Evaluation of Parity in the Federal Employees Health Benefits (FEHB) Program: Final Report. Structural Changes to Plan Benefits


PERT researchers acquired nominal plan benefits information on the 304 FEHB plans with available benefit design information and participating in the FEHB in 1999, the baseline year of the evaluation. This information was obtained from the OPM website for all four years of the evaluation (1999-2002).

To compile information on benefits in each of these plans, a data coding structure, variable definitions, and coding procedures were developed for all benefits elements that could be coded, including:

  • type of plan;
  • enrollment size;
  • geographic region;
  • premiums;
  • beneficiary cost-sharing;
  • deductibles; and
  • day, visit, and dollar limits for general medical, mental health, substance abuse, and pharmacy benefits.

Note that this data set differs from that used for the FEHB Plan PRR in terms of the number of plans and the plan years.

Key Research Questions

To gauge plans’ responses to parity in the FEHB Program, three research questions were posed:

  • Did FEHB plans comply with the parity policy?
  • Did plans exit the FEHB Program in response to the parity policy?
  • To what extent did plans enter into contracts with MBHO carve-out firms in response to the parity policy?

Data Collection

FEHB Plan Benefit Data

Data on the benefit design of health plans participating in the FEHB Program were abstracted from plan brochures publicly available on the OPM website.14 The resulting dataset included health plans’ beneficiary cost-sharing; deductibles; and day, visit, and dollar limits for general medical, pharmacy, mental health, and substance abuse services. It also included information on health plan type, geographic region, and enrollment size over the four-year study period from 1999 to 2002 (two years before and two years after implementing the parity policy). These data were linked to the data from the 2002 FEHB PRR on changes in contracting with carve-out companies. Some variation existed in the years studied in each analysis, as explained below.

Comparison Group Data

Comparison group data were used to account for general trends in benefit design and management. Using a pre-post analysis only, it would be difficult to attribute the increased likelihood of carving out to the parity policy. Rather, FEHB plan contracting might simply reflect industry-wide trends or decision making for all of an insurer’s plans as a general rule.

The comparison group, the Medstat Group MarketScan® Benefit Plan Design Database (i.e., Medstat), provided abstracted benefit information for health plans located around the U.S. Although the total number of health plans included in Medstat in any given year is higher, this analysis used the 35 plans with data available during both the 2000 and 2001 study years. These plans cover employees who work primarily for self-funded, Fortune 200 companies. Medstat benefit data included information on plan type, presence of a carve-out, enrollment, and geographic region.

Because high levels of missing data were reported on MH/SA cost-sharing (54% missing), outpatient visit limits (77% missing), and inpatient day limits (88% missing), comparison group analyses did not include these variables.15

Analytic Methods

Both qualitative and quantitative analytic methods were employed to examine changes in benefit structure and contracting with behavioral health carve-outs after implementing parity. The data from all FEHB plans were also used to assess the comparability of the experience of the eight selected plans with the universe of FEHB plans.

Analyses were divided into three sections:

  • First, researchers described whether FEHB Program health plans complied with the parity policy. Health plans were defined as compliant if MH/SA inpatient day and outpatient visit limits were removed, special annual and lifetime substance abuse dollar limits were removed, and beneficiary cost-sharing was made equivalent for MH/SA and general medical care for the in-network benefit.
  • A second analysis tested for an association between the parity policy and health plan exit from the FEHB Program.
  • Third, descriptive and multivariate analyses were conducted to examine changes in contracting with MBHO carve-out firms in response to the parity policy. Before-and-after analyses with and without a comparison group examined the association between the parity policy and carving out among health plans.

For the analysis with a comparison group, the econometric approach used difference-in-differences estimation to compare the probability of carving out before and after parity among FEHB plans compared with plans that were unaffected by the policy. A before-and-after-only analysis included additional benefit data and provided leverage for interpreting the difference-in-differences results.

Compliance Analysis

Descriptive statistics were compiled to analyze whether and how FEHB health plans complied with the parity policy. We examined data from the 152 health plans that participated continuously in the FEHB Program during two years before (1999, 2000) and two years after parity (2001, 2002).

Plans that exited or entered the FEHB Program in the baseline year were excluded. However, remaining plans included in this analysis covered 95% of beneficiaries from the baseline year. Of these, 14 were FFS plans and 138 were HMO plans.16 However, enrollment was heavily skewed with 72% of beneficiaries in the study population enrolled in FFS plans. For all descriptive results on compliance, the Association’s Standard Option was counted as a single plan since the benefit design was the same across all local Association plans.17 Change in the designation of in-network and out-of-network benefits were also examined descriptively.

Plan Exit Analysis

A second analysis tested for an association between the parity policy and health plan exit from the FEHB Program. Approximately 200 to 300 health plans contract with OPM annually to provide health insurance through the FEHB Program. However, only a relatively small proportion of plans stop and start contracting in any given year.

To qualify as a participating plan, a carrier needs to:

  • be licensed to sell group insurance within the areas of operation,
  • collect performance data,
  • survey beneficiary satisfaction,
  • provide Health Plan Employer Data and Information Set (HEDIS) data,
  • credential providers, and
  • comply with certain regulations (DHHS, 2000).

The 304 health plans that participated in the FEHB Program in the baseline year comprised the study population for this analysis. The model estimated the likelihood of plans exiting in either of the post-parity study years (2001 or 2002) in comparison with the year before parity implementation (2000), controlling for plan-level characteristics.

MH/SA and general medical care benefit design characteristics at baseline were included as covariates to assess how the level of pre-parity benefits might have influenced the exit decision. The unit of analysis was the plan-year (n = 912), i.e., 304 plans x 3 years = 912.

Since data were compiled on characteristics of health plans measured repeatedly over time, it was necessary to adjust for correlation between observations of the same health plan. The Generalized Estimating Equations (GEE) estimator of Liang and Zeger (1986) was used to account for the repeated measurements for each plan and the calculation of appropriate standard errors in the context of a non-Gaussian i.e., dichotomous, outcome variable (plan exit).

The GEE model related the probability of plan exit to year indicator variables and indicators of pre- to post-parity changes in plan limits using logistic modeling, i.e., relating the logit of the plan exit probability to a linear combination of covariates. The detailed model specification is shown in Appendix A: Detailed Model Specification for Plan Exit and Carve-out Analyses.

The model also included covariates describing plan characteristics at baseline, as well as interactions of year and benefit variables. Both limits and cost-sharing variables captured benefit information in the baseline year to determine whether pre-parity benefits affected the probability of plan exit.

Carve-out Analyses

A pre-post analysis with comparison group design was used to study the association between the parity policy and carving out management of MH/SA benefits among FEHB plans. Using Medstat plans as a comparison group, a difference-in-differences estimation allowed comparison of the difference in outcomes before and after parity for affected plans with the difference for unaffected plans.

Medstat plans were matched to the selected FEHB plans to meet the assumption of the difference-in-differences estimation model that the plans were comparable at baseline on observed characteristics that may have affected the likelihood of carving out. This analytic approach provided a way to minimize the possibility that events other than the parity policy explained the results.

Since survey data on carving out were available at only two points in time (before parity in 2000 and after parity in 2001), this analysis used data from only two years of the total four-year study period. Because Association plans participating in the FEHB Program make independent decisions about carving out, this analysis included 62 local Association plans (excluding 3 Association plans due to missing data)18 along with 151 non-Association plans. Thus, a total of 213 FEHB plans and 35 Medstat plans were included in this two-year analysis (n = 496 plan-years).

As in the exit analysis described previously, this carve-out model used a GEE logistic model to relate the logit of the probability of a plan carving out to covariates indicating the post-parity period relative to pre-parity and FEHB plans relative to Medstat plans, plus several plan characteristics (although plan benefit design characteristics were not included in the model due to missing benefit data in Medstat). The detailed model specifications are shown in Appendix A: Detailed Model Specification for Plan Exit and Carve-out Analyses. The GEE method was again used to account for plan-level repeated measurements and to calculate appropriate standard errors.

Next, a pre-post analysis without a comparison group was conducted to enhance understanding of the results obtained through the difference-in-differences model. Without MH/SA benefit information from Medstat comparison plans, the difference-in-differences model provides no information on how plan characteristics or the level of pre-parity MH/SA benefits might have affected the relationship between parity and the carve-out decision. A GEE logistic model was again used to relate the probability of carving out among FEHB plans to an indicator of pre- versus post-parity, as well as more comprehensive plan characteristics. (See Appendix A: Detailed Model Specification for Plan Exit and Carve-out Analyses, for detailed model specifications.)


Plans’ Compliance with the Parity Policy

Results show that health plans within the study population complied fully with the FEHB parity policy.19 FEHB plans removed all inpatient day limits and outpatient visit limits on MH/SA coverage.20 Special annual and lifetime substance abuse dollar limits were also removed. Table III-8 illustrates how plans adjusted beneficiary cost-sharing for MH/SA to comply with parity.

After parity, the median copayment required by plans in the FEHB Program fell from $20 to $10 per visit for outpatient mental health services. Similarly, the median coinsurance rates charged to beneficiaries for these services dropped from 50% to 15%. For inpatient care, median mental health copayments dropped from $40 in 1999 to $0 under the parity policy, and median inpatient mental health coinsurance rates dropped from 30% to 10%. These post-parity cost-sharing levels were found to be on par with the general medical care benefit.

Table III-8. FEHB plan behavioral health cost sharing*

  Pre-parity Post-parity
1999 2000 2001 2002
Median Copayment Outpatient mental health $20 $20 $10 $10
Inpatient mental health $40 $40 -- --
Outpatient substance abuse $10 $15 $10 $10
Inpatient substance abuse $40 $40 -- --
Median Coinsurance Outpatient mental health 50% 50% 15% 15%
Inpatient mental health 30% 30% 10% 10%
Outpatient substance abuse 50% 50% 15% 15%
Inpatient substance abuse 20% 20% 10% 10%
* This table reports median values for plans with cost-sharing greater than zero.

Cost-sharing for general medical care remained relatively stable over the study period (results not shown); thus, plans did not respond to the policy change by simply decreasing general medical benefits. This finding was not surprising in light of spending trends for MH/SA and medical services in the FEHB Program.

Between 1990 and 1997, mental health spending averaged only 2.9% of total paid claims (Kichak, 2001). However, the PERT found that some plans redefined the nature of their medical benefit. For example, in 2001, 23 plans (15%) began distinguishing between general medical cost-sharing for a medical primary care visit and for a medical specialist visit, setting higher cost-sharing for medical specialty care.

These plans required beneficiaries to pay for MH/SA care at a rate equivalent to the higher medical specialty rate rather than the lower medical primary care rate. Table III-9 provides information on how parity affected out-of-network benefits in the FEHB Program.

Table III-9. A national FFS plan

  Pre-parity Post-parity
2000 2001
MH/SA inpatient day limits 45 0 45
MH/SA outpatient visit limits 20 0 20
MH/SA outpatient cost sharing 30% $15 30%

OPM allowed plans to establish higher cost-sharing and special day/visit limits for out-of-network MH/SA services, implicitly recognizing the moral hazard problem facing plans (OPM, 2000). Like all national FFS plans available to FEHB Program beneficiaries, the health plan profiled in this table provided beneficiaries with both in-network and out-of-network options.21

Before parity, national FFS health plans did not distinguish between in-network and out-of-network behavioral health benefits, although most distinguished between in-network and out-of-network general medical benefits. After parity, these FFS plans began differentiating between in-network and out-of-network MH/SA health benefits.

The new out-of-network benefit design typically matched the more limited behavioral health benefit in place prior to the introduction of parity. This pattern is illustrated in Table III-9 where in-network MH/SA benefits on par with general medical benefits were established in 2001, whereas the out-of-network option exactly matched the 2000 behavioral health benefit.

Like the 2000 benefit, MH/SA services were covered with a 30% outpatient cost-sharing requirement and a 20-visit annual limit out-of-network in 2001, while the in-network outpatient medical benefit required only a $15 copayment post-parity. Thus, enrollees could choose either the in-network or out-of-network benefit. If they chose the in-network benefit, then MH/SA parity applied.

While general medical service users enjoy the same option of choosing in-network or out-of-network providers, the benefit differential is even greater for MH/SA services. This is because parity renders MH/SA and general medical care obtained in-network the same with respect to cost-sharing and limits. However, MH/SA benefits for care obtained out-of-network are generally less generous than out-of-network general medical benefits.

Plans’ Exit from the FEHB Program in Response to the Parity Policy

Results from the plan exit analysis suggest that plans did not exit the FEHB Program for reasons related to the parity policy. Of the health plans participating at baseline, descriptive data showed that 15% exited in 2000. After parity, another 14% exited in 2001 and 19% exited in 2002. As shown in regression results in Table III-10, the coefficients on the year dummies were not significant, indicating that health plans were no more likely to exit in either 2001 or 2002 in comparison with 2000, the pre-parity year.

Table III-10. Probability of plan exit*

Health Plan Characteristics Coefficient SE Z-score P-Value OR
Intercept -3.9642 1.1068 -3.58 0.0003 0.02
Year 01 -0.3183 0.7307 -0.44 0.6631 0.73
Year 02 -0.0913 0.6838 -0.13 0.8937 0.91
Visits – less restrictive (31-60 annually) -0.8494 0.7629 -1.11 0.2656 0.43
Visits – more restrictive (20-30 annually) 0.0175 0.6494 0.03 0.9785 1.02
Days – less restrictive (31-60 annually) -0.1241 0.8028 -0.15 0.8771 0.88
Days – more restrictive (20-30 annually) 0.3187 0.6971 0.46 0.6475 1.38
Outpatient mental health cost-sharing 0.1131 0.1592 0.71 0.4775 1.12
Outpatient general medical cost-sharing -0.115 0.1654 -0.69 0.4871 0.89
Plan type 1.9654 1.1114 1.77 0.077 7.14
Plan enrollment size/1000 -0.0319 0.0224 -1.43 0.1539 0.97
Region 1 (Northeast) 0.1356 0.301 0.45 0.6524 1.15
Region 2 (Midwest) -0.0778 0.2976 -0.26 0.7938 0.93
Region 3 (South) 0.888 0.2438 3.64 0.0003 2.43
Region 0 (Nationally-available) 0.6933 0.4984 1.39 0.1642 2.00
Year 01*Visits – less restrictive (31-60) 1.571 0.9773 1.61 0.1079 4.81
Year 01*Visits – more restrictive (20-30) 0.8502 0.8495 1.00 0.3169 2.34
Year 02*Visits – less restrictive (31-60) 2.3231 1.5262 1.52 0.128 10.21
Year 02*Visits – more restrictive (20-30) 0.2979 1.4319 0.21 0.8352 1.35
Year 01*Days – less restrictive (31-60) -0.7704 1.0794 -0.71 0.4754 0.46
Year 01*Days – more restrictive (20-30) -0.6239 0.8864 -0.7 0.4815 0.54
Yea r 02*Days – less restrictive (31-60) 0.2759 1.5742 0.18 0.8609 1.32
Year 02*Days – more restrictive (20-30) -0.9409 1.4384 -0.65 0.513 0.39
* Overall significance of model: Chi-sq = 64.22 (21 DF), p-value = <.0001

Likewise, none of the interactions of time dummies with the MH/SA pre-parity benefit variables significantly affected the plan exit decision. Outpatient medical cost-sharing also did not affect the likelihood of plan exit, although this result might have been expected because of limited variability in outpatient medical copayments and coinsurance across plans.

Indeed, the only factors that appear to have been significantly associated with the plan exit decision were region and plan type. Plans in the South were significantly more likely to exit over the study period compared with national plans or plans in the Northeast, Midwest, or West. The higher rate of exit among FEHB health plans located in the South may be due to regional market characteristics. On its web page providing information to plans interested in joining the FEHB Program, OPM “especially invites applicants” from 14 States determined to be medically underserved; almost half of these designated States are located in the South.

That regional HMOs were significantly (at the 0.1 level) more likely to exit the market compared to national FFS plans is consistent with the tendency of many of the national plans to cater to specific Federal employment groups (e.g., Rural Letter Carriers Plan).

Plan Carve-out in Response to the Parity Policy

Tables III-11 through III-13 present findings on carving out MH/SA care after implementation of the parity policy. Table III-11 presents descriptive information comparing FEHB and Medstat comparison group plans in 2000. While the number of Medstat plans was quite small, a similar proportion of Medstat and FEHB plans carved out in 2000. Likewise, the geographic distribution of plans appeared similar. While a larger proportion of FEHB than Medstat plans were HMO/POS, plan distribution by enrollment was weighted toward FFS/PPO plans in both groups. Medstat plans had a much larger average enrollment compared to FEHB plans. The missing MH/SA benefit data in Medstat precluded benefit design comparison. No additional matching was deemed necessary since Medstat and FEHB health plan characteristics were reasonably similar.

Table III-11. Descriptive data on FEHB and Medstat comparison group plans

Health plan characteristics, 2000 FEHB Medstat
# Plans 213 35
Mean enrollment 18,157 72,964
Plan type
FFS/PPO 29% 57%
HMO/POS 71% 42%
Plan type weighted by enrollment
FFS/PPO 74% 52%
HMO/POS 26% 48%
Northwest 21% 26%
Midwest 29% 28%
South 19% 17%
West 23% 3%
National 8% 26%
Benefit Limits
% of plans with low annual visit limits (1-31 visits) 68% --
% of plans with high annual visit limits (31+ visits) 18% --
% of plans with no annual visit limits 14% --
Carving out
% of plans carving out MH/SA services 47% 49%

Descriptive data in Figures III-1a and III-1b show that more FEHB plans carved out after parity in comparison to the set of health plans not in the FEHB Program and not offering parity MH/SA benefits. As Figure III-1a indicates, across all the FEHB plans, 69% carved out after parity implementation in 2001 compared to 49% at baseline. Likewise, Figure III-1b shows that only 37% of Association FEHB plans opted to carve out in the year before parity, whereas 68% of these plans carved out after implementation. These descriptive results indicate that the one-year increase in the proportion of FEHB plans carving out after parity was substantial, especially for Association plans.

The PRR data (reported in the prior section of this report) showed that a majority of health plans (25 out of 38 plans) that carved out for the first time in 2001 attributed this administrative change directly to the parity policy. In comparison, descriptive data indicate that the Medstat comparison plans did not carve out in greater numbers in 2001. In fact, two plans ended their contracts with MH/SA carve-outs during this two-year period (Figure III-1c).

Figure III-1a-c. Proportion of plans carving out before and after parity

Figure III-1a

Figure III-1a-c. Proportion of plans carving out before and after parity

Figure III-1b

Figure III-1c

Tables III-12a and III-12b present multivariate results on the probability of carving out before and after parity using Medstat plans as a comparison group. Using a difference-in-differences estimation approach, a 29% net increase occurred in the probability of carving out (from pre- to post-parity) among FEHB plans relative to Medstat comparison plans. This result provides evidence of a positive relationship between the parity policy and a health plan’s decision to carve out. Unlike the descriptive results above, this result provides some assurance that the likelihood of carving out was not solely a function of industry-wide changes or insurer-wide changes.

Table III-12a. Carving out behavioral health benefits with comparison group

Health plan characteristics Coefficient SE Z-score P-value OR
Intercept -0.370 0.495 -0.750 0.455 0.69
Post -0.350 0.201 -1.750 0.081 0.70
Treatment (FEHB plans) -0.065 0.382 -0.170 0.864 0.94
Plan type (either FFS/PPO or HMO/POS) 0.337 0.251 1.350 0.178 1.40
Plan enrollment size/1000 0.000 0.002 -0.060 0.950 1.00
Region 1 -0.049 0.396 -0.120 0.902 0.95
Region 2 -0.130 0.363 -0.360 0.720 0.88
Region 3 0.549 0.413 1.330 0.183 1.73
Region 4 0.554 0.459 1.210 0.227 1.74
Post*Treatment 1.240 0.235 5.280 0.000  

Table III-12b. Interaction effect from prior model

Variable Obs Mean SD Min Max
Y(Parity = 1; Post = 1; Parity*Post =1) = Y1,1,1 490 0.683 0.033 0.625 0.799
Y(Parity = 1; Post = 0; Parity*Post =0) = Y1,0,0 490 0.474 0.038 0.409 0.623
Y(Parity = 0; Post = 1; Parity*Post =0) = Y0,1,0 490 0.398 0.036 0.336 0.547
Y(Parity = 0; Post = 0; Parity*Post =0) = Y0,0,0 490 0.483 0.038 0.417 0.631
Dif-in-Dif = (Y1,1,1 – Y1,0,0) – (Y0,1,0 – Y0,0,0)
interaction effect 490 0.294 0.011 0.266 0.301
standard error (of interaction) 490 0.058 0.003 0.054 0.078
z statistic (of interaction) 490 5.123 0.403 3.443 5.506

Tables III-13a and III-13b present multivariate results on FEHB plan characteristics associated with carving out before and after parity.22 Table III-13a shows that before parity, Association plans had a 23% lower predicted probability of carving out compared to other health plans after adjusting for plan characteristics. However, in the post-parity period, Association plans had a 30% greater predicted probability of carving out compared with the pre-period. The predicted probability of Association plans’ carving out after parity was still 9% lower in comparison to other health plans in the FEHB Program.

Table III-13a. Predicted probabilities from pre- post carve-out model*

Variable Before After Predicted Probability
Association Plans 0.3142 0.6163 0.3021
Other Plans 0.5428 0.7039 0.1611
  0.2287 0.0877  
* To review full model, see Appendix A: Detailed Model Specification for Plan Exit and Carve-out Analyses.

Table III-13b. Predicted probabilities from pre-period*

Variable Without limits With limits Predicted Probability
Less restrictive visits (31-60) 0.2742 0.8900 0.6158
More restrictive visits (20-30) 0.2742 0.6433 0.3691
* To review full model, see Appendix A: Detailed Model Specification for Plan Exit and Carve-out Analyses.

Table III-14 shows the pre-post carve-out model used for these analyses. The ability to study this change among Association plans was useful given their reliance on demand-side cost controls in comparison to HMOs in the pre-parity period.

Table III-14. Pre-post carve-out model

Health plan characteristics Coefficient SE Z-score P-value OR
Post (after parity) 1.1212 0.3782 2.96 0.003 3.07
Visits – less restrictive (31-60 annually) 2.5171 0.8453 2.98 0.0029 12.39
Visits – more restrictive (20-30 annually) 1.7163 0.695 2.47 0.0135 5.56
Days – less restrictive (31-60 annually) -0.5332 0.7385 -0.72 0.4703 0.59
Days – more restrictive (20-30 annually) -1.3165 0.6931 -1.9 0.0575 0.27
Outpatient mental health cost sharing 0.0398 0.3577 0.11 0.9114 1.04
Association plan -1.0408 0.4797 -2.17 0.03 0.35
Plan enrollment size/1000 -0.0029 0.0027 -1.08 0.2809 1.00
Post*Association Plan 0.6211 0.3148 1.97 0.0485  
Post*Less restrictive visits -0.2334 0.5311 -0.44 0.6603  
Post*More restrictive visits -0.4497 0.4176 -1.08 0.2816  
* These regression results correspond to the predicted probabilities displayed in Tables III-13a and 13b.

Summary of Findings from Nominal Plan Benefits and Comparison Group Data

Here we summarize the findings related to the three key research questions posed at the start of this analysis, based on the nominal plan benefits data from 304 plans and matched comparison group data.

  • Compliance with the parity policy: All FEHB plans complied with the requirement to implement MH/SA parity.
  • Plan exit from the FEHB Program: No plan left the FEHB Program in response to the parity policy.
  • Plan carve-outs: Comparing the pre-parity 1999 and 2000 period with the post-parity 2001 and 2002 period, FEHB plans were more likely to enter into managed care arrangements through a contract with a carve-out vendor than were a matched set of comparison plans that did not face a parity policy for MH/SA benefits.

View full report


"parity.pdf" (pdf, 5.07Mb)

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®

View full report


"parityA.pdf" (pdf, 61.31Kb)

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®

View full report


"parityB.pdf" (pdf, 194.85Kb)

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®

View full report


"parityC.pdf" (pdf, 134.89Kb)

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®

View full report


"parityD.pdf" (pdf, 147.52Kb)

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®