Studies have documented the significant problems that alcohol and drugs cause for persons with severe and persistent mental illnesses, including elevated rates of homelessness, incarceration, suicide attempts, and other psychiatric symptoms (1) as well as termination of treatment, noncompliance, and increased hospitalization. Among such patients, Shaner and colleagues (2) found that both hospitalizations and elevated blood cocaine concentrations were more frequent at the beginning of the month, when disability benefits were received. By the middle of the month, both blood cocaine concentrations and admissions dropped off as patients ran out of money to buy cocaine. Other studies using large administrative databases have also found that both admissions and substance-related diagnoses were concentrated at the beginning of the month, when welfare or other benefits were received (3).
In an attempt to deal with exacerbation of symptoms and other benefit management problems, the Social Security Administration developed the representative payee program (4), through which a responsible other—often a family member—is appointed as the payee to be in charge of the ill person's finances to ensure that disability payments are used for healthy purposes. Representative payee arrangements have also become part of many assertive community treatment programs, with the treatment team acting as payee and using this function to retain and reward contact with the individual.
Research by Ries and Dyck (5) indicated that about two thirds of mental health centers in Washington State acted as representative payees for some of their patients. Co-occurrence of substance abuse or dependence with mental illness was one of the main factors in this arrangement. However, despite the fact that research has evaluated therapeutic interactional effects (6) and has documented positive outcomes (7,8) related to representative payee arrangements, no published randomized controlled trials have evaluated the clinical effectiveness of the payee intervention.
The study reported here was designed as a randomized controlled clinical trial of contingent versus noncontingent management of Supplemental Security Income (SSI) disability benefits in a sample of patients with both substance abuse and severe mental illness. Contingent management was designed to adjust the type or frequency (not amount) of disability benefits and payment for study participation based on evaluations of substance use, money management, and treatment follow-through. Our goals in this pilot study were to make the contingency intervention clinically practical and then evaluate both feasibility and outcomes (9).
The study was conducted from July 1999 to July 2001. The sample comprised 41 patients with severe and persistent mental illness who were attending an urban mental health center and who were all receiving mental disability payments through SSI or Social Security Disability Insurance (SSDI). These 41 patients were volunteers from a group of approximately 140 patients for whom the mental health center was the representative payee. The study sample was demographically representative of the larger group from which it was drawn. The psychiatric diagnoses of the sample were schizophrenia or schizoaffective disorder (30 patients, or 73 percent), major recurrent depression or bipolar disorder (ten patients, or 24 percent), and other diagnoses (one patient, or 2 percent). All patients had DSM-IV diagnoses of substance dependence with active substance use in the previous six months. The major drugs used by the participants were alcohol (25 patients, or 61 percent), marijuana (16 patients, or 39 percent), and cocaine (12 patients, or 29 percent); a large number of patients were polydrug users (16 patients, or 39 percent).
The study was approved by the institutional review board of the University of Washington in Seattle. The mental health center was the representative payee for all patients. To be included in the study, the patient had to provide written informed consent and his or her treatment team had to concur that the patient was able to safely participate in the study. In addition, a "safety net" category was developed and advocated for by the institutional review board, such that the clinical team could remove patients from the research payee management arrangement and return them to clinical management because of safety concerns but allow the patients to continue to provide data if they so chose.
The study design involved a three-month baseline period, after which patients were randomly assigned to have their benefits contingently or noncontingently managed for a 27-week period, stratified according to baseline substance use and blocked by case manager. Twenty-two patients were assigned to the contingently managed group and 19 to the noncontingently managed group. Stratification and blocking were used because of the relatively small number of patients and the limited number of case managers (N=12), which prevented an overrepresentation of patients with high levels of substance use in one of the intervention groups, or one case manager having all patients in one or the other intervention group. Fidelity recordings were used for monitoring purposes and to ensure that case managers followed their printed protocols.
After the participants had been randomly assigned to receive contingent or noncontingent management, their payee management included payment of their rent plus disbursement of the remaining benefits as cash at a frequency that was one category below the frequency used at baseline. For example, if the patient received benefits five times a week at baseline, the frequency was reduced to three times a week at randomization to avoid a ceiling effect of frequency of cash disbursement. The frequency categories were five times a week, three times a week, once each week, once every two weeks, and once a month. The frequency of benefits remained the same for patients with noncontingently managed benefits but was adjustable for those with contingently managed benefits.
The primary outcome measure was number of weeks of substance use during the 27-week period, defined by a positive weekly urine drug screen or a positive weekly case manager rating of drug or alcohol use. Each participant was monitored with use of a weekly research urine drug screen, to which the clinical team was blinded. Screening was usually done on Mondays (85 percent of all samples) or Tuesdays (14 percent of samples). Missing weekly substance use data (about 25 percent of potential evaluations) were recorded as substance use unless the patient was in the hospital or jail over the previous and current week. Patients were also assessed by their clinical case managers, who regularly spent time with the patients, apartment managers, family members, or other supports; made visits to the patients' homes; and made observations in the field. The relative strengths and weaknesses of case managers' ratings and weekly urine drug screens have been discussed in another report from this study (10).
A weekly payee log form was also completed by case managers for all patients. The form captured weekly attendance, treatment follow-through, money management problems, substance use, and a set of adverse outcomes, such as homelessness, hospitalization, and incarceration. A copy of the form is available from the first author. Within the contingent management weekly paradigm, patients who received a weekly score of 0 on the "sum" component of the form (no substance use, adherence of 80 percent or greater, and money management of 80 percent or greater) received a weekly decision summary score of 0 and a "gold award" ($7 in cash and a motivational message congratulating them for their sobriety and ongoing participation). Those whose weekly sum on the form was 1 or 2 received a decision summary score of 1 and were given a "silver award" ($3 in cash and $4 worth of vouchers for local food as well as a motivational message praising their partial sobriety and continued participation). Those whose weekly sum was 3 or more received a decision summary score of 3 and were given a "red award" ($7 worth of food vouchers and a motivational message to keep trying and participating). All awards were designed by a recovering patient from the center who had a dual diagnosis and who was not participating in the study. The patients with noncontingent management received the same reward each week, regardless of their clinical status. The content of their reward was similar to the silver award.
Within the contingent management monthly paradigm, frequency of benefits was adjusted according to how well the patient had done during the previous month: those who did better were allowed fewer visits and higher cash-voucher ratios, whereas those who did worse were subject to more visits and lower cash-voucher ratios. Frequency was unchanged in the group with noncontingently managed benefits.
All statistical tests of differences between the two groups used the same model: a two-way analysis of covariance with cohort and group as fixed factors and baseline events and scores on the Social and Occupational Functioning Assessment Scale (SOFAS) as covariates.
Compared with the group whose benefits were noncontingently managed, the participants with contingent management used alcohol for significantly fewer weeks (F=7.4, df=1, 34, p<.01) and used both drugs and alcohol for significantly fewer weeks (F=5.26, df=1, 34, p<.05). Drug use declined in both groups from pre- to postrandomization but was not significantly different between groups. Although attendance and medication adherence did not vary by contingency condition (in both groups it started at and remained above 80 percent), the money management rating on the payee log was significantly better in the group with contingent management than in the group with noncontingent management (F=19.27, df=1, 34, p<.001). All five participants for whom the "safety net" was implemented—usually because they had spent money on drugs, leading to at-risk behavior—were in the group with noncontingent management. Adverse events (emergency department visits, hospitalization, incarceration, and danger to others) were significantly higher at baseline and continued to be higher in the group with contingent management throughout the study. However, it should be noted that adverse events were relatively infrequent in both groups—about a third of the rate estimated in the study by Shaner and colleagues (2).
The results of this pilot study must be viewed in light of a number of limitations. One was that the control and experimental payee interventions were more alike than different, which reduced the study's power to demonstrate a result. That is, all participants had the mental health center as their payee, all had their monthly rent paid by the payee, all had the remainder of their money broken into at least weekly disbursements, and all were clinically managed through their long-term case manager. Thus it is clear that a substantial portion of the interventional power of the total representative payee intervention was shared by both groups, leaving only a fraction of potential effectiveness to be influenced by contingencies around substance use, treatment adherence, and money management. This limitation likely explains why we found few differences in adverse outcomes. However, all five patients for whom the safety net was implemented were members of the group whose benefits were noncontingently managed. In terms of contingency management rules, changes in the weekly paradigm were invoked regularly for most patients, but we found that the monthly paradigm was rarely changed by case managers.
Despite the study's limitations in statistical power, we demonstrated that a relatively simple weekly contingency scheme applied to disability benefits could decrease the percentage of weeks in which alcohol and drugs were used and markedly improve money management ratings. Our findings may have application to mental health centers that are currently acting as representative payees or might be adapted by families or other payees. Furthermore, the weekly review of the items in the payee log form used by case managers and patients in this study appeared to have both clinical and documentary benefit in the management of this clinical population. Further controlled research in this area is warranted.
The authors thank Mark Snowden, M.D., M.P.H., for his helpful comments.
Dr. Ries, Dr. Srebnik, and Dr. Comtois are affiliated with the department of psychiatry at the University of Washington and with Harborview Mental Health Services in Seattle. Dr. Dyck and Dr. Short are with Washington Institute for Mental Illness: Research and Training at Washington State University in Spokane. Ms. Fisher is with Harborview Mental Health Services. Send correspondence to Dr. Ries at Harborview Medical Center, 359911, 326 Ninth Avenue, Seattle, Washington 98104 (e-mail, firstname.lastname@example.org).