|Home | About | Journals | Submit | Contact Us | Français|
The Paul Wellstone and Pete Domenici Mental Health Parity and Addiction Equity Act required health plans to provide mental health and substance use disorder (MH/SUD) benefits on par with medical benefits beginning in 2010. Previous research found that parity significantly lowered average out-of-pocket (OOP) spending on MH/SUD treatment of children. No evidence is available on how parity affects OOP spending by families of children with the highest MH/SUD treatment expenditures.
We used a difference-in-differences study design to examine whether parity reduced families’ (1) share of total MH/SUD treatment expenditures paid OOP or (2) average OOP spending among children whose total MH/SUD expenditures met or exceeded the 90th percentile. By using claims data, we compared changes 2 years before (1999–2000) and 2 years after (2001–2002) the Federal Employees Health Benefits Program implemented parity to a contemporaneous group of health plans that did not implement parity over the same 4-year period. We examined those enrolled in the Federal Employees Health Benefits Program because their parity directive is similar to and served as a model for the new federal parity law.
Parity led to statistically significant annual declines in the share of total MH/SUD treatment expenditures paid OOP (−5%, 95% confidence interval: −6% to −4%) and average OOP spending on MH/SUD treatment (−$178, 95% confidence interval: −257 to −97).
This study provides the first empirical evidence that parity reduces the share and level of OOP spending by families of children with the highest MH/SUD treatment expenditures; however, these spending reductions were smaller than anticipated and unlikely to meaningfully improve families’ financial protection.
There is substantial interest in alleviating the economic impact of mental illness on families.1 Mental health and substance use disorders (MH/SUDs) causing significant functional impairment affect an estimated 11% of the pediatric population.2 MH/SUD problems are the number 1 cause of pediatric hospital admissions among those 10 to 14 years of age, and psychiatric admission rates have nearly doubled among children and adolescents over the past decade.3 Research suggests that out-of-pocket (OOP) spending associated with caring for children with MH/SUDs is high, even relative to children with other special health care needs.4,5 One study found, for example, that 43% of parents of children with special health care needs (CSHCN) that involved an MH/SUD condition reported OOP spending exceeding $1000 annually, compared with only 22% of parents of CSHCN that involved only a medical condition.4 Parents of CSHCN involving an MH/SUD condition were also more likely to report reduced labor market participation and a greater time burden associated with providing and arranging for child care in comparison with other parents of special-needs children.4,5
Historically, high OOP spending for a child’s MH/SUD treatment has been due in part to benefit limits imposed by health plans. When families accessed MH/SUD services for their children, they were required to pay higher coinsurance amounts (ie, copays) for MH/SUD services than for non-MH/SUD medical services. Families faced greater caps on the number of outpatient visits or inpatient days covered annually for MH/SUD treatment compared with non-MH/SUD medical services.6,7
The intent of the Paul Wellstone and Pete Domenici Mental Health Parity and Addiction Equity Act (MHPAEA) implemented by Congress in 2010 was to improve insurance coverage for MH/SUD by requiring that insurance benefits be equivalent to those offered for general medical services (eg, if a health plan paid 85% of the cost for an outpatient medical care visit, then the plan would be required to cover the same share of an outpatient MH/SUD visit). A primary goal of parity advocates was to lower the amount spent OOP, especially for those with the most disabling conditions and highest treatment expenses.
Previous research found that, although parity did not affect average total MH/SUD expenditures, it significantly lowered OOP spending on average for both children8 and adults.9–11 These average reductions in OOP spending attributable to parity were relatively small in magnitude, however.8,9 It is unknown how parity affects those at the high end of the spending distribution who may have been disproportionately affected by MH/SUD benefit limits before parity. A primary aim of parity is to protect individuals with severe mental illnesses against the catastrophic costs of seeking treatment, and, in doing so, increase efficiency and fairness in the insurance market. According to this logic, we would expect more substantial financial protection owing to parity among children with the highest treatment costs.
The aim of this study was to examine whether parity reduced families’ (1) share of total MH/SUD treatment expenditures paid OOP or (2) average OOP spending on MH/SUD treatment among children whose total MH/SUD treatment expenditures were at or above the 90th percentile. Because the MHPAEA was implemented nationwide, there is no obvious national comparison group of privately insured children who did not experience parity to allow for a rigorous evaluation of the law’s effects. Rather, we examined the effects of parity on OOP spending by families of children with high MH/SUD treatment expenditures in the Federal Employees Health Benefits (FEHB) Program because it allows us to directly compare enrollees in a health plan that implemented parity with those in plans that did not. The FEHB Program, covering ~8.5 million enrollees, is the largest private health insurer in the United States. In 2001, under a presidential directive, all FEHB Program plans were required to begin offering comprehensive parity for in-network MH/SUD benefits. The FEHB Program parity directive is similar but not identical to the MHPAEA. Understanding the effects of parity in the FEHB Program on the OOP spending for children with the highest MH/SUD treatment expenditures can provide important insights on the possible effects of the newly implemented federal parity law.
We used 4 years of claims data (1999–2002) from 1 large national health plan in the FEHB Program covering 714449 enrollees and from comparison group health plans in the Truven MarketScan database covering 1192769 enrollees. We examined the 2 years preceding (1999 and 2000) and the 2 years after (2001 and 2002) parity implementation in the FEHB Program. Enrollees aged 21 or younger were included in our study population. Because we restricted our study to children continuously covered over the 4-year study period, those older than 18 in our baseline year (1999) were excluded. Our primary study population was restricted to enrolled children with total MH/SUD expenditures at or above the 90th percentile of all children using MH/SUD treatment services within a calendar year in the FEHB Program and in the MarketScan comparison group. To aid interpretation, we also created annual cohorts of children whose total MH/SUD expenditures were at or below the 50th percentile of enrollees using MH/SUD services.
MH/SUD service use included inpatient and outpatient services associated with specified MH/SUD diagnoses and procedure codes and psychotropic drugs.12 MH/SUD diagnoses were defined as those with diagnostic codes 291, 292, 295 through 309 (except 305.1 and 305.8), and 311 through 314 in the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). A child was considered an inpatient user of MH/SUD services if the last primary diagnosis and the majority of all primary diagnoses in the inpatient admission were MH/SUD diagnoses. A child was considered an outpatient user if any of the following was indicated: an MH/SUD primary diagnosis, a procedure specific to MH/SUD treatment, or a face-to-face encounter at a facility specializing in MH/SUD care. Examples of Current Procedural Terminology procedure codes specific to MH/SUD treatment include, for example, 90801 for a psychiatric interview examination, 90802 to 90815 for individual psychotherapy with or without evaluation and management, and 90865 for medication management. To identify use of psychotropic medications, we developed 2 lists: a restricted list of medications that are used only for MH/SUD and an expanded list of medications used for both MH/SUD and other conditions. Expenditures for any medications on the restricted list counted as spending on a child’s MH/SUD care. If a child used any MH/SUD services or incurred any related expenditures during the year, then expenditures for medications on the expanded list counted as MH/SUD spending.
For descriptive purposes, we used diagnosis codes to group children into mutually exclusive hierarchical categories. Accurate psychiatric diagnosis in the pediatric population can be challenging, particularly for younger children who may have a narrower repertoire of speech and behaviors from which to base a diagnosis. In our diagnostic hierarchy, we aimed to achieve distinctions between diagnostic categories that, on average, are expected to be more severe and/or chronic or complex compared with those that are not.13–15 Also, our goal was to apply a clinical rationale for categorizing individuals. For example, if a child had claims for oppositional defiant disorder and major depression, depression would be the diagnostic category in the descriptive table, because the oppositional behaviors may be related to the depression. Children were required to have the same ICD-9 code listed in the first diagnosis field on at least 2 claims on 2 different service dates during a calendar year or an MH/SUD hospitalization with that diagnosis as the reason for admission to be considered for a clinical condition. For those with >1 MH/SUD diagnosis, we applied the following hierarchy: schizophrenia (ICD-9 code: 295); bipolar disorder (ICD-9 codes: 296.0, 296.1, 296.4–296.8, 301.11, 301.13); depression or anxiety disorders (ICD-9 codes: 296.2, 296.3, 300.4, 301.12, 309.1, 311, 300.0, 300.2, 300.3, 309.81); attention-deficit/hyperactivity disorder (ICD-9 code: 314); conduct/oppositional defiant disorders (ICD-9 codes: 312, 313.81); a residual category of other mental disorders that includes autism-spectrum disorders (ICD-9 codes: 290–314, excluding any of the codes specified in the diagnostic algorithm); adjustment disorders (ICD-9 codes: 308, 309.0, 309.2, 309.3, 309.4, 309.82, 309.83, 309.89, 309.9); or SUD (ICD-9 codes: 291, 292, 303, 304, 305.0, 305.2–306.7, 305.9). For example, if a child met the criteria for schizophrenia, then he or she was considered to have schizophrenia in that claims year; or, a child having 2 claims with attention-deficit/hyperactivity disorder diagnoses would be categorized as such only if she or he did not meet the algorithm criteria for schizophrenia, bipolar disorder, or depression or anxiety disorders. We categorized a child as not meeting diagnostic criteria if he or she had a single claim with an MH/SUD diagnosis, 1 or more relevant procedure or specialty MH/SUD provider codes with no diagnosis, or only psychotropic drug claims (ie, no other MH/SUD services) during the calendar year. Finally, we categorized individuals as having co-occurring SUD and mental illness when 1 or more SUD claims appeared in conjunction with any of the psychiatric diagnoses noted in a calendar year.
We examined the effect of parity on 2 OOP spending outcomes: (1) the share of total MH/SUD expenditures paid OOP and (2) average OOP spending on MH/SUD services among children with total MH/SUD expenditures at or above the 90th percentile in a calendar year in the FEHB plan and comparison group plans before and after implementation of parity in the FEHB Program. We also examined the effect of parity on average total spending on MH/SUD among children in the top expenditure decile. We controlled for a child’s gender, age (ie, 0–3, 4–7, 8–11, 12–15, and 16–21), and region (Northeast, West, South, and Midwest).
We calculated summaries (means and SDs) of the demographic and diagnostic characteristics of children with MH/SUD spending at or above the 90th percentile of all children with any MH/SUD service use in each calendar year. For comparison, we calculated similar summaries of children with MH/SUD spending at or below the 50th percentile of those with MH/SUD service use in each calendar year.
We estimated the effect of parity on OOP spending outcomes in the 90th percentile group by using a difference-in-differences (DD) approach. The DD estimate is the average annual difference before and after the implementation of parity in OOP outcomes in the comparison plans subtracted from the average annual difference before and after implementation of parity in the FEHB plan. Negative values indicate reduced spending for children in the FEHB plan relative to the comparison group plans. This analytic strategy allows us to account for any secular trend in outcomes, with any remaining significant differences attributable to parity. Because we use annual cohorts of children in the top decile of total spending for our DD models, a child may be included in 1 or more years depending on whether he or she hit the top decile in each year. (Given the relatively small number of enrollees in the top decile and the volatility of year-to-year MH/SUD spending among children, it was not feasible to restrict models to children hitting the top decile in all study years.)
Two separate models were estimated corresponding to our 2 OOP spending outcomes. To put the effect of parity on OOP spending in context, we used DD estimation to calculate the effect of parity on total MH/SUD spending among this group of children. To estimate the relationship between parity and share of total spending paid OOP, we modeled the log-odds of each share by using a generalized linear quasi-likelihood model assuming an identify link and normal distribution for the variance term. An unstructured covariance matrix to account for correlation among repeated annual observations for children appearing in >1 annual cohort was assumed. In contrast, to estimate the association of parity and average OOP and total spending, we used a generalized linear quasi-likelihood model with a log link and γ distribution for the variance term. This model makes no distributional assumptions about the distribution of spending and directly models the logarithm of average spending without transforming the data. Models were estimated by using the Proc Genmod function in the SAS software system (SAS Institute, Inc, Cary, NC).
We used bootstrapping techniques to construct 95% confidence intervals for the DD estimates by obtaining the mean predicted outcomes produced by the models.16 Specifically, we sampled children with replacement, estimated the models described previously, and obtained the mean predicted OOP spending outcomes (on original share scale or on original dollar scale) for FEHB Program children and MarketScan children in the preparity period and postparity period. For each child, we then computed the DD by subtracting the postparity-preparity difference for the MarketScan children from the postparity-preparity difference from the FEHB children. We repeated this step 2000 times (sampling children with replacement, estimating models, obtaining predication, and calculating the DD). This process yielded 2000 estimated DDs, and we identified the 2.5th and 97.5th percentile of the estimated DDs to construct a 95% confidence interval. This study was approved by the Harvard Institutional Review Board (M16475-102; September 5, 2008).
As expected, MH/SUD benefits were significantly altered from 2000 to 2001, consistent with the requirements of parity (Table 1). Table 1 also confirms that comparison group plans did not have parity-level benefits in the preparity period, and that benefit levels in these plans did not shift during the course of the study period.
Table 2 reports descriptive information on children enrolled in the national FEHB Program plan with MH/SUD spending at the 90th percentile or above, and with MH/SUD spending at the 50th percentile or below in each year of the study period (1999–2002). This descriptive information illustrates high spending levels and a high diagnostic severity within the 90th percentile group. In the baseline year (1999), average total MH/SUD expenditures were $5530, and average OOP spending was $1423. This group of children had high levels of psychoses, depression, anxiety, and other MH/SUD diagnoses, and 34% had an inpatient MH/SUD hospitalization in 1999. In contrast, in 1999, those in the 50th percentile or below group had substantially lower total and OOP MH/SUD spending ($120 and $68, respectively) and a much less severe diagnostic profile. Table 2 also indicates that, as expected, the share of spending paid OOP among FEHB Program plan enrollees drops in both the 90th percentile and the 50th percentile groups from preparity to postparity.
Table 2 provides comparable information for children enrolled in national comparison group plans with spending at the 90th percentile and above and the 50th percentile and below. The relative magnitude of the differences across high and low spenders in the comparison plans are similar to the patterns observed in the FEHB Program plan. Two important differences between comparison group enrollees and FEHB Program enrollees are worth noting. First, although total MH/SUD spending levels are similar to the FEHB Program enrollees, comparison plan enrollees paid a lower share of this spending OOP. Second, in contrast with FEHB Program enrollees, comparison group enrollees did not experience any drop in OOP MH/SUD spending as a share of total MH/SUD spending over the study period. For both FEHB Program children and comparison group children, we observed relatively large shifts in and out of the high-spender category occurring from year to year, suggesting that the population of high-spending children is not stable. For example, among children enrolled in the FEHB Program with spending at the 90th percentile or above in 1999, only 35% fell into this high-spender category in 2000 and only 17% were high spenders in 2002 (results not shown).
Table 3 summarizes the results from DD models estimating the effects of parity on the share of total MH/SUD spending paid OOP and average OOP MH/SUD spending, respectively. We found that parity was associated with a statistically significant 5 percentage-point decrease in the share of total MH/SUD spending paid OOP attributable to parity within this group of high-spending children. There was also a reduction of $178 in average annual OOP MH/SUD spending in this group.
To put this reduction in OOP spending in context, Table 4 reports results from a DD model estimating the effect of parity on average annual total MH/SUD spending in these children. Adjusted results indicate that average annual total MH/SUD spending among both FEHB Program and comparison group children increases before and after parity; however, the comparison group annual average increase after parity was substantially larger ($1115) than the FEHB Program increase ($137). As a result, the DD model indicates a statistically significant decline of $978 in total MH/SUD spending attributable to parity in this group.
This study is the first to examine how parity affects children with high MH/SUD expenditures. We examined the effects of the parity directive implemented in the FEHB Program, a precursor to the new federal parity law, and found that this policy conferred statistically significant but small reductions in the OOP spending for children with high MH/SUD treatment expenditures. Among children with total MH/SUD spending at the 90th percentile or above, parity reduced the share of MH/SUD spending paid OOP annually by 5 percentage points and lowered annual OOP MH/SUD spending by $178 (after adjusting for inflation, $258 in 2011 dollars). Average annual OOP MH/SUD spending among these high-spending FEHB Program children exceeded $1300 in 2002. One goal of parity is to protect against the risk of large financial losses. Spending reductions detected fell short of the level needed to confer meaningful improvements in financial protection for families grappling with how to pay for a seriously ill child’s treatment. An increase in financial protection of this magnitude translates into a relatively small change in most families’ financial well-being and would not meaningfully affect available resources to pay for other health care services that may be needed within this group of more seriously ill children. In fact, OOP spending reductions among children in the top MH/SUD spending decile were similar in magnitude to the reductions in OOP spending found in a previous study examining the effects of parity on all FEHB children using services (−$62 to −$200),9 a group that includes children with much less serious conditions. These findings suggest that OOP spending reductions for severely ill children after implementation of the MHPAEA (and more recent parity provisions included in the Affordable Care Act) might be smaller than anticipated.
In addition, parity led to a significant reduction in total MH/SUD spending among these more severely ill children relative to the comparison group. This finding is suggestive of a strong managed care response to parity such as a more stringent approach to previous authorization or utilization review or more limits on provider networks. As the generosity of MH/SUD benefits expand, we would expect more service use and higher total spending, unless managed care was applied more stringently after parity as a countervailing force.
It is critical to note that the FEHB Program parity directive differs from the new MHPAEA in a number of important respects. First, the FEHB Program directive applied only to in-network benefits. In contrast, the MHPAE law requires that out-of-network (OON) MH/SUD benefits be equal to OON medical care benefits. Second, the federal agency that administers the FEHB Program, the Office of Personnel Management, explicitly encouraged participating health plans to use managed care techniques to implement parity in a manner that would enable them to control costs.17 In contrast, the MHPAEA interim final regulations issued in February 2010 prohibit plans from using so-called nonquantitative treatment limitations (NQTLs) for MH/SUD benefits unless these limits are comparable to those used for medical benefits.18 NQTLs include medical management standards, prescription drug formulary design, standards for provider admission to participate in a network, and provider reimbursement. In direct contrast with the Office of Personnel Management’s approach, the intent of the NQTL regulations is to explicitly prohibit health plans from managing MH/SUD benefits unless the management strategies used are analogous to those used on the medical side. Both this provision and OON provision of the MHPAEA create the potential for the federal law to produce larger effects on OOP spending than those resulting from the FEHB Program parity policy.
There are several important limitations to this study worth noting. First, as is the case with all quasi-experimental studies, there is the concern about equivalence of the comparison group. We do find that although total MH/SUD expenditures at baseline in the treatment and comparison groups are almost identical, comparison group enrollees had lower OOP spending, which likely reflects the cost-sharing differences noted in Table 1. A critical assumption of the DD approach is that the underlying spending trends in the FEHP Program and comparison plans would have been similar in the absence of the parity policy. Unfortunately, we do not have sufficient pre-parity data to confirm long-term spending trends. Second, we included only enrollees who were continuously enrolled over the 4-year study period. Although this approach may limit the study’s generalizability, inclusion of all children would risk confounding the effects of parity with those of changes in plan enrollment. Third, this study is also limited in its focus on a single FEHB Program plan. Although this plan includes over 700000 enrollees nationwide, it likely differs in certain respects (ie, benefit design features) from other FEHB Program plans and from health plans now covered under the MHPAEA. Fourth, there is the generic concern that claims data undercount service use and family financial burden if some MH/SUD treatment is completely paid for OOP (without using insurance). Finally, the FEHB Program parity directive was implemented in 2001, and the total spending and beneficiary cost-sharing levels reported in this study and the underlying patterns of MH/SUD treatment of children have changed in the intervening years; however, all available evidence suggests that the need for MH/SUD services continues to be unmet and rising.19
A key factor in the passage of the MHPAEA was the belief among policymakers, supported by rigorous academic research,20,21 that parity could improve fairness in insurance coverage without dramatically increasing health care expenditures. The federal parity law is rightly viewed as an antidiscrimination measure, and in that regard it achieved its objective. Other possible objectives might not be so clearly met. Although parity offered some protection against the often high costs of caring for a child with an MH/SUD, it did not lower OOP spending as much as expected.
The authors gratefully acknowledge expert statistical programming provided by Hocine Azeni, MA.
Dr Barry conceptualized and designed the study, drafted the initial article, and approved the final article as submitted; Drs Chien, Normand, Busch, Azzone, and Goldman made substantial contributions to study conception, design and interpretation, revised the article critically for important intellectual content, and approved the final article as submitted; and Dr Huskamp made substantial contributions to study conception, design and interpretation, obtained study data, revised the article critically for important intellectual content, and approved the final article as submitted.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Funded by the National Institutes of Mental Health grants R01MH080797 (all authors), R01MH093414 (Drs Barry, Huskamp, and Goldman), K01MH071714 (Dr Busch), and R01MH054693 (Dr Normand); the National Institute on Drug Abuse grant R01DA026414 (Dr Barry); and the Health Services Research Division of Partners Psychiatry and Mental Health (Dr Busch), a Division of Partners HealthCare. Funded by the National Institutes of Health (NIH).
A version of this study was accepted for podium presentation at the AcademyHealth Child Health Services Research Meeting on June 23, 2012, and at the AcademyHealth Annual Research Meeting on June 24, 2012 in Orlando, Florida.