Cost-Effectiveness of Peer-Delivered Interventions for Cocaine and Alcohol Abuse among Women: A Randomized Controlled Trial

Aims To determine whether the additional interventions to standard care are cost-effective in addressing cocaine and alcohol abuse at 4 months (4 M) and 12 months (12 M) from baseline. Method We conducted a cost-effectiveness analysis of a randomized controlled trial with three arms: (1) NIDA's Standard intervention (SI); (2) SI plus a Well Woman Exam (WWE); and, (3) SI, WWE, plus four Educational Sessions (4ES). Results To obtain an additional cocaine abstainer, WWE compared to SI cost $7,223 at 4 M and $3,611 at 12 M. Per additional alcohol abstainer, WWE compared to SI cost $3,611 and $7,223 at 4 M and 12 M, respectively. At 12 M, 4ES was dominated (more costly and less effective) by WWE for abstinence outcomes. Conclusions To our knowledge, this is the first cost-effectiveness analysis simultaneously examining cocaine and alcohol abuse in women. Depending on primary outcomes sought and priorities of policy makers, peer-delivered interventions can be a cost-effective way to address the needs of this growing, underserved population. Trial Registration ClinicalTrials.gov NCT01235091


Introduction
Over 9% of the total US population (23.1 million aged 12 or older) was in need of treatment for an illicit drug or alcohol use problem in 2008 [1]. Among females, the percentage of illicit drug users rose from 5.8% in 2007 to 6.8% in 2010; the number of users increased from 10.4% to 11.2% over the same period among males [1,2]. Alcohol and illicit drug use were the third and ninth leading causes of preventable death in 2000, associated with 85,000 and 17,000 deaths, respectively [3]. Moreover, the economic costs of drug abuse in the US have increased 5.3% annually from 1992 to 2002, reaching a total of $180.9 billion, $16 billion of which were health-related [4]. Despite this rise in the costs of substance abuse, health care spending on substance abuse only rose 4.1% annually, whereas total health care spending rose by 6.5%.
A number of behavioral interventions have shown effectiveness in treating cocaine [5,6] and alcohol abuse [7]. The Women Teaching Women intervention (WTW) was designed to reduce substance use and high-risk sexual behaviors through a holistic, peer-delivered intervention [8]. A randomized, controlled trial (RCT) examined three arms of WTW: (1) a modified National Institute on Drug Abuse (NIDA) Cooperative Agreement Standard Intervention (SI); (2) SI and a field-based Well Woman Exam (WWE); and (3) SI, WWE, and four Educational Sessions (4ES). The RCT took place in St. Louis, where, aside from alcohol, cocaine caused the greatest amount of treatment admissions in 2007 and remains a major drug problem [9,10]. This study examined ten outcomes related to cocaine and alcohol use; a companion study analyzed the cost-effectiveness of these interventions for preventing, STDs and HIV [11]. With cost data from a previously published [12] micro-costing study of WTW conducted from the service provider and societal perspectives, we undertake an incremental cost-effectiveness analysis to determine the cost per outcome gained for each subsequent arm of the trial.
The literature surrounding cost-effectiveness analyses (CEAs) of prevention and treatment for cocaine and alcohol abuse continues to grow [13,14,15]. One article [16] examines the costeffectiveness of a screening and brief intervention for alcohol abuse in a primary care setting. Others [15,17,18,19] address the additional costs of voucher-based and prize-based contingency management programs above and beyond standard treatment for cocaine and opioid dependence. Two CEAs performed analyses alongside RCTs for alcohol treatment, one comparing social behavior and network therapy to motivational enhancement therapy [20], and another examining combined behavioral and pharmacotherapies [13]. An additional analysis illustrated the incremental cost-effectiveness of an enhanced, personalized HIV intervention compared to SI for preventing drug use [21].
Decision and policy makers need to be aware of interventions' impact on both cocaine and alcohol abuse because the two substances are often co-abused. Of the 1,648 cocaine-dependent patients in the national Drug Abuse Treatment Outcomes Studies (DATOS), 51% were also alcohol-dependent [22] and in a study of 158 alcohol-dependent outpatients, 62 of them also manifested cocaine abuse [23]. This study was conducted in accordance with the recommendations of the US Panel on Cost-Effectiveness in Health and Medicine [24].

Methods
The protocol for this study and supporting CONSORT checklist and flow diagram are available as supporting information; see Checklist S1, Protocol S1, and Flow Diagram S1.

Ethics Statement
This study has been approved by IRB Human Studies Committee at Washington University Hilltop (03-20), Washington University Medical Center (04-0285) and Yale University Human Investigation Committee (27312). The randomized controlled trial in the parent study has been registered with ClinicalTrials.gov and the registration number is 01235091. Data from the parent study was analyzed anonymously; written consent was obtained for the effectiveness data and for the cost data.

Recruitment, Design, and Sample
Community health outreach workers recruited study participants from targeted recruitment zones, employing street-outreach methods based on prior work [25]. Eligibility criteria included: (1) being a woman over the age of 18; (2) reporting sexual activity in the prior four months; (3) using cocaine, heroin, amphetamines, or other injection drugs; and (4) residing in the St. Louis metropolitan area during the study period (2000)(2001)(2002)(2003)(2004)(2005)(2006). Of the 501 women in the study, 420 participants received the standard intervention (SI) and completed baseline and both 4-months and 12-months endpoint assessments. They were randomized to one of three intervention conditions: SI alone (n = 135); WWE (n = 144); or 4ES (n = 141). Study participants were 90% racial/ethnic minorities. On average, they were 39 years old, had approximately 11 years of education, and worked for about 7 months when interviewed at baseline. The number of arrests ranged from an average of 4.9 for WWE participants to 9.4 for SI.

Intervention Conditions and Costs
Interventions were administered at two community-based outreach centers shared with the St. Louis City Health Department. A peer facilitator -a female in drug recovery with at least one year's sobriety -conducted the SI, which included: (1) a 20-minute session for HIV pre-test counseling, blood collection, and administration of NIDA SI; [26] and (2) two weeks later, blood test results and HIV post-test counseling. The WWE intervention included an additional breast and pelvic examination with cervical cytological testing (Pap smear) provided by a nurse practitioner, who also obtained a short medical history. 4ES provided the additional opportunity of attending four educational sessions delivered by a peer facilitator paired with a health professional. The interactive sessions (1) were based on the Health Belief Model; (2) were focused on reducing unhealthy behaviors; and (3) employed a holistic approach, emphasizing health and nutrition, stress and coping, substance abuse, and HIV/AIDS.
The intervention costs employed in this analysis are based on a microcosting methodology, results of which are reported elsewhere [12].

Effectiveness Measures
For cocaine use, the outcome measures include: occasions used (number of days used * episodes per day, past 30 days); episodes per day (past 30 days); cocaine free days (past 30 days); participants with negative urine tests; proportion of participants abstaining (past 30 days); and proportion of participants abstaining (past 4 months). For alcohol consumption, we examined the following outcomes: total drinks (past 7 days); drinks per day of the days participants drank alcohol (past 7 days); proportion of participants avoiding heavy drinking (past 7 days; ''heavy drinking'' defined as $4 drinks per day on the days participants drank alcohol); [6] and proportion of participants abstaining (past 30 days).
We estimated incremental effects by a two-step difference-indifferences approach in order to best compare results across trial arms. First, we calculated the difference in effectiveness from 4 months to baseline (4 M) and 12 months to baseline (12 M) within the same intervention. Second, differences within groups were compared to subsequent interventions to obtain the difference in differences, or incremental effects. In cases where a smaller mean from baseline indicates an intervention's effectiveness, the incremental effects were calculated to reflect the greater relative improvement by converting the change to a positive value.

Incremental Cost-Effectiveness Analyses
We estimated the incremental cost-effectiveness of WWE compared to SI and 4ES compared to WWE by calculating the ratio of the difference in intervention costs (incremental costs) to the resulting incremental benefit (incremental effects). This calculation yields an incremental cost-effectiveness ratio (ICER), which provides the cost per obtaining an additional unit of outcome. An intervention is dominated if it is more costly and less effective than the alternative. Moreover, an intervention is extended dominated if the alternative is more costly, more effective and provides a lower ICER [13,27]. An incremental cost-effectiveness analysis is appropriate in this study because each subsequent intervention adds significant components.

Sensitivity Analyses
To examine the robustness of our findings and assess sampling variability, we conducted one-and two-way sensitivity analyses. One-way sensitivity analyses varied the costs and outcomes separately. Incremental costs of WWE were examined at $100, $200, $500 and $1,000 per participant. For 4ES, we evaluated alternative scenarios and their corresponding costs if: building rental/utilities decreased by 50% ($549) and 25% ($777); and, if each session had five people ($759) and three people ($1,233), rather than four [12]. We varied outcomes through statistically significant ranges of 5% intervals, determining when an intervention would become dominated or would no longer be dominated (''switching point''), depending on the base case. WWE and 4ES were also varied to achieve the following clinically significant levels: 12 and 15 occasions of cocaine use (''moderate to high'' use); [28] 1 and 0.5 episodes of cocaine use per day (an approximate range of ''moderate'' use); 3 weeks of cocaine abstinence; [29] 85% cocaine free days (15% days used); [30] and, one and four drink(s) per day on the days participants drank alcohol (a range for ''sporadic'' and ''heavy'' drinkers) [31]. In two-way sensitivity analyses we varied these parameters simultaneously.
We constructed acceptability curves as an alternative to confidence intervals to address uncertainty in the results [32,33,34]. Selecting randomly with replacement (''bootstrapping'') from the base case samples yielded 1,000 new samples. From these new outcome means, 1,000 ICERs for each outcome measure were calculated and plotted as a function of willingness to pay (WTP) per additional abstainer. We analyzed the abstinence results because they are the most clinically meaningful outcome. The indicator of cocaine abstainers for the past 30 days provides a better comparison to alcohol abstainers, for which we only have data for the past 30 days. We chose the comparisons between WWE and SI once it became clear they provided better ICERs for both outcomes. Table 1 presents treatment outcomes by treatment arm. Table 2 reports the mean differences and percentage change in outcome measures by treatment arm. Each intervention achieved improvements above pre-treatment levels across all outcomes. Tables 3 and 4 provide the incremental effects, indicating the improvements in Table 2 relative to the next least costly intervention. For cocaine abstainers, greater incremental gains were generally achieved at 4 M rather than 12 M. For example, both WWE and 4ES achieved incremental improvements at 4 M for cocaine abstainers in the past 30 days and past 4 months, whereas at 12 M only WWE for abstainers in the past 4 months achieved a positive incremental effect. This trend was not universal across all outcomes, however. WWE achieved better outcomes (Table 4) at 12 M than at 4 M for all alcohol outcomes except abstainers. The incremental costs per participant were $144 and $942 for WWE and 4ES, respectively [12].

Incremental Cost-Effectiveness Analyses: Cocaine Outcomes
Columns 7-11 of Table 3 report the incremental costeffectiveness ratios across all cocaine outcomes at 4 M and 12 M. Comparing WWE to SI at 4 M, the cost per additional abstainer (past 30 days) was $14,445, but WWE was dominated by SI at 12 M. In contrast, at 4 M the ICER for WWE compared to SI of abstainers in the past 4 months was $7,223, but only $3,611 at 12 M. To obtain an additional abstainer (past 4 months) at 4 M, WWE compared to SI cost $11,624 less than 4ES compared to WWE. Furthermore, an additional abstainer for the past 30 days at 4 M cost $4,401 less for WWE compared to SI than for 4ES compared to WWE.
The interventions generated lower ICERs, when not dominated, for additional occasion averted, episode reduced per day, and cocaine free day than for the abstinence outcomes. Moreover, 4ES compared to WWE had lower ICERs for the non-abstinence outcomes at 12 M than at 4 M.

Incremental Cost-Effectiveness Analyses: Alcohol Outcomes
Columns 6-9 of Table 4 report the incremental costeffectiveness ratios across all alcohol outcomes at 4 M and

Sensitivity Analyses
One-way sensitivity analyses. Varying the costs per participant changed the ICERs proportionally (Table S1) Varying the effectiveness parameters across statistically ( Table 5) and clinically significant ranges ( Table 6) indicated the sensitivity of each finding. Some outcomes required greater changes than others to achieve a switching point. For example, 4ES required a 115% effectiveness decrease at 12 M and a 30% decrease at 4 M in reducing occasions of cocaine use and a 50% decrease at 12 M and a 15% decrease at 4 M in reducing an episode of cocaine use per day to become dominated by WWE. If WWE obtained 10% fewer cocaine abstainers (past 4 months), it was dominated by SI at 4 M but was still cost-effective at 12 M ($16,050; Table S2) and did not become dominated by SI until its effectiveness was decreased by 15%. Moreover, 4ES remained cost-effective in reducing a drink per day at 4 M up to the point at which its effectiveness decreased by 50%, while WWE at 4 M required a 25% increase in effectiveness to eliminate SI domination.
Conversely, an increase or decrease of as low as 5% revealed a switching point for other cocaine and alcohol outcomes. For example, WWE was dominated by SI at 4 M and 12 M for cocaine free days in the base case; however, a 5% increase in      Two-way sensitivity analyses. We also determined the impact of more than one factor on results (Table S3).
For the cost per additional cocaine abstainer (past 30 days),4ES at 4 M yielded ICERs ranging from $5,490 to $12,330 at incremental costs of $549 to $1,233 per participant and a 10% increase in effectiveness. For cocaine free days, a 5% increase in effectiveness and cost range of $100 to $1,000 per participant for WWE compared to SI yielded an ICER range of $94 to $936 at 4 M and $136 to $1,364 at 12 M (base case: dominated). For alcohol outcomes, relative to SI, WWE would cost $51,282 per additional alcohol abstainer at 4 M if the effectiveness of WWE decreased by 5% and incremental costs increased to $1,000 per participant.
Threshold Analyses. Figure 1 illustrates changes in ICERs relative to changes in costs and effectiveness in a stepwise fashion. For cocaine episodes per day, WWE was dominated by SI with a 15% decrease in effectiveness for WWE, but yielded an ICER under $2,000 with an incremental cost of $100 and decrease in effectiveness of 10%. At incremental costs of $500, WWE was costeffective compared to SI at as large a decrease as 10% in effectiveness, yielded an ICER under $2,000 at base case effectiveness, and under $500 if WWE achieved a mean of 1 Abbreviations: ICER, incremental cost effectiveness ratio, which is the difference in cost divided by the difference in effectiveness as compared with the next least costly intervention and indicates cost per additional outcome achieved; A, SI intervention; B, WWE intervention; C, 4ES intervention. Table S2 includes complete results of ICERs at 5% increments en route to the switching points. a Percentage varied indicates percentage change of base case value for more costly alternative only, whereas base case of its comparator remain the same. A negative percentage change indicates less effective intervention than the base case, which may or may not indicate a mean lower than the base case (see Table 2). A positive percentage change indicates a more effective intervention than base case. Included in parentheses is the corresponding ICER. b Occasions = days used * times per day.   Figure 2 portrays acceptability curves that address the uncertainty of the results, indicating the probability that WWE is cost-effective when compared to SI at a variety of willingness to pay (WTP) values. For example, the WWE intervention has a 0.5 probability of being cost-effective if society's WTP is $20,000 per additional cocaine abstainer (past 30 days) or $2,600 per additional alcohol abstainer. If the WTP is $25,000, it has a roughly 0.72 probability of being cost-effective for alcohol abstainers. Figure 3 presents willingness to pay depending on intervention costs and changes in intervention effectiveness in achieving per additional cocaine abstainer and alcohol abstainer. For example, a 5% increase in effectiveness for 4ES (at a cost of $549) relative to WWE and for WWE (at a cost of $500) relative to SI yields a willingness to pay of less than $10,000 per additional cocaine abstainer and per additional alcohol abstainer, respectively.

Discussion
To our knowledge, this is the first cost-effectiveness analysis simultaneously examining cocaine and alcohol abuse in women. The outcome of preventing heavy drinkers proved highly sensitive to changes in effectiveness values. Moreover, the magnitude of an ICER in the base case did not always predict its sensitivity. Decreasing the effectiveness of WWE for alcohol abstainers by 5% and varying the costs from $100 to $1,000 yielded significantly different ICER ranges at 4 M ($5,128-$51,282) and at 12 M ($40,000-$400,000), despite comparable base case ICERs ($3,611 at 4 M and $7,223 at 12 M). In cases of dominance among cocaine outcomes, the sensitivity to switching points varied considerably. It took only a small improvement (5%) in cocaine abstainers (past 30 days) at 12 M for WWE compared to SI to achieve a low ICER ($8,755); yet the same percentage improvement for 4ES failed to eliminate domination by WWE, and doubling that improvement led to a significantly higher ICER ($78,525). Although SI dominated WWE in the base case for cocaine free days, increasing effectiveness of WWE by 5% and varying incremental costs from $100 to $1,000 yielded low ICERs ranging from $94 to $936 at 4 M and $136 to $1,364 at 12 M.
Our results yielded ICERs that compare favorably to those in the literature, although one needs to be cautious in comparing results across different studies. CEAs alongside randomized controlled trials that examine the cost per obtaining an additional cocaine or alcohol abstainer, aside from analyses of contingency management (CM) programs, are scarce if not non-existent. We identified one study that examined costs and treatment readmission rates [35] and one that looked at costs per abstainer from substance abuse [36], both drawing on calculated ratios of the cost per probability of success. Other studies examined the incremental cost of extending drug abstinence. To extend the longest duration of cocaine and opioid abstinence by an additional week, Olmstead and Petry [15] report costs of $212 and $166 for voucher-based and prize-based CM interventions, respectively, compared to standard outpatient treatment. To reduce an additional day of drug use among out-of-treatment substance abusers at risk for HIV, Zarkin et al. [21] measured ICERs ranging from $36 to $140 for an enhanced intervention compared to SI. Schumacher et al. [37] estimated costs of two addiction interventions compared to usual care to be $1,007 and $1,244 for an additional week of drug abstinence among homeless persons. For non-abstinence outcomes, Zarkin et al. [13] determined that medical management combined with either naltrexone or naltrexone and acamprosate was cost-effective in avoiding heavy drinking (using the same heavy drinking definition as ours), with ICERs of $2,847 and $8,095. Jofre-Bonet et al. [38] examined the cost-effectiveness of disulfiram in addition to methadone maintenance for treating cocaine dependence, yielding an incremental cost of $73 per day of cocaine reduced.
The cost-effectiveness literature on substance abuse contains comparable analyses for treating heroin abuse, where: 7-day abstinence costs under AUD$4,000 (US$2,520) per additional patient for three detoxification methods compared to conventional outpatient treatment, the latter of which was also dominated by buprenorphine; [39] in prison, methadone treatment compared to no treatment costs AUD$38 (US$22) per additional heroin free day; [40] buprenorphine was dominated by methadone mainte-nance for obtaining an additional heroin free day [41]. The cost to prevent a recent quitter from smoking (relapse prevented) among low-income pregnant women was $1,217 [42].
As a benchmark of comparison for determining the worth of a QALY for alcohol treatment in the UK, two studies [14,20] noted a range consistent with National Institute of Clinical Excellence (NICE) valuation of £20,000-30,000 ($37,768-56,652) per QALY gained. Additional studies reported the total benefits gained by society, determining the cost-effectiveness of: alcohol prevention [43] and a screening and brief intervention for excessive alcohol use in a primary care setting [16] in terms of QALYs; and acamprosate therapy in terms of cost savings per additional abstinent alcoholic [44].
We urge caution, however, not only in comparing our results to QALYs, but also in using QALYs as a measure for substance abuse prevention and treatment in an overarching social utility exercise (e.g. league table) because they overlook broader societal benefits and concerns [45]. As a single quantitative scale, QALYs assign less weight to the disabled and elderly in health care rationing and introduce problems of incommensurability [46,47]. Moreover, since incremental CEAs yield ICERs that are reflective of their comparators, a low ICER may either indicate an ineffective and/or costly comparator or highly effective and/or inexpensive intervention. Thus, it is important, as we did in this study, to compare interventions to clinically relevant comparators. Economic considerations should follow clinical input in the decision making process [46].
The need for further research is clear. In 2003, the U.S. spent $21 billion (1.3% of total health care spending) on treatment for substance use disorders, $4.83 billion (23%) of which was spent by Medicare and Medicaid [48]. By comparison, Medicare costs for end stage renal disease (ESRD), which affected over half a million patients in 2007 [49], were an estimated $20.8 billion in 2007 and $18.1 billion in 2004 [50] -close to spending on treatment for substance abuse, which affects a much larger population (about 23.1 million in 2008) [1]. ESRD is very expensive to treat, and the relatively small group of ESRD patients deserve medically necessary and appropriate care; in light of the greater population affected by substance abuse and its broad societal costs ($180.9 billion for drug abuse in 2002) [4], however, it is sensible to direct additional attention and resources toward identifying and implementing effective and efficient substance abuse treatments. Our study has several limitations. First, generalizability is limited since the study was conducted in St. Louis and targeted drug use and high-risk sexual behaviors in women. Second, substance abuse prevention and treatment positively impact housing and employment sectors [51], costs and frequency of crime [52,53,45], and the development of the children of users [54], suggesting our results underestimated the full economic impact of treatment. Third, the intervention contained components designed for HIV prevention, rather than exclusively focusing on substance abuse, which reflects important interactions but may not be as targeted as needed for policy and treatment purposes. Fourth, the analysis did not include the incremental differences between 4 months and 12 months, though that information is derivable from data presented. Fifth, our study did not measure the longest duration of abstinence for either cocaine or alcohol. Sixth, the effectiveness outcomes were measured by self report, and biochemical verification was not available for the relevant endpoints used in our CEA.

Supporting Information
Checklist S1 CONSORT Checklist.