Online-Only Supplement
eMethods eResults
eFigure 1: Sample Size Flowchart
eTable 1: Changes in the time trends of average monthly BH spending and utilization associated with MHPAEA, among all enrollees.
eTable 2: Changes in the time trends of monthly penetration rates and per-user amounts of BH spending and utilization associated with MHPAEA, comparing the post period to the pre-parity period.
eTable 3: Changes in the time trends of average monthly BH spending and utilization associated with MHPAEA, among all enrollees, for members associated with employers continuously enrolled for all months during 2008-2013.
eTable 4: Diagnosis codes included in each diagnostic category. (Please see file
“10_SuppDiagnoses20160429_V02.xls”)
eMethods
Sample considerations
We excluded enrollees younger than 27 to filter out potential effects due to the Affordable Care Act, which could have resulted in newly-covered dependents to enter our sample as early as 2010. MHPAEA did not apply to retiree plans (thus the exclusion of people 65 and older), employers with 50 or fewer employees, employee assistance program (EAP) –only or work/life-only plans, or plans offering fixed indemnity insurance. A plan’s renewal date determined its compliance deadline; thus, we focused on calendar-year plans. Plans subject to collective bargaining agreements were excluded because they could delay compliance until the agreement’s termination. Both SI and FI plans were subject to MHPAEA, but we did not include FI plans in our sample because we had considered using FI plans to define a control group (described below), there were very few FI plans, and we were doubtful employer and plan characteristics were similar between the few FI plans and the SI plans that made up the bulk of our data.
Because our study period spanned 72 months (6 years: 2008-2013), we did not require individuals to be continuously enrolled for each month of the study. However, to see if changing demographics could have confounded our estimates of MHPAEA’s effects on spending and utilization, we used individuals associated with continuously enrolled employers in a sensitivity analysis (described below).
Comparison group and comparison outcome considerations
We considered two comparison groups for our analyses. First, we looked at small groups (employers with 50 or fewer employees), as they were exempt from compliance with MHPAEA. Then, we looked at fully insured (FI) plans in states we classified as enacting strong parity laws prior to MHPAEA. This is because, unlike self- insured (SI) plans, FI plans are subject to state insurance mandates. Thus, in states that already enacted some form of mental health insurance parity, people enrolled in FI plans should have already been exposed to parity provisions before MHPAEA. These people would theoretically have smaller changes in BH spending and utilization in response to MHPAEA, compared to the “intervention group” in SI plans. Unfortunately, however, neither potential comparison group had enough data to reliably model monthly changes in our outcomes.
Furthermore, even if we did have enough data to use these groups as controls, each potential comparison group had theoretical limitations. We had doubts that secular time trends could be similar for our large employers and the potential control group of small employers, as trends in utilization and expenditures among groups with 50 or fewer employees are highly unlikely to generalize to the very large employers in our Optum databases, where groups of fewer than 5,000 employees were already small in relative terms.
We had several concerns about our other potential comparison group of people enrolled in FI plans in strong parity states. First, defining “strong parity states” was problematic; heterogeneity in the populations affected by the laws and in the benefit design features included in the laws made it difficult to draw a clear distinction between states that had “strong” parity laws and states that did not. Furthermore, even states that appeared to have strong parity laws may not have enforced these laws. Most importantly, the NQTL provisions that were such a critical part of MHPAEA were unique to federal parity. Oregon, the one state that mandated parity in NQTLs, anecdotally never enforced this requirement prior to implementation of MHPAEA. Thus, if the comparison group was itself affected by MHPAEA, then formal “difference-in-differences” (DID) estimates would be subject to conservative biases.
service and made a claim received a diagnosis, making the diagnosis category variables endogenous. While the inclusion of the diagnosis category variables could also cause endogeneity issues in the conditional models (among users, higher service use tends to result in a larger number of diagnoses), we wanted to make sure the exclusion of diagnosis category did not bias these models. Secondly, we re-ran our unconditional models, excluding the plan type covariate from the model specification. We were concerned that MHPAEA may have directly affected changes in the types of plans offered during the study period, which could have made the plan type variable a mediator for parity’s effects. Next, we varied the type of regression used (gamma vs. linear for conditional models, and linear regression vs. two-part model predictions for unconditional models). Lastly, we re-ran our main analysis models on the subset of individuals associated with employers continuously enrolled from 2008-2013.
eResults
Changes in penetration rates and per-user amounts of spending and utilization associated with MHPAEA eTable 2 presents the changes in the time trends of monthly penetration rates and per-user amounts of spending and utilization, for the post-parity period. MHPAEA is associated with increases in the penetration rates and per-user amounts of expenditures and outpatient utilization, but declines in the penetration rates and per-user amounts of intermediate care utilization. Specifically, for penetration rates, there is a level increase in the percent of enrollees with any plan expenditures (0.287% [p=0.01]) and outpatient assessment/diagnostic evaluation visits (0.036% [p=0.00]) at the beginning of the post-parity period, relative to the level expected based on the pre-parity trend. The percent of enrollees with any outpatient individual psychotherapy visits starts off the post period with a one-time increase in level of 0.162% (p=0.02) and has faster growth throughout the post period, relative to the pre- parity period (increase in slope of 0.006%/month [p=0.05]). However, the percent of enrollees with any structured outpatient care decreases in level by 0.005% (p=0.02) in January 2011.
Among users, parity is associated with an increase in level of mean total expenditures ($20.01 [p=0.04]), outpatient medication management visits (0.052 [p=0.00]), family psychotherapy visits (0.072 [p=0.04]), and days of inpatient care (0.560 [p=0.04]) at the beginning of the post period. Individual psychotherapy has both an increase in level of 0.119 visits (p=0.00) and an increase in slope of 0.002 visits/month in the post period (p=0.01), relative to the pre-parity period. Conversely, among users parity is associated with decreases for all three intermediate care measures: days of structured outpatient care, day treatment care, and residential care have decreases in level of 0.972 days (p=0.02), 1.552 days (p=0.01), and 6.536 days (p=0.01), respectively, among users of those services.
Among the subsample associated with continuously enrolled employers: changes in per-enrollee means associated with MHPAEA
As shown in eTable 3, parity’s effects were similar for the sample of individuals associated with employers continuously enrolled for all months in 2008-2013. Specifically, when focusing on changes in the post period, MHPAEA was associated with increased total and plan expenditures, though compared to the main analysis, increases in level in January 2011 were smaller in magnitude and no longer statistically significant. Increases in level in January 2011 for outpatient utilization were similar to the main analysis, with immediate increases of 0.00051 assessment/diagnostic evaluation visits (p=0.02) and 0.00570 individual psychotherapy visits (p=0.02).
Results were also similar for structured outpatient days, with the immediate decrease of -0.00070 days (p=0.00) offset by monthly increases of 0.00002 days/month (p=0.03). However, unlike the main analysis, the monthly rate of change for family psychotherapy visits declined (change in slope of -0.00002 visits/month, p=0.02).
eFigure 1. Sample Size Flowchart
People enrolled in any sampled carve-in plan 2008-13: 11,926,924 individuals
Include people with only one plan in the month: 11,914,959 (99.90%)
Include people 27-64 years of age in the given month: 6,869,764 (57.66%)
Include people living in the 50 US states (excluding DC): 6,860,265 (99.86%)
Include standard plans (not retiree, supplemental, or indemnity): 6,685,354 (98.97%)
Include plans that are not collectively bargained: 6,632,941 (99.22%)
Include plans on a calendar-year renewal cycle: 6,017,721 (90.72%)
Include plans associated with a large employer: 6,017,353 (99.99%)
Include self-insured plans: 5,987,776 (99.51%)
Final sample size is 179,506,951 member-months, consisting of 5,987,776 members, 393 employers, and 6,587 plans.
Include people enrolled in a plan with a behavioral health component: 6,754,617 (98.46%)
eTable 1. Changes in the time trends of average monthly BH spending and utilization associated with MHPAEA, among all enrollees.
Transition Period vs. Pre-Parity Period Post Period vs. Pre-Parity Period Outcome Δ Level1 P-value Δ Slope2 P-value Δ Level1 P-value
Expenditures
Total $0.74 0.01 -$0.01 0.76 $1.05 0.02
Plan $0.76 0.01 -$0.02 0.37 $0.88 0.04
Patient Out-Of-Pocket -$0.02 0.81 $0.01 0.11 $0.17 0.20
Outpatient Visits
Assessment/Diagnostic Evaluation 0.00030 0.00 -0.00002 0.09 0.00045 0.00
Medication Management 0.00035 0.23 0.00001 0.88 0.00053 0.27
Individual Psychotherapy 0.00210 0.05 0.00018 0.08 0.00578 0.00
Family Psychotherapy 0.00007 0.43 -0.00001 0.26 0.00026 0.13
Days of Intermediate Care
Structured Outpatient 0.00019 0.14 -0.00002 0.22 -0.00059 0.00
Day Treatment 0.00019 0.01 -0.00001 0.24 -0.00015 0.14
Residential -0.00021 0.38 0.00000 0.66 -0.00056 0.24
Days of Inpatient Care 0.00016 0.09 -0.00002 0.16 0.00004 0.76
Notes: Estimates are from linear regression. Sample is member-months from 2008-2013 (N=179,506,951). Bold denotes significance at p ≤ .05. Interrupted time series segmented regression analysis controlled for a linear monthly time trend, indicators and splines for both the transition and post periods, sex, age group, whether the enrollee was the primary insured person vs. a dependent, employer size category, plan type, state fixed effects, provider supply measures, and seasonality.
1. Discontinuity (change in level) at the beginning of the given period, as measured by an indicator for the given period.
2. Spline (change in slope) for the given period.
eTable 2. Changes in the time trends of monthly penetration rates and per-user amounts of BH spending and utilization associated with MHPAEA, comparing the post period to the pre-parity period.
Change in Penetration Rate1 Change in Mean, Among Users Outcome Δ Level3 P-value Δ Slope4 P-value Δ Level3 P-value
Expenditures
Total 0.181% 0.08 0.006% 0.13 $20.01 0.04
Plan 0.287% 0.01 0.005% 0.30 $4.87 0.68
Patient Out-Of-Pocket 0.207% 0.08 0.001% 0.85 -$0.73 0.85
Outpatient Visits
Assessment/Diagnostic Evaluation 0.036% 0.00 0.000% 0.73 -0.007 0.67
Medication Management -0.001% 0.98 0.002% 0.13 0.052 0.00
Individual Psychotherapy 0.162% 0.02 0.006% 0.05 0.119 0.00
Family Psychotherapy 0.009% 0.35 -0.001% 0.09 0.072 0.04
Days of Intermediate Care
Structured Outpatient -0.005% 0.02 0.000% 0.17 -0.972 0.02
Day Treatment 0.000% 0.70 0.000% 0.33 -1.552 0.01
Residential -0.001% 0.45 0.000% 0.15 -6.536 0.01
Days of Inpatient Care -0.002% 0.08 0.000% 0.27 0.560 0.04
Notes: Sample is member-months from 2008-2013 (N=179,506,951). Bold denotes significance at p ≤ .05. Interrupted time series segmented regression analysis controlled for a linear monthly time trend, indicators and splines for both the transition and post periods, sex, age group, whether the enrollee was the primary insured person vs. a dependent, employer size category, plan type, state fixed effects, provider supply measures, and seasonality.
1. Percent of enrollees with any spending (utilization). Estimates are from logistic regression marginal effects post-estimation; p-values are from logistic regression.
2. Subset of enrollees with any spending (utilization) for the given outcome. Sample size of “users” varies for each outcome. Estimates and p-values are from gamma regression marginal effects post-estimation.
3. Discontinuity (change in level) at the beginning of the post period, as measured by an indicator variable for the post period.
4. Spline (change in slope) for the post period.
eTable 3. Changes in the time trends of average monthly BH spending and utilization associated with MHPAEA, among all enrollees, for members associated with employers continuously enrolled for all months during 2008- 2013.
Transition Period vs.
Pre-Parity Period Post Period vs.
Pre-Parity Period Outcome Δ Level1 P-value Δ Slope2 P-value Δ Level1 P-value
Expenditures
Total $0.58 0.07 $0.02 0.55 $0.81 0.18
Plan $0.62 0.06 $0.01 0.83 $0.71 0.22
Patient Out-Of-Pocket -$0.04 0.61 $0.01 0.08 $0.10 0.58
Outpatient Visits
Assessment/Diagnostic Evaluation 0.00020 0.12 0.00000 0.91 0.00051 0.02
Medication Management 0.00033 0.34 0.00005 0.10 0.00060 0.32
Individual Psychotherapy 0.00282 0.04 0.00022 0.07 0.00570 0.02
Family Psychotherapy 0.00014 0.21 -0.00002 0.11 0.00027 0.25
Days of Intermediate Care
Structured Outpatient 0.00004 0.78 0.00000 0.77 -0.00070 0.00
Day Treatment 0.00015 0.06 -0.00001 0.35 -0.00025 0.08
Residential -0.00035 0.30 0.00000 0.92 -0.00086 0.25
Days of Inpatient Care 0.00012 0.26 0.00000 0.99 0.00009 0.60
Notes: Estimates are from linear regression. Sample is member-months from 2008-2013 (N=117,971,534). Bold denotes significance at p ≤ .05. Interrupted time series segmented regression analysis controlled for a linear monthly time trend, indicators and splines for both the transition and post periods, sex, age group, whether the enrollee was the primary insured person vs. a dependent, employer size category, plan type, state fixed effects, provider supply measures, and seasonality.
1. Discontinuity (change in level) at the beginning of the given period, as measured by an indicator for the given period.
2. Spline (change in slope) for the given period.