The new NAMHC estimate of a 1.4 percent cost increase in total health insurance premiums due to parity is lower than previously reported estimates based on Hay Group actuarial models (all were of similar magnitude). Previous estimates include the 1996 Congressional Budget Office (CBO)/Congressional Research Service (CRS) 4.0 percent increase, the 1998 Substance Abuse and Mental Health Services Administration (SAMHSA) estimate of 3.6 percent, and estimates from the previous NAMHC report, as shown in the Table below. Because each of the Hay Group model-based estimates uses the same basic methodology it is important to understand what drives the latest estimate. Table. Cost of Full Parity Estimates Based on Hay Group Actuarial Models
|
Increase in Total Premium
|
|||
| Estimate | Model Version* | FFS Plans | All Plans |
| CBO/CRS (1996) | HCBVC 6.5 | 5.3% | 4.0% |
| NAMHC (1998) | MHBVC1.7 | 4% to 5% | <1 to 4% |
| SAMHSA (1998) | MHBVC1.7 | 5.0% | 3.6% |
| NAMHC (2000) | MHBVC2.0 | 1.4% | 1.4% |
* Principal model for mental health/substance abuse cost estimates (separate model used to calculate non-mental health costs for HMOS in all but CBO/CRS study).
The different Hay Group models (HCBVC 6.5, MHBVC1.7, MHBVC2.0)
work essentially the same way. The Hay Group uses actuarial methods
to develop an expenditure grid, which represents the extent of
medical, mental health, and substance abuse service use by plan
enrollees if they paid nothing out of pocket (i.e., if the plan
paid 100 percent). Combinations of induction factors (which measure
how responsive enrollees are in reducing their health care use when
faced with higher out-of-pocket costs) and management factors
(which measure how costs are reduced with utilization review and
other managed care practices) are then used to estimate how much
care enrollees would use when faced with a particular set of
copayments. Administrative loading factors are used to multiply
these costs to determine the estimated premium for a particular
health care benefit package.
The cost of a change in the health care benefit package, such as mental health parity, is simulated by first estimating the cost with existing benefits in the Hay model and then estimating the cost with the new benefits. All other assumptions are held constant (except, of course, for benefits). Because both baseline costs and benefits differ in fee-for-service (FFS), preferred-provider organization (PPO), point-of-service (POS), health maintenance organization (HMO), and managed behavioral carve-out plans, the Hay models were used to compute separate estimates of the cost of parity for these different types of health plans. The 1998 SAMHSA and the 2000 NAMHC estimates then combine these separate estimates (using assumptions about the distribution of enrollees across types of plans) to produce a single estimate of the total cost impact of mental health parity. The previous NAMHC (1998) report did not present a single cost estimate, but rather a range of estimates based on the different plan types. The CBO/CRS estimated a typical FFS plan and then made an ad hoc adjustment downwards to reflect managed care.
The major difference between the current NAMHC estimate of 1.4 percent and previous estimates appears to be driven by the evolution of the Hay Group model as it is updated to reflect changes in the underlying distributions of mental health and health care expenditures. The most recent version of the Hay Group model (MHBVC2.0) assumes an average mental health and substance abuse cost per enrollee (with no out-of-pocket costs) that is fully one-half that of the Hay Group model used for the 1998 SAMHSA and the previous NAMHC report estimates, and one-third that of the model used for the 1996 CBO/CRS estimate (NAMHC calculations from Hay Group models). The most recent Hay Model incorporates new data from the FFS/PPO plans of the Federal Employees Health Benefit Program (FEHBP), several large managed behavioral health care companies, and the FFS plan of a large State employees health plan (described in Section III of this report). The experience of the FEHBP program, along with other evidence, suggests that mental health and substance abuse costs have declined substantially during the 1990s, mainly due to sharply reduced inpatient utilization even in FFS/PPO plans. Because mental health costs have declined, especially as a proportion of total health care costs, the estimated impact of a mental health parity benefit in terms of the increase in total premiums has correspondingly declined.
There are other much smaller differences in the assumptions used between current NAMHC estimate and previous estimates. The assumptions are discussed in more detail below, but in general these explain little of the differences between the estimates. (A useful companion to this discussion of the estimates is the report by Sing and Hill (1998) detailing the assumptions used to generate the 1998 SAMHSA estimates.) This is illustrated by the fact that when the NAMHC (2000) assumptions are applied to the previous version of the Hay Model (MHBVC1.7) used to produce the 1998 SAMHSA estimate, the estimated impact of parity is 3.7 percent instead of 1.4 percent (preliminary NAMHC calculations from the MHBVC1.7 model). That is, when the same assumptions are applied to mental health treatment patterns prevailing several years ago, the estimated cost of parity is 3.7 percent, but taking into account recent trends in mental health costs reduces the estimated cost of parity to a current level of 1.4 percent.
Other Differences in Basic Assumptions
Family vs. single premiums: The current NAMHC estimate is for singles only, while the other estimates represent the percent increase in total family premiums. Preliminary calculations using the new Hay model suggest that increase in total family premiums would be approximately 1.6 percent, a slight increase from the 1.4 percent for the single premiums.
Distribution of Health Plans: The 1998 SAMHSA and the current NAMHC estimates assume different distributions of types of health plans, but this difference has no effect on either estimate. That is, applying the 1998 SAMHSA distributions to the current NAMHC estimate produces the same estimate, and vice versa. (The 1998 SAMHSA estimate assumes a distribution in which 20 percent of plans are FFS, 30 percent PPO, 20 percent POS, and 30 percent HMO, compared to a distribution of 14 percent, 40 percent, 20 percent, and 26 percent respectively in the 2000 NAMHC estimate.) The CBO/CRS estimate contains an unknown adjustment for managed care plans from their base FFS estimate.
Carve-outs: The current 1.4 percent NAMHC estimate includes an adjustment that assumes that 20 percent of FFS, PPO and POS plans have carved out their mental health and substance abuse benefits. This 20 percent factor is the same for both the current and parity benefit cost estimates. Several surveys estimate that in at least 20 percent (KPMG Peat Marwick, Hay Group) of these types of plans among medium and large employers, the employer separately contracts (carves-out) with a managed behavioral health firm. This 20 percent is a conservative estimate of carve-out management since FFS, PPO, and POS health plans may themselves subcontract (carve-out) behavioral benefits, but there are no good estimates of the extent of these indirect carve-outs. The 20 percent carve-out adjustment has a negligible impact on the current NAMHC estimate-without it the estimate is 1.5 percent-because the difference between cost increases in carve-out plans and non-HMO plans have decreased considerably. However, applying the 20 percent carve-out adjustment to the 1998 SAMHSA estimate would considerably reduce it from 3.6 percent to around 3.0 percent.
Benefits: All of the estimates use the same basic assumptions about existing mental health and substance abuse with one exception. The CBO/CRS estimates assumed that a $50,000 lifetime maximum applied to mental health benefits, while all the other estimates occurred after the federal parity legislation that removed this limit. There were slight differences across the different estimates in the assumed medical/surgical benefits, but these differences likely had little impact on any of the estimates.
Management Factors: The Hay Group models all incorporate a management factor to account for the effect of use of utilization review and other managed care techniques in reducing costs. The factors were, for the most part, the same across all of the estimates and the same factor was used to estimate both the cost under current benefits and under a parity benefit. The 1998 SAMHSA estimate includes a management factor for FFS, PPO, and POS plans that implies a reduction of 25 percent in mental health and substance abuse costs compared to unmanaged costs, but no reduction for HMO costs because the tighter management in HMOs are already reflected in the HMO expense grids. The current NAMHC estimate uses the same 25 percent management factor adjustment for FFS, PPO, POS, HMO and carve-out plans to estimate costs under current and under parity benefits. Preliminary NAMHC calculations suggest that if the 25 percent adjustment for HMOs and PPO is eliminated, the current NAMHC estimate of 1.4 percent overall increase in premium costs due to parity would change slightly to 1.5 percent.
Induction (Demand Response): All the estimates based on the Hay Group model use the same induction factors, which are derived from the RAND Health Insurance Experiment, with one small exception. The 1998 SAMHSA estimates used slightly lower induction factors for HMO plans. However, this makes little difference in the overall estimates.
Administrative Factors: The Hay Group model multiplies the base claims costs in their models by an administrative loading factor, which represent a fixed percentage of the total claims costs for administrative costs and profit (where applicable), to obtain the total premium cost. The current NAMHC estimate uses the same administrative factor of 1.11 for all types of health care plans (carve-outs, HMO, FFS, PPO, POS). The 1998 SAMHSA estimate used the same administrative factor of 1.11 for FFS, PPO, and POS plans but higher loading factors of 1.15 for medical/surgical expenses and 1.2 for behavioral coverage. However, because these administrative load factors multiply both the numerators and denominators, the effect of differing administrative load factors tends to be canceled out when examining changes in premium costs.
References:
Hay Group (1999): Employer Mental Health Coverage Features and
Compliance with the Mental Health Parity Act. Hay Report for NIMH,
June 1999. (unpublished).
KPMG Peat Marwick, Health Benefits in 1997, June 1997.
Sing M. and Hill S. (1998): Predicted Premium Increases Due to Full and Partial Parity for Mental Health and Substance Abuse Insurance Benefits. Report Submitted to the Substance Abuse and Mental Health Services Administration. Washington, DC: Mathematica Policy Research, Inc.
For information about NIMH and its programs, please email, write or phone:.
NIMH Public Inquiries
6001 Executive Boulevard, Rm. 8184, MSC 9663
Bethesda, MD 20892-9663 U.S.A.
Voice (301) 443-4513; Fax (301) 443-4279
This page was last updated: November 1, 2000.
