Consistency of Large Employer and Group Health Plan Benefits with Requirements of the Paul Wellstone and Pete Domenici Mental Health Parity and Addiction Equity Act of 2008. Overview of Key Data Sources and Methodologies


Milliman Compliance Testing Database. Information from Milliman's MHPAEA compliance testing database was used to evaluate 2010 plan design data for adherence to MHPAEA standards. This database includes detailed quantitative financial requirements and treatment limitations for post-parity, pre-IFR benefit levels for medical/surgical benefits and MH/SUD benefits. It also contains details regarding any NQTLs when they could be identified through SPDs.

Of approximately 1,500 plans available in the database, 124 were analyzed to obtain an unbiased and representative distribution of large group plans by geographic region and industry, including self-insured and fully-insured plans. To obtain sufficient information for testing, detailed plan documents and benefit descriptions were requested to identify any financial requirements or treatment limits by detailed service category. To test plan designs for adherence to the quantitative aspects of the legislation, we utilized Milliman's testing model that completes the "substantially all" and "predominant" tests described in the IFR for quantitative financial requirements and treatment limitations. The actuarial-based model relies on Milliman's Health Cost Guidelines for health plans or employers whose membership is not large enough to be statistically reliable, and it includes specific adjustments for variables that impact health care costs such as geographic area, provider contract arrangements, and degree of health care management. If the health plan's or employer's membership was large enough to be statistically reliable (typically more than 10,000 members), the compliance testing model was based on the health plan's or employer's claim costs, usually on a book-of-business basis.

If plan or group-specific costs were used, detailed health care cost data for the most recent complete plan year were requested from the health plan or offeror. Either total allowed dollars or allowed dollars on a PMPM basis were acceptable. Participating health plans and plan sponsors were provided with a template for the level of detail requested by service category, which align with the service categories in Milliman'sHealth Cost Guidelines. Approximately 50 different medical/surgical categories are included.

Quantitative testing was performed on an allowed claim dollar basis (before application of any financial requirements). After the testing model was set up with the costs by detailed health care service category, each medical/surgical service category is mapped into one of the six classifications as prescribed by the IFR, including the two outpatient sub-classifications. Detailed financial requirements and treatment limits by service category were then entered into the model and calculations were performed to determine which quantitative financial requirements (deductibles, coinsurance, copays, and so forth) and treatment limitations (calendar year limits, lifetime limits, other quantity limits, and so forth) meet the "substantially all" criteria required by the IFR. For those quantitative financial requirements and treatment limitations that met this test, the "predominant" level was identified. The results identified the benefit plan changes that are necessary in each benefit classification to be consistent with MHPAEA requirements. To confirm that the MH and SUD coverage was complete in all classifications, covered MH and SUDs were reviewed to determine if coverage is provided in all classifications where medical/surgical benefits are provided.

When a scope of service issue (such as the exclusion of residential treatment for substance use rehabilitation) was identified, it was discussed with the health plan or plan sponsor as being currently acceptable under the IFR, but potentially capable of becoming non-compliant if rules on required scope of services are enacted. In addition to the quantitative testing, detailed plan documents were reviewed to identify potential compliance problems with NQTLs. The IFR is less specific regarding where the line for non-compliance is drawn for NQTLs. Different interpretations exist among health plans and employers on what is allowable and compliant. Plan documents often contain details for some, but not all, NQTLs. Sometimes, information can be found on precertification requirements, step therapies, prescription drug formulary design, and conditioning benefits on the completion of a course of treatment. When this information is in the plan documents, we determined whether it appeared that the plan applied them in a "comparable" manner and in a manner "no more stringently" than those applied to medical/surgical benefits.

Aon Hewitt Compliance Testing Database. Aon Hewitt plan designs were reviewed to assess compliance with MHPAEA and the IFR standards. The plan design review and compliance testing was conducted in 2010, based on the plan designs each employer expected to implement in the 2011 plan year.

The Aon Hewitt testing database encompasses plan designs from more than 60 employers, ranging in size from 400 to more than 300,000 employees and representing 230 plan options. Each plan option represented a single combination of benefits (a combination of medical/surgical and MH/SUD benefits) that is available to an employer's participants. Plans whose adherence could not be assessed through a review of summary plan documents were subjected to detailed testing procedures. Of the 230 plan options reviewed, 140 required detailed testing to determine the benefit design that would apply to MH/SUD benefits. Plans that used identical coverage criteria for both MH/SUD and medical surgical services were considered to adhere to MHPAEA standards, and did not required detailed testing.

For most employer plans, the benefit type and level within the inpatient in-network and out-of-network, outpatient out-of-network, prescription drug, and emergency care classifications were consistent for both medical/surgical and MH/SUD and, as a result, demonstrated consistency with the parity regulations. For these benefit classifications, detailed testing was not required. Benefit design for the outpatient in-network classification, however, most frequently required detailed testing across employer programs. Within this classification, employer programs typically applied a variety of benefit types (copay or coinsurance) and benefit levels (primary care, specialty care, other). Detailed testing was required within this benefit classification to determine whether benefits met the "substantially all" and "predominant" requirements for MH/SUD services.

For each plan option requiring detailed testing, Aon Hewitt requested the employer's program administrator (vendor) to submit plan costs associated with each covered service category within the classification or sub-classification included in the testing process.

We first conducted the "substantially all" test for each plan option to determine which benefit type represents at least two-thirds of the plan costs in the benefit sub-classification. Plan cost data were grouped according to benefit type (e.g., copay, coinsurance, etc.) and were evaluated to determine the percentage of the total plan costs represented by each type. Once the benefit type representing "substantially all" was determined, we grouped the plan cost data associated with each benefit level (e.g., $15, $20, etc.) within that benefit type to determine the predominant benefit level in that sub-classification.

Aon Hewitt's Plan Design Database. Information obtained from Aon Hewitt's PDD included a review of 2009, 2010, and 2011 plan design data to determine how group health plan and employer-sponsored plan designs have evolved since federal parity was enacted in 2008. The information contained in the PDD allowed us to report on the plan designs that were in place before the implementation of federal parity in 2009 and evaluate how plan designs have changed since the implementation of the MHPAEA and the IFR. For most employers, the MHPAEA legislative requirements were implemented effective January 1, 2010. Further changes were made to employer plan designs effective January 1, 2011, to comply with the February 2010 IFR.

Information obtained from the database allows us to evaluate trends in how employer plan designs have changed since the implementation of the MHPAEA. The 2009 plan year serves as the baseline year, as the MHPAEA was not in effect until October 2009. Plan options in the 2010 plan year reflect plan designs that were in effect after the implementation of the MHPAEA. The plan options included in the 2011 plan year reflect plan designs that were in effect after the release of the IFR, which went into effect for most employers on January 1, 2011.

A total of 12,384 plan options, reflecting 252 employers, were included in the 2009, 2010, and 2011 plan design analysis. Of those options, 2,983 plan options (24.1%) were in the database in all three plan years. Not all plan options are reflected in the database all 3 years for a number of reasons, such as the option was terminated or the option was added in 2010 and 2011.

For many plan options, information on all fields included in this review was available. However, for some plan options certain information was unavailable, the information was unclear, or the information was potentially inaccurate. Therefore, the data for those plan options were excluded from our analyses. Therefore, although 12,384 plan options were included in the database, the actual number of plan options considered valid and used in the analysis for each comparison is much lower. We have reported the size of the sample included in each plan design analysis in Appendix C.

Summary Plan Description Data Provided by BLS. To supplement parity information from large employers that are heavily represented in the Aon Hewitt and Milliman databases, we analyzed a sample of 240 SPDs from midsized employers (establishments between 51 and 500 employees) collected by the BLS between 2008 and 2011 as part of the National Compensation Survey (NCS).65 Under ERISA, employers are required to provide their employees with SPDs of their health, pension, and welfare benefit plans. SPDs must include:

  • Any cost-sharing provisions, including premiums, deductibles, coinsurance, and copayment amounts.
  • Any annual or lifetime maximums or other limits on benefits.
  • The extent to which preventive services are covered.
  • Whether and under what circumstance existing and new drugs are covered.
  • Whether, and under what circumstance, coverage is provided for medical tests, devices or procedures.
  • Any provisions requiring preauthorization or utilization review as a condition of obtaining a benefit or service under the plan.


BLS requests that employers participating in the NCS submit full SPDs. However, many only provide summary tables of benefits, a more circumscribed description of benefits than the complete SPDs. BLS permitted NORC to abstract data from plan documents submitted by midsized employers between 2008 and 2011 to assess changes since the introduction of the MHPAEA and the IFR. The total sample size of abstracted documents was 240. One hundred sixty-seven covered the pre-parity era (plan years 2008-2009), and 73 covered the post-parity era (plan years 2010-2011). Not all documents included every data element of interest, but, when available, information related to the provision of quantitative limits (e.g., copays, coinsurance, and deductibles) was abstracted and analyzed. Observation level characteristics provided by BLS for each SPD was limited to principal industry. In order to increase the generalizability of the information obtained from the SPDs, analysis weights were constructed for each observation.66

To create the analysis weights, the sample was first divided into pre-parity observations (plan year 2008-2009; n = 167) and post-parity observations (plan years 2010-2011; n = 73) subsamples. Each subsample was treated as a separate sample with respect to weight construction. Within each subsample, the observations were assigned to one of seven industry categories based on the observation's North American Industry Classification System (NAICS) code.67

It should be noted that the utility of our analyses is limited by several factors. Many of the documents submitted to BLS were in fact not full SPDs, but brief tables of benefits that lacked many of the elements necessary to carefully track changes in financial requirements and treatment limitations. Our ability to construct weights to analyze the data that was abstracted was further limited by the lack of detailed establishment information available from the plan documents. Ideally, the weights would have been created using information including the number of workers at each establishment, detailed industry classification, and the physical location of the establishment. We were only provided information on basic industry categories. Therefore, we believe the weights as created, and applied in our analyses, are insufficient to remove all potential bias from the sample, and appropriate caution should be exercised when interpreting these results.

Employer Surveys. We reviewed the results of published national employer surveys from the KFF/HRET and Mercer. These surveys provided generalizable information on employers' coverage of MH/SUD. The 2010 KFF/HRET survey included 2,046 randomly selected public and private employers with more than three workers. The sample is randomly selected from a sample frame constructed by Survey Sampling Incorporated from Dun & Bradstreet's listing of public and private employers. KFF/HRET then stratifies the sample by industry and employer size. The 2010 Mercer Health Benefits Survey is also a random survey of employers identified from Dun & Bradstreet. The 2010 survey included 1,977 employers that offered health benefits. The survey uses sampling weights to calculate estimates both nationwide and for four geographic regions. The Mercer survey contains information for large employers (i.e., those with 500 or more employees), and for smaller employers (i.e., those with fewer than 500 employees).

Semi-structured Interviews with Health Plan Representatives. Lastly, we conducted detailed interviews with a non-generalizable sample of senior health plan officials who are responsible for seven major health insurers' compliance with the MHPAEA. The purpose of the interviews was to obtain specific information about plans' disclosure policies and practices required by the MHPAEA. Two behavioral health plan associations, the Association for Behavioral Health and Wellness and the National Behavioral Consortium recruited health plans to participate in the interviews.

Each of the seven individuals interviewed is a senior staff member responsible for leading the company's review of policies and procedures to bring the plan into compliance with MHPAEA and the IFR. The seven companies that participated collectively provide coverage for more than 100 million individuals and are among the largest health plans in the nation. Several of the plans exclusively provide behavioral health care services, and others provide behavioral health services within a larger health plan covering health, disability, and other benefits as well. Collectively, the companies operate in all 50 states, serving self-insured employers and employers purchasing fully-insured group health insurance products. Each interview elicited detailed information about:

  • The use of medical necessity criteria for medical and MH/SUD services.
  • The process for informing beneficiaries of reasons for claim denials for medicaland MH/SUD services and any changes in the processes for informing beneficiaries since implementation of the MHPAEA.
  • The use of utilization management techniques formedical and MH/SUD services and any changes in the use of utilization management techniques.
  • The management of out-of-network care.
  • The presence of any unmet demand for residential and intensive outpatient substance abuse services since the implementation of the MHPAEA.
  • The management of prescription medications, if the company is involved in this service.

  1. Methodology for sample selection and technical specifications at The sample for the NCS is selected on the basis of a three-stage design. The first stage involved the selection of areas consisting of 152 metropolitan and non-metropolitan areas. In the second stage, the sample of establishments is drawn from a sampling frame comprised of state unemployment insurance reports within sampled areas. The third stage is a probability sample of occupations within a sampled establishment. BLS field economists visit sampled establishments or contact them by telephone to collect data for the survey. To reduce the reporting burden, field economists ask respondents to provide Summary Plan Provision documents for health and retirement plans. Approximately 3,300 establishments provide data for each annual NCS.

  2. The analysis weights were calibrated to establishment and worker counts from the 2010 County Business Patterns created by the Department of Census. The end result is two sets of weights -- an establishment and a worker weight for pre-parity and an establishment and a worker weight for post-parity. The weight sums for the respective weights are identical between pre-parity and post-parity. Following describes the detailed process used to construct analysis weights.

    Step 1: Calculate the proportion of observations within each subsample and industry such that:

    rpi =   npi


    Where n = number of observations within subsample p and industry i.

    Step 2: Calculate the share of establishments within each industry such that:

    Ri =   Ni


    Where N = the number of establishments from the 2010 County Business Patterns within industry i.

    Step 3: Calculate the share of workers within each industry such that:

    RWi =   NWi


    Where NW = the number of workers from the 2010 County Business Patterns within industry i.

    Step 4: Calculate the final weights as the ratio of the share of establishments or workers within each industry and the proportion of observations within each subsample and industry such that:

    Establishment Weight =   Ri



    Worker Weight =   RWi


    Where rpi = the proportion of observations within each subsample and industry, Ri = the share of establishments within each industry, and RWi = the share of workers within each industry.

    The sum of both the establishment weight and the worker weight within each subsample equals the sample size within each subgroup. Each weight has a different impact on analyses. For example, the health care industry tends to have more workers as a share of the total workforce than establishments as a share of total establishments. Thus, the worker weight will grant health care observations more influence on an estimate than will the establishment weight. Comparisons of results obtained using both sets of weights demonstrated very minimal differences between the two sets of estimates. The estimates presented in this report were calculated using the establishment weights.

  3. The industry categories are as follows: (1) Agriculture, Mining, Utilities, Construction, and Manufacturing (NAICS = 11, 21, 22, 23, 31, 32, 33); (2) Wholesale (NAICS = 42); (3) Retail (NAICS = 44, 45); (4) Transportation and Information (NAICS = 48, 49, 51); (5) Finance, Real Estate, Professional Services, and Management (NAICS = 52, 53, 54, 55, 56); (6) Health care (NAICS = 62); and (7) Education, Recreation, Food Service and Other (NAICS = 61, 71, 72, 81).

View full report


"mhpaeAct_0.pdf" (pdf, 1.34Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®