        Executive Summary on Substance Abuse Treatment and Savings to state tax payers here. More facts on alcohol and other drug addiction |
 |
CHAPTER 5. EVALUATIONS OF ALCOHOL ABUSE TREATMENT 5.1 Introduction Our original goals for this chapter were straightforward. Based on a review of the existing literature, we hoped to learn the answers to the following questions: (1) What treatments are being used to treat alcoholic abuse? (2) Are these treatments effective? (3) What are the costs of these treatments? (4) Which treatments are the most cost-effective? (5) What are the benefit/cost ratios for these treatments? In addition to learning the answers to these questions for the population of alcohol abusers as a whole, we had hoped to learn which treatments are cost-effective for different subgroups of patients. Unfortunately, we were only able to find partial answers to some of the questions. The outline of the chapter is as follows. Section 5.2 discusses the major types of alcoholism treatments. In the next section, we discuss some important methodological problems involved in estimating the effectiveness of alcoholism treatments. In Section 5.4, we review the literature on the effectiveness of alcoholism treatments. While it is generally recognized that some treatments are more effective than others, at least for some types of alcoholics, a consensus has not been reached on the relative effectiveness of the different treatments. Section 5.5 explains the lack of cost-effectiveness studies of alcoholism treatments. While some information is available on the costs of treating alcoholism by setting or modality, and many researchers have investigated the relative effectiveness of different treatments, no one has estimated the cost-effectiveness of such treatments. A few studies have attempted to indirectly measure the cost effectiveness of alcoholism treatments. As we shall see, this research area is new and relatively unsettled. Some agreement, however, has been reached on the cost effectiveness of alcoholism treatments by settings (i.e., inpatient vs. outpatient) and we shall review this literature in Section 5.6. To date, no full-fledged cost/benefit analysis of alcoholism treatment has been published. There are, however, a growing number of partial cost/benefit studies called “cost-offset” studies of alcoholism treatments. Theses studies will be reviewed in Section 5.7. In the following section, we shall examine the studies of state alcoholism treatment programs cost-offsets. Section 5.9 summarizes our findings and conclusions on alcoholism treatment evaluations. 5.2 Alcohol Abuse Treatments Today alcoholism and alcohol abuse are acknowledge to be multifaceted medical, psychological, and social problems (Saxe 1983). The medical, psychological, and sociocultural views of the causes of alcoholism are associated with a number of alternative treatment approaches. Miller and Hester (1986) set forth nine major classes of alcoholism treatments, although they recognized that these methods frequently overlap or are combined. These include the following: (1) Pharmacotherapy The conception of alcoholism as a disease has fostered investigations of a large number of medications as potential therapeutic agents including: (a) antidipsotropic drugs, (b) psychotropic medications, and (c) hallucinogens. (2) Psychotherapy and Counseling Various types of counseling and psychotherapy have been proposed as appropriate for alcoholics. (3) Alcoholics Anonymous Some experts regard the AA as a non-professional self help group and not a treatment, per se. (4) Alcoholism Education Usually consists of a series of lectures, films, readings, or discussions on the topic of alcohol and alcoholism. (5) Marital and Family Therapy Recognizing that alcohol problems both influence and are influenced by the family, programs have increasingly included the spouse and other family members in the treatment process. (6) Aversion Therapies The aversion therapies have as their common goal the altering of an individual’s attraction for alcohol. Through counter-conditioning procedures, alcohol is paired with any of a variety of unpleasant experiences. If the conditioning is successful, the individual shows an automatic negative response when later exposed to alcohol alone. (7) Controlled Drinking Controlled drinking is not a treatment method, per se, but rather an outcome or goal of a treatment. Treatments are designed to teach moderate and non-problem drinking.
(8) Operant Methods Operant conditioning techniques alter behavior through modification of its consequences. With alcoholics, reinforcement and punishment contingencies have been used to influence drinking and drinking related behaviors. (9) Broad-Spectrum Approaches The premise of this approach is that drinking behavior is functionally related to other problems in the person’s life, and that an approach addressing this broader spectrum of problems is more effective than one that focuses on drinking alone. Under this approach individuals are provided social skills training, stress management, and other training. The community reinforcement approach (CRA) is included in the broad-spectrum approach. Based on their review of controlled outcome research, Miller and Hester concluded that the following treatment methods are effective in reducing alcoholics drinking: (1) aversion therapies, (2) behavioral self-control training, (3) community reinforcement approach, (4) marital and family therapy, (5) social skills training, and (6) stress management. Unfortunately, according to Miller and Hester these are not the treatments most widely used in America. The most widely used techniques are, in fact, not very effective. These include: (1) AA, which many experts do not regard as a treatment technique, (2) alcoholism education, (3) confrontation, (4) disulfiram, (5) group therapy, and (6) individual counseling. Miller and Hester do not attempt to explain why this is true. In a subsequent analysis Holder, Longabaugh, Miller and Rubonis (1991) devised a scheme to rank the relative effectiveness of 33 different treatment modalities in terms of abstinence and reduced use outcome. They reviewed the literature of controlled studies with drinking outcome measures to assess the cumulative evidence for effectiveness of specific modalities of treatment for alcohol abuse problems. Each treatment within a controlled study was classified as yielding either a positive or a negative finding. Positive findings were counted whenever a modality was observed to produce incremental effectiveness over (1) no treatment, (2) minimal alternative intervention, (3) a placebo intervention or (4) another treatment modality. In additive designs, a positive finding was counted when a treatment package including a specific modality improved outcomes relative to the same treatment package without the additional modality. Results were counted as negative when incremental effectiveness was not observed. They were not satisfied with a simple plus or minus frequency measure of positive or negative results, so they devised a weighted evidence index (WEIn). The WEIn was calculated by subtracting the number of negative (N) from the number of positive studies (P), then adding an extra point for each positive finding greater than two. Their WEIn rankings of treatment models grouped by the extent of evidence are presented in Table 5.2. The Holder et al. study has been criticized by Howard (1993) for failing to account for the fact that the treatment modules being compared were applied to persons with alcohol problems of widely varying severity. Also the length of the follow-up periods varied across the studies making comparisons difficult. Finally, Howard objected to the selection of studies reviewed and the interpretation of their results. Finney and Monahan (1996) extended the work of Holder, et al. (1991) by creating an alternative index to rank the relative effectiveness of alcoholism treatments. They examined 41 comparative treatment studies and determined whether or not each found at least one statistically significant positive effect on a drinking outcome variable for the modalities examined in a paired contrast with one another. Next they calculated the predicted probability of each study yielding at least one statistically significant treatment effect based on the number of tests for treatment effects conducted. Following that, for each treatment evaluated, the strength of the “weakest competitor” against which the modality had been compared was determined. For each modality, they used the average predicted probability of the relevant study finding a significant effect and the average
Table 5.2 Ranking of Alcoholism Treatment Modality Effectiveness
Source: Holder, Longabaugh, Miller, Rubonis (1991, p. 526). effectiveness of the weakest competitor to predict the modality’s effectiveness. Their Adjusted Effectiveness Index (AEIn) for each modality was the difference in predicted and actual effectiveness score. Table 5.3 compares Finney and Monahan’s. AEIn rankings of treatment effectiveness with Holder et al.’s rankings for 24 common treatments. Some of the same treatment modalities rank high on both indexes (e.g. social skills training, the community reinforcement approach, behavioral marital therapy and stress management training). These modalities were also rated highly by Miller and Hester (1986), so we can have some confidence in their relative effectiveness in reducing drinking problems. Some treatments are rated as relatively unaffected on both indexes (residential milieu treatment, general counseling, and metronidazole). The relative effectiveness of the other treatments continues to be in doubt. 5.5 Cost Effectiveness Analysis of Alcoholism Treatments As noted in Section 2.2, the central purpose of cost effectiveness analysis (CEA) is to compare the relative efficiency of different interventions (i.e., alcoholism treatments) in creating better outcomes. CEA can help government decision makers decide how to allocate their scarce resources across different treatment programs to get more value for their money. CEA analysis of treatment program components helps managers to redesign programs to improve its overall efficiency. For meaningful comparisons of alcoholism treatment CEAs to be made, the studies must use the same methodology, have the same analysts perspective, define and measure costs and outcomes similarly, and treat similar groups of patients.
Table 5.3 Comparative Rankings of Alcoholism Treatment Modality Effectiveness
Source: Finney and Monahan (1996, p. 239).
5.3 Methodological Problems in Evaluating Alcoholism Treatments This Section discusses nine major methodological problems the authors faced while trying to interpret the literature on evaluations of alcoholism treatments. The problems are discussed in no particular order. 5.3.1 Problem 1. Standardizing Treatment Protocols. As noted in Section 3.4, for meaningful economic evaluations to be made both the primary program and the alternative must be fully described in terms of who does what to whom, where, and how often. As of yet, there is little agreement among researchers or clinicians as to what the common active ingredients of alcoholism treatments are. Many studies only briefly describe the treatments provided in a general way. For our understanding of the effectiveness of alcoholism treatments to advance, a required next step would be the codification of procedures necessary to provide protocols for treatment implementation (Holder, Longabaugh, et al. (1991). 5.3.2 Problem 2. Standard Outcome Measures. Alcoholism treatment does not have a commonly accepted standard of effect or output measure. In many treatment situations, the treatment goal is abstinence, but there is no uniform agreement of what constitutes abstinence, how it should be reliably measured or over what time period it should be measured (Holder, Longabaugh, et al. 1991). It is not unusual for a study to report six to ten different measures of alcohol use, health care utilization or expenditure measures, and various social outcomes. Since the reported effectiveness of the treatment varies over these alternative measures, it is difficult to draw any conclusions on its overall effectiveness. 5.3.3 Problem 3. Patient Variation. Until very recently, research on alcoholism treatments assumed that alcoholics are all alike. In recent years, we have become aware that alcoholics are heterogeneous and that subgroups may differentially respond to different treatments (Holder, Longabaugh, et al. 1991). Researchers have begun tracking the types and amounts of treatments provided to different types of alcoholics to determine what works best for whom under the so-called “patient-treatment matching hypothesis”. The National Institute on Alcohol Abuse and Alcoholism (NIAAA) conducted research that addressed the “patient-treatment matching hypothesis” (PROJECT MATCH RESEARCH GROUP 1997). In this study, clients were randomly assigned to one of three 12-week, manual-guided, individually delivered treatments: Cognitive Behavioral Coping Skills Therapy, Motivational Enhancement Therapy or Twelve-Step Facilitation Therapy. Clients were then monitored over a 1-year post-treatment period. Individual differences in response to treatment were modeled as a latent growth process and evaluated for 10 primary matching variables and 16 contrasts specified a priori. The primary outcome measures were percent days abstinent and drinks per drinking day during the 1-year post-treatment period. Significant and sustained improvements in drinking outcomes were achieved from baseline to 1-year post-treatment by clients assigned to each of the three treatments. Importantly, there were no significant findings in 15 of the 16 matching hypotheses tested. The one significant finding was that clients with little or not psychopathology were more likely to maintain abstinence in the Twelve Step Facilitation treatments than in the Cognitive Behavioral Coping Skills Therapy. This is only a single study and more research needs to be conducted on the patient-treatment matching hypothesis in the future. In the interim, we need to be aware of differences in the client populations, when discussing the relative effectiveness of alcoholism treatments and settings. 5.3.4 Problem 4. Costs of Treatments. Treatment costs are typically reported in unit costs by the type of facility, setting, or provider. They are not broken down by the types of treatments provided and they are not reported on an individual client basis. It is assumed that all clients use the same resources and thus have the same costs of treatments. In fact, we know that some clients receive more treatments than others and outcomes should reflect this (see Section 3.5). It should not be surprising to learn that outcomes improve with the length of the treatment period.
5.3.5 Problem 5. Research Design Problem. As noted in Section 3.7, the preferred research design for determining the effectiveness of alcoholism treatments is a randomized clinical trial (RCT). Under an RCT, patients are randomly assigned to an “experimental group” which receives treatment services and to a “control group” which does not receive treatment services. The advantage of this research design, in comparison to a nonrandom design, is that it allows differences in outcomes to be attributed more confidently to the treatment, and not to pre-existing differences in the samples tested. Due to the random assignment process, the “experimental” and “control” groups of alcoholics should be nearly identical in terms of motivation (i.e., they both applied for treatment), severity of drinking problems, and other personal characteristics that might affect outcomes. Random assignment designs including “no-treatment” control groups have not been used in evaluations of alcoholism treatments for ethical reasons (Holder, 1987 and McLellan, Woody, and Metzger, 1996). Thus, as Holder (1987) concludes with respect to the effects of alcoholism treatments, “we have not had in practice the basis for determining directly what total health care costs would have been under a no-treatment condition.” There have been a number of evaluations of alcoholism treatments that utilize the treatment group as its own control group and time series pre-/post-treatment generated data to measure the effectiveness of the treatments. As we shall see in our discussion of the “ramp-up” effect and the “regression-to-the-mean” problem below, there are limits to the use of own control groups. Another group of studies use matched samples drawn from the non-alcoholic population as a control group. The outcome measures used to measure treatment effectiveness with this control group is health care utilization or health care expenditures rather than measures of alcohol use. From these studies, we learn that alcoholics spend considerably more money on health care than non-alcoholics in the period prior to treatment, whereas the ratio of health care spending of alcoholics to non-alcoholics is greatly reduced in the post-treatment period. Although this result is consistent with the argument that alcoholism treatments are effective, it is not conclusive. It is always possible that alcoholic patients would have reduced their health care expenditures over this same period of time without treatment. In fact, we shall see that such a reduction is likely when we discuss the problem of the “ramp up effect” and associated “regression-to-the-mean problem” presented below. These so-called “cost-offset” studies will be reviewed in Section 5.6. Finally, a number of studies have developed control groups of either “low to minimum” treatment to measure the effectiveness of different treatments (Holder 1987). Such studies may help us to understand the relative effectiveness of alcoholism treatments, but they do not measure the absolute effectiveness of such programs. Some treatments may work better for some types of patients than for others. To determine the absolute effectiveness of an alcoholism treatment, treated patients must be compared to patients randomly assigned to a no-treatment option as noted above. In comparative treatment studies, patients are sometimes randomly assigned to the different treatment groups and sometimes they are not. Random assignment is the preferred research design for measuring the relative effectiveness of alcoholism treatments. If patients are randomly assigned to treatment groups, it is likely that the treatment groups are similar in terms of motivation, the severity of the drinking problem and other personal characteristics that might affect outcomes. Thus, differences in the outcomes can be attributed to treatments rather than some confounding factors. The comparative treatment studies will be reviewed in the next section. Finally, it should be noted that the early alcoholism treatment literature included a large number of uncontrolled case studies and group designs. We have ignored these studies in our review because as Miller and Hester (1986) point out, positive uncontrolled reports can be found for virtually every treatment that has ever been tried for alcoholism. The observed effects of such studies may be attributable not only to the treatment offered, but to a host of confounding factors including patient selection criteria, expectancies, additional treatment components and post treatment factors. 5.3.6 Problem 6. The “Ramp Up Effect” and “Regression-to-the-Mean” Problem It has been observed that alcoholic’s health care costs tend to rise rapidly just prior to treatment (the so-called “ramp up effect”) and then fall rapidly following treatment. As Holder (1987, p. 66) explains, the difference between untreated alcoholics and non-alcoholics’ health care spending increases over time prior to alcoholism treatment. During the 25-36-month pretreatment period, the alcoholic on the average incurs costs that are about 130% higher than those incurred by comparable non-alcoholics. In the 13-24 month period before treatment, the alcoholic’s costs are around 180% higher, and in the last 12 months before treatment, the alcoholic’s costs are close to 300% higher than costs of comparable non-alcoholics. Most of this difference is attributable to inpatient utilization resulting from substantially higher inpatient days per month per person for alcoholics. Most studies show a statistically significant reduction of health care costs following treatment, usually in the first 12 months after treatment. Does this constitute proof of the effectiveness of alcoholism treatments? The answer is not necessarily. As Holder (1987) notes, the observed expenditure pattern can occur as a result of natural cyclic patterns or the random behavior of a time series. That is, it is natural, all other things equal, for a high level of a measure to be followed by a lower level or vice versa. The sharp up-and-down pattern of health care utilization and costs around the point in time when alcoholism treatment begins could be a consequence of this “regression-to-the-mean” phenomenon rather than the effects of treatments. We shall return to this point below. 5.3.7 Problem 7. Follow-Up Analysis Problems Outcome data are usually obtained through follow-up interviews with patients that have undergone treatments. There are a number of potential problems associated with such interviews. The first is the issue of who is included in the follow-up sample. It has long been recognized that a high rate of patient follow-up contact is necessary to ensure representative information from the treated sample (McLellan, Woody and Metzger, 1996). Studies have shown that the patients who are more difficult to find at follow-up typically have worse outcomes. For this reason, the Food and Drug Administration requires a minimum of 70 percent contact at follow-up in their studies. Many of the alcoholism treatment studies in the literature contain follow-up samples of far less than 70 percent. These studies are likely to overestimate the effects of treatment and therefore should be regarded critically. A second problem in evaluating alcoholism studies is that the time interval at which outcomes are compared also varies widely across studies. According to Holder, Longabraugh et al. (1991), follow-ups conducted shortly after treatment are more likely to indicate treatment effects than follow-ups conducted at more extended points. Holder (1987) recommends extended follow-ups as a way of minimizing the “regression-to-the-mean” problem. He says that in the typical study, which includes a 12 month pre- and 12 month post-period, differences may only be the result of “regression-to-the-mean” and not treatment. Extended pre-periods (24 months or longer) will reduce the “ramp-up effect” and extended post-periods (24 months or longer) will minimize the “regression-to-the-mean problem.” Unfortunately, as we shall see in Section 5.6, only a few studies have used such extended pre- and post-periods. 5.3.8 Problem 8. The Relapse Issue We know that a significant number of alcoholics who successfully complete treatments will relapse at some point in the future. Furthermore, many of these will reenter treatment at a later date. Because so many relapsed patients return to treatment, later follow-up evaluations of a single treatment episode may become contaminated by the effects of previous treatments (McLellan, Woody, and Metzger, 1996). As noted in Section 3.7, if first-time patients and readmitted patients use different amounts of treatments and have different expected outcomes, then treatment evaluations could be a function of their mix of patients. Most evaluations of alcoholism treatments ignore the issue of readmission so we have no idea what effect it has on the estimated effectiveness of treatments. Clearly the long-term effectiveness of treatments depends on future relapse rates. Undoubtedly, some treatments have longer-term effects than others. Extrapolating long-term outcomes from short (one year or less) follow-ups is impossible unless we model relapse rates and this has not been done to our knowledge. Most authors either ignore the relapse issue and simply note that the long-term benefits of treatments must be greater than the short-term benefits captured by the short follow-up period (unless, of course, the relapse rate is 100 percent). 5.3.9 Problem 9. Spontaneous Recovery As discussed in Section 3.8, an unknown number of treated alcoholics would have recovered spontaneously from their addiction without treatment, and therefore their benefit should not be counted in the effectiveness of treatment (Cartwright 1998). We have almost no data on alcoholics who do not seek treatments, so it is hard to judge the significance of the spontaneous recovery problem. A recent study by Estee and Nordlund (2001) of SSI recipients in the State of Washington sheds some light on this issue. In that study SSI recipients were placed in three separate groups for purposes of comparison. First, based on their medical diagnoses and procedures, receipt of alcohol or drug abuse (AOD) treatment, and arrests for drug- or alcohol-related offenses, individuals were placed into “need” and “do not need” treatment groups. The need treatment group was divided into two groups: those who were treated and those who remained untreated. The authors tracked the average monthly medical costs (including treatment costs) for the No Need, Treated, and Untreated groups in the pre- and post-identification periods. The identification period for the Treated and Untreated groups refers to the point in time when it first became known that the recipient needed AOD treatment. The length of the pre- and post-treatment identification periods varies for individuals in the Treated and Untreated groups depending on when they were identified as needing AOD treatment. The identification month for No Need recipients was arbitrarily set at the midpoint of their observation period. On average, the Treated group had 12.0 months in the pre-period and 25.6 months in the post identification period. The Untreated had an average of 15.1 months in the pre-period and 20.7 months in the post-period. The No Need group had an average of 18.5 months in both the pre- and post-period. Table 5.1 presents the average medical costs in pre- and post-identification periods for the three groups of SSI recipients. For both the treated and untreated groups, their costs after need for AOD treatment was identified were substantially higher than their medical costs before that seminal event. From the pre- to post-identification periods the average monthly medical costs rose from $387 to $740 for Treated recipients and from $648 to $1,445 for Untreated recipients. No Need recipients experienced only a small increase from $453 to $525 over the two periods. Since these figures are in constant December 2000 dollars, this increase suggests a growth in their medical expenses due to either general worsening of their medical conditions over time or rises in medical costs over and above inflation (measured by the CPI). The numbers reported in Table 5.1 are a bit misleading. Both the Treated and Untreated alcoholic groups experienced a sharp rise in medical costs right before they were identified as needing AOD treatments as shown in Figure 5.1. This sharp rise in medical costs is consistent with the so-called “ramp-up effect” discussed above. Figure 5.1 indicates that both the Treated and Untreated groups experienced a sharp decline in spending in the post-identification period almost to the pre-identification spending levels. The sharp decline in spending by the Untreated group cannot be attributed to treatments since they did not receive any. One interpretation of the data is that individuals in both the Treated and Untreated groups were more or less out of control
Table 5.1 Average Medical Costs in Pre- and Post-Event Periods for SSI Recipients July 1997 – December 2000a
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
