Participation in a parent training program for methadone clients

Addictive Behaviors, Vol. 20, No. 1, pp. 117-125, 19% Copyright 0 1995Elsevier Science Ltd Printed in the USA. All rights reserved 0306-4603/95$9.SOt .OO

PARTICIPATION

RANDY R. GAINEY, Social Development

IN A PARENT TRAINING PROGRAM METHADONE CLIENTS

FOR

RICHARD F. CATALANO, KEVIN P. HAGGERTY, and MARILYN J. HOPPE

Research Group, School of Social Work, University of Washington.

Seattle

Abstract - Programs for drug abusers are plagued by high rates of dropout. Because of the strong relationship between longer treatment and positive outcome, researchers have begun to study individual and program-specific factors that influence premature termination of treatment. For the most part, these studies have focused on dichotomous measures of dropout or number of sessions attended. In this article, we extend this line of research in two ways. First, we develop and measure a number of indicators of treatment participation based on therapist ratings. Second, we develop a model of treatment participation that employs both individual and program-specific factors. The data show that tremendous variation in participation occurred even among those who attended a majority of sessions, which highlights the importance of obtaining more elaborate measures of treatment participation. The model predicting treatment participation suggests that initiation of heroin use later in life, continued use of marijuana, and behavioral indicators of motivation are the strongest predictors of program participation. Research and practical implications of the findings are discussed.

Treatment programs for drug abusers are plagued by high dropout rates (Catalan0 et al., 1988; DeLeon, Wexler, & Jainchill, 1982). Consistent empirical evidence indicates that continued use or relapse is associated with exposure, usually measured by number of sessions attended (Allison & Hubbard, 1985; DeLeon, Wexler, & Jainchill, 1982; Hubbard et al., 1989; McLellan, Luborsky, & Woody, 1983; Roffman et al., 1993; Simpson & Sells, 1982). Measures of treatment participation and compliance have not been adequately addressed. Clients may attend a relatively large number of sessions but not pay close attention, actively participate, or comply with the structure of the program. We argue that although program dropout and attendance are important, the addition of client participation measures (e.g., involvement in program activities) will provide a more reliable predictor of treatment outcome, especially among frequent attenders. Dropout also reduces the statistical power to detect experimental effects, hampering the assessment of treatment efficacy (Boruch & Gomez, 1977: Cook & Poole, 1982; Wells et al., 1994). Treatment participation data can increase statistical power in experimental designs, although this introduces its own limitations and biases. Cook and Poole (1982, p. 426) note that “even in well designed studies, persons can self-select into different levels of service receipt.” Because self-determined levels of treatment exposure are likely to be associated with individual characteristics such as interest or motivation, artificial biases in favor of the experimental condition may

This research was supported by grant 5 R01 DA05824-02 from the National Institute on Drug Abuse. An earlier draft of this article was presented at the Annual Meetings of the American Society of Criminology in New Orleans, Louisiana, November. 1992. Requests for reprints should be sent to Randy Gainey, Social Development Research Group. 146 North Canal, Suite 211, Seattle, WA 98103. 117

118

R. R. GAINEY

et al.

result from including measures of participation in experimental analyses (Mark, 1983). One approach to control for these biases is to develop models specifying the determinants of treatment exposure and participation and include these variables in regression equations predicting outcome. This approach, suggested by Mark (1983), has it roots in the econometric and sociological literature (Berk, 1983; Heckman, 1979), although we have not seen systematic use of such techniques in treatment outcome studies. The literature on treatment exposure and client participation in drug treatment is both theoretically and empirically underdeveloped, although research has identified a few consistent predictors of treatment exposure (Allison & Hubbard, 1985; Gainey et al., 1993; Mark, 1983; Siddel & Conway, 1988). Demographic characteristics, client motivation, drug history and continued drug use, and structural or programspecific factors are some of the general categories of variables associated with treatment attendance and dropout (for reviews see Allison & Hubbard, 1985; Baekeland & Lundwall, 1975; Siddel & Conway, 1988). Demographic predictors of treatment exposure include age, marital status, gender, and race/ethnicity (Allison & Hubbard, 1985; Baekelund & Lundwall, 1975). In general, little theoretical relevance is attributed to these variables, although Baekeland and Lundwall’s review provides some insights. Client motivation would seem to be a theoretically important predictor of treatment exposure. In Baekeland and Lundwall’s (1975) review, 34 out of 41 studies indicated poor motivation as an important predictor of dropout. However, many of these studies relied on inferences made by treatment professionals or research staff. In fact, these authors note that “motivation has been rightly criticized as a diffuse, poorly defined and partly circular concept” (Baekeland & Lundwall, 1975, p. 767). More recent studies show that self-reported “willingness or intent” to attend has not predicted treatment retention (Miller, 1985). Behavioral indicators of internal motivation (e.g., previous treatment attempts or active attempts to participate in ancillary treatments) may be alternative measures (Gainey et al., 1993). Extent of drug use prior to treatment predicts relapse following treatment (Catalano et al., 1988) and premature termination of treatment (Gainey et al., 1993; Hser, Anglin, & Liu, 1990-1991). Drug-using lifestyles can offer social and economic incentives for behaviors that are inconsistent with treatment expectations and reduce client attendance. Studies of predictors of treatment exposure have largely ignored structural characteristics of the treatment program (Craig, Rogalski, & Veltri, 1982). Knowledge of program characteristics that promote or impede attendance may be directly useful to clinicians attempting to modify their programs to increase participation. The analyses presented in this article examine levels of participation in an experimental parent training program for clients receiving methadone treatment. The program, Focus on Families, seeks to improve the parenting and relapse prevention skills of these parents and reduce their children’s risk of drug abuse (Catalano, Haggerty, & Gainey, in press). First, we examine several measures of treatment exposure and participation to assess variation in treatment participation across levels of exposure. Second, we develop a model predicting program participation that includes both individual and structural characteristics. Results of these analyses should inform those concerned with treatment attrition from either a clinical or a research standpoint.

Methadone

METHODS

client

participation

AND

in parent

training

119

SAMPLE

Focus on Families is a 5-year field experiment to examine the effectiveness of parent skills training and case management services to reduce parent and child drug use. The program addresses risk factors for relapse among opiate addicts and family and school risk factors for drug abuse among their children. Clients were eligible for participation if they had children under 15 and had been enrolled for 90 days in one of two methadone clinics in the Seattle area. Those volunteering for the project had to agree to random assignment to (1) the parent skills program with case management plus continued methadone treatment or (2) methadone treatment only. Parent skills training was carried out in group sessions at the clinics twice weekly for 16 weeks, following a structured curriculum with flexibility to address each family’s needs. Parents practiced skills during the sessions and were provided with additional home practice activities. ’ Parent trainers kept records of group attendance and of individual levels of participation at each session. Client charmctrristics

Most subjects were white (83%) and female (73%). The average age was 35, and few of the clients had ever been married. Average age of opiate initiation was the later teens, although variation ranged from 10 to 38 years. Abstinence from illegal substance use is a criterion for methadone treatment, and subjects had been in treatment at least 90 days prior to the baseline interview. Nevertheless, a sizable proportion admitted to using such substances during the previous month. The most often cited illegal substances were marijuana, heroin. and other narcotics. Dependent

meusures

A variety of measures of treatment exposure and participation were collected. The first three variables are standard measures based solely on attendance, whereas the latter are based on parent trainer ratings following each of the group sessions. “Initiation” is a dichotomous variable indicating client attendance at one or more sessions. “Significant participation” is a dichotomous measure indicating attendance at 16 or more sessions (half of all sessions offered). “Total sessions” is the percent of the 32 sessions attended by the client. “Attention” is the percent of total sessions attended at which therapists assessed participants as “paid close attention” (opposed to “somewhat distracted” or “did not pay attention”). Parent trainers also rated the clients’ level of participation, on a five-point scale, at each of the parent is the percent of total sessions attended at which training sessions. “Participation” the client was rated as “participated/helpful” or “very active.” “Role plays” is the number of sessions in which clients practiced skills in role-play situations. Therapists also rated the overall performance of each client in role-play situations on a five-point Likert scale. “Effective role plays” was the average score the client is the percent of total homework assignments received on role plays. “Homework” the client completed. “Children” is the percent of appropriate sessions that the children attended (children were included in 12 sessions). ‘Families also receive 34 weeks of case management service\. Case managers work with the families in their homes to reduce risk factors for drug relapse by parents and drug abuse by their children. Information on exposure to case managers is being collected: only the parent training sessions are examined here. For more information on the Focus on Families intervention. see Catalano. Haggerty. and Gainey tin press).

R. R. GAINEY et al.

120

Independent

measures

We examined four sets of variables hypothesized to affect program participation: (1) motivation; (2) drug history, continued use, and self-efficacy regarding abstinence; (3) structural factors; and (4) demographic characteristics. We hypothesize that persons compliant with the research protocol have more interest and motivation to participate in the parent training sessions. A behavioral indicator of internal motivation was developed from the total number of contact attempts required for the data collection staff to complete the baseline interview. A dichotomous measure of previous treatment was included as a second behavioral indicator of motivation. In addition, a scale indicating self-reported motivation to stay off heroin was developed from the work of Curry, Wagner, and Grothaus (1990). The 13-item scale (alpha = .68) included reasons such as “to become a better parent, ” “so others will not be upset with me,” and “to feel in control of my life.” Drug use history and current use variables included age at first narcotic use, selfreported efficacy in resisting drugs following problem situations (alpha = .83), and two dichotomous measures indicating current use of marijuana and current use of other illegal drugs. We expected early initiation and continued use to be associated with greater drug dependence and less participation. Self-efficacy in terms of resisting drugs was expected to be positively associated with program participation. Because the program was delivered separately to seven groups, the possibility of group or compositional effects exists. Multivariate analyses of variance techniques (MANOVA) showed little variation across groups, with two notable exceptions. First, the racial mix at the two sites varied, in that there were no eligible African American clients at one of the treatment sites. A dichotomous variable indicating treatment location was included to control for this. Second, measures of client participation were exceptionally low in one of the cohorts. One client’s methadone dose was lost or stolen at the first session of this group, which caused internal conflict, loss of trust in the group, and consequently a severe lack of attendance by members of this group. A dichotomous variable was included to control for this particular cohort. Finally, several demographic characteristics were examined, including education,* gender (coded 1 for female, 0 for male), race (coded 1 for African American, 0 all others), marital status (coded 1 for currently married, 0 not currently married), and age (coded in years). Analysis

To examine variation in participation among high attenders, the first set of analyses organizes descriptive statistics of the measures of program participation by level of attendance (all assigned clients, those attending at least one session, and those attending at least half of the sessions). The second set of analyses uses bivariate correlation and multivariate regression techniques to develop a model predicting treatment participation. RESULTS

There was tremendous variation in participation across the measures of treatment exposure (Table 1). When all those assigned to the experimental condition were ZEducation was coded as follows: 1 = less than eighth grade, 2 = some high school, 3 = high school completion or GED, 4 = some trade or business school, 5 = some college, 6 = college graduation.

Methadone client participation

in parent training

121

Table 1. Levels of program participation: Means and standard deviations for all clients, initiators, and high attenders Percent of sessions N Total assigned Total initiating Completed 16+ sessions

Total initiating Completed l6+ sessions

a2 71 42

Attended 45.0 51.9 75.0

(34.5) (31.8) (15.8)

Paid attention

56.7 69.2

(30.3) (20.8)

N

‘Z Homework completed

71 42

45.7 64.3

(30.0) (20.4)

Actively participated 38.4 46.1

(27.0) (22.8)

# Role plays 10.0 14.1

(6.4) (4.2)

Children attended 32.2 46.2

(29.7) (27.9)

Mean rating of role plays ( l-5 scale) 3.0 3.5

(1.2) (4.2)

included, attendance averaged only 45% of the sessions. However, when the I I persons (13%) who never attended a single session were excluded, the average attendance (those who attended at least one session) was just over half (52%) of the sessions (range 3%-100%). On average, those attending at least one session paid close attention to 57% of the sessions (range O-100%) and participated actively in 38% (range O-94%). Among those attending at least half the sessions, we found more active participation and less distraction. These persons were rated as paying attention to an average of over two thirds of the sessions they attended and actively participating in just under half. The variation in participation found among high attenders may be more important than the average. Percent of sessions attended in which the high attenders were rated as “paying attention” ranged from 6% to 100%: sessions in which they participated actively ranged from 0 to 94%. These clients completed an average of 64% of their homework assignments, but the range was 13% to 100%. They brought their children to an average 46% of the 12 child-attended sessions, with a range of 0 to 100%. Of the 22 sessions in which role plays were practiced, high attenders came to an average of 64%, with a range of 18% to 100%. Therapist ratings of skill levels in the role-play situations ranged from 2.6 to 4.5 (scale of l-5), with an average of 3.5. These data reveal considerable variation in participation even among those who attended frcquently. Factor analyses suggest that the indicators reflect a single underlying dimension. Although each of the indicators was selected to represent unique aspects of treatment participation, the variables were highly intercorrelated (Pearson’s Y ranged from .56 to .94 among all subjects and .47 to .92 including just initiators). Analyses of only the high attenders suggest that there are two factors distinguishing attendance (including “children,” “ homework” and “roles”) from active participation (including “% pay attention,” “% participate,” and “average role-play score”). This may be an important distinction, the first factor representing treatment compliance (e.g.. bringing their children to appropriate sessions and completing homework) and the second representing active participation in parent training sessions. This subsample is small, however, and the results should be interpreted with caution. Because thccc measures of participation seemed to reflect a single dimension for the majority of

122

R. R. GAINEY et al.

cases, the items were standardized and averaged to compute a composite measure of participation (standardized alpha = .93). To develop a mode1 of treatment participation, we employed bivariate correlation and multivariate regression techniques (Table 2). Column 1 shows the bivariate relationship between each of the independent variables and the composite measure of program participation. African Americans and persons who reported continued use of marijuana in the month prior to the baseline interview were less likely to participate in the parent training groups. Motivation, as indicated by the number of contact attempts made by the data collection staff, was also negatively related to participation. Levels of participation did not vary across clinic locations, with the exception of the disrupted cohort, which had significantly less participation. Age of first use, self-efficacy, and being married were moderately correlated with treatment participation; self-reported motivation, use of illegal drugs other than marijuana during the month prior to parent program involvement, and previous drug treatment were unrelated to program participation. Because of the relatively small sample size, we could not include all of the variables of interest in a single regression equation. To build a model, we examined each block of variables separately (motivation, drug use, program variables, and demographic variables). We then included all variables significant at an alpha level of .I0 and removed insignificant variables to build a parsimonious model of treatment participation. The results are presented in columns 2-5 of Table 2. The block consisting of drug use history and recent drug use variables explained the most variance. This was largely a function of the influence of current use of marijuana and age at first narcotic use.

Table 2. Bivariate correlations

Program variables Disrupted cohort Clinic Demographics Education Female African American Married Age Drug use Use marijuana Other drug use Self-efficacy Age first use Motivation Contact attempts Self-reported motivation Previous treatment R square Adj. R square F statistic %ignificant at . 10. bSignificant at .05. cSignificant at .Ol.

and standardized r

(1)

-.25b .07

-.24b .02

.07 -.Ol -.32’ .20” -.02

.15 .06 -.3lC .I6 -.03

regression coefficients predicting total participation

(2)

(3)

(4)

-.I9

-.19”

-.27b - .07 .09 -.24b

- .26b -.I4 .19” .21” -.26b -.lO -.02 .06 .05 2.55”

(5)

.15 .lO 2.53b

.15 .lO 3.35s

- .23b -.37’ -.26b -.lO .02 .08 .04 2.17”

-.27b .32 .27 7.01’

Methadone

client participation in parent training

123

The final multivariate model, presented in column 6, shows that the strongest predictor of program participation is age at first narcotic use. Those who began their use of narcotics at an earlier age were less likely to participate. Marijuana use in the month prior to the parent training was also detrimental to program participation. The behavioral indicator of motivation (number of contact attempts) was strongly predictive of program participation, although self-reported motivation and previous treatment were unrelated to participation. The negative bivariate relationship between being African American and program participation was only marginally significant when other variables were controlled. It was explained in part by the high proportion, of African Americans present in the disrupted cohort (38%). These variables shared common variance (Y = .29, p < .OS)and both were only marginally associated with participation in the final model. African Americans in this study required more contact attempts to complete their baseline interview (v = .35, p < .05),which suggests that they may have been more difficult to reach and/or less motivated to participate in the project. The final model explains 32% of the variation in total participation. D I S C: U S S I 0 N

This research sought (1) to develop original measures of program participation and use them to explore variation in levels of participation among high attenders; and (2) to develop a model predicting program participation. We developed a number of indicators of program exposure and participation based on parent trainers’ observations. We found tremendous variation in levels of participation across levels of exposure, even among those who attended a relatively large number of sessions. We also found that the indicators of participation used in this study were highly intercorrelated and formed a single underlying dimension. The data presented here suggest the value of even more detailed observations of client participation, especially in situations in which attendance is high. To develop a model predicting total program participation, we tested the impact of client motivation, drug use history and continued drug use, demographic characteristics, and program variables. The strongest predictors of program participation were drug use history and continued use, which is consistent with the literature and with our clinical observations. This suggests that continued use of drugs impedes clients’ ability to focus on other issues, such as parenting, and should be addressed first. As expected, self-reported measures of client motivation did not predict treatment participation. However, the behavioral indicator of contact attempts for data collection was significantly associated with less participation. An implication of this is that problems in recruiting participants should be documented. Those who decline or participate minimally can be monitored and given additional encouragement or incentives, or aspects of the program may be modified to serve them better. There were no differences in levels of participation across clinic locations, with the exception of low participation in the one cohort that had interpersonal problems. This would have been difficult to explain without consistent and continued communication between the clinical and research staff, which highlights the power of collaborative clinical-research efforts and the need to collect data on group process. The relationship we found between race and participation is largely consistent with other research. Nonwhites and persons of lower socioeconomic status tend to drop out earlier than do white and middle-class participants (Baekelund & Lundwall,

R. R. GAINEY et al.

124

1975; Hser, Anglin, & Liu, 1990-1991). This statistical relationship, however, was significantly attenuated when structural factors (e.g., cohort) and motivation (e.g., contact attempts) were controlled. Two limitations of the study should be mentioned. First, these data come from a unique experimental program, and it is not clear to what extent the results can be generalized. This is a recurring problem in the research on program participation (Noel et al., 1987; Roffman et al., 1993). However, until large-scale studies focusing explicitly on participation are designed and implemented, many studies of unique programs must be reviewed to assess the generalizability of the findings. This requires explicit conceptualization and measurement of independent and dependent measures. Second, all of our measures were based on parent trainers’ observations. These are better measures than those found in current literature but could lack objectivity. Use of trained observers with no clinical contact would have been ideal. The results of this study are nevertheless an important contribution to the literature on program exposure and participation. The considerable variation in treatment participation found among attenders suggests that studies focusing exclusively on exposure, especially using dichotomous measures of dropout, ignore important factors. Although there is a consistent empirical relationship between length of drug program exposure and outcome, measures that tap variation in participation among high attenders may better predict outcomes. The results have implications for drug treatment clinicians interested in increasing participation among their clients and for program evaluators interested in measuring and predicting program participation. Continued research should incorporate a variety of measures of program participation in addition to program exposure; examine individual as well as structural factors that may affect total program participation; and continue to assess the impact of participation on individual outcomes.

REFERENCES Allison, M., & Hubbard, R. L. (1985). Drug abuse treatment process: A review of the literature. The International Journal of the Addictions, 20, 1321-1345. Baekeland, F., & Lundwall, L. (1975). Dropping out of treatment: A critical review. Psychological Bulletin, 82, 738-783.

Berk, R. A. (1983). An introduction

to sample selection bias in sociological data. American Sociological

Review, 48, 386-398.

Boruch, R. F., & Gomez, H. (1977). Sensitivity, Psychology,

bias, and theory in impact evaluations.

Professional

8, 41 l-434.

Catalano, R. F., Haggerty, K. P., & Gainey, R. R. (in press). Prevention approaches in methadone treatment settings: Children of drug abuse treatment clients. In W. J. Bukoski & Z. Amsel (Eds.), Drug abuse prevention:

Sourcebook

on strategies

and research.

Catalano, R. F., Howard, M. O., Hawkins, J. D., & Wells, E. A. (1988). Relapse determinants, and promising relapse prevention strategies. Incorporated Health, The health consequences of smoking: Nicotine addiction: A report Washington, DC: U.S. Government Printing Office. Cook, T. J., & Poole, W..K. (1982). Treatment implementation and statistical Evaluation

in the addictions: Rates, in Office of Smoking & of the Surgeon

General.

power: A research note.

Quarterly, 6, 425-430.

Craig, R. J., Rogalski, C., & Veltri, D. (1982). Predicting treatment dropouts from a drug abuse rehabilitation program. The International Journal of the Addicfions, 17, 641-653. Curry, S., Wagner, E. H., & Grothaus, L. C. (1990). Intrinsic and extrinsic motivation for smoking cessation. Journal of Consulting and Clinical Psychology, 58, 3 10-3 16. DeLeon, G., Wexler, H. K., 6%Jainchill, N. (1982). The therapeutic community: Success and improvement rates five years after treatment. The International Journal of the Addictions, 17, 603-747. Gainey, R. R., Wells, E. A., Hawkins, J. D., & Catalano, R. F. (1993). Predicting treatment retention among cocaine abusers. The Znternational Journal of the Addictions, 28, 487-505.

Methadone client participation

in parent training

125

Heckman, J. (1979). Sample selection bias as specification error. Econometrica, 45, 153-161. Hser, Y., Anglin, D. M.. & Liu, Y. (1990-1991). A survival analysis of gender and ethnic differences in responsiveness to methadone maintenance treatment. The InternutionuI Jownal of the Addictions. 25. 1295-1315. Hubbard, R. L., Marsden, M. E.. Rachal, J. V., Harwood, H. J.. Cavanaugh, E. R.. & Ginzburg, H. M. (1989). Drug abuse treutment: A nutionu/ study of effectiueness. Chapel Hill, NC: University of North Carolina Press. Mark. M. M. (1983). Treatment implementation. statistical power and Internal validity. EucrhcttionR+ view.

7, 543-549.

McLellan, A. T., Luborsky, L., & Woody, G. E:. (1983). Predicting response to alcohol and drug abuse treatments: Role of psychiatric severity. Archives ofGeneral Psyc.hiut~~, 40. 620-625. Miller. W. R. (1985). Motivation for treatment: A review with special emphasis on alcoholism. P.c~c~/zo/o~kul

Bulletin,

98, 84-107.

Noel, N. E.. McCrady, B. S.. Stout, R. L.. & Fisher-Nelson. H. (19871. Predictors of attrition from an outpatient alcoholism treatment program for couples. Jorrrncrl of.Sttrdies on Alcohol. 48, 229-235. Roffman. R. A., Klepsch, R., Wertz. J. S., Simpson. E. E., & Stephens. R. S. (1993). Predictor\ of attrition from an outpatient marijuana dependence counseling program. Addictiue Behuuiors. 18. 553-566. Siddel. J. W.. & Conway, G. L. (1988). Interactional variables associated with retention and succe\j in residential drug treatment. The InfernutionuI Journul ofthe Addictions. 23, 1241-1254. Simpson, D. D.. & Sells, S. B. (1982). Effectiveness of treatment for drug abuse: An overview of the DARP research program. Advances in Alcoholism and Substunce Abuse. 2, 7-29. Wells. E. A., Peterson, P. L., Gainey. R. R., Hawkins. J. D., & Catalano, R. F. (1994). Outpatient treatment for cocaine abuse: A controlled comparison of relapse prevention and Twelve Step approaches. Americun Journul of Drug und Alcohol Abu.te. 20. l-17.