The Impact of HDHPs on Service Use and Spending for Substance Use Disorders

ABSTRACT

Objectives: Although high-deductible health plans (HDHPs) reduce health care spending, higher deductibles may lead to forgone care. Our goal was to determine the effects of HDHPs on the use of and spending on substance use disorder (SUD) services.

Study Design: We used difference-in-differences models to compare service use and spending for treating SUD among enrollees who were newly offered an HDHP relative to enrollees offered only traditional plan options throughout the study period.

Methods: We used deidentified commercial claims data from OptumLabs (2007-2017) to identify a sample of 28,717,236 person-years (2.2% with a diagnosed SUD). The main independent measure was an indicator for being offered an HDHP. The main dependent measures were the probability of (and spending associated with) using SUD services and specific treatment types.

Results: Enrollees were 6.6% (P < .001) less likely to use SUD services after being offered an HDHP relative to the comparison group. Reductions were concentrated in inpatient, intermediate, and ambulatory care, as well as medication use. Being offered an HDHP was associated with a decrease of 21% (P < .001) on health plan spending and an increase of 14% (P < .01) on out-of-pocket spending.

Conclusions: Offering an HDHP was associated with a reduction in SUD service use and a shift in spending from the plan to the enrollee. In the context of the US drug epidemic, these study findings highlight a concern that the movement toward HDHPs may be exacerbating undertreatment of SUD.

Am J Manag Care. 2022;28(10):530-536. https://doi.org/10.37765/ajmc.2022.89250

Takeaway Points

• Although high-deductible health plans (HDHPs) reduce health care spending, higher deductibles may lead to forgone care.
• This is of concern for individuals with substance use disorders (SUDs) because these conditions are undertreated and often co-occur with other chronic conditions.
• We find that enrollees were 6.6% less likely to use SUD services after being offered an HDHP relative to the comparison group. Reductions were concentrated in intermediate care, ambulatory care, and medication use.
• Being offered an HDHP was associated with a decrease of 21% on health plan spending and an increase of 14% on out-of-pocket spending.
• In the context of the US drug epidemic, these study findings highlight a concern that the national movement toward HDHPs may be exacerbating undertreatment of SUD.

Enrollment in high-deductible health plans (HDHPs) is increasing rapidly, with 22% of covered workers enrolled in a plan with an annual deductible exceeding $2000 in 2019.¹ Proponents of HDHPs argue that higher cost sharing incentivizes enrollees to “shop” for care. Detractors argue that cost sharing is a blunt instrument and reduces utilization across the board. Indeed, HDHPs have been shown to reduce health care spending,^2,3 but higher deductibles encourage consumers to forgo care in the short term,^4-6 leading to adverse outcomes in the long term. Previous research on HDHPs suggests that vulnerable populations, including those with chronic conditions whose health outcomes are tied to continuing, uninterrupted care, are at the greatest risk of negative consequences in response to higher cost sharing.⁷

This evidence on HDHPs raises concerns about potential negative effects for individuals with substance use disorders (SUDs) because these conditions are most effectively managed as chronic conditions with longer time in treatment leading to better outcomes,^7,8 often co-occur with mental illness and other chronic medical conditions,⁹ and can be costly.¹⁰ Furthermore, given that SUD is vastly undertreated,¹¹ that only about 10% of individuals with SUD treatment needs receive treatment,¹² and that access to evidence-based SUD treatments (eg, medications for opioid use disorder) is severely limited in many communities,^13-15 the shift toward HDHPs might be creating further barriers to SUD diagnosis and treatment. Prior descriptive work has found suggestive evidence that HDHPs are associated with lower rates of emergency department and hospital use, as well as higher out-of-pocket (OOP) costs for individuals with SUD.^16,17 However, selection bias makes it difficult to draw causal inferences from comparisons of individuals enrolling in HDHPs with enrollees of traditional, low-deductible plan choices because the decision to enroll may be nonrandom: HDHP enrollees are known to be, on average, younger and healthier.¹⁸

Importantly, increases in drug- and alcohol-related mortality have made connecting individuals with evidence-based SUD treatment a national priority. From 1999 to 2017, more than 700,000 Americans died from drug overdoses,¹⁹ and drug-related deaths have contributed to declines in life expectancy over the past 3 years.^20,21 In this context, it is critical to understand how health insurance design choices may be facilitating or deterring access to lifesaving treatments for SUD. We aimed to evaluate whether the decision of an employer to offer an HDHP was associated with use of health care services to treat SUD and spending on these services. Our use of an intent-to-treat design to examine how a firm’s decision to offer an HDHP affects SUD service use and spending (rather than directly examining outcomes for those choosing to enroll in HDHPs) minimizes concerns about individual nonrandom selection into HDHPs vs traditional health plans. We also examine whether any changes in SUD spending attributable to HDHPs are borne primarily by the health plan or by enrollees in the form of OOP payments.

METHODS

Data

We used deidentified administrative claims data from the OptumLabs Data Warehouse from 2007 to 2017. The database includes enrollment records for commercial enrollees, medical (including behavioral health) and pharmacy claims, benefit design information, and a blinded firm identifier. Each firm identifier distinguished groups within a firm with a shared plan offering. Firms that offered the same choice set of plans to all employees had a single identifier, whereas firms that offered different choice sets to different groups of employees were assigned different identifiers for each group.

Sample Identification

We constructed our sample in several steps. First, we began with individuals with a valid deductible and medical, pharmacy, and mental health coverage (ie, behavioral health coverage falls under the same deductible). We required enrollees to be enrolled for at least 11 months of their plan year and limited enrollees to be aged between 12 and 64 years. Next, we selected continuous spans of time for each firm, eliminating firms with large changes in firm size (ie, changed by ± 50% between consecutive years; this resulted in an exclusion of 2% person-years) given that large swings in enrollment from year to year might indicate that many enrollees switched to a different insurance carrier that is unobserved in our data.

To define the treatment group, we categorized 369,239 firms as those offering HDHPs by calculating the percentage of enrollees at that firm enrolled in an HDHP in that year. Treatment firms were characterized by having at least 2 years with less than 5% HDHP enrollment immediately followed by having greater than 5% HDHP enrollment in all subsequent years (11% of firms). This threshold is arbitrary by nature but has support in prior literature.

We chose it as a number that would be high enough to signal that a nontrivial number of enrollees actually had the opportunity to enroll, but low enough not to exclude firms with low uptake. This threshold allows us to identify a more plausibly generalizable treatment group, with enrollment in an HDHP in the postperiod treatment group ranging from 5% enrolled to 100% enrolled. Comparison firms were defined as those having 0% HDHP enrollment for all years in which the firm is found in the data (35% of firms). We omitted enrollees affiliated with other types of firms, such as those always offering an HDHP (eAppendix A [eAppendices available at ajmc.com]).

Identification of Enrollees With SUD

For our spending analyses, we focused on a population coded as having an SUD based on having at least 1 medical claim with an SUD diagnosis in any of the diagnosis code positions, following prior research.^22,23 Enrollees having claims meeting this criterion during the time in which they were at a treatment or comparison firm were designated as having an SUD for the remainder of the study period. This particularly sensitive SUD sample selection criterion was intended to minimize potential bias due to reduced use of services associated with HDHPs.^2,24 We identified SUD diagnosis codes with International Classification of Diseases, Ninth Revision, Clinical Modification codes 291, 292, 303, 304, and 305 (excluding 305.1 [tobacco use disorder] and 305.8 [antidepressant abuse]) and Tenth Revision codes F10-F19 (excluding F17.2x [tobacco use disorder]).

Measures

The key independent variable in our analysis was a binary indicator of whether an enrollee was associated with a firm that offered an HDHP. For this measure, we defined a plan as an HDHP if its shared medical and pharmacy deductible (or sum of medical and pharmacy deductible) met the Internal Revenue Service definition in that calendar year. This cutoff changes by calendar year but averaged $1232 over the study period for individual plans and $2427 for family plans.²⁵ Traditional plans were defined as plans that did not meet this threshold.

The main dependent variables were measures of SUD service use and spending conditional on use overall and disaggregated into specific treatment types including inpatient, emergency department, intermediate (inclusive of residential, partial hospitalization, and intensive outpatient), ambulatory, and medication. Nonmedication services were identified using procedure (Current Procedural Terminology and Healthcare Common Procedure Coding System), revenue, and American Medical Association place-of-service codes having an SUD diagnosis code (defined above) as the primary diagnosis on the medical claim. SUD emergency department use also included overdoses. SUD inpatient use required 50% or more of hospital facility claims to have a primary diagnosis of SUD or a diagnosis of an overdose. Intermediate and ambulatory SUD services were identified if the SUD diagnosis was the primary diagnosis or the second diagnosis when accompanied by a primary diagnosis for a mental health condition. Ambulatory SUD services also included Screening, Brief Intervention, Referral to Treatment and medication administration procedure claims, agnostic of diagnosis codes. Medication use was identified from pharmacy claims for alcohol use disorder (AUD) and opioid use disorder (OUD) medications and medical claims for administration of AUD and OUD medications.

To avoid double-counting, nonmedication services occurring on the same day were attributed to the highest level of care on that day (eg, claims meeting criteria for an ambulatory SUD service but occurring on the day of SUD intensive outpatient treatment were attributed to the intermediate category). Medication administration spending was attributed to the highest level of care on that day, whereas pharmacy spending was captured separately (eAppendix B).

Spending was categorized by OOP, health plan, and total spending (the sum of OOP and health plan spending) by aggregating values within an enrollee-year. Negative dollar amounts were coded as zeroes (0.001% of enrollee-years) and spending was top-coded at the 99.9th percentile in each calendar year. Spending was examined conditional on use in the associated category.

Enrollee-level covariates included age, sex, race/ethnicity (Asian, Black, Hispanic, White, or unknown), family size, Census block household income (ie, < $40,000; ≥ $40,000 and < $75,000; ≥ $75,000 and < $125,000; ≥ $125,000 and < $200,000; > $200,000; or unknown), Census block–level education (ie, less than high school, high school, some college, bachelor’s or more, or unknown), and Census division. Income, race/ethnicity, and education were estimated based on personal identifying information linked to a mix of proprietary sources for OptumLabs. The Chronic Conditions Warehouse was used to construct 47 condition indicators included as covariates.²⁶

Statistical Analyses

We analyzed the effects of offering an HDHP on SUD service use and spending outcomes using a 2-way fixed effects difference-in-differences study design. We included preperiod data before a firm began offering an HDHP to control for differences in individual characteristics that varied across firms but might be correlated with our outcome measures. To control for secular trends, we also included, as a comparison group, enrollees at firms that never offered an HDHP. Our empirical approach compared changes in SUD service use and spending over time between treatment group enrollees who are offered an HDHP and comparison group enrollees who have not yet been offered an HDHP. The unit of analysis was the person-year. The HDHP variable indicated whether the enrollee’s firm offered an HDHP in a given calendar year and was coded as zero for all years in firms that never offered an HDHP. Models included the individual-level covariates listed above and calendar year and firm fixed effects. For probability of service use outcomes (entire sample), we estimated ordinary least squares (OLS) regressions. For spending outcomes (SUD sample), we estimated OLS regressions on a logged dependent variable, to account for skew. In all models, SEs were clustered at the firm level to account for unobservable correlations within each firm. All analyses were conducted in Stata 16 MP (StataCorp LLC). This study was approved by the institutional review board of the Johns Hopkins Bloomberg School of Public Health.

The key assumption for our analyses was that, absent the firm offering an HDHP, enrollees at firms in the treatment group would have had trends in SUD service use and spending consistent with the trends seen in enrollees at firms in the comparison group. Although this assumption is ultimately untestable, 3 pieces of evidence gave us confidence in this assumption. First, we evaluated differences across treatment and comparison groups and across study year trends using standardized mean differences (SMDs) and found that 98% of covariate SMDs fell below 0.1, a commonly used threshold for evaluating covariate balance (eAppendix C).²⁷ Firm-level characteristics were also similar across groups (eAppendix D), and we observed no differential selection into the sample (eAppendix E). Second, unadjusted rates of our outcome variable (SUD service use) had fairly similar levels and trends across groups in the preperiod (eAppendix F.1). Third, when we explicitly tested for preperiod differences in trends prior to HDHP offer, we found no statistically significant differences (eAppendix F.2). Given new advances in difference-in-differences methodology, we also estimated stratification models analyzing treatment heterogeneity over time (eAppendix G).

Our analytic approach relied on discrete changes in the fraction of a firm’s enrollees enrolled in an HDHP, requiring a sharp cutoff to identify when a firm began offering an HDHP. As described above, we used a 5% threshold in our main analyses based on prior research.^2,5 In sensitivity analyses, alternative thresholds were used. As noted above, we restricted our analysis to firms that had a stable size year to year, using a greater than 50% turnover cutoff. In sensitivity analyses, alternative cutoffs were used.

RESULTS

Table 1 displays unadjusted enrollee characteristics in the treatment and comparison groups across the pre- and post periods. Table 2 shows unadjusted rates of SUD use and spending across service type. Overall, unadjusted rates of SUD use and spending increased from the preperiod to the post period but at a slower rate in the treatment group relative to the comparison group. This is consistent with prior evidence showing that HDHPs do not decrease spending but, instead, slow the growth in spending.^2,3 Mean annual deductibles showed similar trends prior to HDHP offer, with a large, expected increase in the treatment group after HDHP offer (eAppendix H).

Table 3 indicates that the adjusted probability of using any SUD services increased after treatment group enrollees were offered an HDHP; however, this increase was smaller relative to the change in the comparison group (–0.04 percentage points [PP]; P < .001), implying a 6.6% reduction in the probability of using SUD services after the treatment group is offered an HDHP relative to otherwise similar enrollees who continued to be offered only traditional health plan choices by their employer (full model results in eAppendix I).

Among the study population with an SUD diagnosis, we found no differences in mean total spending on SUD services among treatment group enrollees offered an HDHP relative to comparison group enrollees who continued to be offered only traditional health plan options. However, enrollees in the treatment group had larger increases in OOP spending on SUD services ($168 to $215, pre- to post) relative to the comparison group ($169 to $190, pre- to post), resulting in higher annual OOP spending averaging $24 (P < .01). In contrast, mean spending on SUD services paid for by health plans decreased for treatment group enrollees ($655 to $588, pre- to post) relative to comparison group enrollees ($662 to $746, pre- to post), resulting in $138 (P < .001) lower mean annual per-enrollee spending by HDHPs (full model results in eAppendix I).

Figure 1 shows that there was a decrease in the probability of enrollees with SUD diagnoses using inpatient SUD services (–0.30 PP; P = .01), intermediate SUD services (–0.50 PP; P = .006), ambulatory SUD services (–1.70 PP; P < .001), and guideline-recommended medications to treat SUD (–0.38 PP; P = .029) attributable to being offered an HDHP.

Similarly, as shown in Figure 2 (focused on the population with SUD diagnoses), both emergency department spending and ambulatory SUD spending by health plans decreased by annual means of $137 and $23, respectively, in HDHPs relative to traditional plans.

Our key robustness checks found no qualitative changes when we varied the HDHP enrollment threshold used to identify the treatment group or when we varied the maximum allowed change in year-to-year firm size (eAppendix J).

DISCUSSION

Offering an HDHP was associated with a 0.04-percentage-point lower probability of using SUD services, implying a 6.6% reduction from baseline service use. These reductions were driven by reductions in use of intermediate and ambulatory services and medication treatment for SUD. Overall, we did not find differences in total spending among SUD service users but did find a shift in costs from the insurer to the patient. These changes in SUD treatment rates and spending were consistent with prior literature on HDHPs in other clinical contexts.^2-6 Although small in magnitude, these estimates are clinically relevant because SUD is vastly undertreated.¹¹

Our results have several important implications. First, the financial barriers imposed by HDHPs were associated with lowered SUD treatment rates. This is especially concerning given the already low rates of treatment in this population and the high morbidity and mortality associated with SUD. When evaluating the costs and benefits associated with switching employees to an HDHP, individuals should consider the potential increased costs for SUD treatment, and plan benefit managers should be cognizant of the financial implications for their enrollees with SUD.

Second, our results suggest that the effects are driven by reductions in inpatient, intermediate, ambulatory, and medication use, with no effects on emergency department use. This is not surprising, as demand for emergency care is more inelastic than intermediate, ambulatory, or medication use. However, these findings are in contrast to the public health policy goal for improving SUD care—that is, more robust use of ambulatory care (rather than just receiving episodic inpatient or emergency care) and also of evidence-based AUD and OUD medication.^15,28 Thus, even in the context of HDHPs, financial and administrative barriers to accessing ambulatory care and SUD medications should be as low as possible. For OUD, for example, all 3 FDA-approved medications should be offered and be included in preferred formulary tiers without prior authorization or cost sharing.^29-31

Limitations

This study is subject to several important limitations. First, although our analysis uses a rigorous analytic approach, our results are still based on observational data and not a randomized controlled trial. We took steps to mitigate this concern, including taking advantage of the exogenous HDHP offer (rather than relying on individual selection decisions) and by carefully analyzing preperiod trends (eAppendix F). Second, although the data have broad national coverage, they have a higher proportion of enrollees in the South and Central regions and thus our results may not be generalizable to the broader employer-sponsored market. Although prior studies have used similar data to evaluate the effects of HDHPs, this remains a limitation.^32,33 Our analysis focuses on stable employers with stable enrollment over time, but we are unable to observe what happens to employees who switch to a plan offered by a different insurer and may be missing important variation. Third, our data do not include important information about networks or consumer-facing tools that may have been rolled out to employees at the same time as HDHP adoption. Some studies on the effects of HDHPs have had more institutional knowledge about other tools, but the fact that they analyze a single employer may limit generalizability.²⁴

CONCLUSIONS

Offering an HDHP led to a 6.6% reduction in the probability of using SUD services and a shift in spending from the plan to the enrollee. In the context of the US drug epidemic, our findings highlight a concern that the national movement toward enrollment in HDHPs may be exacerbating the undertreatment of SUD.

Author Affiliations: Department of Health Policy and Management (MDE, AK-H, CS, EAS, MKM) and Department of Mental Health (AK-H, EAS), Johns Hopkins Bloomberg School of Public Health, Baltimore, MD; Johns Hopkins Center for Mental Health and Addiction Policy Research (MDE, AK-H, CS, EAS), Baltimore, MD; OptumLabs Visiting Fellow, OptumLabs (MDE), Cambridge, MA; Department of Health Care Policy, Harvard Medical School (ABB, HAH), Boston, MA; McLean Hospital (ABB), Boston, MA; Cornell Jeb E. Brooks School of Public Policy (CLB), Ithaca, NY.

Source of Funding: This project was supported by grant number R01DA044201 from the National Institute on Drug Abuse (NIDA). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIDA.

Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (MDE, AK-H, ABB, HAH, EAS, MKM, CLB); acquisition of data (MDE, CLB); analysis and interpretation of data (MDE, AK-H, CS, ABB, HAH, EAS, CLB); drafting of the manuscript (MDE, MKM, CLB); critical revision of the manuscript for important intellectual content (MDE, AK-H, CS, ABB, HAH, EAS, MKM, CLB); statistical analysis (CS); obtaining funding (MDE, CLB); administrative, technical, or logistic support (MDE, ABB, CLB); and supervision (MDE, CLB).

Address Correspondence to: Matthew D. Eisenberg, PhD, Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, 624 N Broadway, Hampton House 406, Baltimore, MD 21205. Email: eisenberg@jhu.edu.

REFERENCES

1. 2019 Employer Health Benefits Survey. Kaiser Family Foundation. September 25, 2019. Accessed May 15, 2020. https://www.kff.org/health-costs/report/2019-employer-health-benefits-survey/

2. Haviland AM, Eisenberg MD, Mehrotra A, Huckfeldt PJ, Sood N. Do “consumer-directed” health plans bend the cost curve over time? J Health Econ. 2016;46:33-51. doi:10.1016/j.jhealeco.2016.01.001

3. Bundorf MK. Consumer-directed health plans: a review of the evidence. J Risk Insur. 2016;83(1):9-41. doi:10.1111/jori.12141

4. Wharam JF, Landon BE, Galbraith AA, Kleinman KP, Soumerai SB, Ross-Degnan D. Emergency department use and subsequent hospitalizations among members of a high-deductible health plan. JAMA. 2007;297(10):1093-1102. doi:10.1001/jama.297.10.1093

5. Eisenberg MD, Haviland AM, Mehrotra A, Huckfeldt PJ, Sood N. The long term effects of “consumer-directed” health plans on preventive care use. J Health Econ. 2017;55:61-75. doi:10.1016/j.jhealeco.2017.06.008

6. Beeuwkes Buntin M, Haviland AM, McDevitt R, Sood N. Healthcare spending and preventive care in high-deductible and consumer-directed health plans. Am J Manag Care. 2011;17(3):222-230.

7. Schuckit MA. Treatment of opioid-use disorders. N Engl J Med. 2016;375(4):357-368. doi:10.1056/NEJMra1604339

8. McLellan AT, Lewis DC, O’Brien CP, Kleber HD. Drug dependence, a chronic medical illness: implications for treatment, insurance, and outcomes evaluation. JAMA. 2000;284(13):1689-1695. doi:10.1001/jama.284.13.1689

9. Mertens JR, Lu YW, Parthasarathy S, Moore C, Weisner CM. Medical and psychiatric conditions of alcohol and drug treatment patients in an HMO: comparison with matched controls. Arch Intern Med. 2003;163(20):2511-2517. doi:10.1001/archinte.163.20.2511

10. Frank RG, Glazer J, McGuire TG. Measuring adverse selection in managed health care. J Health Econ. 2000;19(6):829-854. doi:10.1016/s0167-6296(00)00059-x

11. Barry CL, Epstein AJ, Fiellin DA, Fraenkel L, Busch SH. Estimating demand for primary care–based treatment for substance and alcohol use disorders. Addiction. 2016;111(8):1376-1384. doi:10.1111/add.13364

12. Key substance use and mental health indicators in the United States: results from the 2018 National Survey on Drug Use and Health. Substance Abuse and Mental Health Services Administration. August 2019. Accessed January 10, 2021. https://www.samhsa.gov/data/sites/default/files/cbhsq-reports/NSDUHNationalFindingsReport2018/NSDUHNationalFindingsReport2018.pdf

13. Knudsen HK, Abraham AJ, Roman PM. Adoption and implementation of medications in addiction treatment programs. J Addict Med. 2011;5(1):21-27. doi:10.1097/ADM.0b013e3181d41ddb

14. Mark TL, Kassed CA, Vandivort-Warren R, Levit KR, Kranzler HR. Alcohol and opioid dependence medications: prescription trends, overall and by physician specialty. Drug Alcohol Depend. 2009;99(1-3):345-349. doi:10.1016/j.drugalcdep.2008.07.018

15. Volkow ND, Frieden TR, Hyde PS, Cha SS. Medication-assisted therapies—tackling the opioid-overdose epidemic. N Engl J Med. 2014;370(22):2063-2066. doi:10.1056/NEJMp1402780

16. Parthasarathy S, Campbell CI. High-deductible health plans: implications for substance use treatment. Health Serv Res. 2016;51(5):1939-1959. doi:10.1111/1475-6773.12456

17. Eisenberg MD, Du S, Sen AP, Kennedy-Hendricks A, Barry CL. Health care spending by enrollees with substance use and mental health disorders in high-deductible health plans vs traditional plans. JAMA Psychiatry. 2020;77(8):872-875. doi:10.1001/jamapsychiatry.2020.0342

18. McDevitt RD, Haviland AM, Lore R, Laudenberger L, Eisenberg M, Sood N. Risk selection into consumer-directed health plans: an analysis of family choices within large employers. Health Serv Res. 2014;49(2):609-627. doi:10.1111/1475-6773.12121

19. Drug overdose: understanding the epidemic. CDC. Accessed May 15, 2020. https://www.cdc.gov/drugoverdose/epidemic/index.html

20. Dowell D, Arias E, Kochanek K, et al. Contribution of opioid-involved poisoning to the change in life expectancy in the United States, 2000-2015. JAMA. 2017;318(11):1065-1067. doi:10.1001/jama.2017.9308

21. White AM, Castle IP, Hingson RW, Powell PA. Using death certificates to explore changes in alcohol-related mortality in the United States, 1999 to 2017. Alcohol Clin Exp Res. 2020;44(1):178-187. doi:10.1111/acer.14239

22. Barry CL, Stuart EA, Donohue JM, et al. The early impact of the ‘Alternative Quality Contract’ on mental health service use and spending in Massachusetts. Health Aff (Millwood). 2015;34(12):2077-2085. doi:10.1377/hlthaff.2015.0685

23. Busch AB, Frank RG, Lehman AF, Greenfield SF. Schizophrenia, co-occurring substance use disorders and quality of care: the differential effect of a managed behavioral health care carve-out. Adm Policy Ment Health. 2006;33(3):388-397. doi:10.1007/s10488-006-0045-3

24. Brot-Goldberg ZC, Chandra A, Handel BR, Kolstad JT. What does a deductible do? the impact of cost-sharing on health care prices, quantities, and spending dynamics. Q J Econ. 2017;132(3):1261-1318. doi:10.1093/qje/qjx013

25. 26 CFR 601.602: Tax forms and instructions. Internal Revenue Service. Accessed January 10, 2021. https://www.irs.gov/pub/irs-drop/rp-14-30.pdf

26. Chronic conditions. Chronic Conditions Data Warehouse. Accessed January 10, 2021. https://www2.ccwdata.org/web/guest/condition-categories-chronic

27. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28(25):3083-3107. doi:10.1002/sim.3697

28. HHS Office of the Surgeon General. Facing Addiction in America: The Surgeon General’s Report on Alcohol, Drugs, and Health. HHS; 2016.

29. National Academies of Sciences, Engineering, and Medicine. Medications for Opioid Use Disorder Save Lives. The National Academies Press; 2019.

30. Samet JH, Botticelli M, Bharel M. Methadone in primary care — one small step for Congress, one giant leap for addiction treatment. N Engl J Med. 2018;379(1):7-8. doi:10.1056/NEJMp1803982

31. Polsky D, Arsenault S, Azocar F. Private coverage of methadone in outpatient treatment programs. Psychiatr Serv. 2020;71(3):303-306. doi:10.1176/appi.ps.201900373

32. Zhang X, Trish E, Sood N. Financial burden of healthcare utilization in consumer-directed health plans. Am J Manag Care. 2018;24(4):e115-e121.

33. Zhang X, Haviland A, Mehrotra A, Huckfeldt P, Wagner Z, Sood N. Does enrollment in high-deductible health plans encourage price shopping? Health Serv Res. 2018;53(suppl 1):2718-2734. doi:10.1111/1475-6773.12784