Effect of High-Deductible Insurance on Health Care Use in Bipolar Disorder

High-deductible health plan members with bipolar disorder experienced a reduction in nonpsychiatrist mental health provider visits but no changes in other utilization.


Objectives: To determine the impact of high-deductible health plans (HDHPs) on health care use among individuals with bipolar disorder.

Study Design: Interrupted time series with propensity score—matched control group design, using a national health insurer’s claims data set with medical, pharmacy, and enrollment data.

Methods: The intervention group was composed of 2862 members with bipolar disorder who were enrolled for 1 year in a low-deductible (≤$500) plan and then 1 year in an HDHP (≥$1000) after an employer-mandated switch. HDHP members were propensity score matched 1:3 to contemporaneous controls in low-deductible plans. The main outcomes included out-of-pocket spending per health care service, mental health—related outpatient visits (subclassified as visits to nonpsychiatrist mental health providers and to psychiatrists), emergency department (ED) visits, and hospitalizations.

Results: Mean pre— to post–index date out-of-pocket spending per visit on all mental health office visits, nonpsychiatrist mental health provider visits, and psychiatrist visits increased by 21.9% (95% CI, 15.1%-28.6%), 33.8% (95% CI, 2.0%-65.5%), and 17.8% (95% CI, 12.2%-23.4%), respectively, among HDHP vs control members. The HDHP group experienced a –4.6% (95% CI, –11.7% to 2.5%) pre- to post change in mental health outpatient visits relative to controls, a –10.9% (95% CI, –20.6% to –1.3%) reduction in nonpsychiatrist mental health provider visits, and unchanged psychiatrist visits. ED visits and hospitalizations were also unchanged.

Conclusions: After a mandated switch to HDHPs, members with bipolar disorder experienced an 11% decline in visits to nonpsychiatrist mental health providers but unchanged psychiatrist visits, ED visits, and hospitalizations. HDHPs do not appear to have a “blunt instrument” effect on health care use in bipolar disorder; rather, patients might make trade-offs to preserve important care.

Am J Manag Care. 2020;26(6):248-255. https://doi.org/10.37765/ajmc.2020.43487

Takeaway Points

  • High-deductible health plan (HDHP) members with bipolar disorder experienced a moderate decrease in nonpsychiatrist mental health outpatient visits, but rates of psychiatrist visits, medication use, emergency department visits, and hospitalizations did not change.
  • HDHP members with bipolar disorder might elect to pay more out of pocket to maintain psychiatrist care and associated medication use but not nonpsychiatrist mental health provider care.
  • HDHPs do not appear to have a “blunt instrument” effect on health care use in bipolar disorder; rather, patients might make trade-offs to preserve important care.
  • Policy makers, employers, and health plans could use these findings to construct highly efficient value-based or tailored health insurance designs that optimize health care use and spending; for example, plans might reduce out-of-pocket costs for nonpsychiatrist mental health provider visits to enhance use.

Bipolar disorder is a serious mental illness characterized by acute episodes of mania, hypomania, and depression. Bipolar spectrum disorders include subtypes bipolar I, bipolar II, and cyclothymia. These often have an early age of onset, high risk of suicide, and high rates of co-occurring psychiatric conditions such as substance use disorders.1,2 In the United States, the 12-month prevalence of bipolar disorder is 2.8% and the lifetime prevalence is 4.4%.3 Although persons with bipolar disorder can have asymptomatic periods, episodes of clinical instability and impairment are common.2 In addition to managing acute mood episodes, guideline-recommended care for bipolar disorder requires a chronic care model of evidence-based medications, ongoing follow-up, and often evidence-based psychotherapies during nonmanic phases of care, such as individual and group psychoeducation, cognitive behavior therapy, interpersonal social rhythm therapy, and family-focused therapy, to reduce depressive symptoms or prevent future manic or depressive episodes.4,5

In an effort to control rising health care costs, payers and employers are increasingly adopting high-deductible health plans (HDHPs)6 that require high out-of-pocket payments.6,7 HDHP advocates believe that providing patients with information about the quality of medical services while exposing them to greater costs will create “activated healthcare consumers”8 who will seek higher-value health care, adopt healthy behaviors, and reduce future costs.

HDHPs have proliferated over the past decade. In 2018, 58% of covered workers had deductibles of $1000 or more and 26% had deductibles of $2000 or more.9 Based on findings going back to the RAND Health Insurance Experiment (HIE)10 of the 1970s and 1980s, high cost sharing is generally considered a “blunt instrument” that reduces all health care utilization among all types of patients. However, more recent studies have found that such reductions do not occur in all clinical situations11-14 or for all subgroups.13,14 HDHP effects on patients with serious mental illness including bipolar disorder are unknown.

We chose to study patients with bipolar disorder because it is a serious mental illness that is common enough among commercially insured persons to study robustly. Bipolar disorder is also a chronic condition and requires consistent access to medications and expensive specialist care. This population could be “tipped over the edge” by even small reductions in access given that suboptimal adherence to bipolar medications can result in debilitating episodes of mania or depression. We hypothesized that the higher out-of-pocket spending for specialist care under HDHPs would reduce mental health outpatient visits to both psychiatrists and nonpsychiatrist mental health providers and that reduced psychiatrist visits would reduce bipolar medication fills. The expected direction of emergency department (ED) visits and hospitalizations was uncertain.


Data Source and Population Setting

We drew our commercially insured study population from a large national claims database with members enrolled between January 2003 and December 2012. Data included medical, pharmacy, and outpatient claims from members of a large national health plan.

Research Design and Study Groups

We used an interrupted time series with propensity-matched comparison series research design.15-17 We included individuals with bipolar disorder who were enrolled in low-deductible plans during a baseline year. Some experienced an employer-mandated switch to HDHPs and were followed for a subsequent year. Others continued in low-deductible plans because their employers offered only these arrangements, and we followed these members as the control group. Thus, our study groups were not offered a choice of deductible level, minimizing self-selection.

Employer selection and study group identification. Our process for identifying study groups involved first identifying eligible employers and then eligible patients with bipolar disorder at those employers. In a given benefit year (the 12 months when an employer provides certain health insurance benefits), we classified employers as offering low- or high-deductible coverage based on offering exclusively annual deductibles of $0 to $500 or at least $1000, respectively (eAppendix section I[A] and eAppendix Table 1 [eAppendix available at ajmc.com]). We then defined the pool of potential HDHP group employers as those with at least 1 year of low-deductible coverage followed by at least 1 year of mandated HDHP coverage.

We defined the index date for employers that switched to HDHPs as the first day of the month when the switch occurred. We defined the index date for employers that did not switch plans as the first day of the month when their yearly account was renewed. Members had index dates as early as January 1, 2004, and as late as January 1, 2012. For each person, we defined the beginning of the baseline period as 12 months before the employer’s index date.

Bipolar cohort selection. Study cohort selection began with 53 million overall members up to age 64 years enrolled during 2000-2012 (eAppendix Figure 1). Similar to prior research,18-21 we used a hierarchical algorithm and evidence of outpatient and inpatient bipolar disorder diagnoses captured in International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes to classify patients as having bipolar I, bipolar II, or uncertain type (eAppendix section I[B]).

After applying this algorithm, the bipolar cohort included 333,780 members. After restricting to members who were continuously enrolled for at least 2 years in the previously defined employers, who were aged 12 to 64 years during the calendar year of the index date, who had a most recent bipolar diagnosis that occurred from 5 years to 4 months prior to the index date, and whose employers did not carve out mental health benefits, we had a prematch pool of 2950 HDHP members and 26,911 controls with bipolar disorder.

Outcome Measures

Primary measures included mental health outpatient visits, overall and subclassified into (1) visits to psychiatrists and (2) visits to nonpsychiatrist mental health providers (eg, psychologists, counselors, mental health social workers). We chose this categorization because among mental health specialists, psychiatrists typically prescribe medications for bipolar disorder, whereas nonpsychiatrist mental health specialists commonly provide psychotherapy. In secondary analyses, we also examined the remaining small subset of mental health visits to (3) non—mental health specialist clinicians such as primary care providers. To better understand changes in use of health care that was not specifically related to mental health, we also measured non–mental health–related office visits, overall ED visits, and overall hospitalizations. Non–mental health–related visits comprised all office visits (including hospital outpatient visits) not classified as mental health related using the approach described earlier. That is, they included visits to non–mental health specialists that had evaluation and management codes for office visits (except those that were mental health specific) and a primary or secondary diagnosis that was not related to mental health.

As a secondary measure, we assessed bipolar medication use to determine whether this key component of outpatient bipolar disorder therapy was affected. Our measure was the number of days per study year that the member had at least 1 medication that is recommended for the treatment of bipolar disorder (lithium, atypical antipsychotics, and select anticonvulsants) on hand. We also assessed out-of-pocket spending on these health services per member and per visit, event, or prescription. Details about measure construction are included in the eAppendix.


Using 2000 US Census block group data and validated methods,22,23 we classified members according to the income and education levels of their neighborhood (Table 1 [part A and part B]).22-24 We applied the Johns Hopkins ACG algorithm, a validated measure that predicts mortality,25,26 to members’ baseline year to estimate general morbidity, including both mental and physical conditions. We matched study groups on morbidity level because patients with differing levels of morbidity will have differing cost exposure and contact with the health system. We first used geocoding to classify patients as being from white, black, Hispanic, or mixed neighborhoods and then, when available, overwrote a participant’s classification as Hispanic or Asian using the E-Tech (Ethnic Technologies) system,27 which analyzes full names and geographic locations of individuals. This validated approach of combining surname analysis and Census data has positive and negative predictive values of approximately 80% and 90%, respectively.28 Other covariates included age category (12-18, 19-29, and 30-64 years) and US region of residence (West, Midwest, South, Northeast).

Member-level clinical covariates included bipolar disorder type (type I, type II, or uncertain type) (eAppendix section I[B]), number of baseline outpatient psychiatrist and nonpsychiatrist mental health provider visits per quarter, substance use disorder diagnosis (ICD-9-CM codes 291.x-292.x, 303.xx-305.xx), number of mental health ED visits per quarter, number of baseline mental illness (bipolar disorder or depression) inpatient days per quarter, number of baseline 30-day equivalent bipolar medication fills per quarter, and number of months from first observed bipolar diagnosis to the index date.

Derived employer-level characteristics included employer size (1-49, 50-99, 100-249, 250-499, 500-999, or ≥1000 enrollees); index month/year; percentage of women; percentage of enrollees in each of 4 US regions and in income, education, age, and race/ethnicity categories; employer baseline cost level and trend derived from a standardized cost variable; and mean employer ACG score.


To further minimize potential selection effects, we used a 2-level (employer- and member- level) propensity score matching approach29,30 and estimated propensity scores predicting the likelihood of a mandated HDHP switch based on the employer-level covariates defined above. Within these quartiles, we performed the member-level match between contemporaneous HDHP and control group members in the match pool with index dates between January 1, 2004, and January 1, 2012 (matching details in eAppendix section I[D]). After propensity score matching, the final study sample included 2862 HDHP members with bipolar disorder and 7705 matched controls.


We compared baseline characteristics of our study groups using a standardized differences approach.31 To determine whether HDHP members might be dropping out of their employers’ health plans at different rates compared with control members, we compared characteristics of the groups that had at least 12, 13, 18, and 24 months of enrollment, counting from the beginning of the baseline period.

After analyzing interrupted time series data (eAppendix section I[E]) and finding that all primary outcome measures had parallel baseline trends (data not shown), we used generalized estimating equations in a difference-in-differences analytic framework to compare changes in outcomes among members in the year before and after the mandated HDHP switch vs controls. Regression models for outpatient visits and corresponding out-of-pocket spending used a negative binomial distribution and were adjusted for employer size, calendar year of index date, gender, age category, race, ethnicity, and neighborhood income level and education level. The term of interest was the 2-way interaction between an indicator of HDHP vs control group and an indicator of the year before vs after the index date. Using terms from the regression model, we then used marginal effects methods to calculate mean adjusted baseline and follow-up outpatient visit rates and out-of-pocket spending, as well as absolute and relative changes.32 We used a similar approach to model ED visits, hospitalizations, and corresponding out-of-pocket spending, but we used a zero-inflated negative binomial distribution to account for patients with zero utilization and spending in these areas.


After matching, all standardized differences between HDHP and control group characteristics were well below 0.2, indicating minimal differences (Table 1).31 The average age of HDHP and control members was approximately 37 years, and 62% in each group were women. Approximately 26% lived in low-income neighborhoods, 16% lived in low-education neighborhoods, 4% were Hispanic, and the mean (SD) ACG morbidity score was 2.1 (3.0). About half of members (HDHP, 45.4%; control, 48.0%) were enrolled through midsize employers with 100 to 999 enrollees. Differential dropout was minimal and thus should not bias our results (eAppendix Tables 4 and 5).

In adjusted difference-in-differences analyses, out-of-pocket spending from before to after the index date for all medical services increased by 20.9% (95% CI, 15.7%-26-3%) among HDHP members relative to controls (eAppendix Table 6 and eAppendix Figure 2). HDHP members experienced a 21.9% (95% CI, 15.1%-28.6%) increase in mental health visit prices (spending per visit) in the follow-up year compared with baseline relative to control group members (Table 2) but no detectable changes in overall mental health outpatient visits (relative change, —4.6% [95% CI, –11.7% to 2.5%]) (Figure). The HDHP group experienced a 13.1% (95% CI, 4.1%-22.2%) relative pre- to post increase in out-of-pocket spending per member for outpatient mental health visits (eAppendix Figure 3).

Analyses by visit type revealed that, in the context of a 33.8% (95% CI, 4.3%-63.3%) relative increase in visit price, HDHP members reduced nonpsychiatrist mental health provider visits by 10.9% (95% CI, —20.6% to –1.3%) (Table 2 and Figure). In contrast, psychiatrist visits were unchanged (0.6% [95% CI, –9.3% to 10.6%]) despite a 17.8% (95% CI, 12.2%-23.4%) increase in visit price. Per-member out-of-pocket spending changes were 4.9% (95% CI, –9.0% to 18.9%) for nonpsychiatrist mental health provider visits and 15.5% (95% CI, 4.7%-26.4%) for psychiatrist visits (eAppendix Figure 3). Non–mental health provider visits for mental health reasons and corresponding out-of-pocket spending were unchanged (eAppendix Table 7). HDHP members experienced a —9.0% (95% CI, –23.5% to 5.5%) nonsignificant relative change in non–mental health visits (Table 2).

HDHP members experienced no detectable changes in the proportion of days with any bipolar medication on hand after the HDHP switch relative to controls (Table 2 and eAppendix Figure 4) despite an increase in 30-day fill price of 16.8% (95% CI, 10.3%-23.3%) at follow-up vs baseline. Similarly, ED visits and hospitalizations were unchanged (Table 3), whereas per-event prices increased by 40.1% (95% CI, 20.5%-59.8%) and 38.7% (95% CI, 3.7%-73.8%), respectively.


After an employer-mandated transition to HDHPs, commercially insured members with bipolar disorder maintained stable overall mental health outpatient visit rates relative to similar patients who remained in low-deductible health plans. However, nonpsychiatrist mental health provider visits declined by 11% among HDHP members, whereas psychiatrist visits were unchanged despite increased prices per visit. We did not find statistically significant changes in use of bipolar medications, ED visits, or hospitalizations.

A potential interpretation is that HDHP members with bipolar disorder elected to pay more out of pocket to maintain psychiatrist care and associated medication use but not nonpsychiatrist mental health provider care. HDHP members with bipolar disorder might be trying to preserve visits that they needed for medication refills, changes, or dose adjustments. Or, given that visits to nonpsychiatrist mental health providers generally occur more frequently, patients might view reducing the frequency of such visits as a necessary trade-off to manage their out-of-pocket expenses. It is difficult to interpret the lack of ED visit changes, in part because visits could simultaneously decrease because of HDHP financial barriers and increase because of inappropriately deferred care, leaving overall rates unchanged. In addition, HDHP members might not be able to reduce this key source of care given that research has demonstrated that individuals with bipolar disorder use the ED to obtain urgent mental health care.33

Previous research has not addressed effects of HDHPs on outpatient care among patients with mental health conditions. Earlier studies examined the broad impact of cost sharing on mental health care utilization. For example, the RAND HIE of the 1970s and 1980s found that higher cost sharing led to lower mental health care utilization compared with the demand for general medical services.10,34 In contrast, a more recent study found that patients’ price sensitivity for mental health and general medical care was comparable.35 Similar to our results, that study found significant variation in price sensitivity within types of mental health care (eg, visits, medications). However, these older studies did not examine populations with serious mental illness and were unable to distinguish cost-sharing effects on patient visits to different provider types. We are also not aware of research regarding the association between HDHP enrollment and changes in nonpsychiatrist mental health care visits, emergency care, and hospitalizations. In a context of reduced nonpsychiatrist mental health care visits, we found that rates of ED visits and hospitalizations did not change for HDHP members.

Substantially more is known about the impact of cost sharing on non—mental health care use and outcomes. The RAND HIE10 found that high out-of-pocket costs reduced almost all utilization to a similar degree. However, similar to our results among individuals with bipolar disorder, more recent studies have found that such reductions do not occur consistently in all clinical situations12-14 or for all demographic subgroups.13,14,36 Recent research has generally found that when services are subject to the deductible, utilization decreases more among vulnerable members (eg, low-income, high-morbidity members) than less vulnerable members.13,14,37,38

The RAND study also predicted that utilization decreases would not worsen health outcomes except among vulnerable members. Few modern studies have assessed health outcomes, but 2 detected evidence of increased adverse outcomes among low-income and high-morbidity members with diabetes.13,14

Our study thus adds the novel insight that commercially insured patients with a serious mental illness might not experience cost sharing in HDHPs as a “blunt instrument” but rather make trade-offs to preserve certain types of mental health care. Policy makers, employers, and health plans could therefore use our findings to construct highly efficient value-based39 or tailored40 health insurance designs that optimize health care use and spending; for example, plans might reduce out-of-pocket costs for nonpsychiatrist mental health provider visits to enhance use while maintaining cost sharing for psychiatrist visits. Findings could also help clinicians recognize high-deductible insurance as a risk factor for patients with serious mental illness to reduce their use of certain mental health—related services.

Further research should investigate whether the reduced visits to nonpsychiatrist mental health providers that we detected cause adverse mental health or clinical events and should examine detailed medication adherence measures. Future studies should also address effects of HDHP transitions on patients with bipolar disorder who have lower incomes, higher morbidity, and higher levels of cost sharing under HDHPs with health savings accounts.


Our study has several limitations. We studied enrollees with employer-sponsored health insurance enrolled through a single national claims database. In addition, this study did not focus on the most vulnerable subset of members with bipolar disorder who might have public insurance or be uninsured. Nevertheless, given that HDHPs are almost exclusively a feature of commercial health insurance, our results should be generalizable to many individuals with bipolar disorder in employer-sponsored plans. Although we measured changes in ED visits and hospitalizations, these are imperfect health outcomes measures and our study was unable to assess important outcomes such as mania, depression,41 or suicide attempts. Our data did not allow us to classify members as the beneficiary, spouse, or dependent. Observational studies are subject to bias from unmeasured confounders. However, the controlled interrupted time series design that we used guards against most threats to validity.15 Finally, the patients included in our study were in different phases of their condition with regard to duration of recognized disease and disease control but unlikely to be imbalanced by study group.


HDHP members with bipolar disorder who faced substantial increases in cost sharing preserved psychiatrist visits while reducing nonpsychiatrist mental health provider visits. Further research should determine how such reduced care affects mental health and clinical outcomes such as suicide, mania, and depression.


Research reported in this publication was funded through a Patient-Centered Outcomes Research Institute (PCORI) award (IHS-1408-20393). The Harvard Pilgrim Healthcare Institutional Review Board approved the research protocol. We thank Carina Araujo-Lane, MSW, and Francesca Napolitano, PharmD. We would like to acknowledge the project team’s stakeholder advisory panel for consistent engagement with the project team during the development and execution of the study: Kimberly Allen, MS, LCDC, PRS, CPSS; Gregory E. Simon, MD, MPH; Francisca Azocar, PhD; Denise D’Aunno, MBA; Kenneth Dolan-Del Vecchio, MSW; Kristin A. Olbertson, JD, PhD; Ken Duckworth, MD; and James Sabin, MD. We would also like to thank hundreds of individuals in the Depression and Bipolar Support Alliance social media and advocacy community who have assisted us with insights on living with bipolar disorder. The statements in this publication are solely the responsibility of the authors and do not necessarily represent the views of PCORI, its Board of Governors, or its Methodology Committee.Author Affiliations: Division of Health Policy and Insurance Research, Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Healthcare Institute (JFW, FZ, MC, RFL, SS, DR-D, CYL), Boston, MA; McLean Hospital (ABB), Belmont, MA; Department of Healthcare Policy, Harvard Medical School (ABB), Boston, MA; Department of Pharmacy and Health Systems Sciences, School of Pharmacy, Bouvé College of Health Sciences, Northeastern University (JM), Boston, MA; Depression and Bipolar Support Alliance (PF), Chicago, IL.

Source of Funding: Research reported in this publication was funded through a Patient-Centered Outcomes Research Institute Award (IHS-1408-20393).

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 (JFW, ABB, MC, SS, DR-D, CYL); acquisition of data (JFW, MC, PF, CYL); analysis and interpretation of data (JFW, ABB, JM, FZ, MC, RFL, PF, SS, DR-D, CYL); drafting of the manuscript (JFW, MC, SS, CYL); critical revision of the manuscript for important intellectual content (JFW, ABB, JM, FZ, MC, PF, SS, DR-D, CYL); statistical analysis (JFW, FZ, MC, RFL); provision of patients or study materials (JFW, MC, PF); obtaining funding (JFW, JM, FZ, MC, PF, SS, CYL); administrative, technical, or logistic support (JFW, JM, MC, RFL, PF); and supervision (JFW, MC).

Address Correspondence to: J. Frank Wharam, MD, MPH, Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Healthcare Institute, 401 Park St, Ste 401, Boston, MA 02215. Email: jwharam@post.harvard.edu.REFERENCES

1. Judd LL, Akiskal HS, Schettler PJ, et al. A prospective investigation of the natural history of the long-term weekly symptomatic status of bipolar II disorder. Arch Gen Psychiatry. 2003;60(3):261-269. doi:10.1001/archpsyc.60.3.261

2. Judd LL, Akiskal HS, Schettler PJ, et al. The long-term natural history of the weekly symptomatic status of bipolar I disorder. Arch Gen Psychiatry. 2002;59(6):530-537. doi:10.1001/archpsyc.59.6.530

3. Merikangas KR, Akiskal HS, Angst J, et al. Lifetime and 12-month prevalence of bipolar spectrum disorder in the National Comorbidity Survey replication. Arch Gen Psychiatry. 2007;64(5):543-552. doi:10.1001/archpsyc.64.5.543. Published correction appears in Arch Gen Psychiatry. 2007;64(9):1039. doi:10.1001/archpsyc.64.9.1039

4. National Collaborating Centre for Mental Health. Bipolar Disorder: The Management of Bipolar Disorder in Adults, Children and Adolescents, in Primary and Secondary Care. British Psychological Society; 2006.

5. The Management of Bipolar Disorder Working Group. VA/DoD clinical practice guideline for management of bipolar disorder in adults. U.S. Department of Veterans Affairs. 2009. Accessed August 9, 2017. https://www.healthquality.va.gov/bipolar/bd_305_full.pdf

6. 26 CFR 601.602: tax forms and instructions. Internal Revenue Service. Accessed May 31, 2019. https://www.irs.gov/pub/irs-drop/rp-18-30.pdf

7. Health savings accounts and other tax-favored health plans. Internal Revenue Service. January 30, 2020. Accessed April 26, 2020. http://www.irs.gov/pub/irs-pdf/p969.pdf

8. Herzlinger RE. Let’s put consumers in charge of health care. Harv Bus Rev. 2002;80(7):44-50, 52-45, 123.

9. Claxton G, Rae M, Long M, Damico A, Foster G, Whitmore H. 2017 employer health benefits survey. Kaiser Family Foundation. September 19, 2017. Accessed October 28, 2017. https://www.kff.org/health-costs/report/2017-employer-health-benefits-survey/

10. Newhouse JP. Free for All? Lessons From the RAND Health Insurance Experiment. Harvard University Press; 1993.

11. 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. Published correction appears in JAMA. 2008;299(2):171. doi:10.1001/jama.299.2.171

12. Selby JV, Fireman BH, Swain BE. Effect of a copayment on use of the emergency department in a health maintenance organization. N Engl J Med. 1996;334(10):635-641. doi:10.1056/NEJM199603073341006

13. Wharam JF, Zhang F, Eggleston EM, Lu CY, Soumerai SB, Ross-Degnan D. Effect of high-deductible insurance on high-acuity outcomes in diabetes: a Natural Experiment for Translation in Diabetes (NEXT-D) study. Diabetes Care. 2018;41(5):940-948. doi:10.2337/dc17-1183

14. 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):358-368. doi:10.1001/jamainternmed.2016.8411

15. Shadish WR, Cook TD, Campbell DT. Experimental and Quasi-experimental Designs for Generalized Causal Inference. 2nd ed. Houghton Mifflin; 2001.

16. Hallberg K, Cook TD, Steiner PM, Clark MH. Pretest measures of the study outcome and the elimination of selection bias: evidence from three within study comparisons. Prev Sci. 2018;19(3):274-283. doi:10.1007/s11121-016-0732-6

17. St. Clair T, Cook TD, Hallberg K. Examining the internal validity and statistical precision of the comparative interrupted time series design by comparison with a randomized experiment. Am J Eval. 2014;35(3):311-327. doi:10.1177/1098214014527337

18. Busch AB, Yoon F, Barry CL, et al. The effects of mental health parity on spending and utilization for bipolar, major depression, and adjustment disorders. Am J Psychiatry. 2013;170(2):180-187. doi:10.1176/appi.ajp.2012.12030392

19. Busch AB, Huskamp HA, Neelon B, Manning T, Normand SL, McGuire TG. Longitudinal racial/ethnic disparities in antimanic medication use in bipolar-I disorder. Med Care. 2009;47(12):1217-1228. doi:10.1097/MLR.0b013e3181adcc4f

20. Busch AB, Huskamp HA, Landrum MB. Quality of care in a Medicaid population with bipolar I disorder. Psychiatr Serv. 2007;58(6):848-854. doi:10.1176/ps.2007.58.6.848

21. Busch AB, Ling D, Frank RG, Greenfield SF. Changes in the quality of care for bipolar I disorder during the 1990s. Psychiatr Serv. 2007;58(1):27-33. doi:10.1176/ps.2007.58.1.27

22. Krieger N. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology. Am J Public Health. 1992;82(5):703-710. doi:10.2105/ajph.82.5.703

23. Krieger N, Chen JT, Waterman PD, Rehkopf DH, Subramanian SV. Race/ethnicity, gender, and monitoring socioeconomic gradients in health: a comparison of area-based socioeconomic measures—the public health disparities geocoding project. Am J Public Health. 2003;93(10):1655-1671. doi:10.2105/ajph.93.10.1655

24. Census 2000 Gateway. United States Census Bureau. Accessed June 2, 2018. https://www.census.gov/main/www/cen2000.html

25. The Johns Hopkins ACG System. Accessed April 26, 2020. https://www.hopkinsacg.org

26. Reid RJ, Roos NP, MacWilliam L, Frohlich N, Black C. Assessing population health care need using a claims-based ACG morbidity measure: a validation analysis in the Province of Manitoba. Health Serv Res. 2002;37(5):1345-1364. doi:10.1111/1475-6773.01029

27. Frequently asked questions. Ethnic Technologies. Accessed October 24, 2018. https://www.ethnictechnologies.com/faq

28. Fiscella K, Fremont AM. Use of geocoding and surname analysis to estimate race and ethnicity. Health Serv Res. 2006;41(4 pt 1):1482-1500. doi:10.1111/j.1475-6773.2006.00551.x

29. Cook EF, Goldman L. Performance of tests of significance based on stratification by a multivariate confounder score or by a propensity score. J Clin Epidemiol. 1989;42(4):317-324. doi:10.1016/0895-4356(89)90036-x

30. Schreyögg J, Stargardt T, Tiemann O. Costs and quality of hospitals in different health care systems: a multi-level approach with propensity score matching. Health Econ. 2011;20(1):85-100. doi:10.1002/hec.1568

31. Yang D, Dalton JE. A unified approach to measuring the effect size between two groups using SAS. Paper presented at: SAS Global Forum 2012; April 22-25, 2012; Orlando, FL. Accessed July 24, 2017. http://support.sas.com/resources/papers/proceedings12/335-2012.pdf

32. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986;42(1):120-131. doi:10.2307/2531248

33. Waxmonsky J, Verchinina L, Kim HM, et al. Correlates of emergency department use by individuals with bipolar disorder enrolled in a collaborative care implementation study. Psychiatr Serv. 2016;67(11):1265-1268. doi:10.1176/appi.ps.201500347

34. Manning WG, Newhouse JP, Duan N, Keeler EB, Leibowitz A, Marquis MS. Health insurance and the demand for medical care: evidence from a randomized experiment. Am Econ Rev. 1987;77(3):251-277.

35. Meyerhoefer CD, Zuvekas SH. New estimates of the demand for physical and mental health treatment. Health Econ. 2010;19(3):297-315. doi:10.1002/hec.1476

36. Kozhimannil KB, Law MR, Blauer-Peterson C, Zhang F, Wharam JF. The impact of high-deductible health plans on men and women: an analysis of emergency department care. Med Care. 2013;51(8):639-645. doi:10.1097/MLR.0b013e31829742d0

37. 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

38. Wharam JF, Zhang F, Landon BE, LeCates R, Soumerai S, Ross-Degnan D. Colorectal cancer screening in a nationwide high-deductible health plan before and after the Affordable Care Act. Med Care. 2016;54(5):466-473. doi:10.1097/MLR.0000000000000521

39. Fendrick AM, Chernew ME. Value-based insurance design: a “clinically sensitive” approach to preserve quality of care and contain costs. Am J Manag Care. 2006;12(1):18-20.

40. Wharam JF, Ross-Degnan D, Rosenthal MB. The ACA and high-deductible insurance—strategies for sharpening a blunt instrument. N Engl J Med. 2013;369(16):1481-1484. doi:10.1056/NEJMp1309490

41. Busch AB, Frank RG, Sachs G. Bipolar-I depression outpatient treatment quality and costs in usual care practice. Psychopharmacol Bull. 2008;41(2):24-39.