Abstract
Objective:
Determine whether employer-mandated transitions from low- to high-deductible health plans (HDHPs) are associated with delays in opioid use disorder (OUD)-related care presentations. Cost-sharing may negatively impact timely diagnosis and treatment of OUD.
Methods:
Using 2003–2017 national commercial insurance claims data, we employed a matched time-to-event and difference-in-differences design to examine the association between employer-mandated transitions from low to HDHPs on OUD-related care presentations. Study group included 574,058 adults aged 18–64 years continuously enrolled in low-deductible (<$500) health plans during a baseline year followed by up to 4 years in HDHPs (≥$1000) after an employer-mandated transition (exposure). Control group included 4,386,636 adults contemporaneously enrolled in low-deductible plans matched on employee and employer characteristics. Outcomes included first OUD-related office visit, buprenorphine pharmacy fill, and OUD-related high-acuity visit. The secondary outcome was the yearly number of high-acuity care days.
Results:
After an employer-mandated HDHP transition, there were no differences in time-to-first OUD-related office visit (HR, 1.02, 95% CI 0.94, 1.11) or buprenorphine fill (HR, 1.05, 95% CI 0.97–1.13) in the HDHP versus control cohort. In contrast, the HDHP transition was associated with delays in time-to-first OUD-related high-acuity visits compared to control members (HR 0.86, 95% CI, 0.79–0.93). HDHP members experienced a 37.4% (95% CI −57.8, −17.0) relative reduction in high-acuity care days relative to the control group from baseline to follow-up.
Conclusions:
Employer-mandated transitions to HDHPs were associated with delays and reductions in OUD-related high-acuity presentations. Such delays and reductions in timely OUD care could lead to adverse health outcomes.
Key words (MeSH terms): Opioid-Related Disorders; Substance-Related Disorders; Opioid Use Disorder; Insurance, Health; Cost-sharing; Health Services Utilization; Health Policy
INTRODUCTION
Opioid overdose is a leading cause of death in the US, killing 80,000 people in 2023.1 People with opioid use disorder (OUD) experience barriers to timely diagnosis and treatment with medications for opioid use disorder (MOUD), including buprenorphine or methadone. Although most OUD-related care is delivered in outpatient settings, emergency department (ED) visits and hospitalizations are also critical touchpoints2 for identifying OUD, initiating MOUD, and linking patients to follow-up care after discharge.3–6
Survey and qualitative studies have identified cost as a barrier to OUD treatment.7–9 Furthermore, out-of-pocket costs have been associated with lower MOUD retention.10,11 Yet, the impact of cost-sharing on initial OUD-related care presentations remains largely unexplored.
Nearly 1 in 3 adults with OUD are commercially insured and subject to benefit design changes.12 High-deductible health plans (HDHPs), the most common benefit designs in the employer-sponsored health insurance market, require members to pay approximately $1000-$7000 annually for most non-preventive care.13 The RAND Health Insurance Experiment (HIE) found that cost-sharing reduced both effective and ineffective health care use, contributing to adverse events among the poorest and sickest participants.14 HDHPs have been associated with delays in breast cancer and diabetes care, particularly among low-income populations.15,16 Few studies using rigorous, quasi-experimental designs have examined how HDHPs impact OUD-related care. Prior work suggests that employer-offered HDHPs are associated with reduced substance use disorder (SUD) service utilization, but not with changes in buprenorphine adherence or rates of non-fatal overdose.17–19
However, no studies have examined the impact of employer-mandated transitions to HDHPs, in which all enrollees are required to adopt HDHP coverage, on initial presentations for OUD care. This question is especially important given the proliferation of HDHPs during a period of rising overdose deaths and the growing likelihood that Medicaid coverage losses from unwinding and federal work requirements will shift some individuals with OUD onto employer-sponsored plans.20,21
Leveraging a natural experiment in which some individuals experience an employer-mandated transition to a HDHP, we examined the association between cost-sharing and timing of initial OUD-related care.22 We hypothesized that employer-mandated transitions to HDHPs are associated with delays in OUD-related office visits and buprenorphine initiation. Drawing on the RAND HIE and previous HDHPs studies, we hypothesized that these associations are concentrated among low-income people. For OUD-related high-acuity care (ED and hospital visits), we acknowledged a priori uncertainty regarding the direction of effect: transitions to HDHPs could lead to relative accelerations in OUD-related high-acuity visits due to delayed outpatient visits and subsequent health deterioration, or to delayed high-acuity visits due to increased out-of-pocket costs discouraging care. We further examined the impact of HDHPs on the volume of OUD-related high-acuity care in secondary analyses. We acknowledge that this study occurred before and at the onset of the synthetic opioid overdose epidemic. However, understanding the extent to which high cost-sharing impacts OUD-related care remains important today given the predominance of HDHPs in employer-sponsored health insurance markets, rising healthcare costs nationally, and persistent care and treatment gaps that people with OUD experience.
METHODS
Study Population:
We conducted a longitudinal, quasi-experimental study of commercially insured adults aged 18–64 years using claims data from a large US commercial (and Medicare Advantage) health plan from 2003–2017. More recent data spanning the proliferation of synthetic opioids was unavailable. These adults were continuously enrolled for 12 baseline months in a low deductible health plan (annual deductible of <=$500). After 12 baseline months, the intervention group (a portion of the full sample of adults) experienced an employer-mandated switch to a HDHP with an annual deductible of >=$1,000 while the control group (the remaining sample of adults) remained in a low deductible health plan (technical appendix in the Supplemental Digital Content). These groups were then followed for 1–48 months. The first month of the follow-up period was considered the index date. We defined the index date for the intervention group as the month of the employer-mandated HDHP switch. For the control group, potential index dates comprised the month employers renew their health plans. This HDHP identification strategy has been described in previous work15,16,23. The Internal Revenue Service (IRS) sets the minimum deductible threshold for health savings account eligible HDHPs: we chose the $1,000 minimum threshold initially set by the IRS in 2005 because this remains a common deductible level. We excluded adults with baseline evidence of cancer or palliative care.
Because lower-income populations are likely to reduce healthcare use following increases in cost-sharing,14,24 a population of interest was members living in higher-poverty neighborhoods. We used 2010–2014 American Community Survey census tract-level data and validated cut points to create 4 neighborhood level poverty categories (0–4.9%, 5.0–9.9%, 10.0–19.9%, and >=20% of households below the federal poverty line, FPL)25,26. Members were considered to live in a higher-poverty or lower-poverty neighborhood if >=10.0% or <10% lived above or below the FPL, respectively. This study was approved by the Harvard Pilgrim Health Care institutional review board. Informed consent was waived because data were deidentified. This analysis followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
Study Design:
To assess timing of OUD-related presentations and buprenorphine initiation among HDHP compared to control group members, we used a matched time-to-event design during the baseline and follow-up periods. Enrollees in both groups were followed for 12 months at baseline and 1–48 follow-up months after the index date. This approach allows for outcome measurement in two periods: a baseline period where all members are enrolled in a low-deductible health plan, and a follow-up period where some members experience an employer-mandated HDHP transition. Outcome differences in the follow-up period are assumed to be associated with the HDHP transition if there are otherwise no differences in the baseline period.
Given the uncertainty about how HDHP might influence timing of high-acuity presentations, we performed secondary analyses using a matched pre-post difference-in-differences design to examine changes in OUD-related days spent in the ED, observation unit, and hospital (termed “high-acuity care days”) among HDHP compared to control group members. In these analyses, all members were followed for 12 baseline and 12 follow-up months to examine changes in the volume (instead of timing) of high-acuity care days. Unlike time-to-event designs, which accommodate loss to follow-up and censoring, pre-post designs require a balanced panel to ensure differential drop-out does not bias estimates.
Outcomes:
Our three primary outcomes were time-to-first: (1) OUD-related office visit, (2) buprenorphine pharmacy fill, and (3) OUD-related high-acuity visit. Our claims-based definition for an OUD-related office visit included the first of two or more office encounters with opioid abuse or dependence ICD9 or 10 codes based on established claims-based definitions (technical appendix in the Supplemental Digital Content).27 Buprenorphine fills were identified at the pharmacy claims-level using National Drug Codes. Methadone maintenance therapy was not covered by the insurer during the study period and was not included as an outcome. Our claims-based definition for an OUD-related high-acuity visit included the first ED, observation unit, or hospitalization encounter with opioid abuse, dependence, or poisoning ICD9 or 10 codes. Our secondary outcome was the number of high-acuity care days for OUD- or opioid overdose-related ED, observation unit, or hospitalization encounters per member per year (technical appendix in the Supplemental Digital Content).
Matching Strategy:
To minimize confounding, we employed Stata’s kmatch procedure to balance groups on key employee- and employer-level characteristics. We employed the procedure’s combined multivariate-distance kernel and propensity score matching followed by entropy balancing to generate matching weights.28 Matching weights equivalently balanced study groups across characteristics. Members were matched on age category (18–24, 25–34, 35–44, 45–54, and 55–64 years old), gender, the index date calendar year, baseline prescription opioid fills, and indicators for baseline benzodiazepine fills and long-term opioid therapy (technical appendix in the Supplemental Digital Content). We additionally matched on employee out-of-pocket and standardized total costs, and baseline quarter indicators for a first OUD-related office visit, buprenorphine fill, and OUD-related high-acuity visit. We used the Johns Hopkins ACG System (version 11.1) to create and match on indicators for baseline diagnoses of any mental illness (unipolar/bipolar depression, anxiety, schizophrenia), substance use disorder, tobacco use, and low back pain.29 We stratified matches by lower- and higher-poverty neighborhood level and performed subgroup analyses on these strata. We also matched on employer firm size (1–49, 50–249, and >=250 employees) given overrepresentation of HDHP offerings among small firms. Cohort selection details including pre-match sample sizes are provided in Figure S1 in the Supplemental Digital Content.
Statistical analysis:
We estimated standardized differences in covariates to assess covariate balance between study groups before and after matching. For the time-to-event design, we generated match-weighted cumulative event rate curves to visually assess outcome differences by group. For our main analyses, we used Cox Proportional Hazards models with match-generated weights to analyze the three time-to-event outcomes in the baseline and follow-up periods separately. In baseline models, patients were censored at baseline period ending. In the follow-up model, patients were censored if they disenrolled, reached age 65, had cancer or palliative care diagnostic codes, or reached the follow-up period end. The term of interest for each study period was a HDHP membership indicator (yes/no) which yielded a hazard ratio (HR) of the HDHP relative to the control group. A statistically significant HR <1.0 was interpreted as the HDHP group experiencing a delay or reduction in the outcome relative to the control group. We analyzed baseline HRs to confirm that the timing of outcomes did not differ between groups before the employer-mandated HDHP transition. No differences in baseline timing (HR ~1.0) strengthens the assumption that outcome differences in the follow-up HRs are attributable to the employer-mandated HDHP transition. We considered P-value < 0.05 to be significant. We tested 9 different models in total (3 outcomes across the full, higher-poverty, and lower-poverty samples). For primary analyses, we additionally report the Bonferroni-corrected threshold for statistical significance (α=0.05÷9) to compare with model-derived P-values.30 Statistical analyses were performed using Stata, version 15.1 (StataCorp), and SAS Studio software, release 3.7, enterprise edition (SAS Institute).
In secondary analyses, we used a person-level difference-in-differences framework to assess for changes in the number of high-acuity care days across the HDHP and control group from baseline to follow-up. We used models with generalized estimating equations and a negative binomial family, log link function, and robust standard errors. The term of interest was an interaction between HDHP membership and period (baseline versus follow-up). We used marginal effects and non-linear combination of estimators to calculate absolute and relative differences in high-acuity care days per year between HDHP and control group members from the baseline to follow up period. We assessed the parallel trends assumption by plotting marginally adjusted quarter and study group interactions in the baseline period to assess for trend differences in high-acuity care days (Figure S2 in the Supplemental Digital Content). There were no significant baseline differences between groups. Finally, we stratified analyses by higher- versus lower-poverty census tract residence then used the same approaches as above.
Sensitivity Analyses:
First, we restricted our sample to those without evidence of long-term opioid therapy in the baseline period given that this population might receive a diagnostic code for opioid dependence but not have OUD.27 Second, time-to-event analyses additionally adjusted for geographic region to account for regional differences in clinical practice and substance use patterns, and for employer-level shares of female employees and employees aged 50 years or older to account for employer propensity to transition to a HDHPs based on employee health care utilization.
RESULTS
Baseline Characteristics:
Our final weighted sample included 574,058 HDHP and 4,386,636 matched control members (Table 1). The mean (SD) age of HDHP and control members was 41.7 (12.6) and 41.7 (12.7) years, respectively. Among both groups, 50.2% lived in higher-poverty neighborhoods, 5.2% had any mental illness diagnoses, and 1.1% had any SUD diagnoses during the baseline period. Among HDHP and control members, 216 (0.04%) and 1,651 (0.04%) attended an initial OUD-related office visit, 313 (0.05%) and 2392 (0.05%) filled an initial buprenorphine prescription, and 202 (0.04%) and 1,544 (0.04%) experienced an initial OUD-related high-acuity visit during the baseline period, respectively.
Table 1:
Baseline Characteristics of the Study Population Before and After Matching and Weighting
| Unmatched (%) | Unmatched (%) | Matched (%) | Matched (%) | |||
|---|---|---|---|---|---|---|
| Control (n=4417515) | HDHP (n=578,029) | Std. diff.a | Control (n=4386636) | HDHP (n=574058) | Std. diff. | |
| Age at index date, year | ||||||
| 18–24 | 647173 (14.7) | 70330 (12.2) | 0.09348 | 533847 (12.2) | 69862 (12.2) | 0 |
| 25–34 | 920992 (20.8) | 113933 (19.7) | 8634689 (19.7) | 112998 (19.7) | ||
| 35–44 | 1082178 (24.5) | 141493 (24.5) | 1073922 (24.5) | 140539 (24.5) | ||
| 45–54 | 1030945 (23.3) | 150581 (26.1) | 1143979 (26.1) | 149707 (26.1) | ||
| 55–64 | 736227 (16.7) | 101692 (17.6) | 771420 (17.6) | 100952 (17.6) | ||
| Femaleb | 2250400 (50.9) | 280600 (48.5) | 0.04811 | 2128572 (48.5) | 278556 (48.5) | 0 |
| Neighborhood residents below poverty line, % | ||||||
| <5.0 | 1066960 (24.3) | 126222 (22) | 0.06285 | 1010056 (23) | 126125 (22) | 0.0408 |
| 5.0–9.9 | 1241384 (28.2) | 159800 (27.8) | 1173634 (26.8) | 159644 (27.8) | ||
| 10.0–19.9 | 1301801 (29.6) | 180759 (31.5) | 1392840 (31.8) | 180563 (31.5) | ||
| >20 | 784200 (17.8) | 107825 (18.8) | 810106 (18.5) | 107726 (18.8) | ||
| Predominant race or ethnicity at neighborhood levelc | ||||||
| White | 2452906 (55.5) | 362361 (62.7) | 0.16034 | 2585785 (59.0) | 361979 (63.1) | 0.1323 |
| Black | 99432 (2.3) | 8274 (1.4) | 91262 (2.1) | 8271 (1.4) | ||
| Mixed | 1103854 (25) | 125864 (21.8) | 1076640 (24.5) | 125730 (21.9) | ||
| Hispanic | 524097 (11.9) | 59436 (10.3) | 455788 (10.4) | 58961 (10.3) | ||
| Asian | 219745 (5) | 19303 (3.3) | 177160 (4.0) | 19117 (3.3) | ||
| Year of index date | ||||||
| 2004 | 582034 (13.2) | 68565 (11.9) | 0.17491 | 547262 (12.5) | 68139 (11.9) | 0.0034 |
| 2005 | 549183 (12.4) | 94475 (16.3) | 696768 (15.9) | 94154 (16.4) | ||
| 2006 | 592273 (13.4) | 69836 (12.1) | 530234 (12.1) | 69547 (12.1) | ||
| 2007 | 497079 (11.3) | 53274 (9.2) | 408148 (9.3) | 53274 (9.2) | ||
| 2008 | 488008 (11) | 54013 (9.3) | 408607 (9.3) | 53757 (9.4) | ||
| 2009 | 383623 (8.7) | 57117 (9.9) | 432900 (9.9) | 56912 (9.9) | ||
| 2010 | 258028 (5.8) | 42322 (7.3) | 323709 (7.4) | 42196 (7.4) | ||
| 2011 | 221110 (5) | 36524 (6.3) | 273565 (6.2) | 36292 (6.3) | ||
| 2012 | 204713 (4.6) | 22875 (4) | 178049 (4.1) | 22756 (4) | ||
| 2013 | 187141 (4.2) | 21243 (3.7) | 161545 (3.7) | 21137 (3.7) | ||
| 2014 | 143707 (3.3) | 18997 (3.3) | 139109 (3.2) | 18638 (3.3) | ||
| 2015 | 110888 (2.5) | 14460 (2.5) | 112047 (2.6) | 14324 (2.5) | ||
| 2016 | 105003 (2.4) | 13236 (2.3) | 95167 (2.2) | 12741 (2.2) | ||
| 2017 | 94725 (2.1) | 11092 (1.9) | 79528 (1.8) | 10407 (1.8) | ||
| Region | 0.2765 | 0.2136 | ||||
| Northeast | 567422 (12.8) | 35220 (6.1) | 517887 (11.8) | 34714 (6.1) | ||
| South | 1740078 (39.4) | 259999 (45.0) | 1806577 (41.2) | 259020 (45.1) | ||
| Midwest | 1108658 (25.1) | 175003 (30.3) | 1219909 (27.8) | 174419 (30.1) | ||
| West | 987639 (22.4) | 104072 (18.0) | 833354 (19.0) | 103257 (18.0) | ||
| Missing | 13718 (0.3) | 3735 (0.7) | 8909 (0.2) | 2648 (0.5) | ||
| Substance use disorderd | 44704 (1) | 6511 (1.1) | 0.01113 | 46124 (1.1) | 6036 (1.1) | 0 |
| Any mental illnessd | 237645 (5.4) | 30281 (5.2) | 0.00629 | 228319 (5.2) | 29879 (5.2) | 0 |
| Tobacco use disorderd | 120139 (2.7) | 19236 (3.3) | 0.03553 | 144974 (3.3) | 18972 (3.3) | 0 |
| Low back paind | 348889 (7.9) | 47122 (8.2) | 0.00936 | 356695 (8.1) | 46679 (8.1) | 0 |
| Long-term opioid therapy | 58110 (1.3) | 8861 (1.5) | 0.01836 | 65999 (1.5) | 8637 (1.5) | 0 |
| Benzodiazepine prescription fills | 319845 (7.2) | 46214 (8) | 0.02845 | 349360 (8) | 45719 (8) | 0 |
| Opioid prescription fills, no. | ||||||
| 0 | 3637787 (82.3) | 469620 (81.2) | 0.03555 | 3564294 (81.3) | 466442 (81.3) | 0 |
| 1 | 528847 (12) | 71055 (12.3) | 540488 (12.3) | 70731 (12.3) | ||
| 2 to 3 | 174434 (3.9) | 25673 (4.4) | 194475 (4.4) | 25450 (4.4) | ||
| 4 to 5 | 43139 (1) | 6643 (1.1) | 49914 (1.1) | 6532 (1.1) | ||
| 6 or more | 33308 (0.8) | 5038 (0.9) | 37466 (0.9) | 4903 (0.9) | ||
| First buprenorphine fill, quarter | ||||||
| No fill | 4414158 (99.92) | 577465 (99.90) | 0.00762 | 4384244 (99.95) | 573745 (99.95) | 0 |
| 1 | 1977 (0.05) | 342 (0.06) | 1887 (0.043) | 247 (0.043) | ||
| 2 | 476 (0.01) | 67 (0.01) | 199 (0.005) | 26 (0.005) | ||
| 3 | 462 (0.01) | 76 (0.01) | 176 (0.004) | 23 (0.004) | ||
| 4 | 442 (0.01) | 79 (0.01) | 130 (0.003) | 17 (0.003) | ||
| First OUD-related office visit, quarter | ||||||
| No visit | 4414583 (99.93) | 577541 (99.92) | 0.00679 | 4384985 (99.96) | 573842 (99.96) | 0 |
| 1 | 1353 (0.03) | 235 (0.04) | 1139 (0.026) | 149 (0.026) | ||
| 2 | 572 (0.01) | 89 (0.02) | 199 (0.005) | 26 (0.005) | ||
| 3 | 530 (0.01) | 80 (0.01) | 183 (0.004) | 24 (0.004) | ||
| 4 | 477 (0.01) | 84 (0.02) | 130 (0.003) | 17 (0.003) | ||
| First OUD-related high-acuity visit, quarter | ||||||
| No visit | 4414631 (99.3) | 577618 (99.93) | 0.00358 | 4385092 (99.97) | 573856 (99.97) | 0 |
| 1 | 709 (0.02) | 95 (0.02) | 313 (0.007) | 41 (0.007) | ||
| 2 | 711 (0.02) | 120 (0.02) | 489 (0.011) | 64 (0.011) | ||
| 3 | 712 (0.02) | 100 (0.02) | 344 (0.008) | 45 (0.008) | ||
| 4 | 752 (0.02) | 96 (0.02) | 398 (0.009) | 52 (0.009) | ||
| Employee OOP costs, quintiles | ||||||
| 1 | 876530 (19.8) | 108798 (18.8) | 0.14636 | 823703 (18.8) | 107794 (18.8) | 0 |
| 2 | 889924 (20.1) | 102851 (17.8) | 781163 (17.8) | 102227 (17.8) | ||
| 3 | 890329 (20.2) | 107051 (18.5) | 813455 (18.5) | 106453 (18.5) | ||
| 4 | 890240 (20.2) | 110633 (19.1) | 840567 (19.2) | 110001 (19.2) | ||
| 5 | 870492 (19.7) | 148696 (25.7) | 1127748 (25.7) | 147583 (25.7) | ||
| Employee standardized total costs, quintiles | ||||||
| 1 | 878308 (19.9) | 113275 (19.6) | 0.01327 | 858563 (19.6) | 112356 (19.6) | 0 |
| 2 | 881201 (19.9) | 116792 (20.2) | 887150 (20.2) | 116097 (20.2) | ||
| 3 | 883847 (20) | 117840 (20.4) | 895013 (20.4) | 117126 (20.4) | ||
| 4 | 886279 (20.1) | 114924 (19.9) | 872982 (19.9) | 114243 (19.9) | ||
| 5 | 887880 (20.1) | 115198 (19.9) | 872929 (19.9) | 114236 (19.9) | ||
| Employer size, No. of employees | ||||||
| 0–49 | 673131 (15.2) | 280047 (48.4) | 1.08402 | 2124736 (48.4) | 278054 (48.4) | 0 |
| 50–249 | 916173 (20.7) | 191572 (33.1) | 1455597 (33.2) | 190487 (33.2) | ||
| >249 | 2828211 (64) | 106410 (18.4) | 806303 (18.4) | 105517 (18.4) |
Abbreviation: HDHP, high-deductible health plan; OUD, opioid use disorder; OOP, out-of-pocket.
Closer to zero indicates greater similarity.
We report gender as directly obtained from the data vendor. All other covariates and outcomes derived from variables provided by the data vendor.
We classified members as from predominantly White, Black, or Hispanic neighborhoods if they lived in a census tract with at least 75% of members of the respective race or ethnicity. We then applied a superseding ethnicity assignment using flags created by the E-Tech system (Ethnic Technologies), which analyzes full names and geographic locations of individuals. We classified remaining members as from mixed race/ethnicity neighborhoods.
Based on Johns Hopkins ACG software definition.
Time-to-Event Measures:
At baseline, the study groups had no statistically significant differences in time-to-first OUD-related office visit, buprenorphine fill, or OUD-related high-acuity visit (Table 2). At follow-up, there were no differences in time-to-first OUD-related office visits (HR, 1.02, 95% CI 0.94, 1.11; Figure S3 in the Supplemental Digital Content) or buprenorphine fills (HR, 1.05, 95% CI 0.97–1.13) in the HDHP versus control cohort. In contrast, the HDHP transition was associated with a delay in time-to-first OUD-related high-acuity visits (HR 0.86, 95% CI, 0.79–0.93; Figure 1A). Among members living in higher-poverty neighborhoods, there were no differences observed in time-to-first OUD-related office visits or buprenorphine fills at follow-up. However, HDHP members living in higher-poverty neighborhoods experienced a delay in time-to-first OUD-related high-acuity visits compared to control members at follow-up (HR 0.81, 95% CI 0.72–0.91; Figure 1B). We observed no significant differences across all three outcomes at follow-up among members living in lower-poverty neighborhoods (Figure 1C).
Table 2:
Analyses of Time to First OUD-related Events Before and After an Employer-Mandated Transition to High-Deductible Health Plans Compared With a Matched Control Group Remaining in Traditional Health Plans
| Time to First:a | Baseline HR (95% CI) | Follow-Up HR (95% CI)b |
|---|---|---|
| Buprenorphine Fill | 1.00 (0.89, 1.13) | 1.05 (0.97, 1.13) |
| Higher-Poverty | 1.00 (0.84, 1.20) | 1.06 (0.95, 1.19) |
| Lower-Poverty | 1.00 (0.86, 1.17) | 1.03 (0.92, 1.14) |
| OUD-Related Office Visit | 1.00 (0.87, 1.15) | 1.02 (0.94, 1.11) |
| Higher-Poverty | 1.00 (0.81, 1.24) | 1.11 (0.98, 1.25) |
| Lower-Poverty | 1.00 (0.83, 1.21) | 0.96 (0.85, 1.08) |
| OUD-related High-Acuity Visit | 1.00 (0.86, 1.16) | 0.86 (0.79, 0.93)c |
| Higher-Poverty | 1.00 (0.82, 1.23) | 0.81 (0.72, 0.91)c |
| Lower-Poverty | 1.00 (0.81, 1.23) | 0.91 (0.81, 1.02) |
Abbreviations: HR, hazard ratio.
Cox proportional hazard models were adjusted with weights generated from matching on key baseline employee and employer level covariates listed in the manuscript. A statistically significant HR <1.0 suggests a delay in the HDHP relative to the control group.
Bonferroni-corrected threshold for statistical significance equal to P<0.006.
P<0.0001
Figures 1A-C:
Weighted Cumulative Event Rates of First OUD-Related Emergency Department, Observation Unit, or Inpatient Events in the High-Deductible Health Plan (HDHP) Group Compared With a Matched Control Group Remaining in a Traditional Health Plan
The results of our sensitivity analyses were similar in magnitude and direction compared with our main analyses (Tables S1 and S2 in the Supplemental Digital Content). We found no differences in time-to-first OUD-related office visit or buprenorphine fill but identified significant delays in time-to-first OUD-related high-acuity visit when excluding patients with long-term opioid therapy and when adjusting for geographic region and additional employer-level variables. These findings were more pronounced among members living in higher-poverty neighborhoods.
Number of High-Acuity Care Days:
From baseline to follow-up, HDHP compared to control members experienced a significant reduction in the number of high-acuity care days (relative change: −37.4%, 95% CI: −57.8%, −17.0%; Table 3 and Figure 2). The magnitude of reduction was more pronounced among members living in higher-poverty neighborhoods (relative change: −44.7%, 95% CI: −71.6%, −17.8%) than lower-poverty neighborhoods (relative change: −30.4%, 95% CI: −60.7%, −20.4%).
Table 3:
Yearly Rates of OUD-Related High-Acuity Care Days Before and After an Employer-Mandated Transition to a High-Deductible Health Plan Compared with a Matched Control Group Remaining in a Traditional Health Plana
| Baselineb | Follow-Upc | Absolute Difference-in-Difference Estimate (95% CI)d | Relative Difference-in-Difference Estimate, % (95% CI)e | |||
|---|---|---|---|---|---|---|
| Control (se) | HDHP (se) | Control (se) | HDHP (se) | |||
| High-acuity care days per 10,000 members | 16.9 (2.0) | 20.6 (2.4) | 38.8 (2.4) | 29.7 (1.9) | −17.7 (−32.9, −0.1)** | −37.4 (−57.8, −17.0)**** |
| Higher- poverty | 16.4 (2.6) | 18.6 (3.0) | 39.3 (2.6) | 24.7 (2.3) | −19.9 (−40.7, 0.8)* | −44.7 (−71.6, −17.8%)*** |
| Lower- poverty | 17.6 (3.1) | 22.9 (3.7) | 38.2 (4.0) | 34.5 (3.1) | −15.0 (−36.3, 6.2) | −30.4 (−60.7, −20.4)** |
Abbreviations: HDHP, high-deductible health plan.
Among our HDHP and matched control group, we calculated the number of high-acuity days members experienced per year during presentations for OUD- or opioid overdose-related emergency department, observation unit, or hospitalization encounters. For example, a 1 day emergency department visit followed by a same day admission to the hospital for a 5-day hospitalization with an admission or discharge OUD diagnosis represents 5 OUD-related acute care days.
Represents the mean number of OUD-related high-acuity care days per 10,000 members in the baseline year where both the HDHP and control group are enrolled in plans with an annual deductible <$500.
Represents the mean number of OUD-related high-acuity care days per 10,000 members in the follow-up year in which the HDHP group undergoes an employer-mandated switch to a plan with an annual deductible ≥$1000.
Absolute difference-in-differences estimates represent the marginally adjusted difference in OUD-related high-acuity care days between the HDHP group’s predicted and observed follow-up rates, using the control group’s baseline-to-follow up relative change to estimate the predicted HDHP rate.
Relative difference-in-differences estimates represent the marginally adjusted relative baseline-to-follow up change in OUD-related high-acuity care days in the HDHP group adjusted for the relative baseline-to-follow up change in the control group, expressed as a percentage change.
p<0.1,
p<0.05,
p<0.01,
p<0.001
Figure 2:
Mean Weighted High-Acuity Care Days Per Quarter Among High-Deductible Health Plan Members and a Matched Control Group Continuously Enrolled in a Traditional Health Plan
DISCUSSION
In this time-to-event and pre-post study with a matched control group of commercially insured adults, we found that employer-mandated HDHP transitions were associated with delays in first high-acuity OUD presentations. The magnitude was more pronounced among members living in higher-poverty neighborhoods. We also found large relative reductions in our related measure of high-acuity care days among the full sample and those living in higher- and lower-poverty neighborhoods. Our findings suggest that members with OUD, especially those living in higher-poverty neighborhoods, are affected by increases in out-of-pocket costs after HDHP enrollment. Employer-mandated HDHP transitions were not associated with delays in OUD-related office visits or buprenorphine fills.
The reduction in high-acuity OUD care among HDHP members could be interpreted in several ways. This could suggest a beneficial effect: although we found no corresponding improvement in OUD outcome mediators, near-statistically significant increases in first outpatient visits and buprenorphine fills among higher-poverty HDHP members could suggest that this population substituted office-based care for ED care. Larger studies are needed to evaluate such patterns. Another interpretation is that reduced high-acuity visits represented declines in discretionary care; such reductions could have several effects including (1) reductions in important touchpoints for a vulnerable population, (2) unchanged health, and/or (3) reductions in iatrogenic harm. Finally, reduced high-acuity care could imply that HDHP members with time-sensitive or concerning conditions inappropriately avoided care to minimize out-of-pocket costs. However, we did not detect an accompanied increase in high-acuity days, suggesting delays were not followed by longer-than-expected hospital admissions, which might occur after delaying medical care.
Although we are unable to distinguish between these scenarios, reduced ED and hospital contact raise concerns given these settings are critical touchpoints for people with OUD.31,32 ED and hospital staff increasingly initiate treatment and link patients to addiction medicine care following discharge.3,5,6,33 Therefore, our findings could represent concerning avoidance of important medical care or OUD treatment touchpoints due to a higher level of cost-sharing.
This is the first study to examine the association between employer-mandated HDHP transitions and changes in OUD-related presentation timing, suggesting the novel finding that high cost-sharing levels are associated with lower use of high-acuity OUD care. Prior studies have examined employer-offered HDHPs in which only some beneficiaries enroll, a scenario analogous to an employer-level “intent-to-treat” analysis but with attenuated insights about HDHP effects, and have reported mixed associations with SUD-related oucomes.17–19 Similar to our findings, one study detected lower hospital visits for people with SUDs.17 However, this analysis also found lower ambulatory care visits, contrasting with our finding of no changes in time-to-first office visits.17 Such discrepancies may reflect different populations (OUD versus SUD broadly) and study designs (employer-offered versus employer-mandated HDHP enrollment). Another study found no differences in non-fatal overdoses associated with employer-offered HDHPs.19 We are unable to make comparisons to this study given overdose events were rare, limiting disaggregation from our high-acuity measure. Employer-offered HDHPs were also not associated with changes in buprenorphine adherence,18 similar to our finding of no change in initial presentations for buprenorphine fills. Further research is needed to examine how HDHPs impact MOUD engagement and retention after initiating buprenorphine. While a cross-sectional study found HDHPs were associated with lower MOUD initiation but not subsequent fills,34 these findings may be susceptible to selection bias.
Our findings are consistent with economic theory that service use reductions are proportional to cost-sharing, and that lower-income individuals are more sensitive to out-of-pocket costs.14,35 We found a greater magnitude delay and reduction in OUD-related high-acuity care among members from higher-poverty than lower-poverty neighborhoods, suggesting low-income members avoid OUD-related care they presume will be costly.
In contrast to our high-acuity findings, we did not detect differences in time-to-first ambulatory care visit or buprenorphine fill among members living in higher-poverty neighborhoods. This may be due to modest cost-sharing for office visits and buprenorphine fills in the HDHP group compared with substantially higher out-of-pocket costs for ED or hospital visits. Members may prioritize buprenorphine given its critical role in reducing overdose risk. This interpretation is consistent with prior evidence regarding HDHP effects in bipolar disorder in which cost-related medication nonadherence was rare for medications perceived to be of “life or death” importance.36,37 These findings also differ from surveys and qualitative studies that cite cost as a barrier to OUD treatment.7–9 One potential explanation for this discrepancy is that commercially insured individuals, especially those continuously enrolled for one year, may have greater financial resources, less severe OUD, and greater access to care than publicly insured and uninsured populations represented in survey studies. Future research should examine how cost-sharing impacts OUD-related care and MOUD initiation—including methadone maintenance therapy—among publicly insured and uninsured populations.
To address the observed delays and reductions in high-acuity OUD-related care, clinicians and health systems could ensure availability of low-barrier, same-day outpatient OUD care, while adopting evidence-based, approaches in the ED and hospital, including rapid MOUD initiation protocols, addiction medicine consult services, and care management teams that help with financial assistance. Policymakers at state, federal, and health plan levels could consider reducing cost-sharing for high-acuity OUD care. Policymakers could consider extending legislation such as New Mexico’s No Behavioral Cost-Sharing laws to include high-acuity OUD care. This law eliminated cost-sharing for behavioral health medications (including those for SUDs) leading to a 85% decrease in out-of-pocket spending per dispensed medication.38 At the federal level, legislation could support cost-sharing reductions for OUD-related high-acuity care. Health plans could bundle payments for episodes spanning high-acuity and outpatient care. Cost-sharing should be lower in outpatient versus high-acuity settings to incentivize contact with primary care and MOUD access.
Our study has potential limitations. Our findings do not generalize to Medicaid-enrolled or uninsured people living with OUD. Our empiric approach utilizes insurance claims which cannot disentangle whether changes in OUD-related outcomes are attributable to cost-sharing or to contemporaneous, unobserved employer-level interventions (e.g., offering employee assistance programs). The 12-month continuous enrollment necessary for this study excluded patients with OUD with enrollment gaps.39 We could not observe OUD-related outcomes after disenrollment or reasons for disenrollment (e.g., loss of coverage, death). Differential disenrollment related to unmeasured outcome risk could bias estimates. We are unable to assess whether delays in OUD-related high-acuity care were associated with changes in general health, mortality, or other OUD-related adverse outcomes such as endocarditis. Our algorithm for identifying OUD-related health care might have misclassified or missed events.27 However, we detected similar findings in magnitude and direction when excluding patients prescribed long-term opioid therapy, suggesting findings are robust to different claims-based OUD definitions. We identified more buprenorphine fills than OUD-related office visits; this counterintuitive finding has been previously described among commercially insured individuals with OUD and may be explained clinician and patient reluctance to label patients with having an OUD in the electronic health record.40
Employer-mandated transitions to HDHPs were associated with delays and reductions in OUD-related high-acuity presentations, particularly among members living in higher-poverty neighborhoods, but not with delays in OUD-related office visits or buprenorphine fills. Policymakers and health plans should consider programs that reduce cost-sharing for high-acuity OUD-related care. Future studies should examine the impact of high out-of-pocket costs on opioid-related health outcomes among uninsured and publicly insured adults.
Supplementary Material
Funding/Support:
Dr. Shuey was supported for this work by the Department of Population Medicine in the Harvard Pilgrim Health Care Institute, which is connected to the Institutional National Research Service awards T32HP32715 and T32HP42013 led by Drs. Ed Marcantonio and Jennifer Haas, and by the National Institute on Drug Abuse (NIDA) grants K12DA050607 and K23DA062822. The Institutional National Research Service Award and NIDA had no direct role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Footnotes
Conflicts of Interest: No potential conflicts of interest exist.
This is a pre-copyedited, author-produced version of an article accepted for publication in Journal of Addiction Medicine. The published version of record “Shuey B, Zhang F, Argetsinger S, Costa R, Wen H, Wharam JF. Opioid Use Disorder Care Presentations After High-Deductible Health Plan Enrollment. J Addict Med. 2026. Published online ahead-of-print.” Is available online at: doi: 10.1097/ADM.0000000000001662
References
- 1.CDC. Understanding the Opioid Overdose Epidemic. Overdose Prevention. July 12, 2025. Accessed January 19, 2026. https://www.cdc.gov/overdose-prevention/about/understanding-the-opioid-overdose-epidemic.html [Google Scholar]
- 2.Larochelle MR, Bernstein R, Bernson D, et al. Touchpoints – Opportunities to predict and prevent opioid overdose: A cohort study. Drug Alcohol Depend. 2019;204:107537. doi: 10.1016/j.drugalcdep.2019.06.039 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Liebschutz JM, Crooks D, Herman D, et al. Buprenorphine Treatment for Hospitalized, Opioid-Dependent Patients: A Randomized Clinical Trial. JAMA Intern Med. 2014;174(8):1369–1376. doi: 10.1001/jamainternmed.2014.2556 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Solomon KT, O’Connor J, Gibbons JB, et al. Association Between Hospital Adoption of an Emergency Department Treatment Pathway for Opioid Use Disorder and Patient Initiation of Buprenorphine After Discharge. JAMA Health Forum. 2023;4(3):e230245. doi: 10.1001/jamahealthforum.2023.0245 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wilson JD, Altieri Dunn SC, Roy P, Joseph E, Klipp S, Liebschutz J. Inpatient Addiction Medicine Consultation Service Impact on Post-discharge Patient Mortality: a Propensity-Matched Analysis. J Gen Intern Med. 2022;37(10):2521–2525. doi: 10.1007/s11606-021-07362-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.D’Onofrio G, O’Connor PG, Pantalon MV, et al. Emergency Department–Initiated Buprenorphine/Naloxone Treatment for Opioid Dependence: A Randomized Clinical Trial. JAMA. 2015;313(16):1636–1644. doi: 10.1001/jama.2015.3474 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Mackey K, Veazie S, Anderson J, Bourne D, Peterson K. Barriers and Facilitators to the Use of Medications for Opioid Use Disorder: a Rapid Review. J Gen Intern Med. 2020;35(Suppl 3):954–963. doi: 10.1007/s11606-020-06257-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Saini J, Johnson B, Qato DM. Self-Reported Treatment Need and Barriers to Care for Adults With Opioid Use Disorder: The US National Survey on Drug Use and Health, 2015 to 2019. Am J Public Health. 2022;112(2):284–295. doi: 10.2105/AJPH.2021.306577 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Anyanwu P, Olateju OA, Tata V, et al. Understanding the financial barriers to treatment among individuals with opioid use disorder: a focus group study. Harm Reduct J. 2024;21(1):220. doi: 10.1186/s12954-024-01133-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Bowser D, Bohler R, Davis MT, Hodgkin D, Horgan C. Payment-related barriers to medications for opioid use disorder: A critical review of the literature and real-world application. J Subst Use Addict Treat. 2024;165:209441. doi: 10.1016/j.josat.2024.209441 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Dunphy C, Peterson C, Zhang K, Jones CM. Do out-of-pocket costs influence retention and adherence to medications for opioid use disorder? Drug Alcohol Depend. 2021;225:108784. doi: 10.1016/j.drugalcdep.2021.108784 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Key Facts about Uninsured Adults with Opioid Use Disorder. KFF. July 15, 2019. Accessed December 16, 2025. https://www.kff.org/uninsured/key-facts-about-uninsured-adults-with-opioid-use-disorder/ [Google Scholar]
- 13.2023 Employer Health Benefits Survey. Kaiser Family Foundation; 2023. Accessed August 1, 2024. https://www.kff.org/health-costs/report/2023-employer-health-benefits-survey/ [Google Scholar]
- 14.Newhouse JP. Free for All? Lessons from the RAND Health Insurance Experiment. Harvard University Press; 1993. [Google Scholar]
- 15.Wharam JF, Zhang F, Wallace J, et al. Vulnerable And Less Vulnerable Women In High-Deductible Health Plans Experienced Delayed Breast Cancer Care. Health Aff (Millwood). 2019;38(3):408–415. doi: 10.1377/hlthaff.2018.05026 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Wharam JF, Zhang F, Eggleston EM, Lu CY, Soumerai S, Ross-Degnan D. Diabetes Outpatient Care and Acute Complications Before and After High-Deductible Insurance Enrollment: A Natural Experiment for Translation in Diabetes (NEXT-D) Study. JAMA Intern Med. 2017;177(3):358. doi: 10.1001/jamainternmed.2016.8411 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Eisenberg MD, Kennedy-Hendricks A, Schilling C, et al. The impact of HDHPs on service use and spending for substance use disorders. Am J Manag Care. 2022;28(10):530–536. doi: 10.37765/ajmc.2022.89250 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Kennedy-Hendricks A, Schilling CJ, Busch AB, et al. Impact of High Deductible Health Plans on Continuous Buprenorphine Treatment for Opioid Use Disorder. J Gen Intern Med. 2022;37(4):769–776. doi: 10.1007/s11606-021-07094-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Eddelbuettel JCP, Barry CL, Kennedy-Hendricks A, et al. High-deductible Health Plans and Nonfatal Opioid Overdose. Med Care. 2023;61(9):601. doi: 10.1097/MLR.0000000000001886 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Constantin J, Kenney GM, Simon K, Chua KP. Medicaid Unwinding and Changes in Buprenorphine Dispensing. JAMA Netw Open. 2025;8(5):e258469. doi: 10.1001/jamanetworkopen.2025.8469 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Hodgkin D, Stewart M, Daily SM, Huskamp HA, Frank RG. Imposing Medicaid Work Requirements Could Cause Many Adults With Substance Use Disorders To Lose Access To Treatment. Health Aff Forefr. doi: 10.1377/forefront.20250226.421765 [DOI] [Google Scholar]
- 22.Williams AR, Johnson KA, Thomas CP, et al. Opioid use disorder Cascade of care framework design: A roadmap. Subst Abuse. 2022;43(1):1207–1214. doi: 10.1080/08897077.2022.2074604 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Lu CY, Busch AB, Zhang F, et al. Impact of High-Deductible Health Plans on Medication Use Among Individuals With Bipolar Disorder. Psychiatr Serv. 2021;72(8):926–934. doi: 10.1176/appi.ps.202000362 [DOI] [PubMed] [Google Scholar]
- 24.Wharam JF, Zhang F, Landon BE, Soumerai SB, Ross-Degnan D. Low-Socioeconomic-Status Enrollees In High-Deductible Plans Reduced High-Severity Emergency Care. Health Aff (Millwood). 2013;32(8):1398–1406. doi: 10.1377/hlthaff.2012.1426 [DOI] [PubMed] [Google Scholar]
- 25.Krieger N, Waterman PD, Spasojevic J, Li W, Maduro G, Van Wye G. Public Health Monitoring of Privilege and Deprivation With the Index of Concentration at the Extremes. Am J Public Health. 2016;106(2):256–263. doi: 10.2105/AJPH.2015.302955 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.US Census Bureau. American Community Survey (ACS). Accessed January 20, 2022. https://www.census.gov/programs-surveys/acs
- 27.Scherrer JF, Sullivan MD, LaRochelle MR, Grucza R. Validating opioid use disorder diagnoses in administrative data: a commentary on existing evidence and future directions. Addict Sci Clin Pract. 2023;18:49. doi: 10.1186/s13722-023-00405-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Jann B KMATCH: Stata module module for multivariate-distance and propensity-score matching, including entropy balancing, inverse probability weighting, (coarsened) exact matching, and regression adjustment. Published online 2020. Accessed August 12, 2023. https://ideas.repec.org/c/boc/bocode/s458346.html
- 29.Johns Hopkins University. ACG System Version 11.1: Technical Reference Guide. Accessed June 15, 2023. https://www.hopkinsacg.org/document/acg-system-version-11-1-technical-reference-guide/
- 30.VanderWeele TJ, Mathur MB. SOME DESIRABLE PROPERTIES OF THE BONFERRONI CORRECTION: IS THE BONFERRONI CORRECTION REALLY SO BAD? Am J Epidemiol. 2019;188(3):617–618. doi: 10.1093/aje/kwy250 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Lewer D, Freer J, King E, et al. Frequency of health-care utilization by adults who use illicit drugs: a systematic review and meta-analysis. Addiction. 2020;115(6):1011–1023. doi: 10.1111/add.14892 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Rhee TG, Rosenheck RA. Use of Drug Treatment Services Among Adults With Opioid Use Disorder: Rates, Patterns, and Correlates. Psychiatr Serv. 2019;70(11):992–999. doi: 10.1176/appi.ps.201900163 [DOI] [PubMed] [Google Scholar]
- 33.McNeely J, Wang SS, Rostam Abadi Y, et al. Addiction Consultation Services for Opioid Use Disorder Treatment Initiation and Engagement: A Randomized Clinical Trial. JAMA Intern Med. 2024;184(9):1106–1115. doi: 10.1001/jamainternmed.2024.3422 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Morgan JR, Quinn EK, Chaisson CE, et al. Variation in initiation, engagement, and retention on medications for opioid use disorder based on health insurance plan design. Med Care. 2022;60(3):256–263. doi: 10.1097/MLR.0000000000001689 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Finkelstein AN, Taubman SL, Allen HL, Wright BJ, Baicker K. Effect of Medicaid Coverage on ED Use — Further Evidence from Oregon’s Experiment. N Engl J Med. 2016;375(16):1505–1507. doi: 10.1056/NEJMp1609533 [DOI] [PubMed] [Google Scholar]
- 36.Madden JM, Araujo-Lane C, Foxworth P, et al. Experiences of health care costs among people with employer-sponsored insurance and bipolar disorder. J Affect Disord. 2021;281:41–50. doi: 10.1016/j.jad.2020.10.033 [DOI] [PubMed] [Google Scholar]
- 37.Wharam JF, Busch AB, Madden J, et al. Effect of high-deductible insurance on health care use in bipolar disorder. Am J Manag Care. 2020;26(6):248–255. doi: 10.37765/ajmc.2020.43487 [DOI] [PubMed] [Google Scholar]
- 38.Golberstein E, Campbell JM, Maclean JC, Harris SJ, Saloner B, Stein BD. Prescription Drug Dispensing and Patient Costs After Implementation of a No Behavioral Health Cost-Sharing Law. JAMA Health Forum. 2024;5(3):e240198. doi: 10.1001/jamahealthforum.2024.0198 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Christine PJ, Goldman AL, Morgan JR, et al. Insurance Instability for Patients With Opioid Use Disorder in the Year After Diagnosis. JAMA Health Forum. 2024;5(7):e242014. doi: 10.1001/jamahealthforum.2024.2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Shen K, Barrette E, Dafny LS. Treatment Of Opioid Use Disorder Among Commercially Insured US Adults, 2008–17. Health Aff (Millwood). 2020;39(6):993–1001. doi: 10.1377/hlthaff.2019.01041 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.