Group Versus Standard Behavioral Couples’ Therapy for Alcohol Use Disorder Patients: Cost-Effectiveness

Abstract

Objective:

The purpose of this study was to evaluate the costs and cost-effectiveness of two treatments for 101 alcohol use disorder patients and their intimate partners—group behavioral couples’ therapy plus individual-based treatment (G-BCT), or standard behavioral couples’ therapy plus individual-based treatment (S-BCT).

Method:

We estimated the per-patient cost of each intervention using a microcosting approach that allowed us to estimate costs of specific components in each intervention as well as the overall total costs. Using simple means analysis and multiple regression models, we estimated the incremental effectiveness of G-BCT relative to S-BCT. Immediately after treatment and 12 months after treatment, we computed incremental cost-effectiveness ratios (ICER) and cost-effectiveness acceptability curves for percentage days abstinent, adverse consequences of alcohol and drugs, and overall relationship functioning.

Results:

The average per-patient cost of delivering G-BCT was $674, significantly less than the cost of S-BCT ($831). However, 12 months after treatment, S-BCT participants performed better on all outcomes compared with those in G-BCT, and the calculated ICER moving from G-BCT to S-BCT ranged from $10 to $12 across these outcomes. The current findings indicated that, except at very low willingness-to-pay values, S-BCT is a cost-effective option relative to G-BCT when considering 12-month posttreatment outcomes.

Conclusions:

As expected, G-BCT was delivered at a lower cost per patient than S-BCT; however, S-BCT performed better over time on the clinical outcomes studied. These economic findings indicate that alcohol use disorder treatment providers should seriously consider S-BCT over G-BCT when deciding what format to use in behavioral couples’therapy.

Alcohol misuse, including abuse and dependence, is associated with serious health and social consequences (e.g., losses in work productivity, criminal justice involvement, and motor vehicle accidents). The social costs associated with excessive alcohol use have been estimated to exceed $249 billion yearly in the United States (Sacks et al., 2015). These substantial costs have prompted considerable interest in implementing treatments that are effective at lessening the negative consequences of problematic drinking.

Behavioral couples’ therapy (BCT) for adults with alcohol use disorder and their intimate partners is one approach that has been developed over the past few decades (Epstein & McCrady, 1998). BCT recognizes that the behaviors of the drinker’s partner and the interactions between the drinker and his or her partner can trigger problematic drinking. BCT also recognizes that promoting a positive relationship between the patient and partner and fostering the support of the partner in the patient’s recovery may help achieve desired changes in drinking behavior (O’Farrell & Fals-Stewart, 2006). Research has shown that BCT produces greater abstinence, fewer substance-related problems, and better relationship functioning than standard individual-based therapy, and it reduces domestic violence and emotional problems of the couples’ children (e.g., Meis et al., 2013; O’Farrell & Clements, 2012; Powers et al., 2008).

Previous studies have examined the cost-effectiveness or cost-benefit for group-based therapies in substance use treatment (e.g., French et al., 2008; Goorden et al., 2016), with some findings suggesting that group-based therapies may be cost-effective. However, few studies have examined the cost-effectiveness or cost-benefit of BCT for substance using adults relative to individual treatment. Fals-Stewart and colleagues (2005) examined the cost-effectiveness of brief BCT, and they found that brief BCT was significantly more cost-effective in reducing heavy drinking than standard BCT or individual treatment. In two cost-benefit studies of BCT for men with alcohol use disorders and their partners, O’Farrell et al. (1996a, 1996b) estimated the cost savings attributable to declines in alcohol-related hospital treatment and jail stays 12 months (O’Farrell et al., 1996a) and 24 months (O’Farrell et al., 1996b) following treatment. They found statistically significant reductions in social costs with an average cost savings per patient ranging from U.S. $5,000 to $6,700. Further, in a cost-benefit study, Fals-Stewart et al. (1997) found that BCT for male substance users resulted in statistically significant reductions in social costs that were associated with an average cost savings of $6,600 per patient in the year following treatment compared with $1,900 per patient for those patients who received individual treatment only. Despite these findings, the adoption of BCT in substance use treatment programs continues to be slow, and costs are often cited as a barrier to its implementation (Gifford, 2012; Schonbrun et al., 2012).

In response to perceived cost implementation barriers, O’Farrell and colleagues (2016) developed a group BCT (G-BCT). G-BCT incorporates a couples’ group format with rotating content and rolling admissions in which couples receive group sessions of BCT. In contrast, standard conjoint BCT (S-BCT) is delivered to one couple at a time. The expectation was that G-BCT would produce similar outcomes as S-BCT, but at a lower per-patient cost. In this article, we present a cost-effectiveness analysis (CEA) for the O’Farrell et al. (2016) randomized controlled clinical trial that examined G-BCT relative to S-BCT.

Method

Study design

The study design has been described previously by O’Farrell et al. (2016). In brief, married or cohabiting male and female patients seeking treatment for an alcohol use disorder at a large treatment center in the northeastern United States were randomly assigned to either G-BCT or S-BCT. All patients were expected to attend 12 weekly 12-step–oriented group individual-based treatment (IBT) sessions, which were based on the group drug counseling manual (Daley et al., 2002), slightly modified to focus on alcohol dependence. Patients and their partners also were expected to attend the treatment-specific couples’ treatment (i.e., multi-couple group BCT or single-couple standard BCT) for 11 consecutive weeks during the 12-week period. The G-BCT-specific couples’ treatment consisted of two 60-minute conjoint introductory counseling sessions and nine 90-minute group BCT sessions with three to five couples each. The S-BCT-specific couple treatment consisted of two 60-minute conjoint introductory counseling sessions and nine 60-minute conjoint standard BCT sessions. All sessions were conducted by study-trained therapists who were either masters-level, licensed addiction counselors or a doctoral-level psychologist. Urine tests were collected weekly during the treatment period. The larger study from which data used in the economic evaluation were collected was approved by institutional review boards at Harvard Medical School and at the Veterans Affairs Boston (O’Farrell et al., 2016).

Participants were patients (N = 101) with alcohol dependence and their heterosexual relationship partners without substance use disorder, mostly White, in their forties, and 30% of patients were women. Participants underwent assessments before treatment (baseline), after treatment, and at 3, 6, 9, and 12 months following the scheduled end of treatment. For more details, see O’Farrell et al. (2016).

Cost analysis

We used a microcosting approach to compute service level costs of G-BCT and S-BCT. Costs were calculated from the provider perspective as it is the most relevant for providers looking to implement these types of therapies in real-world practice. We included the labor time and nonlabor costs incurred by the treatment site to deliver these services to the patients and their partners. We did not include the value of the patient’s or partner’s time or other costs that they may have incurred to undergo this treatment (e.g., travel costs to/from treatment). In addition, because this was a research study, we wanted to estimate treatment costs that would be incurred in real-world settings (as opposed to those required to implement a clinical trial research protocol). Therefore, we identified the clinical activities and services that constituted each treatment. Treatment activities included planned information and counseling sessions, unplanned crisis sessions (i.e., sessions to offer immediate, short-term help to patients experiencing distress or problems), and urine testing. We did not include nontreatment activities related to implementing the clinical research trial such as randomization and conducting follow-up assessments because these costs would not be incurred in real-world clinical practice. All costs were converted to 2017 dollars using the Bureau of Labor Statistics’ Consumer Price Index (www.bls.gov/cpi).

We used a study-specific treatment attendance form to track the amount of services received by each patient each week and the duration of each service. This information was summed across services to calculate the total direct treatment time each patient received. Study intervention staff also estimated the average time spent on support activities (e.g., session preparation, note taking, administrative work) for each type of direct care service. Support time was calculated for each direct care service based on the quantity of services received by that patient. By summing across the direct care services, we calculated the total support time for each patient. Information on therapists’ wages was collected from the study’s principal investigator based on the grant’s financial records. Labor costs were calculated by taking the average therapist hourly wage (including fringe benefit rate) and multiplying by the total time reported (time spent providing direct care plus support time) for each session attended by the patient. Estimates for the building space size and associated costs used for the therapy sessions were provided by the administrative office of the clinic where the study was implemented. Last, the material cost of urine testing supplies was provided by the study’s principal investigator and was included in the costs for a urine test along with the labor time to provide that service.

The total cost of a service was simply the sum of costs for labor time, building space, and materials. Summing across all services received by the patient yielded the total cost of the treatment for that patient, and summing across all patients yielded the total cost of the treatment (i.e., G-BCT or S-BCT). To get the cost per patient for group counseling sessions, we divided the total session costs over the number of patients in the G-BCT sessions. Finally, we derived the average costs per patient for each treatment by dividing the total costs of the treatment by the number of patients assigned to that treatment.

Effectiveness measures

Following the analytic approach of the main findings article (O’Farrell et al., 2016), the CEA examined four effectiveness measures: (a) percentage days abstinent (PDA) from the Timeline Followback (TLFB) interview (Sobell & Sobell, 1996), (b) the Inventory of Drug Use Consequences (InDUC; Tonigan & Miller, 2002), (c) patient-reported Dyadic Adjustment Scale (DAS; Spanier, 2001) score, and (d) partner-reported DAS score. We focused on the immediate posttreatment and 12-month posttreatment assessments, which allowed us to examine the immediate effect, if any, of the treatments and whether this effect was sustained 12 months later.

PDA was calculated by dividing the number of days during which the patient was not in a hospital or jail for alcohol-related reasons and they remained abstinent from alcohol and other drugs by the total days during a given assessment period. To reduce possible underreporting of the patients’ substance use, the lowest reported PDA was used when both partners’ data were available. The InDUC measured the adverse consequences of the patient’s alcohol and drug use, with higher scores indicating more adverse consequences. To reduce possible underreporting of the patients’ substance-related problems, the higher report (i.e., worse outcome) was used when both partners provided responses to an InDUC item. Both patient and partner completed the TLFB and InDUC assessments with reference to the patient’s behavior. The DAS measured overall relationship adjustment as reported by the patient and by the partner, with higher scores indicating a better functioning relationship.

Using the methodology employed in the main clinical findings article (O’Farrell et al. 2016), we calculated predicted values for each of the outcomes with generalized estimating equations (GEE; Hall et al, 2001) analyses that included a time effect¹, the baseline outcome of interest, baseline days of nonstudy treatment, assigned study treatment, and a time by treatment interaction. Although we had a low degree of missing data, using GEE modeling for outcomes has the advantage of allowing the inclusion of covariate-dependent missing data.

Cost-effectiveness analysis

The CEA compared G-BCT to S-BCT by calculating an incremental cost-effectiveness ratio (ICER) for each of the outcomes. The first step in the CEA was to rank the treatments in increasing order of average per-patient cost. The ICER for each outcome was then computed, defined as the difference in average cost divided by the difference in average effectiveness. If one treatment is both less expensive and more effective, it strictly dominates the alternative intervention. When neither therapy is strictly dominated, the cost-effectiveness ratio is calculated regardless of statistical significance.

Choosing the optimal treatment after strictly dominated options are removed depends on a decision maker’s willingness to pay (WTP). WTP refers to the value that a person is willing to pay to achieve a given outcome. We calculated a cost-effectiveness acceptability curve (CEAC) to show the probability that a treatment is the cost-effective option as a function of the decision maker’s WTP for each of the outcomes. The use of CEACs as well as sensitivity analysis, described below, is important in reporting CEAs, especially given the combination of smaller sample sizes, skewed costs, and outcomes present in this study, which can introduce variability in the average estimates. The CEAC incorporates the inherent variability of the cost and effectiveness estimates and allows us to better capture the variability in our CEA in lieu of confidence intervals for the ICERs (Fenwick et al., 2001, 2006). We used a nonparametric bootstrap method to calculate the CEAC that compared the G-BCT and S-BCT across the four outcomes.

In addition to the use of CEACs, we performed sensitivity analyses examining the impact of changes in unit costs. We examined the effect of both increasing and decreasing the hourly wage (i.e., unit cost of labor) and nonlabor costs. The sensitivity analyses revealed that our point-estimate ICERs increased as labor costs increased because S-BCT used proportionally greater staff time relative to G-BCT. Conversely, the ICERs decreased as nonlabor costs increased because G-BCT had proportionally greater nonlabor costs. The interpretation of the cost-effectiveness results was not greatly affected by these analyses, although the overall costs of G-BCT and S-BCT changed. S-BCT was still economically dominated by G-BCT for the PDA and InDUC outcomes after treatment. S-BCT remained the more likely optimal choice when the WTP exceeded the estimated ICER, which still supported the overall findings of this study (see the Appendix). (The supplemental appendix appears as an online only addendum to the article on the journal’s website.)

Results

Table 1 shows the average time per session for each of the treatment activities and the average number of sessions received per patient. On average, G-BCT patients received more time per patient across all treatment activities (approximately 40 hours) compared with S-BCT patients (36 hours). The variance in average total time between the two treatments was primarily due to the treatment-specific couples’ counseling sessions. G-BCT counseling sessions averaged 1.47 hours per session and the S-BCT counseling sessions averaged 0.982 direct hours per session. This difference was part of the study design with the group format of the G-BCT counseling specifying 90-minute sessions, whereas the individual-couple format of the S-BCT counseling specified 60-minute sessions. Patients in both treatments received an average of 19.5 sessions during their treatment episode.

Table 1.

Mean characteristics of service utilization

	Average per session				Average per patient
Treatment	Average no. of patientsa (SD)	Direct treatment time (face-to face) (hours)b (SD)	Support activities time (hours)b (SD)	Total time (direct + support) (hours)b (SD)	Average no. of sessions (SD)	Average total time (hours) (SD)
G-BCT arm
IBT activities
Information session	1.000 (0.000)	0.897 (0.174)	0.480 (0.000)	1.377 (0.174)	1.000 (0.000)	1.377 (0.174)
Introductory counseling sessions	1.000 (0.000)	0.658 (0.174)	0.480 (0.000)	1.137 (0.174)	0.980 (0.140)	1.115 (0.234)
Counseling sessions	7.547 (1.743)	1.490 (0.070)	0.889 (0.109)	2.379 (0.132)	8.333 (3.615)	19.692 (8.390)
Crisis sessions	1.000 (0.000)	0.683 (0.137	0.230 (0.000)	0.913 (0.137)	0.157 (0.543)	1.534 (1.048)
G-BCT activities
Introductory counseling sessions	1.000 (0.000)	0.927 (0.126)	0.480 (0.000)	1.406 (0.126)	1.941 (0.580)	2.840 (0.673)
Counseling sessions	3.446 (0.772)	1.467 (0.091)	0.632 (0.048)	2.099 (0.113)	6.922 (3.463)	16.063 (5.736)
Crisis sessions	1.000 (0.000)	0.782 (0.375)	0.230 (0.000)	1.011 (0.375)	0.255 (0.523)	1.258 (0.798)
Total G-BCT arm					19.588	39.821*
(IBT plus G-BCT activities)					(7.261)	(15.790)
S-BCT arm
IBT activities
Information session	1.000 (0.000)	0.877 (0.167)	0.480 (0.000)	1.357 (0.167)	1.000 (0.000)	1.357 (0.167)
Introductory counseling sessions	1.000 (0.000)	0.653 (0.174)	0.480 (0.000)	1.132 (0.200)	0.980 (0.140)	1.132 (0.200)
Counseling sessions	7.525 (1.786)	1.500 (0.000)	0.887 (0.112)	2.387 (0.112)	8.549 (3.466)	20.765 (7.838)
Crisis sessions	1.000 (0.000)	0.708 (0.358)	0.230 (0.000)	0.938 (0.358)	0.157 (0.543)	1.367 (0.769)
S-BCT activities
Introductory counseling sessions	1.000 (0.000)	1.033 (0.096)	0.480 (0.000)	1.512 (0.096)	4.027 (0.666)	4.907 (0.710)
Counseling sessions	1.000 (0.000)	0.982 (0.106)	0.480 (0.000)	1.462 (0.106)	7.098 (3.590)	11.457 (4.073)
Crisis sessions	1.000 (0.000)	0.996 (0.259)	0.230 (0.000)	1.226 (0.259)	0.333 (0.766)	1.893 (1.118)
Total S-BCT arm					19.549	35.823*
(IBT plus S-BCT activities)					(7.018)	(13.270)

Notes: G-BCT = group behavioral couples’ therapy plus individual-based treatment; IBT = individual-based treatment; S-BCT = standard behavioral couples’ therapy plus individual-based treatment; no. = number.

^aAverage no. of patients per session is a conditional mean calculated only for those patients receiving the service indicated;

^breported average times per session (i.e., direct, support, and total) are conditional means calculated only for those patients receiving the service indicated.

^*Significantly different at p < .05 (two-sample Wilcoxon rank-sum [Mann–Whitney] test).

Table 2 shows the average costs per patient by treatment arm. As expected, patients in S-BCT had greater costs per patient than patients in G-BCT ($831 vs. $674; z = -4.007, p = .0001). This cost difference was driven by the more costly, on a per-patient basis, couples’ counseling sessions provided in S-BCT. G-BCT couples’ counseling sessions were in a group format that averaged 3.4 couples per session rather than the S-BCT’s individual-couple session with one couple per session.

Table 2.

Average costs per patient, by arm

Variable	G-BCTa ($, 2017) (SD)	S-BCTa ($, 2017) (SD)
Labor costs	604.19 (218.43)	778.66 (276.68)
Nonlabor costs	70.00 (50.17)	52.65 (35.11)
Total costs	674.19b (248.17)	831.31b (291.57)

Notes: G-BCT = group behavioral couples’ therapy plus individual-based treatment; S-BCT = standard behavioral couples’ therapy plus individualbased treatment.

^aThe unit cost for labor was the same across both arms at $38.15. The space costs varied by activity, with space costs for group drug counseling sessions (for both arms) and G-BCT session of $7.16 per hour and space costs for S-BCT sessions and support services (for both arms) of $1.33 per hour. Last, the unit cost for urine screens was $6.79 per screen.

^bThe total costs of G-BCT and S-BCT are statistically significant (z = -4.007, p = .0001). All costs are reported in 2017 dollars.

Table 3 shows the results of the CEA immediately after treatment for each outcome. G-BCT economically dominated S-BCT on the PDA and InDUC outcomes as patients in G-BCT had slightly better outcomes that were achieved with a lower per-patient cost. G-BCT patients had a slightly higher average PDA (85.3%) compared with S-BCT (83.7%). G-BCT patients also reported a slightly lower InDUC score (i.e., better outcome) on adverse consequences due to alcohol or drug use compared with S-BCT (23.5 vs. 26.0, respectively).

Table 3.

Cost-effectiveness analysis for immediate posttreatment outcomes

		% days abstinent		Patient report of adverse consequences of alcohol and drug use (InDUC)		Patient reported overall relationship functioning (DAS)		Patient reported overall relationship functioning (DAS)
Treatment	Mean cost per patient ($, 2017) (SD)	Mean effectivenessa (SD)	ICERb (ΔC/ΔE, $) (SD)	Mean effectivenessa (SD)	ICERb (ΔC/ΔE, $) (SD)	Mean effectivenessa (SD)	ICERb (ΔC/ΔE, $)	Mean effectivenessa (SD)	ICERb (ΔC/ΔE, $)
G-BCT arm	674.19 (248.17)	85.27 (0.640)	–	23.51 (0.946)	–	110.18 (1.794)	–	107.32 (1.474)	–
S-BCT arm	831.31 (291.57)	83.74 (0.758)	Economically dominated	26.00 (0.957)	Economically dominated	111.88 (1.374)	92.42	110.66 (1.486)	47.05

Notes: G-BCT = group behavioral couples’ therapy plus individual-based treatment; S-BCT = standard behavioral couples’ therapy plus individual-based treatment; InDUC = Inventory of Drug Use Consequences; DAS = Dyadic Adjustment Scale; ICER = incremental cost-effectiveness ratio.

^aMeans are predicted outcomes;

^bICERs are calculated from the point estimates of mean costs and outcomes presented in the table.

For the DAS scores on overall relationship adjustment we found that, on average, patients in S-BCT achieved a slightly better DAS outcome (i.e., a higher DAS score) compared with G-BCT patients (111.9 vs. 110.2, respectively, for the patient-reported DAS and 110.7 vs. 107.3, respectively, for the partner-reported DAS). Because S-BCT performed slightly better on the DAS outcome but was also more expensive to deliver on a per-patient basis compared with G-BCT, we calculated the ICER moving from G-BCT to S-BCT. This ICER is approximately $92 per 1-point increase in the patient-reported DAS score and $47 for a 1-point increase in the partner-reported DAS score. For example, the incremental cost of S-BCT relative to G-BCT for improving a patient’s DAS score from 104 to 112 (8% increase on the scale) would be an average of $736 ($92 × 8 percentage points).

As shown in Table 4, the CEA for the 12-month posttreatment outcomes reveals that over time the average effectiveness of G-BCT declined more than that of S-BCT. Patients in S-BCT had a higher average PDA than patients in G-BCT (81.1% vs. 68.1%, respectively). The ICER moving from G-BCT to S-BCT for PDA is $12 for a percentage point increase in days abstinent. For example, the incremental cost of S-BCT relative to G-BCT for increasing a patient’s PDA from 30% (approximately 9 days in a 30-day period) to 80% (approximately 24 days) would be an average of $600 (50 percentage points × 12). S-BCT patients also performed better on the InDUC outcome compared with G-BCT, reporting an average InDUC score of 16.5 compared with 31.9 for patients in G-BCT. The associated ICER moving from G-BCT to S-BCT is $10 for a 1-point decrease in the InDUC score.

Table 4.

Cost-effectiveness analysis for 12 months posttreatment outcomes

		% days abstinent		Patient report of adverse consequences of alcohol and drug use (InDUC)		Patient reported overall relationship functioning (DAS)		Patient reported overall relationship functioning (DAS)
Treatment	Mean cost per patient ($, 2017) (SD)	Mean effectivenessa (SD)	ICERb (ΔC/ΔE, $) (SD)	Mean effectivenessa (SD)	ICERb (ΔC/ΔE, $) (SD)	Mean effectivenessa (SD)	ICERb (ΔC/ΔE, $)	Mean effectivenessa (SD)	ICERb (ΔC/ΔE, $)
G-BCT arm	674.19 (248.17)	68.10 (0.640)	–	31.917 (6.427)	–	102.60 (1.794)	–	96.27 (1.474)	–
S-BCT arm	831.31 (291.57)	81.13 (0.758)	12.06	16.493 (6.858)	10.19	116.51 (1.374)	11.30	111.19 (1.486)	10.53

Notes: G-BCT = group behavioral couples’ therapy plus individual-based treatment; S-BCT = standard behavioral couples’ therapy plus individual-based treatment; InDUC = Inventory of Drug Use Consequences; DAS = Dyadic Adjustment Scale; ICER = incremental cost-effectiveness ratio.

^aMeans are predicted outcomes;

^bICERs are calculated from the point estimates of mean costs and outcomes presented in the table.

For the DAS scores, patients in S-BCT had a higher average score of 116.5 compared with 102.6 for patients in G-BCT. Similarly, partners in S-BCT had a higher average DAS score of 111.2 compared with 96.3 for partners in G-BCT. The ICER moving from G-BCT to S-BCT is approximately $11 for a 1-point increase in the patient’s DAS score and $11 for a 1-point increase in the relationship partner’s DAS score.

Cost-effectiveness acceptability curve results

Figures 1a through 1d include the CEACs for the posttreatment outcomes. The CEACs show the probability that S-BCT is the cost-effective option at different values of WTP for a unit improvement for the specified outcome. We present alternative dollar amounts for the WTP because there is no established standard in the field for the value of any of the selected outcomes. As shown, the probability that S-BCT is the cost-effective option for the DAS score outcomes (Figures 1c and 1d) is greater than 50% at WTP values of $45. This probability is greater than 80% once the WTP reaches $125 for the patient-reported DAS score and $80 for the partner-reported DAS score. However, the probability that S-BCT is the cost-effective option for the PDA and InDUC score (Figures 1a and 1b) is low, which supports the finding that S-BCT is economically dominated for these two outcomes.

Figure 1a.

Patient-reported percentage days abstinent, after treatment (2017 dollars). S-BCT = standard behavioral couples’ therapy plus individual-based treatment.

Figure 1d.

Partner-reported DAS score, after treatment (2017 dollars). DAS = Dyadic Adjustment Scale; ICER = incremental cost-effectiveness ratio; S-BCT = standard behavioral couples’ therapy plus individual-based treatment.

Figure 1c.

Patient-reported DAS score, after treatment (2017 dollars). DAS = Dyadic Adjustment Scale; ICER = incremental cost-effectiveness ratio; S-BCT = standard behavioral couples’ therapy plus individual-based treatment.

Figure 1b.

Patient-reported InDUC, after treatment (2017 dollars). InDUC = Inventory of Drug Use Consequences; S-BCT = standard behavioral couples’ therapy plus individual-based treatment.

Figures 2a through 2d present the CEACs for the 12-month posttreatment outcomes. As shown, the probability that S-BCT is the cost-effective option 12 months after treatment reaches greater than 50% at very low WTP values (approximately $11) for each outcome. This probability is almost 100% once the WTP reaches the $20 to $30 range.

Figure 2a.

Percentage days abstinent, 12 months after treatment (2017 dollars). S-BCT = standard behavioral couples’ therapy plus individual-based treatment; ICER = incremental cost-effectiveness ratio.

Figure 2d.

Partner-reported DAS score, 12 months after treatment (2017 dollars). DAS = Dyadic Adjustment Scale; ICER = incremental cost-effectiveness ratio; S-BCT = standard behavioral couples’ therapy plus individualbased treatment.

Figure 2b.

Patient-reported InDUC score, 12 months after treatment (2017 dollars). InDUC = Inventory of Drug Use Consequences; S-BCT = standard behavioral couples’ therapy plus individual-based treatment.

Figure 2c.

Patient-reported DAS score, 12 months after treatment (2017 dollars). DAS = Dyadic Adjustment Scale; ICER = incremental cost-effectiveness ratio; S-BCT = standard behavioral couples’ therapy plus individualbased treatment.

Discussion

This study is the first cost-effectiveness study to compare an innovative BCT that used a group couples’ counseling format for alcohol use disorder patients and their partners with a more traditional standard individual-couple format for delivering BCT. If G-BCT with its group format can achieve similar outcomes compared to S-BCT but for a lower perpatient cost, then G-BCT may be a more attractive option for providers trying to control costs while providing quality care.

Costs were calculated from the provider perspective because this perspective is the most relevant for providers looking to implement these therapies in real-world practice. Costs are often cited as a barrier to BCT implementation (Gifford, 2012; Schonbrun et al., 2012), although providers actually have limited information on the resources needed to implement such therapies. We found that, although patients in G-BCT received, on average, more hours of care than patients in S-BCT, because G-BCT used a group counseling format it was less costly to deliver on a per-patient basis compared with S-BCT.

Given that G-BCT was less costly on a per-patient basis, the question remains as to whether G-BCT can achieve desired levels on the selected outcomes—that is, is G-BCT a cost-effective treatment option relative to the more standard S-BCT? The results of the CEA showed that, immediately after treatment, G-BCT economically dominated S-BCT by producing slightly better average patient outcomes for PDA and InDUC at a lower cost. We should note that here we have performed pre-specified pairwise comparisons for the CEA to examine G-BCT relative to S-BCT. The pairwise comparisons presented here were not formal statistical tests. However, the statistical tests performed in O’Farrell et al. (2016) on the PDA and InDUC outcomes showed that, although G-BCT and S-BCT were not significantly different after treatment, they were also not statistically equivalent. Therefore, for posttreatment PDA and InDUC outcomes, the potential appeal of G-BCT as a viable option for providers is driven by its lower per-patient cost.

The results also showed that S-BCT performed slightly better on both the average patient DAS score and the average partner DAS score, but S-BCT was also more expensive. As indicated by the CEACs, the probability that S-BCT is the optimal treatment choice for these two posttreatment outcomes relative to G-BCT exceeds 50% at low WTP values, suggesting that, if a decision maker is willing to pay at least $50 to achieve a 1-point improvement in the DAS score, then S-BCT is likely a cost-effective option over G-BCT.

Our results for outcomes at 12 months after treatment showed that S-BCT performed better on all outcomes compared with G-BCT. The statistical tests performed in O’Farrell et al. (2016) showed that S-BCT had significantly more favorable outcomes 12 months after treatment compared with G-BCT. An exception was PDA, for which no statistically significant difference was found, but for which a numerically substantial (and, perhaps, clinically meaningful) difference was observed. Based on the CEACs, the probability that S-BCT is the cost-effective option 12 months after treatment is almost 100% once the WTP reaches $20 for all four outcomes. These findings indicate that, except at very low WTP values, S-BCT is a cost-effective option relative to G-BCT when considering 12-month posttreatment outcomes.

The main contributions of this study are the cost estimates for G-BCT and S-BCT and the CEA that combines costs and effectiveness to compare G-BCT to the more traditional S-BCT. However, this study highlights another key issue—sustainability of outcomes may be an important factor in considering cost-effective treatment options. In this study, G-BCT appeared to perform slightly better on PDA and InDUC outcomes immediately following treatment and at a lower per-patient cost, suggesting it is a cost-effective option compared with S-BCT. However, 12 months later, G-BCT’s dominance on these outcomes was no longer apparent. Twelve months after treatment, S-BCT performed better on all four outcomes, but these outcomes were achieved at a higher per-patient cost. Whether S-BCT is the optimal treatment choice here depends on the decision maker’s WTP for a given outcome. Last, different outcome measures may be of interest to different decision makers, and the prioritization that a decision maker puts on these outcomes is also important.

Our study had some limitations that should be noted. First, our cost analysis relied on the judgment of the study team as to which activities were primarily research related and which would be used in best clinical practice. To minimize this issue, the study team consisted of alcohol use disorder clinicians and researchers who provided expertise about how such interventions would be implemented in realworld practice. Second, although we attempted to identify activities and costs that would be part of real-world treatment practice from the provider perspective, in reality this study was part of a randomized controlled trial and followed the study’s rigorous protocol. We expect that patients’ treatment adherence may be better in a study trial compared with real-world practice, and, therefore, patients may have been seen more frequently during their treatment in the study; thus, we expect our cost estimates to be an upper bound of the actual treatment costs providers may incur for delivering these services. We should note that less treatment adherence may also lead to poorer treatment outcomes than we observed in this trial; however, we do not expect that the impact on treatment adherence would affect treatments differently; therefore, the differential effect across treatments should not affect the CEA. Third, we limited our study to the provider perspective and did not include costs incurred by patients or other affected stakeholders (e.g., third-party payers). We had limited data on nonprovider costs, and, therefore, we were unable to include these other perspectives in our analyses. However, future work may want to examine cost-effectiveness from other perspectives. Fourth, our PDA outcome relied on self-reports of alcohol use by the patient or their partner, which may be subject to recall bias. However, the study used the TLFB (Sobell & Sobell, 1996) to help minimize reporting bias. Last, our CEA was limited to the two interventions included in this study—G-BCT and S-BCT. The inclusion of additional interventions or an alternative set of interventions would provide different comparisons between cost and effectiveness, and different cost-effectiveness results. Furthermore, our CEA was limited to four selected outcomes. The study was not designed to examine more global health outcomes (e.g., quality-adjusted life years). However, PDA is a common outcome measure in the substance use treatment literature, allowing for comparability across other treatment studies. Furthermore, DAS and InDUC scores are important outcomes for this population.

Despite these limitations, this study makes an important contribution to the literature because it provides an economic analysis of an innovative approach to BCT for alcohol treatment—using a lower cost group counseling format. The study also demonstrates the impact that the erosion of effectiveness over time may have on cost-effectiveness results. S-BCT appeared to perform better 12 months after treatment on the selected outcomes compared with G-BCT. The choice of S-BCT as economically optimal depends on the value placed on these outcomes by the decision maker. The cost-effectiveness results, when combined with the overall clinical findings from the same study (O’Farrell et al., 2016), converge to indicate that providers should seriously consider favoring S-BCT over G-BCT when deciding which format to use in BCT.