Sociogeographic determinants of rapid opioid reduction or discontinuation among patients on high-dose long-term opioid therapy in North Carolina, 2006–2018

Abstract

Objective

Rapid opioid reduction or discontinuation among patients on high-dose long-term opioid therapy (HD-LTOT) is associated with increased risk of heroin use, overdose, opioid use disorder, and mental health crises. We examined the association of residential segregation and health care access with rapid opioid reduction or discontinuation among patients on HD-LTOT and examined effect measure modification of individual-level characteristics.

Methods

Using 2006–2018 North Carolina private insurance claims data, we conducted a retrospective cohort study of patients who were 18–64 years of age and on HD-LTOT (≥90 morphine milligram equivalents for 81 of 90 consecutive days), with 1-year follow-up. The outcome was rapid opioid reduction or discontinuation (versus maintenance, increase, or gradual reduction/discontinuation). Individual-level characteristics included age, sex, and clinical diagnoses (post-traumatic stress disorder [PTSD], depression, anxiety, and substance use disorder). Neighborhood-level characteristics included health care access (measured as geographic distance to health care facilities) and residential segregation (operationalized with the Index of Concentration at the Extremes). We conducted bivariate linear regression to estimate 1-year risk differences and 95% confidence intervals.

Results

Of 13 375 patients on HD-LTOT, 48.6% experienced rapid opioid reduction or discontinuation during 1-year follow-up. Female patients and those diagnosed with PTSD who lived in areas of least racial and economic privilege had higher risks of rapid opioid reduction or discontinuation than did those living in areas with the most racial and economic privilege.

Conclusion

Health care providers need to address potential biases toward patients living in underserved and marginalized communities, as well as intersectionality with mental health stigma, by prioritizing training and education in delivering unbiased care during opioid tapering.

Introduction

The opioid overdose epidemic has been a major public health crisis in the United States for more than 2 decades.¹ In 2016, the US Centers for Disease Control and Prevention (CDC) published opioid prescribing guidelines for chronic non-cancer pain that encouraged clinicians to prescribe the lowest effective dosage of opioids and consider discontinuation if benefits did not outweigh risks.² The 2016 CDC guidelines recommended gradual reduction or discontinuation (10% per week) to minimize symptoms and signs of opioid withdrawal.² Subsequently updated in 2022, these guidelines specifically advised against rapid opioid reduction or discontinuation, especially for patients on higher dosages.³ In response, legislative efforts to limit opioid prescriptions have significantly increased across the United States since 2016, with 38 states implementing laws by 2021 to prevent opioid overdoses.⁴ However, these prescribing policies have unintended consequences, including an increase in rapid opioid reduction or discontinuation among patients on high-dose, long-term opioid therapy (HD-LTOT), which is associated with increased risk of heroin use, nonfatal or fatal overdose, opioid use disorder, and mental health crises.^5–16

Examining the complexities of disparities in opioid prescribing, a study in a large, urban academic health care system found that Black patients on opioid therapy were more likely to have their opioids reduced than were White patients.¹⁷ Although extensive research has focused on individual-level factors in understanding racial and ethnic disparities in pain management, individual-level factors alone are insufficient in explaining the processes through which neighborhood characteristics can worsen health inequities.^18,19 Prior research has found that residential segregation and downstream inequities in access to health care are fundamental drivers of health disparities.^20,21 To date, no studies have investigated the impact of residential segregation and health care access within the context of rapid opioid reduction or discontinuation. Furthermore, research suggests that female patients and patients with mental health conditions, such as depression, have higher risks of rapid opioid reduction or discontinuation.^6,12,17 However, the extent to which these associations might vary in underserved and marginalized communities is unknown.

To address these knowledge gaps, we aim to (1) examine how residential segregation and health care access are associated with the risk of rapid opioid reduction or discontinuation among patients on HD-LTOT and (2) examine how the associations of individual-level characteristics, such as age, sex, and mental health diagnosis, and neighborhood-level characteristics, such as health care access, with rapid opioid reduction or discontinuation are modified by residential segregation.

Methods

Study data and population

Using deidentified insurance claims data from a large private health insurance provider in North Carolina, we conducted a retrospective cohort study of adults (18–64 years of age) prescribed HD-LTOT between January 1, 2006, and September 30, 2018. HD-LTOT was defined as ≥90 morphine milligram equivalents (MME) for at least 90% of 90 consecutive days (ie, 81 out of 90 days). We created a closed cohort, where eligible individuals could enter the cohort on an ongoing basis but must have been followed up for at least 1 year after meeting HD-LTOT eligibility. Patients were excluded if they disenrolled or died within the 1-year follow-up. The HD-LTOT cohort has been described in greater detail in a prior study.¹⁰

Individual-level characteristics

Individual-level demographic characteristics included age (categorized as 18–24, 25–34, 35–44, 45–54, and 55–64 years) and sex (male, female; determined from enrollment information). Individual clinical characteristics included diagnosis of mental health conditions (post-traumatic stress disorder [PTSD] with or without depression or anxiety [hereafter PTSD], depression and anxiety but not PTSD, depression only, anxiety only, no PTSD, depression, or anxiety [hereafter no diagnosis]), diagnosis of opioid use disorder (OUD) or substance use disorder (SUD) (SUD but no OUD, OUD, neither), and history of nonfatal opioid overdose (yes, no). Individual-level clinical characteristics were identified via the International Classification of Diseases (ICD), Ninth Revision, Clinical Modification (ICD-9-CM) or ICD, Tenth Revision, Clinical Modification (ICD-10-CM) (Supplementary Material, eTable S1) during a 1-year look-back period. All individual-level characteristics were time-fixed and assessed at baseline.

Neighborhood-level characteristics

Neighborhood-level characteristics were assessed with 5-digit ZIP codes of residence at baseline (obtained from member insurance files) and residential segregation and distances to the nearest pharmacy, mental health provider, substance use treatment program, opioid treatment program, and hospital.^22–24

Residential segregation was conceptualized as a proxy for structural racism and operationalized with the Index of Concentration at the Extremes (ICE).^20,23 ICE measures the extent to which an area’s residents are concentrated at the extremes of privilege and disadvantage (Supplementary Material, eMethods 1). ICE scores range from −1 to 1, where −1 means 100% of the area’s residents are concentrated at the extremes of disadvantage and 1 means 100% of the area’s residents are concentrated at the extremes of privilege.²³ We included 3 ICE measures for segregation by income only, race/ethnicity only, and income and race/ethnicity jointly. ICE income measures economic segregation, comparing high income (above 20th income percentile) versus low income (below 80th income percentile). ICE race and ethnicity measures racial segregation, comparing high racial privilege (White, non-Hispanic) versus low racial privilege (Black, non-Hispanic). ICE income+ race and ethnicity jointly measures racial and economic segregation, comparing high racial privilege and high income (White, non-Hispanic high income) versus low racial privilege and low income (Black, non-Hispanic low income).²³ Race and ethnicity were conceptualized as a social construct that measures exposure to racism.²⁵ We calculated ICE scores by residential ZIP code using American Community Survey (Census) data and merged them with the HD-LTOT cohort using a 1-year lag to establish temporality, considering a latency period between residing in a segregated area and experiencing adverse health outcomes.²⁶ ICE scores were categorized into 5 quantiles as per the score distributions, where quantile 1 (Q1) was the least privileged and quantile 5 (Q5) was the most privileged²³ (Supplementary Material, eTable S2).

Distances to the nearest pharmacy, mental health provider, substance use treatment program, opioid treatment program, and hospital were obtained from the Opioid Environment Policy Scan data warehouse and conceptualized as proxies for health care access.^24,27 Pharmacies included drugstores that retailed prescription and nonprescription medication.²⁸ Substance use treatment programs included facilities that offered medical and mental health services, including behavioral counseling, medication, treatment for withdrawal symptoms, and continuous monitoring to prevent long-term relapse.²⁹ Mental health providers included facilities that offered therapy and medication to treat mental health conditions and increase behavioral functioning.²⁹ Opioid treatment programs included facilities that dispensed US Food and Drug Administration (FDA)–approved medications for OUD treatment.²⁹ Hospitals included short-term acute care hospitals, critical access hospitals, rehabilitation hospitals, and psychiatric hospitals.³⁰ Distance to the nearest health care facility was defined as the Euclidean distance from the ZIP code centroid to the nearest health care facility in miles.³¹ Distance measures were categorized into 5 quantiles as per the value distributions, where quantile 1 (Q1) was the most geographic access and quantile 5 (Q5) was the least geographic access (Supplementary Material, eTable S3). All neighborhood-level characteristics were time-fixed and assessed at baseline.

Outcome

We created a dichotomous measure of ever rapid tapering or discontinuation of opioids (versus maintenance or increase or gradual reduction or discontinuation) during the 1-year follow-up period after HD-LTOT eligibility was met. Using a previously developed algorithm, we classified each 30-day period of a patient’s prescription trajectories as dose maintained, increased, gradually reduced, rapidly reduced, gradually discontinued, or rapidly discontinued.¹⁰ The algorithm compared the mean dose of the current month with the mean dose of the previous month and the 6-month rolling average dose to classify monthly prescription trajectories. Gradual tapering was defined as opioid dose reduction of less than 34% per month (10% per week), whereas rapid tapering was defined as dose reduction exceeding 34% per month.² Rapid discontinuation was defined as stoppage of opioid prescriptions when the daily MME before discontinuation was ≥30 MME.¹⁰

Statistical analysis

We conducted bivariate linear regression analyses to estimate 1-year risk differences (RDs) and 95% confidence intervals (CIs) for the association between individual- and neighborhood-level characteristics and rapid opioid reduction or discontinuation. We reported confidence limit differences (CLDs) as a measure of precision.³²

We examined effect measure modification (EMM) by racial segregation by using stratum-specific RDs and 95% CIs. Although overlap in stratum-specific 95% CIs might suggest lack of statistical significance (P > .05), we reported and interpreted EMM estimates with overlapping CIs to identify and highlight clinically meaningful differences in the risk of rapid opioid reduction or discontinuation.^33–36 We considered differences of at least 2.0% between strata-specific estimates as clinically meaningful, whereas for main effects, we highlighted estimates that were at least 1.0%.

IRB statement

This study was reviewed and approved by the University of North Carolina at Chapel Hill Institutional Review Board (#23-1712).

Results

The cohort consisted of 13 375 patients on HD-LTOT with 1 year of follow-up (Table 1). Over the 1-year risk period, 6502 patients (48.6%) had their opioids rapidly reduced or discontinued. More than half of the patients on HD-LTOT (51.9%) were male, and the majority (63.0%) were 45–64 years of age. At baseline, 1.7% had a PTSD diagnosis, 12.0% had comorbid depression and anxiety without PTSD, 13.1% had depression alone, and 12.0% had anxiety alone. Additionally, 11.1% of patients had SUD or OUD (5.5% with SUD but no OUD and 5.6% with OUD).

Table 1.

Outcome distribution and sample characteristics: patients on high-dose long-term opioid therapy in North Carolina, 2006–2018.

	n	%
Outcome
Maintenance or increase or gradual tapering or discontinuation	6873	51.4
Rapid reduction or discontinuation	6502	48.6
Sex
Male	6935	51.9
Female	6440	48.1
Age, years
18–24	204	1.5
25–34	1506	11.3
35–44	3235	24.2
45–54	5049	37.7
55–64	3381	25.3
Mental health diagnosis a
PTSD b	231	1.7
Depression and anxiety but no PTSD	1610	12.0
Depression only	1748	13.1
Anxiety only	1609	12.0
No diagnosis c	8177	61.1
SUD/OUD diagnosis a
SUD but no OUD	736	5.5
OUD	743	5.6
No SUD	11 896	88.9

Abbreviations: OUD opioid use disorder; PTSD= post-traumatic stress disorder; SUD= substance use disorder.

^aAssessed with a 12-month look-back period.

^bPTSD with or without depression or anxiety.

^cNo PTSD, depression, or anxiety.

We found that patients who lived in areas with the most racial privilege had 2.9% lower risk (95% CI: −5.6 to −0.2), whereas patients who lived in areas with the most economic privilege had 3.7% higher risk (95% CI: 1.0 to 6.3), than that of patients who lived in areas with the least privilege (Table 2). However, patients who lived in areas with the highest levels of both racial and economic privilege had risks comparable to those in areas with the least privilege. Patients who lived in areas with the least geographic access to health care facilities had lower risks of rapid opioid reduction or discontinuation than those of patients who lived in areas with the most geographic access, with the largest magnitude of risk difference observed for geographic access to substance use treatment (RD −3.2% [95% CI: −5.9 to −0.5]).

Table 2.

Bivariate associations between neighborhood-level characteristics and rapid opioid reduction or discontinuation among patients on high-dose long-term opioid therapy in North Carolina, 2006–2018.

	Total (a + b) (%)	Maintenance or increase or gradual tapering or discontinuation (a)	Rapid reduction or discontinuation (b)	1-Year risk [b/(a + b)]	1-Year RD (95% CI); CLD
Racial segregation (ICE race)
Least racial privilege (Q1)	2682 (20.1%)	1307	1375	51.3%	REF
Most racial privilege (Q5)	2666 (20.0%)	1376	1290	48.4%	−2.9% (−5.6 to −0.2); 5.4
Economic segregation (ICE income)
Least economic privilege (Q1)	2672 (20.0%)	1377	1295	48.5%	REF
Most economic privilege (Q5)	2669 (20.0%)	1278	1391	52.1%	3.7% (1.0 to 6.3); 5.3
Racial and economic segregation (ICE race+ income)
Least racial and economic privilege (Q1)	2673 (20.0%)	1320	1353	50.6%	REF
Most racial and economic privilege (Q5)	2653 (19.9%)	1294	1359	51.2%	0.6% (−2.1, 3.3); 5.4
Distance to nearest pharmacy
Least access (Q5; 4.4–28.4 miles)	2641 (19.7%)	1379	1262	47.8%	−2.2% (−4.9, 0.4); 5.3
Most access (Q1; ≤0.9 miles)	2713 (20.3%)	1356	1357	50.0%	REF
Distance to nearest mental health provider
Least access (Q5; 13.0–70.4 miles)	2657 (19.9%)	1395	1262	47.5%	−1.9% (−4.6, 0.8); 5.4
Most access (Q1; ≤2.9 miles)	2675 (20.0%)	1353	1322	49.4%	REF
Distance to nearest substance use treatment
Least access (Q5; 8.0–35.3 miles)	2669 (20.0%)	1422	1247	46.7%	−3.2% (−5.9,−0.5); 5.4
Most access (Q1; ≤1.9 miles)	2693 (20.1%)	1349	1344	49.9%	REF
Distance to nearest opioid treatment program
Least access (Q5; 15.1–54.4 miles)	2669 (20.0%)	1386	1283	48.1%	−1.6% (−4.2, 1.1); 5.3
Most access (Q1; ≤4.0 miles)	2713 (20.3%)	1366	1347	49.6%	REF
Distance to nearest hospital
Least access (Q5; 12.4–66.8 miles)	2661 (19.9%)	1404	1257	47.2%	−2.3% (−5.0, 0.4); 5.4
Most access (Q1; ≤3.5 miles)	2698 (20.2%)	1361	1337	49.6%	REF

Abbreviations: CI= confidence interval; CLD= confidence limit difference; ICE= Index of Concentration at the Extremes; RD= risk difference; REF= reference.

We observed clinically meaningful EMM of the associations between individual characteristics and rapid opioid reduction or discontinuation by racial and economic segregation (Table 3). Among patients living in areas with the least racial and economic privilege, female sex (versus male sex) was associated with 5.5% higher risk (95% CI: 1.7 to 9.3) of rapid opioid reduction or discontinuation, whereas, among patients living in areas with the most racial and economic privilege, this association was weaker (RD: 3.0% [95% CI: −0.8 to 6.8]). Furthermore, among patients living in areas with the least racial and economic privilege, age 18–24 years (versus 35–64 years) was associated with 19.1% higher risk (95% CI: 4.2 to 31.9), with a stronger association observed among patients living in areas with the most racial and economic privilege (RD 28.3% [95% CI: 16.1 to 38.1]). In addition, among patients living in areas with the least racial and economic privilege, PTSD diagnosis (versus no diagnosis) was associated with 11.0% higher risk (95% CI: −5.2 to 26.0) of rapid opioid reduction or discontinuation, whereas, among patients living in areas with the most racial and economic privilege, this association was null (RD −0.2% [95% CI: −13.1 to 12.9]). Moreover, among patients living in areas with the least racial and economic privilege, depression and anxiety but no PTSD diagnosis (versus no diagnosis) was associated with 8.8% higher risk (95% CI: 2.8 to 14.7), with stronger association observed among patients living in areas with the most racial and economic privilege (RD 15.8% [95% CI: 10.1 to 21.4]).

Table 3.

Effect measure modification by racial segregation in the bivariate associations between individual- and neighborhood-level characteristics and rapid opioid reduction or discontinuation.

	Least racial and economic privilege (Q1)		Most racial and economic privilege (Q5)
	1-Year risk	1-Year RD (95% CI); CLD	1-Year risk	1-Year RD (95% CI); CLD
Individual characteristics
Sex
Male	47.9%	REF	49.7%	REF
Female	53.4%	5.5% (1.7 to 9.3); 7.6	52.8%	3.0% (−0.8 to 6.8); 7.6
Age
18–24 years	69.0%	19.1% (4.2 to 31.9); 27.7	78.2%	28.3% (16.1 to 38.1); 22.0
25–34 years	53.6%	3.7% (−2.5 to 9.9); 12.4	56.2%	6.4% (0.5 to 12.1); 11.6
35–64 years	49.9%	REF	49.9%	REF
Mental health diagnosisa
PTSDb	59.5%	11.0% (−5.2 to 26.0); 31.2	47.4%	−0.2% (−13.1 to 12.9); 26.0
Depression and anxiety but no PTSD	57.2%	8.8% (2.8 to 14.7); 11.9	63.4%	15.8% (10.1 to 21.4); 11.3
Depression only	49.8%	1.4% (−4.5 to 7.4); 11.9	54.2%	6.6% (1.0 to 12.2); 11.2
Anxiety only	55.3%	6.9% (0.9 to 12.8); 11.9	53.3%	5.7% (−0.3 to 11.7); 12.0
No diagnosisc	48.4%	REF	47.6%	REF
SUD/OUD diagnosisa
SUD but no OUD	53.4%	3.6% (−4.0 to 11.1); 15.1	60.6%	10.7% (2.6 to 18.4); 15.8
OUD	59.8%	10.0% (2.1 to 17.6); 15.5	59.7%	9.7% (2.1 to 17.1); 15.0
No SUD	49.8%	REF	50.0%	REF
Neighborhood characteristics
Distance to nearest pharmacy
Least access (Q5; 4.4–28.4 miles)	50.7%	4.7% (−2.1 to 11.4); 13.5	50.5%	−4.1% (−10.4 to 2.3); 12.7
Most access (Q1; ≤0.9 miles)	46.1%	REF	54.6%	REF
Distance to nearest mental health provider
Least access (Q5; 13.0–70.4 miles)	50.7%	2.3% (−3.1 to 7.7); 10.8	49.0%	−7.3% (−13.8 to −0.7); 13.1
Most access (Q1; ≤2.9 miles)	48.4%	REF	56.3%	REF
Distance to nearest substance use treatment
Least access (Q5; 8.0–35.3 miles)	46.4%	−3.9% (−9.6 to 1.9); 11.5	49.7%	−5.4% (−11.3 to 0.5); 11.8
Most access (Q1; ≤1.9 miles)	50.3%	REF	55.1%	REF
Distance to nearest opioid treatment program
Least access (Q5; 15.1–54.4 miles)	50.4%	−1.3% (−6.8 to 4.2); 11.0	47.1%	−6.9% (−14.6 to 0.8); 15.4
Most access (Q1; ≤4.0 miles)	51.7%	REF	54.0%	REF
Distance to nearest hospital
Least access (Q5; 12.4–66.8 miles)	49.0%	−0.2% (−5.7 to 5.3); 11.0	48.1%	−6.7% (−13.0 to −0.3); 12.7
Most access (Q1; ≤3.5 miles)	49.2%	REF	54.7%	REF

Abbreviations: CI = confidence interval; CLD = confidence limit difference; ICE = Index of Concentration at the Extremes; OUD = opioid use disorder; PTSD = post-traumatic stress disorder; RD = risk difference; REF = reference; SUD = substance use disorder.

^aAssessed with a 12-month look-back period.

^bPTSD with or without depression or anxiety.

^cNo PTSD, depression, or anxiety.

When examining the associations between neighborhood-level characteristics and rapid opioid reduction or discontinuation by racial and economic segregation, we found that among patients living in areas with the least racial and economic privilege, living in areas with least geographic access to a pharmacy or a mental health provider was associated with 4.7% (95% CI: −2.1 to 11.4) and 2.3% (95% CI: −3.1 to 7.7) higher risks, respectively, whereas, among patients living in areas with the most racial and economic privilege, living in areas with least geographic access to a pharmacy or mental health provider was associated with 4.1% (95% CI: −10.4 to 2.3) and 7.3% (95% CI: −13.8 to −0.7) lower risks, respectively. Moreover, among patients living in areas with the least racial and economic privilege, living in areas with the least geographic access to a substance use treatment program or opioid treatment program was associated with 3.9% (95% CI: −9.6 to 1.9) and 1.3% (95% CI: −6.8 to 4.2) lower risks, with stronger associations observed among patients living in areas with the most racial and economic privilege (substance use treatment: RD 5.4% [95% CI: −11.3 to 0.5]; opioid treatment program: RD 6.9% [95% CI: −14.6 to 0.8]). Furthermore, among patients living in areas with the least racial and economic privilege, living in areas with the least geographic access to a hospital was not associated with rapid opioid reduction or discontinuation (RD −0.2% [95% CI: −13.1 to 12.9]), whereas, among patients living in areas with the most racial and economic privilege, living in areas with the least geographic access to a hospital was associated with 6.7% (95% CI: −13.0 to −0.3) lower risk.

Discussion

In a cohort of privately insured patients prescribed HD-LTOT in North Carolina, we found that patients who lived in areas with the least racial privilege had higher risk of rapid opioid reduction or discontinuation than did those who lived in areas with the most racial privilege. This finding is consistent with a large body of literature that suggests that racial segregation is associated with adverse health outcomes, with racially minoritized people disproportionately experiencing the negative effects of segregation.^20,37,38 Additionally, we found that patients who lived in areas with the least economic privilege or had least geographic access to health care facilities had lower risk of rapid opioid reduction or discontinuation. These findings are counterintuitive and require further investigation, as research suggests that income inequality and barriers to health care access are associated with adverse health outcomes.^39,40

Furthermore, we found that although racial and economic residential segregation alone is not associated with rapid opioid reduction or discontinuation, when neighborhood-level residential segregation interacts with individual-level sociodemographic characteristics, we observe evidence of effect measure modification. Prior research suggested that female patients have higher risk of rapid opioid reduction or discontinuation.^6,17 Our research extends this research and finds that female patients who live in areas with the least racial and economic privilege, generally communities with greater proportions of people of color, have higher risk than those who live in areas with the most racial and economic privilege. Although disparities in pain management and care among females are well documented, there is also evidence that these inequities disproportionately impact women of color.^41–43 For example, in postpartum care settings, Hispanic and non-Hispanic Black women received fewer inpatient MMEs/day, despite reporting higher pain scores, and were less likely to receive an opioid prescription at discharge than were non-Hispanic White females.⁴² Empowering health care providers with knowledge about potential implicit biases toward patients living in poorer neighborhoods and communities of color, particularly how these biases could be further exacerbated by gender biases in pain management, is essential for establishing equitable practices in opioid tapering.

Additionally, we found that the risk of rapid opioid reduction or discontinuation of opioids is disproportionately higher for patients diagnosed with PTSD in areas with the least racial and economic privilege than for those in areas with the most racial and economic privilege. Research suggests that Black patients in the VA were less likely to receive a minimal trial of pharmacotherapy and were less likely to receive any minimal trial of treatment after being diagnosed with PTSD.⁴⁴ Although our study does not directly examine provider bias in clinical decision-making, these findings, interpreted in the context of prior research, potentially suggest the presence of implicit biases toward pain patients residing in areas with lower racial and economic privilege, which might be further exacerbated by stigma toward patients with PTSD.⁴⁵ Addressing these disparities requires health care providers to acknowledge potential stigma or biases and undertake training to ensure the delivery of equitable care when considering opioid tapering in underserved and marginalized communities.⁴⁶

Furthermore, we observed that patients living in areas with the least racial and economic privilege have higher risk of rapid opioid reduction or discontinuation when they have poor access to pharmacies or mental health providers. However, patients living in areas with more racial and economic privilege have lower risks of rapid opioid reduction or discontinuation despite having poor access to pharmacies, mental health providers, substance use treatment, opioid treatment programs, or hospitals, possibly because of better health care quality and infrastructure that mitigate the barriers associated with limited geographic access. Future research should consider how opioid prescribing guidelines are interpreted and implemented across different neighborhoods and regions in the United States to ensure that these policies do not inadvertently harm people who live in underserved and marginalized communities.

Limitations

The findings of this study should be interpreted in light of limitations. First, we did not examine the impact of the 2016 CDC guidelines that recommended gradual opioid reduction or discontinuation to minimize symptoms and signs of opioid withdrawal.² Second, we have not applied covariate adjustment to any of the associations presented in this study; therefore, the findings should not be interpreted as causal estimates. Future studies should consider covariate adjustment and apply other causal inference methods. Third, the generalizability of the study is limited to privately insured patients on HD-LTOT who are 18–64 years of age and in North Carolina and might not extend to patients on HD-LTOT who are publicly insured, uninsured, or older. Fourth, clinical diagnoses, such as depression and SUD, can be vulnerable to measurement errors, as they are frequently underreported in health care data.^47,48 Fifth, ZIP codes are spatial units that are both large and heterogenous, and they might not accurately align with social or neighborhood boundaries.⁴⁹ Sixth, distances to nearest health care facilities are used as proxies for health care access and estimated by using Euclidean distance between ZIP code centroids, which although commonly used in health service research, tends to underestimate distances.⁵⁰ Lastly, we present CLDs in the results tables as measures of precision. Although several estimates have low precision due to small cell sizes in some stratifications, CLDs should not be misinterpreted as measures of statistical significance. The magnitude of estimates is the only measure of clinical significance.

Conclusion

In a cohort of privately insured patients prescribed HD-LTOT in North Carolina, we found that female patients and those diagnosed with PTSD who live in areas of least racial and economic privilege have higher risks of rapid opioid reduction or discontinuation than do those living in areas with the most racial and economic privilege. To ensure equitable care across underserved and marginalized communities, health care providers must actively confront biases, considering their intersectionality with mental health stigma, and prioritize training and education in the delivery of unbiased care during opioid tapering.

Supplementary material

Supplementary material is available at Pain Medicine online.

Funding

This work was supported by grant funding from the National Center for Injury Prevention and Control, Centers for Disease Control and Prevention (R01CE003009, PI: S.I.R.) and the National Institute of Health’s National Institute of Mental Health (R01MH124752, MPIs: B.W.P. and S.I.R.). In addition, the data used for this work was developed with funding from the National Institute of Health’s National Institute for Drug Abuse (R21DA046048, PI: B.W.P.). The database infrastructure used for this project was supported by the Cecil G. Sheps Center for Health Services Research and the CER Strategic Initiative of UNC’s Clinical and Translational Science Award (5-UL1-TR002489-05).

Conflicts of interest: None declared.