Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Ment Health Policy Econ. Author manuscript; available in PMC 2011 April 25.
Published in final edited form as:
PMCID: PMC3081606

The Effect of Employee Assistance Plan Benefits on the Use of Outpatient Behavioral Health Care

Dominic Hodgkin, Ph.D.,1 Elizabeth L. Merrick, Ph.D., MSW,1 Deirdre Hiatt, Ph.D.,2 Constance M. Horgan, Sc.D.,1 and Thomas G. McGuire, Ph.D.3


Many US workers have access both to an employee assistance plan (EAP) and to employer health insurance that includes behavioral health services. We examine whether EAP services substitute for outpatient behavioral health care services covered by the health plan, in settings where the employer has purchased both products jointly. We analyze administrative data for 26,464 patients enrolled with a managed behavioral health organization in 2005. General linear models are used to predict visits and spending for regular outpatient care. We also use instrumental variable analysis to test for endogeneity of the number of EAP visits covered. Having more generous EAP coverage predicts fewer regular outpatient visits, and lower spending for outpatient care. This supports the idea that the two types of care are to some extent perceived as substitutes, although patients also appear to perceive that EAP services offer something distinct from regular outpatient care.


In recent decades, many large employers in the US and elsewhere have begun offering employee assistance plans (EAPs) to their workforces. EAPs are workplace-based programs that aim to address behavioral health and other problems that affect employees’ well-being or job performance. EAPs often provide 3–8 visits to employees and family members, with no copayment required, for assessment and short-term counseling. Case-finding, early intervention, high acceptability, and easy access are among the typical goals of an EAP. In 2007, 42% of all US workers had access to an EAP, with access being particularly common at firms with at least 100 workers.1

At the same time, most large US employers also purchase health benefits for their employees, and these benefits packages typically include behavioral health services. In 2005, 90% of workers were covered for outpatient mental health care, 84% for outpatient alcohol rehabilitation, and 83% for outpatient drug rehabilitation.2 In some cases these behavioral health services are managed by the health plan itself, while in others this is done by a managed behavioral health organization (MBHO) that is hired either by the health plan or directly by the employer.3 In 2002, 66% of privately insured Americans were enrolled in some type of managed behavioral health program.4 There is some potential overlap in services covered by the EAP and the health plan, as is discussed further below.

Initially, most EAPs were operated by employers themselves as an on-site service, but over time most employers have contracted with external EAP vendors that use networks of contracted off-site providers.5 This shift generally reduced cost and expanded access, while potentially addressing some employees’ concerns regarding the confidentiality of care at an on-site facility. More recently, some MBHOs have begun offering an ‘integrated product’ which folds EAP services into the standard behavioral health benefit. In this model, the first few visits are provided free under the EAP benefit, not only for employees explicitly seeking EAP services but also for those seeking outpatient behavioral health care more generally. In some integrated EAP/MBHO plans, individuals who receive assessment and/or short-term counseling through the EAP may continue specialty treatment with the same provider. Referral to non-EAP specialty treatment when necessary may be facilitated, because a single vendor administers the EAP and managed behavioral health care (MBHC) benefits.


In this paper, we ask how access to EAP sessions might affect the use of regular outpatient behavioral health services. We first consider two alternative hypotheses, which differ according to enrollees’ attitude to the free initial EAP services:

  • Substitution hypothesis: Suppose that enrollees view the EAP visits as a good substitute for regular outpatient visits. In this case, access to free EAP visits is effectively a price reduction for a substitute service, and patients with more generous EAP coverage would be expected to use fewer regular outpatient visits.
  • Independence hypothesis: Suppose instead that patients view the initial EAP visits differently than the regular outpatient visits, such that their demand for regular outpatient care is unaffected. This could be because the EAP-covered visits are genuinely different in nature or perceived to be different, or because they are free to the patient. As a result, there is no reduction in regular outpatient visits, and total treatment length (including the EAP visits) is longer.

The substitution hypothesis reflects a perception among some in the industry that EAP services are becoming less distinguishable from routine outpatient behavioral health treatment. Surveying one vendor’s network, Sharar found that 63% of the contracted individual EAP providers reported little or no familiarity with the ‘EAP core technology’, and 74% said they treated EAP clients either ‘completely the same’ or ‘more or less the same’ as other clients.6 On the other hand, the providers in that survey also noted that for EAP clients, they were more likely to assess the impact of the client’s problem on job performance. Nevertheless, lack of clear distinctions has caused the National Business Group on Health to be concerned that EAP services may be “duplicative with services provided by the employer’s mental health benefit plan”.7 Some EAP professionals have expressed concern that this would lead employers to eliminate or reduce EAP benefits, particularly if they see behavioral health spending increasing due to the new federal parity law.8

The independence hypothesis would be more plausible if the EAP pathway attracts different users into treatment, who might not have entered if regular outpatient treatment were the only entry point. This may be because they perceive the EAP as involving less stigma, more convenient provider locations, or other reasons. The EAP is often promoted to enrollees as a practical way of dealing with work-life issues, which may lead enrollees to view EAP visits differently than specialty behavioral health care. Finally, even if the user could not distinguish EAP from regular visits, the absence of an out-of-pocket price could stimulate demand. If vulnerability to dropout is highest early in treatment, then making those visits free could be a way to reduce that vulnerability.

It is worth noting some implications of these hypotheses for health plans and for employers. In the case of employers, many would prefer to see at least some substitution occur, with the EAP serving as a low-cost way to reduce use of other, costlier therapies, thereby reducing employers’ total spending on health care. However, a free-standing EAP may under-invest in prevention, because the savings it generates accrue elsewhere in the system, in health plan treatment costs and employer costs of lost productivity. This is a classic case of a positive externality resulting in under-production, in this case of preventive services. Integrating the EAP with the health plan could be a way to address one part of this externality, as the health plan may become more motivated to undertake additional preventive activities, now that they would reduce its own costs. Against this interpretation, one should note that the higher the turnover in health plan enrollment, the less incentive plans have to invest in prevention.9

An additional consideration is that some employers may be concerned that managed health plans and MBHOs are over-managing initial access, and discouraging employees from getting into treatment. This type of employer might see an EAP as a good way to check health plans’ potential over-management of initial access, while leaving them fully in charge of treatment duration. This might persuade such employers to keep the EAP contract separate rather than integrating it into the MBHO contract.

From the MBHOs’ perspective, they may be less concerned about whether or not EAP visits substitute for regular outpatient care, if they are paid for administrative services only, or otherwise not fully at risk for treatment costs (as is typical in contracts with large employers).10 The MBHOs’ contracted providers may be even less concerned, if they are not at risk for the number of visits.

This paper seeks to examine whether the number of EAP visits covered by an integrated EAP affects the use of regular outpatient behavioral health care. We test for effects on spending as well as on the number of visits. The reason is that even if a particular approach had no effect on the number of regular outpatient visits, it could still alter total spending by changing the mix of provider types being used. For example, a shift to lower-cost provider types would lead to a larger proportional reduction in outpatient spending than in outpatient visits.

Some prior studies have tested whether having an EAP is associated with lower use of health care in general, with most suggesting that reductions did occur, according to reviews.1112 However, most of these studies did not examine the relationship to use of behavioral health care specifically. One exception was Merrick et al’s study, which found that enrollees in an integrated (MBHC+EAP) plan were more likely to use behavioral health services in general, and specifically substance abuse services, compared to enrollees with MBHC coverage only.13 In addition, Cuffel and Regier found that treated prevalence of behavioral health problems was lower in plans that had an EAP integrated with the MBHO, compared to those with a stand-alone EAP.14 Given this design, they were testing the effect of regular outpatient coverage on EAP use, a related but different question to the one addressed by this paper.

The present study seeks to fill these gaps in the research literature, by examining how use of outpatient behavioral health care is affected by the number of EAP visits covered, and by the number used.


Study setting

We examine use of regular outpatient services at Managed Health Network (MHN), a large national MBHO which contracts with employers and other payers to manage and deliver specialty behavioral health and EAP services, through both integrated and stand-alone products. MHN has 11 million members. This paper examines MHN’s integrated product only, as this allows us to observe both EAP and regular outpatient care for the study sample.


We used 2005 administrative data, including de-identified claims, eligibility and benefits files. Claims included specialty mental health and substance abuse services covered by MHN. For EAP claims, only clinical services were included, not assistance such as legal or financial consultation. The study received Institutional Review Board approval at Brandeis University.

Treatment Entry Process

In the plan’s integrated product, accessing services involves calling a phone center for authorization. Authorization is a routine process in which eligibility is verified, brief intake is performed, and enrollees are approved to see a network provider. In the integrated product, enrollees call a single toll-free number to access care for either EAP or managed behavioral health services. Following a brief intake, enrollees assessed as needing regular outpatient care are typically offered the opportunity to use the EAP portion of the benefit first. (Approval of at least one visit is automatic). Typically 3–5 EAP visits are covered at no cost to the enrollee, although some plans compute this on a per-year rather than a per-incident basis. (An incident is a treatment episode). Enrollees electing to use the EAP benefit are preferentially referred to providers with an EAP qualification or specific training/expertise in EAP work. When an enrollee reaches the EAP visit limit and needs more services, the MBHC portion of the benefit is accessed, after re-authorization. Enrollees continuing outpatient treatment can usually choose to remain with the same network provider seen through the EAP portion of the benefit. Some enrollees, such as those initially needing a higher level of care or requesting a medication evaluation, would automatically bypass the EAP and access services under the MBHC part of the benefit.

Previous research with this organization found that among integrated-product enrollees with any outpatient mental health visits, 43% had at least one EAP visit. For those with any outpatient substance abuse visits, 36% had an EAP visit.15


Our sample reflected enrollment from employers in the services, sales and government sectors. From this sample, we selected enrollees who used either regular outpatient behavioral health services or clinical EAP services during 2005. Finally, we excluded 6,007 members who were enrolled in a model with 8 EAP sessions as they were all with a single employer that had unusual benefits provisions.

The various criteria applied resulted in 27,397 members for the analytic sample. For 933 of these members, data on their copayment level were not available. The results we present exclude those members, resulting in a sample of 26,464 members. However, we ran the same models without the copayment variable on the larger sample of 27,397, and found qualitatively similar results for the key variables discussed.


Behavioral health utilization

Outpatient behavioral health care visits were identified using service category and procedure codes, and the number of visits in 2005 was computed for each member. These services were provided by behavioral health specialists, and therefore do not include treatment by primary care or other medical providers. Spending for these visits was computed from the allowed charge variable in the claims, which includes payments by both plan and member.

Independent variables

The key explanatory variable is the EAP model the member is enrolled with, in terms of the number of sessions allowed (3 or 4–5) and the nature of the limit (per incident or per year). A 3-visit limit is substantially more generous if that limit applies per incident rather than per year, as the enrollee could then have more than one incident per year (e.g. 2 incidents would allow 6 visits). The various combinations serve as regressors, with 3 visits per year, the least generous model, serving as reference category.

Other covariates included gender, age category (reference group = 36–55 years old), relationship to subscriber (employee versus spouse or dependent), and the enrollee’s Census region of residence (reference group = West). In addition, ICD-9 diagnoses were grouped into behavioral health categories using AHRQ’s Clinical Classification Software,16 and indicator variables were created for each category. We included the proportion of the year the member was enrolled as a covariate, since members enrolled for less than 12 months were likely to make fewer visits. Finally, we used the benefits design file to identify the copayment each enrollee would be charged for a first regular outpatient visit. This controls for a potentially important dimension of benefit generosity.

Data Analytic Procedures

Regression analyses investigated the determinants of the number of regular outpatient visits, and spending for regular outpatient care. In both cases we found substantial skewness in the distribution of the dependent variable, creating problems for standard linear regression. Furthermore, when we tried a logarithmic transformation, the log version of each variable showed generalized heteroskedasticity, making the commonly used ‘Duan smear’ approach problematic.17 We therefore opted instead to use generalized linear models with a log link.18 The resulting coefficient estimates are still on the log scale, but any predictions derived from them do not require correction for retransformation bias.19

An additional concern was the potential endogeneity of diagnosis variables. This could occur because the more claims a user of behavioral health care has, the more likely it is that any particular diagnosis will be found. To test the sensitivity of our results to the inclusion of diagnosis variables, we ran alternative models without them. Finally, we estimated instrumental variables models to address possible endogeneity of the EAP visit allowance.


The study sample was somewhat concentrated in the southern and western regions of the US, reflecting the enrollment base of the MBHO (Table 1). The most common diagnoses reported in claims were for adjustment disorders (40%) and mood disorders (39%).

Table 1
Characteristics of the study sample

Nearly half the enrollees in this sample were in employer plans that allowed 4–5 visits per incident, the most generous of the four possibilities (Table 2). Another 31% were allowed 3 visits per year, the least generous. The average sample member had 5.8 regular outpatient visits during 2005, incurring total spending of $467. Members with per-incident EAP benefits had fewer regular outpatient visits than those with per-year EAP benefits, although the differences were small.

Table 2
Use of regular outpatient care, by EAP allowance and utilization

Tables 3 and and44 present results of regressions using the number of EAP visits allowed as the key explanatory variable. The reference group contains enrollees with an allowance of 3 sessions per year (the least generous benefit). Compared to that group, having an EAP visit allowance of 3 per incident predicts fewer regular outpatient visits (Table 3) and lower spending (Table 4). Similar results are found for an allowance of 4–5 sessions per incident, but not for an allowance of 4–5 sessions per year. Each table includes a specification without diagnosis dummies (Model 2) to address the potential endogeneity of those variables. The results were qualitatively similar, although some coefficients changed substantially. All EAP variables that were statistically significant in Model 1 remained so in Model 2, with the same sign. It appears that our main findings were not sensitive to the inclusion of diagnosis variables.

Table 3
Effects of the EAP visit allowance on regular outpatient visits
Table 4
Effects of the EAP visit allowance on regular outpatient payments

The results are also informative as to how the other covariates affect the use of regular outpatient care. Younger enrollees had fewer regular outpatient visits and lower spending. Enrollees in the Northeast region (omitted category) had more outpatient visits and spending than those in the other three regions. The copayment level had a small but statistically significant negative effect on use of regular outpatient care. The diagnosis dummies had large positive effects on the quantity of visits and spending, except in the case of substance abuse.

Specification checks

The results presented in Table 3 and Table 4 assume exogeneity of the EAP visit allowance variables. This assumption might be incorrect if unobserved characteristics of firms or their environment affect both employers’ choice of EAP benefit level and their enrollees’ use of regular outpatient care. For example, suppose that higher levels of behavioral health need in the covered population lead some employers both to experience higher use of behavioral health care and to choose more generous EAP benefits. In this case, the coefficient on ‘EAP visit allowance’ would be biased upward in cross-sectional models. We pursued an instrumental variables (IV) analysis to address this issue.20

We considered using state legislation such as parity laws and benefit mandates as instrumental variables, since they potentially influence benefit offerings. However, this option proved infeasible, as 95% of our sample was enrolled through employers that are self-insured, and therefore exempt from state insurance legislation. In addition, the few employers that were not self-insured all had the same EAP benefit, making self-insurance itself not a feasible instrument. The only two variables available as instruments were the employer’s size and industry. Both variables are likely to affect benefit offerings, as required for an IV. However, we had some concern how plausibly exogenous the employer’s industry might be, given documented differences in behavioral health need across industries,2122 which could affect outpatient use through channels other than the EAP benefit level. These two variables were missing for 4,561 observations (17% of the sample), so using them reduced the sample size to 21,903.

Although the employer size variable had 4 categories and the industry variable had 12, in most of these categories only some levels of EAP benefit were represented. As a result, in order to successfully use them in a first stage model predicting EAP benefit level, each variable had to be collapsed into two categories (employer size above/below 10,000 employees; and whether industry is in a combined category of manufacturing/ utilities/ trade, versus all other). The resulting variables proved to be jointly significant in a model predicting EAP benefit level, satisfying the first requirement for use as instruments. However, since there were only two instruments, we had to reduce the number of potentially endogenous explanatory variables from the original three, in order to achieve identification. In order to be able to test the exogeneity of the instruments, we chose to estimate models with only one potentially endogenous explanatory variable: whether the EAP benefit was per-incident (versus per-year). The first-stage model was therefore estimated using binary logit. The second stage models were estimated with OLS rather than GLM, given the use of instruments. The instruments did pass the test of over-identifying restrictions (the second requirement for consistency of IV).

Table 5 presents the coefficients for the per-incident variable from the OLS and IV versions of this model. In each case, the IV estimate has the same sign and significance level as the OLS estimate, although for payments the IV estimate is considerably larger. As is typical, the standard errors are much larger with IV estimation. The results are not directly comparable to those in Table 4 because a) 17% of the sample had to be dropped, b) the EAP benefit variable has been simplified, and c) the second-stage models were estimated with OLS rather than GLM. Nevertheless, the results suggest that there is unlikely to be a large endogeneity problem with the benefit allowance variable, which should improve confidence in the findings in Tables 3 and and44.

Table 5
Effects of the per-incident benefit

Finally, there was concern that the behavioral health copayment level, like the EAP model, might be somewhat endogenous, for example if it was chosen in response to enrollee preferences. We tested the sensitivity of results to this variable by rerunning the main models without it. In the models without a copayment variable, the coefficient on ‘4–5 visits per year’ increased substantially in magnitude (to -0.126), becoming newly significant in the model to predict outpatient visits. This suggests that the copayment variable is indeed related to EAP visit allowance in one of our four models. However, coefficients on the other EAP allowance variables did not change much and retained significance. On the whole, our main findings are not sensitive to inclusion or exclusion of the copayment variable.


We find that greater availability of EAP benefits does reduce utilization of regular outpatient care, supporting the idea that the two types of care are to some extent substitutes. To our knowledge, no prior studies have measured how the number of EAP visits allowed affects individuals’ use of regular outpatient behavioral health care. The generosity of EAP visit allowances affects both the number of regular outpatient visits and total spending on them.

From the employer’s point of view, it is probably encouraging that EAP benefits reduce the use of regular outpatient behavioral health care, rather than merely adding to their total costs for behavioral health care. At the same time, patients appear to perceive that EAP services are still offering something distinct, which should discourage employers from eliminating EAP benefits as duplicative. Conversely, these results should not be interpreted as encouraging replacement of behavioral health benefits with an expanded EAP, as some employers have reportedly considered since the recent expansion of federal parity legislation.8 The EAP approach may not be well suited to handling more severely ill patients or those needing higher levels of care, who were not included in this study. In fact such services are always excluded from EAP benefits.

A number of studies have tried to address the wider question of whether EAP programs ‘pay for themselves’, in the sense of generating societal benefits greater than the cost of implementation and operation.11 A variant of this would be to consider how to structure the relative prices so that any given amount of mental health is produced at minimum cost by choosing the right mix of EAP and regular outpatient services (a ‘productive efficiency’ approach). The present study does not attempt to address such questions, since treatment and workplace outcomes were not reported in the administrative data available. As a result we cannot answer wider questions about the net benefits of delivering behavioral health care through EAP versus other models. However, our findings may be helpful in designing future studies on this topic. In particular, it may be important for such studies to include plans with a wider range of EAP visit allowances, in order to further investigate the effects of the generosity of the EAP allowance. Another useful extension would be to examine the sequential patterns of use of EAP and regular outpatient care within treatment episodes.


In conclusion, several limitations of this study should be noted. First, the study is cross-sectional in nature, and the relationships observed could reflect the effect of other unmeasured variables. These include enrollee attitudes to EAP and other services; severity of illness; and other services used, such as pharmaceuticals. Second, the data are for a single year, so we may be missing some patterns (for example, utilization in the year following EAP use). Third, the data are from a single managed behavioral health organization, limiting generalizability somewhat, although many employers are represented in the data. Finally, our results come from an integrated EAP/MBHC setting, and may not generalize as well to settings where the EAP benefit is managed by a separate entity than the regular outpatient benefits. It would be interesting to test whether the relationships observed here would be replicated in that different setting.


This study was funded by the National Institute on Drug Abuse grant # P50-DA-010233. The authors thank Nancy Pun and Kikumi Usui for analytic file preparation, Galina Zolotusky for programming assistance, Vanessa Azzone and Grant Ritter and two anonymous reviewers for helpful comments.


1. U.S. Bureau of Labor Statistics. [Accessed 4/15/2009];National Compensation Survey: Employee Benefits in Private Industry in the United States. 2007a March;
2. U.S. Bureau of Labor Statistics. [accessed Oct 28, 2010];National Compensation Survey: Employee Benefits in Private Industry in the United States, 2005. 2007b Bulletin 2589. Internet:
3. Horgan CM, Garnick DW, Merrick EL, et al. Changes in how health plans provide behavioral health services. J Behav Health Serv Res. 2009;36(1):11–24. [PMC free article] [PubMed]
4. Oss M, Jardine EL, MJP . Open Minds Yearbook of Managed Behavioral Health & Employee Assistance Program Market Share in the United States, 2002–2003. Behavioral Health Industry News, Inc; 2002.
5. Merrick ES, Volpe-Vartanian J, Horgan CM, McCann B. Revisiting employee assistance programs and substance use problems in the workplace: key issues and a research agenda. Psychiatr Serv. 2007;58(10):1262–4. [PubMed]
6. Sharar DA. Do employee assistance programs duplicate with services offered through mental health benefit plans? Compens Benefits Rev. 2009;41:67–74.
7. Finch RA, Phillips K. An Employer’s Guide to Behavioral Health Services: A Roadmap and Recommendations for Evaluating, Designing, and Implementing Behavioral Health Services. Washington, DC: Center for Prevention and Health Services, National Business Group on Health; 2005.
8. Maynard J, London M. Employee Assistance Professionals Association; 2009. [Accessed Nov 18, 2009]. The new parity law: What it means for EAPs. Internet:
9. Herring B. Suboptimal provision of preventive healthcare due to expected enrollee turnover among private insurers. Health Econ. 2009;19(4):438–448. [PubMed]
10. Hodgkin D, Horgan CM, Garnick DW, Merrick EL, Goldin D. Why carve out? Determinants of behavioral health contracting choice among large US employers. J Behav Health Serv Res. 2000;27(2):178–193. [PubMed]
11. Csiernik R. A review of EAP evaluation in the 1990s. Employee Assistance Quarterly. 2005;19(4):21–37.
12. Blum TC, Roman PM. DHHS pub no RP-0907. Rockville, MD: Substance Abuse and Mental Health Services Administration, Center for Substance Abuse Prevention, 1995; Cost-Effectiveness and Preventive Implications of Employee Assistance Programs.
13. Merrick EL, Hodgkin D, Horgan CM, Hiatt D, et al. Integrated employee assistance program/managed behavioral healthcare benefits: relationship with access and client characteristics. Adm Policy Ment Health. 2009;36(6):416–423. [PMC free article] [PubMed]
14. Cuffel BJ, Regier D. The relationship between treatment access and spending in a managed behavioral health organization. Psychiatr Serv. 2001;52:949–952. [PubMed]
15. Merrick EL, Hodgkin D, Hiatt D, Horgan CM, et al. Patterns of service use in two types of managed behavioral health care plan. Psychiatr Serv. 2010;61(1):86–89. [PMC free article] [PubMed]
16. Agency for Healthcare Research and Quality. [accessed Oct 28, 2010];Clinical Classification Software for ICD-9-CM Fact Sheet. 2003 Internet
17. Duan N. Smearing estimate: a nonparametric retransformation method. J Am Stat Assoc. 1983;78(383):605–610.
18. McCullagh P, Nelder JA. Generalized linear models. Chapman and Hall; London: 1989.
19. Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ. 2001;20(4):461–94. [PubMed]
20. Wooldridge J. Introductory Econometrics. 3. Thomson; Mason, Ohio: 2006.
21. Substance Abuse and Mental Health Services Administration, Office of Applied Studies. The NSDUH Report: Worker Substance Use, by Industry Category. Rockville, MD: 2007.
22. National Institute for Occupational Safety and Health. [accessed June 25, 2010];Worker Health Chartbook, 2004. Internet: