 31.3% had a substance use disorder diagnosis, 4.2% were prescribed MOUD before admission, 26.3% were prescribed opioids before admission, and 31.8% were admitted with moderate or severe pain. The rates of MOUD prior to admission varied among racial and ethnic groups (White, 5.6%; Black, 6.5%; Asian, 1.4%; and Latino, 2.6%). The average hospital stay was 6.8 days (SD 13.7).

Primary Analysis

The results of the main analysis, showing the relationship between race/ethnicity and average daily opioid administration in MMEs, both unadjusted and adjusted, are presented. In the unadjusted analysis, significant differences were found across racial/ethnic groups, with Black or African American patients receiving the highest average MME/day of 81.3 (SD 192.7) and Asian patients receiving the lowest at 12.9/day (SD 65.3).

For the adjusted analysis, after considering factors such as demographics, medical conditions, substance use, and pain characteristics, and accounting for multiple hospitalizations by the same patient, significant differences in opioid administration were still found across racial/ethnic groups. Every defined racial/ethnic minority group, except for “Other” and “Unknown/Declined,” received significantly fewer opioids compared to White patients. The largest racial/ethnic minority groups all received fewer adjusted opioids, including Asian patients (61.3 MME/day less), Black patients (44.9 MME/day less), and Latino patients (48.8 MME/day less). The largest differences were found for Native American/Alaska Native patients (80.4 MME/day less) and Native Hawaiian/Pacific Islander patients (72.9 MME/day less). Several other factors were linked to higher opioid administration, such as having a substance use disorder diagnosis, being prescribed MOUD or other opioids before admission, and receiving a consultation from the inpatient pain or palliative medicine service. However, patients with higher levels of pain on admission did not receive more opioids compared to those without higher pain levels.

Subgroup Analysis 1: Patients with SUD

Compared to the overall group, this specific subgroup of patients with a substance use disorder (2,846 patients across 4,446 hospitalizations) had a higher percentage of Black or African-American and male patients, and fewer Asian patients. The most common substance-related disorders were nicotine/tobacco, alcohol, and opioids. For the adjusted analysis of this subgroup, the overall findings were similar, with most racial/ethnic minority groups receiving fewer adjusted opioids compared to White patients. For the largest racial/ethnic minority groups, Asian patients (124.4 MME/day less), Black patients (68.7 MME/day less), and Latino patients (110.8 MME/day less) all received fewer opioids than White patients. These differences were even larger than those seen in the overall analysis.

Subgroup Analysis 2: Excluding All Patients on Methadone and/or Buprenorphine

This subgroup included a total of 8,612 patients across 12,153 hospitalizations. The group was very similar in makeup to the overall study group. For the adjusted analysis of this subgroup, the findings were similar to the overall analysis, where most racial/ethnic minority groups received fewer adjusted opioids compared to White patients. For the largest racial/ethnic minority groups, Asian patients (19.9 MME/day less), Black patients (19.3 MME/day less), and Latino patients (13.4 MME/day less) all received fewer opioids than White patients. Notably, the average MMEs were lower in this subgroup compared to the overall group.

Discussion

In this study, which looked back at over 13,000 hospitalizations at a teaching medical center, researchers found significant racial/ethnic unfairness in the amount of opioids given for pain management in the hospital. This was true even after adjusting for substance use disorders, personal background, medical conditions, and pain characteristics. The study revealed compounded unfairness for patients with multiple marginalized identities, specifically those who are both from a racial/ethnic minority and have a substance use disorder.

Consistent with the initial prediction, racial/ethnic minority groups received significantly fewer opioids compared to White patients in the overall study group. Large differences were seen for Asian, Black, Latino, Native American/Alaska Native, Native Hawaiian/Pacific Islander, Southwest Asian/North African, and Multiracial patients. These findings remained consistent and were even more pronounced when the analysis was limited to patients with a substance use disorder diagnosis. Similar and substantial decreases in opioids were also found in the analysis that excluded all patients who received methadone or buprenorphine before or during their hospital stay. These are important findings: even after controlling for key factors like having a substance use disorder, previous opioid and MOUD prescriptions, average self-reported pain scores, and significant pain at admission, minority patients still received significantly fewer opioids during their hospital stays.

The observed differences were not just statistically significant but also substantial in a clinical sense. For example, a common opioid dose prescribed by a doctor is a 5 mg oxycodone tablet, which equals 7.5 MMEs. In the subgroup of patients with substance use disorder, Asian patients received 124 fewer average daily MMEs than White patients, which is like receiving 16 fewer oxycodone tablets per day.

The study’s results align with previous research showing unfairness in pain management for vulnerable groups in various healthcare settings, including racial/ethnic minority groups and those with a substance use disorder. However, this study extends those findings to the specific inpatient population of racial/ethnic minority patients who also have a substance use disorder, a challenging clinical situation where appropriate pain management is particularly complex. This study is also groundbreaking because it examined and found unfairness for often underreported racial/ethnic minority groups, such as Native American/Alaska Native, Native Hawaiian/Pacific Islander, Southwest Asian/North African, and Multiracial patients. Several factors could explain these findings, which can be understood by looking at the clinician, patient, and broader systemic levels. At the clinician level, biases are likely present when providing care to patients with substance use disorders and those from minority racial/ethnic groups. These biases have been identified in previous studies for both populations, and this patient group faces overlapping biases and stigma because of their race/ethnicity and substance use. Clinicians may be especially reluctant to prescribe opioids due to concerns about misuse or worsening substance use disorders, even though evidence suggests that undertreated pain can lead to worse outcomes.

At the patient level, it is notable that Asian patients received the fewest opioids in both the overall and subgroup analyses. It is possible that there are differences among racial/ethnic groups in how they experience and express pain, their ability or willingness to communicate a specific pain level to a provider, and their willingness to accept opioids in general, or higher doses for a given pain level. Despite these possibilities, research in this area still indicates that patient concerns about opioids are more often shaped by inadequate pain treatment, rather than being a cause of it. Between clinicians and patients, good communication is crucial. Patients from minority racial groups are more likely to have limited English proficiency. While the study did not find language status to be a significant predictor of opioid administration, cultural factors in clinician-patient communication may still affect overall pain assessment and management. At the systems level, there has been a growing national focus on reducing opioid prescriptions. The study location was San Francisco, a city greatly affected by the opioid epidemic with record overdose levels during the study period. This context likely influenced doctors' decisions on how aggressively to treat pain with opioids, but it does not fully explain the racial/ethnic differences in pain management.

Several other notable findings emerged. In the study’s models, the most significant predictors of receiving higher inpatient opioid doses were, not surprisingly, having a prescription for MOUD and/or other opioids before admission. These patients likely had higher opioid tolerance, the potential for increased pain sensitivity, and might have been experiencing withdrawal. The subgroup analysis that removed these patients showed lower doses of inpatient opioids but persistent racial/ethnic disparities. In the overall model, the average pain score was not linked to opioid administration, but this connection was positive in the subgroup models. As supported by other research, reported pain is just one of many factors that influence the decision to prescribe opioids, especially for those with co-occurring substance use disorders. This study had limitations to consider. First, it was a single-center study, meaning doctors at this hospital might practice differently than elsewhere, which limits how broadly the findings can be applied. Second, the study used administrative billing codes to identify patients with substance use disorders, which could either understate or overstate the true clinical prevalence. Third, the calculation of inpatient MMEs for the overall model included methadone and buprenorphine, which can be used for both pain control and for treating opioid use disorder. The study could not determine the specific reason for these medications (for pain versus for opioid use disorder). Therefore, the second subgroup model, which excluded all patients who received these medications, was performed to minimize the influence of methadone and buprenorphine prescriptions.

Nevertheless, the study’s findings are new and important for general medicine, health equity, and substance use fields. Patients from minority groups who also have a substance use disorder are particularly vulnerable to unfair inpatient pain management. Future studies that specifically note the reason for each opioid medication, use clinician-based diagnoses of substance use disorders, and conduct more detailed analyses comparing different substance use disorders are needed to fully understand the reasons behind these disparities. Researchers also plan to examine the detailed relationship between individual pain assessments and subsequent medication administration in various clinical situations across different patient groups. This would require complex analysis to account for multiple prescribers, varying medication durations, and how pain scores and prescribing decisions are related over time. The consistency of all findings across the overall group, the substance use disorder subgroup, and the subgroup without buprenorphine/methadone highlights the importance of developing in-hospital interventions to promote fair, culturally sensitive pain care for marginalized populations. Possible strategies include educating providers about biases, improving patient-provider communication tools, standardizing pain assessment and management protocols, and implementing hospital policies that track and promote fair pain management practices. Ultimately, addressing these unfair practices is vital for improving the quality of care and patient outcomes for everyone, regardless of their race, ethnicity, or history of substance use.

Introduction

Fair and proper pain treatment for patients from groups often treated unfairly, especially those with substance use disorders, presents special difficulties. Patients with substance use disorders, particularly opioid use disorder, may need more opioids for relief or feel pain more strongly. They might also be experiencing withdrawal symptoms during their hospital stay. The rates of certain substance use disorders vary for patients from minority groups, and research shows that these groups often receive unfair pain treatment. Past studies have also shown that there is a lack of trust between doctors and these patients, partly because patients feel their pain is not taken seriously. Doctors may be hesitant to give enough opioids to these patients due to worries about drug misuse, selling drugs, patients leaving the hospital, or breathing problems. It is not well understood if and how much unfairness exists in pain management for hospitalized patients with substance use disorders who identify as a racial or ethnic minority. This study aimed to understand the connection between a patient's racial or ethnic background and the amount of opioids they received in the hospital, even when considering their personal details, health conditions, the presence of a substance use disorder, and pain characteristics. This study helps to describe opioid prescribing for a patient group that is hard to treat and where fair care is becoming more important.

Methods: Study Design and Participants

This study looked back at adult patient records (aged 18 or older) from January 1, 2021, to December 31, 2022. The patients were hospitalized and discharged from the general medicine department at a large teaching hospital that serves a diverse patient population. All information was gathered from the hospital's electronic health records. Only hospital stays where patients fully reported their pain levels were included. Patients who spent time in the intensive care unit or who received end-of-life care were not included because these patients have different pain needs and specialists often manage their care. A special committee approved this study without needing patient permission.

Methods: Data Collection and Measurements

The main factor examined was the patient's self-reported race or ethnicity. Following official guidelines, this study included categories such as Asian, Black or African American, Latino, Multi-Race/Ethnicity, Native American or Alaska Native, Native Hawaiian or Pacific Islander, Southwest Asian or North African, "Other," and "Unknown/Declined." White patients were used as the comparison group. These groups are based on social ideas, not genetics, and help to understand how race and ethnicity connect with fairness in healthcare. The main result measured was the average daily amount of opioids given to patients during their hospital stay, measured in morphine milligram equivalents (MMEs). This was calculated using standard methods to convert different opioid medications to MMEs. All analyses were adjusted for other factors, including patient age, self-reported sex, whether patients had trouble speaking English, and their insurance type. Other health-related factors included a score showing how many other health problems a patient had and how long they stayed in the hospital. Substance use factors included having a diagnosis for any substance use disorder and whether a patient was already taking medication for opioid use disorder (MOUD) upon admission. Pain-related factors included the patient's average self-reported pain score during their hospital stay, if they were admitted with moderate to severe pain, if a pain or comfort care specialist team was asked for help, if they had an opioid prescription before admission, and the average daily amounts of acetaminophen and ibuprofen given. Pain levels were regularly checked by nurses throughout a patient's hospital stay.

Methods: Statistical Analysis

All calculations were performed using Stata software. To see differences in pain management, average daily MMEs were first calculated without any adjustments across racial/ethnic groups and other factors. For the adjusted analysis, a specific statistical method was used because the daily MME numbers varied widely. This method adjusted for all patient characteristics, health conditions, substance use, and pain factors, and it accounted for patients who had more than one hospital stay. Results were considered meaningful if there was less than a 5% chance they happened by random chance. The study also checked how race/ethnicity and substance use disorder might interact. If no strong interaction was found, a simpler model was used. White patients were used as the comparison group. The results showed the average difference in daily opioid amounts between minority groups and White patients. Two specific additional analyses were done. First, the same analysis was repeated only for patients with a substance use disorder diagnosis to focus on how race/ethnicity affects opioid use in this group. Second, the analysis excluded all patients who were taking methadone or buprenorphine before or during admission. This was done to reduce potential confusion from medications that can treat both addiction and pain, and to account for possible differences in how these medications are prescribed to different racial/ethnic groups.

Results Overview

The study included 9,102 patients over 13,058 hospital stays from January 1, 2021, to December 31, 2022, all discharged from the general medicine department. The average patient age was 62.7 years, and about half were male. The racial/ethnic makeup of the group was 43.3% White, 23.2% Asian, 13.0% Latino, 12.3% Black or African American, with the remaining groups making up 8.4%. At the individual patient level, 31.3% had a substance use disorder diagnosis, 4.2% were prescribed MOUD before admission, 26.3% were prescribed opioids before admission, and 31.8% were admitted with moderate or severe pain. The use of MOUD before admission varied among racial and ethnic groups. The average hospital stay was 6.8 days.

Results: Primary Analysis

Table 2 shows the results, both before and after adjusting for other factors, regarding the link between a patient's race/ethnicity and the average daily opioids they received in MMEs. Without any adjustments, significant differences were found across racial/ethnic groups. Black or African American patients received the highest average daily MMEs (81.3), while Asian patients received the lowest (12.9).

For the adjusted analysis, even after considering personal details, health conditions, substance use, pain characteristics, and accounting for multiple hospital stays per patient, significant differences in opioid use were still found across racial/ethnic groups. Nearly every defined racial/ethnic minority group received significantly fewer opioids compared to White patients, except for "Other" and "Unknown/Declined" categories. The largest minority groups also received fewer adjusted opioids: Asian patients received 61.3 fewer MMEs per day, Black patients received 44.9 fewer MMEs per day, and Latino patients received 48.8 fewer MMEs per day. The biggest differences were seen for Native American/Alaska Native patients (80.4 fewer MMEs per day) and Native Hawaiian/Pacific Islander patients (72.9 fewer MMEs per day). Several other factors were linked to receiving more opioids, including having a substance use disorder diagnosis, being prescribed MOUD or opioids before admission, and getting help from a pain or comfort care specialist. Patients who reported high pain levels when admitted did not necessarily receive more opioids compared to those with lower pain levels.

Results: Subgroup Analysis 1 (Patients with SUD)

Compared to the overall study group, this specific group of patients with a substance use disorder (2,846 patients across 4,446 hospitalizations) had a higher percentage of Black or African American and male patients, and fewer Asian patients. The most common substance-related disorders were related to nicotine/tobacco, alcohol, and opioids. For the adjusted analysis within this group, the overall findings were similar, with most racial/ethnic minority groups receiving fewer adjusted opioids compared to White patients. For the largest minority groups, Asian patients received 124.4 fewer MMEs per day, Black patients received 68.7 fewer MMEs per day, and Latino patients received 110.8 fewer MMEs per day, all receiving fewer opioids than White patients. These differences were even bigger than in the main analysis.

Results: Subgroup Analysis 2 (Excluding Methadone/Buprenorphine Patients)

This group included 8,612 patients across 12,153 hospitalizations. The makeup of this group was very similar to the main study group. For the adjusted analysis of this group, the findings were similar to the main analysis: most racial/ethnic minority groups received fewer adjusted opioids compared to White patients. Asian patients received 19.9 fewer MMEs per day, Black patients received 19.3 fewer MMEs per day, and Latino patients received 13.4 fewer MMEs per day, all receiving fewer opioids than White patients. It was noted that the average MMEs were lower in this group compared to the overall study group.

Discussion

This study, which looked back at over 13,000 hospitalizations at a teaching hospital, found important unfair differences in how opioids were given to patients for pain management. These differences were present even after considering substance use disorders, personal details, health conditions, and pain levels. The study found even greater unfairness for patients who had more than one disadvantaged identity, such as being a racial/ethnic minority with a substance use disorder.

As expected, in the overall group, racial/ethnic minority groups received significantly fewer opioids compared to White patients. Large differences were seen for Asian, Black, Latino, Native American/Alaska Native, Native Hawaiian/Pacific Islander, Southwest Asian/North African, and Multiracial patients. These findings remained true and were even more noticeable in the analysis focused only on patients with a substance use disorder diagnosis. Similar and substantial drops in received opioids were also found in the analysis that excluded all patients who received methadone or buprenorphine. These are major findings: even after accounting for important factors, including the presence of a substance use disorder, previous opioid prescriptions, average self-reported pain scores, and significant pain on admission, minority patients still received significantly fewer opioids during their hospital stay.

These differences were not just statistically noticeable; they were also important in real-life patient care. For example, a common opioid dose prescribed by a doctor is a 5 mg oxycodone tablet, which equals 7.5 MMEs. In the subgroup with substance use disorder, Asian patients received 124 fewer average daily MMEs than White patients, which is like 16 fewer oxycodone tablets each day. These results match earlier studies showing unfairness in pain management for vulnerable groups in different healthcare settings. This study adds to these findings by specifically looking at the unique group of hospitalized minority patients who also have a substance use disorder, which is a difficult medical situation where managing pain correctly is very complicated. This study is also new because it found unfairness for minority groups that are not often studied, including Native American/Alaska Native, Native Hawaiian/Pacific Islander, Southwest Asian/North African, and Multiracial patients.

Several potential reasons for these findings can be understood by looking at doctors, patients, and the healthcare system. At the doctor level, there is likely unconscious bias when providing care for patients with substance use disorders and those from minority racial/ethnic groups. These biases have been identified in past research for both groups, and this patient population faces combined biases and negative views because of their race/ethnicity and substance use. Doctors may be slow to prescribe opioids due to worries about misuse, selling them, or making substance use worse, even though not treating pain enough can make things worse for patients. At the patient level, it is worth noting that Asian patients received the fewest opioids in both the main analysis and the special group analyses. There might be differences among racial/ethnic groups in how pain is felt or shown, how willing or able patients are to tell a doctor about their pain, and if they want to take opioids or higher doses. However, research in this area has found that patient worries about opioids are more likely caused by not getting enough treatment, rather than being the cause of undertreatment. Communication between doctors and patients is very important. Patients from minority racial groups are more likely to have trouble speaking English. While the study did not find language ability to be a strong factor linked to opioid use, there might still be cultural issues in how doctors and patients talk that affect pain checks and treatment. At the healthcare system level, there has been a national push to reduce opioid prescriptions. The study hospital was located in a city severely affected by the opioid crisis, with a record number of overdoses during the study period. This background likely influenced doctor's choices on how strongly to treat pain with opioids, but it does not fully explain why pain treatment differs for different racial/ethnic groups.

Several other important findings were noted. In the study models, the strongest indicators of receiving higher opioid doses in the hospital were, as expected, having a prescription for MOUD or other opioids before admission. These patients likely needed more opioids because they were used to them, or felt pain more strongly, or were experiencing withdrawal. The subgroup analysis that removed these patients still showed lower opioid doses in the hospital but persistent unfair differences among racial/ethnic groups. In the overall model, the average pain score was not linked to getting opioids, but it did show a positive link in the subgroup models. As supported by research, a patient's reported pain is only one of many things that affect the decision to prescribe opioids, especially for those with a substance use disorder. There are limitations to consider. First, this was a study at only one hospital, so the results might not apply everywhere. Second, billing codes were used to identify patients with substance use disorders, which might not always be perfectly accurate in showing the true number of cases. Third, in the main model, MMEs included methadone and buprenorphine, which can be used for both pain control and addiction treatment. It was not possible to tell the exact reason these medications were given from the available data. Therefore, the second special group analysis, which removed all patients who received these medications, was performed to reduce this issue. Despite these limitations, the findings are new and important for general medicine, fair health care, and substance use. Patients from minority groups who also have a substance use disorder are especially at risk for unfair pain treatment in the hospital. Future studies should look at the exact reason each opioid was given and use doctor diagnoses for substance use disorders. More detailed analyses comparing different substance use disorders are needed to fully understand why these differences exist. There are also plans to examine the detailed connection between individual pain checks and when medicine was given in various situations, which would require complex statistical analysis over time to account for different doctors, how long medicines were taken, and the timing between pain scores and prescribing decisions. The consistency of all the findings across the overall group, the substance use disorder subgroup, and the subgroup without buprenorphine/methadone highlights the importance of developing hospital programs to encourage fair, culturally sensitive pain care for disadvantaged groups. Possible ways to do this include teaching doctors about their unconscious biases, creating better communication tools for patients and doctors, setting standard ways to check and treat pain, and putting in place hospital rules that track and promote fair pain care practices. Ultimately, fixing these unfair differences is crucial for improving the quality of care and patient outcomes for everyone, no matter their race, ethnicity, or history of substance use.

INTRODUCTION

Managing pain fairly for adults with substance use problems, especially those from minority groups, can be difficult. Patients who use substances, like opioids, often need more pain medicine or may feel more pain than others. They might also be going through withdrawal in the hospital. Research shows that people from minority groups do not always get the same pain care.

Past studies found that doctors and these patients sometimes do not trust each other. Doctors might worry about giving too much opioid medicine, fearing problems like drug misuse. This study aimed to see if patients from minority groups get less pain medicine in the hospital, even when they have a substance use problem. Researchers believed this was true and looked at how much opioid medicine different racial and ethnic groups received while in the hospital.

METHODS

Study Population

This study looked back at medical records of adult patients aged 18 and older. The records were from hospital stays between January 1, 2021, and December 31, 2022. All patients were cared for by the general medicine team at a large university hospital. Information came from the hospital's electronic health records. Only hospital stays with full pain reports from patients were included. Patients were not included if they were in intensive care or getting comfort care, as their pain needs are different. The study was approved by the ethics board.

Predictor/Exposure

The main thing looked at was the patient's self-reported race or ethnicity. The study used categories like Asian, Black or African American, Latino, and others, with White patients as a comparison group. These groups are used to understand how race and ethnicity connect to fair healthcare and can help guide future studies on why differences exist.

Outcome

The main result measured was the average amount of opioid medicine a patient received each day in the hospital. This was measured in a standard unit that compares different opioid medicines.

Covariates

The study also looked at other factors that could affect pain medicine amounts. These included:

• Patient information: age, sex, if they needed a language interpreter, and type of health insurance.

• Health information: how sick the patient was overall and how long they stayed in the hospital.

• Substance use information: if the patient had a medical record of a substance use problem. This also included if they were already taking medicine for opioid use disorder (like methadone) before coming to the hospital.

• Pain information: the patient's average pain score during their stay, if they had strong pain when admitted, if a pain doctor was called in, if they took opioids before coming to the hospital, and how much over-the-counter pain medicine they received.

Nurses recorded pain levels throughout a patient's hospital stay.

Statistical Analysis

All data was analyzed using a special computer program. To see how different groups managed pain, the study first looked at the average daily opioid amounts for each race and ethnicity and other factors. For a more complete look, the study used a special math method to account for all factors at once. This method also considered that some patients might have had more than one hospital stay. The goal was to see if race/ethnicity alone was linked to different amounts of opioid medicine, after considering everything else.

Subgroup Analyses

Two extra analyses were done:

1. The first looked only at patients who had a substance use problem. This helped focus on how race/ethnicity affected opioid medicine for this specific group.

2. The second left out patients who were taking specific medicines (methadone or buprenorphine) before or during their hospital stay. This helped to remove any confusion from these medicines, which can be used for both pain and opioid use disorder.

RESULTS

The study included records from 9,102 patients and 13,058 hospital stays between 2021 and 2022. The average age of patients was about 63 years, and just over half were male. About 43% of patients were White, 23% Asian, 13% Latino, and 12% Black or African American. Almost a third of patients had a substance use problem. The average hospital stay was about 7 days.

Primary Analysis

The study first looked at the amounts of opioid medicine without adjusting for other factors. It found clear differences among racial and ethnic groups. Black or African American patients received the most opioid medicine, while Asian patients received the least.

After considering other factors like patient age, health problems, substance use, and pain levels, there were still important differences in opioid medicine amounts among racial and ethnic groups. Most minority groups received less opioid medicine compared to White patients. For example, Asian, Black, and Latino patients all received less. Native American/Alaska Native patients and Native Hawaiian/Pacific Islander patients received even less. Other factors linked to getting more opioid medicine included having a substance use problem, taking certain medicines before the hospital, or getting a pain doctor to help. Patients with high pain when admitted did not always get more opioid medicine.

Subgroup Analysis 1: Patients with SUD

This part of the study looked only at the 2,846 patients who had a substance use problem. This group had more Black or African American patients and fewer Asian patients compared to the overall study. Common substance use problems included tobacco, alcohol, and opioids. When looking at only this group, the results were similar: most minority groups received less opioid medicine compared to White patients. For example, Asian, Black, and Latino patients with substance use problems received even less opioid medicine than in the main study.

Subgroup Analysis 2: Excluding All Patients on Methadone and/or Buprenorphine

This second analysis looked at 8,612 patients, leaving out those who received methadone or buprenorphine. The patients in this group were similar to the main study group. The findings were also similar: most minority groups received less opioid medicine compared to White patients. For example, Asian, Black, and Latino patients all received less. The overall amounts of opioid medicine were lower in this group, but the differences among racial and ethnic groups remained.

DISCUSSION

This study, which looked at over 13,000 hospital stays, found clear differences in how pain medicine was given based on a patient's race or ethnicity. These differences remained even after considering substance use problems, other health issues, and how much pain a patient reported. The study also showed that patients with more than one reason for being overlooked, such as being from a minority group and having a substance use problem, faced even bigger differences in care.

The study's findings support the idea that minority groups received much less opioid medicine than White patients. This was true for Asian, Black, Latino, and other groups. These differences were even greater when looking only at patients with substance use problems. The study also found similar, large differences even when certain pain medicines (methadone or buprenorphine) were not counted. These are important findings: even when many factors were considered, minority patients still received less pain medicine in the hospital.

The differences found were not just small. For example, Asian patients with substance use problems received 124 units less of daily opioid medicine than White patients, which is like getting 16 fewer strong pain pills each day. These results match other studies that have shown unfair pain care for minority groups and those with substance use problems. This study adds to that knowledge by focusing on hospitalized patients who have both.

There are several reasons why these differences might happen. Doctors may have biases when treating patients with substance use problems or those from minority groups. These patients often face unfair treatment because of both their race and their substance use. Doctors might be careful about giving opioids, even if not enough pain medicine can make problems worse. Patients from minority groups might also express or report pain differently, or be less willing to take strong pain medicine. However, research suggests that patient concerns about opioids are often due to not getting enough treatment, rather than causing it.

This study showed that having certain pain medicines (like methadone or buprenorphine) or prior opioid prescriptions was linked to getting more opioid medicine in the hospital. However, even when these patients were removed from the study, the racial and ethnic differences still existed. It is important to know that this study only looked at one hospital, so the results might be different elsewhere. Also, the study used billing codes to identify substance use problems, which might not always be exact.

Despite these points, the study's findings are important. They show that patients from minority groups who also have substance use problems are especially at risk for unfair pain management in hospitals. These findings highlight the need for hospitals to improve how they provide pain care. Future efforts should focus on helping doctors understand their biases, improving how doctors and patients talk about pain, creating clear rules for pain care, and tracking if pain care is fair for all patients. Making care fair for everyone, no matter their race, ethnicity, or history with substance use, is key to better health for all.