Medicaid, Private Insurance, and the Availability of Smoking Cessation Interventions in Substance Use Disorder Treatment

Rates of cigarette smoking among individuals in the United States with substance-related diagnoses typically exceed 70%, according to epidemiological data (1–3) and reports from treatment organizations (4,5). Much of the excess mortality experienced by individuals with substance-related disorders is attributable to tobacco use (6,7). Among individuals receiving substance abuse treatment, continued smoking is a significant risk factor for relapse (8,9).

These observations regarding prevalence, mortality, and relapse have led to calls for greater integration of smoking cessation into substance abuse treatment services (10–13). Such integration is consistent with the Preventive Health Service’s practice guideline Treating Tobacco Use and Dependence (14). Many of the guideline’s recommended practices draw upon clinical skills that parallel the counseling delivered in specialty treatment. However, research has shown that counseling-based smoking cessation programs are rare in these settings (15,16), and few counselors explicitly address tobacco use with their clients (17–20). Data also show relatively low use of smoking cessation medications (5,21,22), despite strong evidence regarding effectiveness, particularly when these medications are combined with counseling (23).

Numerous reports have examined the diffusion of smoking cessation interventions while considering organizational structure, cultural norms, and perceived barriers to adoption (5,15,16,24–27). Measures of organizational structure provide important information about where smoking cessation services are more likely to be delivered, and cultural research highlights potential strategies to increase the uptake of smoking cessation services (28).

Less is known about whether patterns of treatment funding are associated with the availability of smoking cessation services. The omission of financial variables in prior research is a distinct oversight because of the importance of external resources in shaping organizational behavior. Resource dependence theory suggests that an organization’s strategic decisions frequently reflect its efforts to manage relationships with entities that provide financial resources (29–31). Specialty treatment programs vary considerably in funding: some organizations rely heavily on federal block grant funding, others receive substantial revenue from Medicaid and private insurance, and many juggle numerous public and private funding streams (32–34).

The literature on the adoption of medications to treat addiction documents relationships between funding and availability. Medicaid revenues are positively associated with both the adoption (that is, any use) and the implementation (that is, extent of use) of such medications (35,36). Comparisons of privately funded treatment programs with those that rely on governmental grants and contracts have revealed sizable differences in adoption of medications (36,37), and the percentage of patients insured by health maintenance organizations has been shown to be positively associated with the adoption of naltrexone (38). These studies suggest that receipt of Medicaid and private insurance revenues may be associated with smoking cessation pharmacotherapy. It is less clear whether these funding streams are relevant for the availability of counseling-based smoking cessation programs.

In this study, Medicaid and private insurance were examined as correlates of smoking cessation services. Such an analysis is timely given the ongoing changes in the financing of health care associated with the Affordable Care Act (ACA). Provisions in the ACA that allow young adults to remain on their parents’ insurance have already reduced the percentage who are uninsured and who lack access to care (39). About half of the states are expanding Medicaid (40), and it is anticipated that individuals with substance use disorders will represent a sizable percentage of those who become insured via Medicaid (41). The ACA’s individual insurance mandate is likely to further increase the percentage of individuals who are insured.

The ACA is likely to affect the substance abuse treatment system. However, major efforts to enroll individuals in the health insurance exchanges began in late 2013 (42), and it will thus be some time before these impacts can be analyzed. These measurement challenges are not unique to the treatment system. Recent commentary has noted the difficulties and time lags in measuring changes in insurance coverage under the ACA (43). In the interim, this study offers a method of considering the potential impact of health reform by examining whether Medicaid and private insurance are associated with two types of interventions: counseling-based smoking cessation programs and cessation medications.

Methods

Sample and Data Collection

This research drew on a nationally representative cross-section of U.S. treatment organizations offering specialty treatment for alcohol use disorders. The initial sample was constructed by using the Substance Abuse and Mental Health Services Administration’s online Substance Abuse Treatment Services Locator to randomly sample treatment organizations. Telephone screening identified eligible organizations, defined as treatment facilities that were open to the general public, reported that at least 25% of their patients had a primary diagnosis of an alcohol use disorder, employed at least two full-time-equivalent employees, and offered treatment that was equivalent to or greater than structured outpatient services. Additional details regarding the sampling methods and the data collected from 2009–2012 have been published elsewhere (44).

This study drew on more recent data collected from October 2011 to December 2013. The prior sample was recontacted by telephone to assess eligibility. To address attrition based on organization closures, ineligibility, and refusal to participate, the original sample was supplemented with randomly selected replacement organizations. This process identified 437 eligible organizations by using the eligibility criteria described above. Trained interviewers visited administrators and clinical directors from 372 treatment organizations who were willing to participate in face-to-face interviews (85% response rate). All participants provided written informed consent. The institutional review boards of the University of Georgia and the University of Kentucky approved this research. The sample was compared with the prior sample on the independent variables and control variables; the distributions of these variables were highly similar. Differences in instrumentation did not allow for similar comparisons of the dependent variables.

Measures

Four smoking cessation services were examined. First, administrators were asked whether the center offered a formal smoking cessation program. Additional questions addressed whether that program included individual or group counseling sessions (or both) dedicated to smoking cessation. Centers with dedicated sessions were coded 1 for having a counseling-based smoking cessation program (all others were coded 0). Administrators indicated (dichotomous measures) whether the center currently used sustained-release bupropion, varenicline, or nicotine replacement therapy (NRT) in the form of the nicotine patch or gum. These are frontline medications for treating tobacco dependence (14).

Questions about the two funding variables were e-mailed to participants before the face-to-face interviews so that they could consult financial records. During the interview, administrators were asked to report the percentage of past-year revenues accounted for by Medicaid and the percentage accounted for private insurance.

Organizational structure, staffing, and treatment culture were controlled in the models because previous studies have suggested that these variables are associated with service delivery. Structural variables included government ownership, profit status, accreditation by the Joint Commission or the Commission on the Accreditation of Rehabilitation Facilities, and whether the organization offered only outpatient care. Staffing measures included the number of counselors employed, the percentage of counselors holding a master’s-level degree or greater, and a typology of access to physicians. This typology categorized organizations on the basis of whether they had any physicians on staff or any physicians on contract (if none on staff) or lacked access to staff or contract physicians. Four dichotomous variables measured treatment culture. Emphasis on the 12-step model and emphasis on the medical model of addiction were measured by two variables in which Likert responses (0, no extent; 5, very great extent) were dichotomized such that 4 and 5 were coded as the strong-emphasis group and all others were in the reference category (weak emphasis). The other two cultural variables were current use of motivational enhancement therapy and current use of cognitive-behavioral therapy. Similar to Aarons and colleagues (45), these dichotomous measures reflect treatment organizations that report current use and the presence of at least one counselor trained in the intervention. Region was coded as the four primary geographic regions defined by the U.S. Census Bureau (46).

Analysis

Descriptive statistics were computed for all measures. Prior to estimating multivariate models, multiple imputation by chained equations was implemented to address missing data and avoid the limitations of complete-case analysis (47). Using complete case analysis would have reduced the sample by 24% (N=88). Rates of missing values ranged from 1% to 8%. Our use of “mi impute chained” in Stata 13.1 included the four dependent variables, the two independent variables, and the control variables. Each variable was imputed by using the appropriate link function based on the level of measurement (for example, logistic regression if dichotomous and Poisson regression if a count). Twenty data sets were generated.

All dependent variables were dichotomous, and thus logistic regression was used for model estimation (48). All models included the “cluster” option in Stata, with clusters defined by the state in which the organization was located. The cluster option yields robust standard errors, which reduces potential bias from data clustering (49). This adjustment is important because states vary in how their Medicaid programs are structured, particularly in terms of which services are covered (50).

Results

Descriptive statistics are presented in Table 1. The most commonly adopted smoking cessation medication was NRT. Only 31% (N=109) of the 354 treatment organizations providing medication-related data had adopted at least one medication. The availability of counseling-based smoking cessation programs was low (29%). Only 15% of organizations (N=54) offered at least one medication and had a counseling-based smoking cessation program. Other organizational characteristics are summarized in Table 1. The average program received only a small portion of its past-year revenues from Medicaid and private insurance.

Significant associations between the smoking cessation services were noted. Offering a counseling-based smoking cessation program was positively correlated with current use of varenicline (Pearson’s r=.17, p<.01), bupropion (r=.15, p<.01), and NRT (r=.32, p<.001). There were also strong correlations between the medications. Current use of varenicline was positively correlated with use of bupropion (r=.66, p<.001) and NRT (r=.52, p<.001), and current use of bupropion was positively associated with NRT (r=.48, p<.001).

Table 2 presents results of a multivariate analysis of variables as potential predictors of the availability of a counseling-based smoking cessation program. Greater reliance on Medicaid revenue was positively associated with the presence of a counseling-based smoking cessation program. A 10-percentage point increase in Medicaid revenues was associated with a 12% increase in the odds of offering a smoking cessation program [12=100(e^(.011)(10)−1)]. Private insurance revenues were not associated with the availability of a smoking cessation program. Of the other organizational variables, only three were significantly associated with the availability of a counseling-based smoking cessation program: outpatient-only treatment, the percentage of master’s-level counselors, and region (organizations in the Midwest were less likely than those in the Northeast to offer a counseling-based smoking cessation program).

Results of three multivariate analyses of variables as potential correlates of current use of bupropion, varenicline, and NRT are summarized in Table 3. Medicaid revenues were positively associated with the three medications. For a 10-percentage point increase in Medicaid revenues, there was a 25% increase in the odds of currently using bupropion [25=100(e^(.022)(10)−1)], a 35% increase in the odds of using varenicline [35=100(e^(.030)(10)−1)], and a 13% increase in the odds of using NRT [13=100(e^(.012)(10)−1)].

Reliance on private insurance was also positively associated with the three medications. A 10-percentage point increase in private insurance revenues was associated with a 39% increase in the odds of currently offering bupropion [39=100(e^(.033)(10)−1)], a 62% increase in the odds of offering varenicline [62=100(e^(.048)(10) −1)], and a 40% increase in the odds of offering NRT [40=100(e^(.034)(10)−1)].

A limited number of associations were found between the organizational control variables and the three medications. The number of counselors was positively correlated with the odds of offering bupropion, varenicline, and NRT. Government-owned organizations were significantly more likely than privately owned organizations to offer bupropion and varenicline. For varenicline and NRT, organizations with at least one physician on contract were less likely to offer the medication than those with at least one physician on staff. Organizations with no physicians were less likely than organizations with at least one physician on staff to offer NRT. Accreditation was positively correlated with offering bupropion. Outpatient-only programs were less likely to offer NRT, and hospital-based programs were more likely to offer NRT. Profit status, the percentage of master’s-level counselors, the four treatment culture variables, and region were not associated with offering any of the medications.

Discussion

Although the literature on the delivery of smoking cessation services in substance abuse treatment has grown considerably over the past decade, few studies have examined the associations between financing and availability of these services. This study found that greater reliance on Medicaid was positively associated with the availability of counseling-based smoking cessation programs and medications. Greater reliance on private insurance was positively associated with the current use of bupropion, varenicline, and NRT but was not associated with availability of counseling-based smoking cessation programs.

Examining the role of financing is particularly timely given the ongoing implementation of health reform under the ACA. Despite ongoing court challenges, the ACA is likely to have an impact on the financing of specialty substance abuse treatment, particularly in states that expand Medicaid. An empirical question remains: to what extent do state-level decisions about the Medicaid expansion affect the financing of treatment and, subsequently, the strategic decisions made by organizations regarding service delivery? Such a question points to two issues for future research—namely, the need for ongoing longitudinal studies of treatment organizations and for multilevel models that estimate the impact of the Medicaid expansion on these organizations.

Future research using multilevel modeling is also needed to address the variability within state Medicaid plans regarding coverage of smoking cessation interventions. Recent data suggest fairly widespread coverage of medications by state Medicaid programs in all Medicaid plans for all populations: programs in 38 states cover varenicline, and programs in 45 states cover the nicotine patch (51). However, coverage for counseling is much more variable, with 27 states covering individual counseling and just 8 states covering group counseling in all Medicaid plans. What is striking is that, at least in the context of specialty substance abuse treatment, the overall impact of Medicaid funding on availability of smoking cessation services was found to be positive despite these variations. Some variability may decrease because as of 2014 the ACA mandates that state Medicaid programs cannot exclude from coverage medications approved for smoking cessation by the Food and Drug Administration (FDA). States are still expected to vary in terms of limiting the number of annual quit attempts, limiting the duration of treatment, and requiring prior authorization (51). Thus continued research on the intersection between state Medicaid programs and service delivery in specialty treatment organizations is warranted.

Although both types of revenue were associated with the three medications, Medicaid revenues were positively associated with offering a counseling-based smoking cessation program, whereas private insurance revenues were not. This lack of association may reflect variability and uncertainty regarding coverage of cessation counseling in private health plans. An analysis of 39 private plans found that ten plans excluded group counseling from coverage and another 20 plans did not provide sufficient detail to determine whether group cessation counseling was covered (52). Treatment organizations commonly rely on the modality of group counseling, and thus this heterogeneity may partly explain why private insurance revenues were not associated with counseling-based cessation programs. The ongoing implementation of the ACA may improve coverage because many health plans are now required to cover smoking cessation counseling (53). Recent guidance from the U.S. Department of Labor, prepared in collaboration with the Department of Health and Human Services, notes that “nongrandfathered” group health plans as well as policies sold in the individual and group markets must cover at least four counseling sessions during two cessation attempts per year plus FDA-approved smoking cessation medications (54).

An additional contribution of this study is the estimation of models by using the same covariates across multiple smoking cessation interventions. In general, relatively few of the organizational characteristics, beyond the two financial measures, were significantly correlated with smoking cessation services. The associations between organizational size and availability of pharmacotherapy are consistent with our prior work (21). The null findings for treatment culture, specifically 12-step orientation, align with recent work by Muilenburg and colleagues (24).

One puzzling finding was that the typology of access to physicians did not show an expected difference in medication availability between organizations with a physician on staff and those without a physician. Additional models that excluded the two financial variables (not shown) indicated significant differences in the odds of medication availability between organizations with a staff physician and those without a physician. Further analyses indicated that greater reliance on Medicaid and private insurance reduced the odds that organizations lacked access to physicians, which may explain why the comparison between organizations with a physician on staff and those without access to physicians was not significant in the multivariate models of medication availability.

A second unexpected finding was that the percentage of master’s-level counselors, an indicator of professionalism, was inversely associated with the availability of a counseling-based smoking cessation program. A previous study of counselors found no association between educational attainment and the delivery of smoking cessation counseling (17,18). One possible explanation is that more highly educated counselors are less cognizant of the importance of cessation counseling because they are less likely to have a personal smoking history (55). Additional research is needed to clarify the relationship between the education of the counseling workforce and availability of counseling-based smoking cessation services.

Several limitations are notable. First, the analyses relied on cross-sectional data, and, therefore, causality cannot be determined. Second, the dependent variables measured availability of cessation programs and medications, not the level of implementation within treatment programs. Measuring the extent to which tobacco-using individuals actually engage in counseling-based smoking cessation programs or receive cessation medications is important for future research. Third, these findings may not generalize to organizations that were excluded from participating, such as treatment programs that treat only individuals with drug disorders (for example, methadone programs), corrections-based programs, or programs in the Veterans Health Administration. Finally, all data were self-reported by program leaders, which may be subject to social desirability or recall bias. The relatively limited availability of services may assuage some of those concerns.

Conclusions

Although smoking cessation services in specialty substance use disorder treatment remain limited, these findings suggest that the major changes in financing under health care reform may be relevant for the diffusion of smoking cessation services. Reliance on Medicaid and on private insurance were both significantly associated with the availability of smoking cessation services, and these sources of funding are anticipated to expand through the implementation of the ACA. Currently, these funding sources are rather limited in most treatment centers. Responding to the changing funding landscape will represent a substantial shift for many treatment organizations. Longitudinal research is needed to fully understand whether financing changes under the ACA result in the expanded delivery of smoking cessation interventions.