Copyright 2001 by the American Psychological Association, Inc. 1040-3590/01/S5.00 DOI: 10.1037//1040-3590.13.2.189
Psychological Assessment 2001, Vol. 13, No. 2, 189-198
Assessing Readiness for Change in the Eating Disorders: The Psychometric Properties of the Readiness and Motivation Interview Josie Geller, Sarah J. Cockell, and Danae L. Drab St. Paul's Hospital Eating Disorders Clinic, and University of British Columbia This study examined the psychometric properties of the Readiness and Motivation Interview (RMI), a symptom-specific measure of readiness and motivation for change in the eating disorders. For 4 symptom domains, the RMI assesses the extent to which individuals are in precontemplation, contemplation, and action/maintenance, and the extent to which change is made for internal versus external reasons. Ninety-nine individuals with eating disorders completed the RMI and measures to assess convergent, divergent, and criterion validity. RMI profiles revealed differences in readiness and motivation across symptom domains. The RMI demonstrated good reliability and construct validity, and RMI scores predicted anticipated difficulty of recovery activities, completion of recovery activities, decision to enroll in an intensive symptom-reduction program, and treatment dropout. The RMI may have important clinical applications by providing much-needed information on client readiness for action-oriented treatment.
ing to prevent relapse). Stages of change have been measured using two self-report measures: a continuous measure that produces separate scores for each dimension (McConnaughy, DiClemente, Prochaska, & Velicer, 1989; McConnaughy, Prochaska, & .Velicer, 1983), and a discrete categorical measure that assesses stage of change from a series of mutually exclusive questions (DiClemente et al., 1991). Processes of change have been assessed using a self-report questionnaire. Principle component analyses have produced a 10-process model with two secondary factors. The cognitive-affective factor includes consciousness raising, dramatic relief, self-reevaluation, environmental reevaluation, and social liberation, and the behavioral factor includes stimulus control, helping relationships, counterconditioning, reinforcement management, and self-liberation (Prochaska et al., 1988). Extensive research among individuals who are attempting to modify addictive behaviors has supported the utility of the transtheoretical model (e.g., Prochaska et al., 1992). That is, a set of consistent relationships between categorical stage of change and recovery activities has been demonstrated. For instance, in a study comparing individuals in precontemplation, contemplation, and preparation with regard to smoking cessation (Fava, Velicer, & Prochaska, 1995) across all 10 processes, individuals in precontemplation reported using significantly fewer change strategies than did individuals in contemplation and preparation (univariate effect sizes ranged from .01 to .21). Stage and process of change have also been shown to predict clinical outcome variables (DiClemente et al., 1991; Prochaska, DiClemente, Velicer, Ginpil, & Norcross, 1985). For instance, in a study of 1,466 smokers, stage of change was related to quit attempts and indicators of cessation success at 6-month follow up (DiClemente et al., 1991). That is, 6 months following the initial assessment, the proportions of individuals in preparation who made a quit attempt and who reported not smoking at all were 80% and 21 %, respectively. In contrast, for individuals in precontemplation, the proportions of individuals falling into these two categories were 26% and 8%, respectively.
Individuals with eating disorders have been described as ambivalent about recovery, and increasing attention has focused on the valued function eating disorder symptoms perform in these individuals' lives (Crisp, 1980; Serpell, Treasure, Teasdale, & Sullivan, 1999; Szmukler & Tantam, 1984; Vitousek, DeViva, Slay, & Manke, 1995). It has been hypothesized that treatment recidivism and dropout, commonly observed in this population, may be due to programmatic attempts to bring about symptom reduction in individuals who are not yet ready to change. Given that individuals differ in their readiness for change and consequent engagement in treatment, a method for identifying readiness status would allow for a better understanding of clients' experience and for a tailoring of treatments to individual needs. A framework for conceptualizing readiness for change in treatment-resistant individuals is provided in the transtheoretical model of change (Prochaska, 1979; Prochaska, DiClemente, & Norcross, 1992). According to this model, change occurs along two interrelated dimensions: stage and process (Prochaska & DiClemente, 1992). Stage refers to an individual's readiness status at a particular moment in time, and process refers to what an individual is doing to work on the problem and bring about change (DiClemente & Prochaska, 1985; Prochaska, Velicer, DiClemente, & Fava, 1988). The stages of change include precontemplation (being unaware of or unwilling to change symptoms), contemplation (seriously thinking about change), preparation (having the intention of changing soon), action (actively modifying behavior and experiences to overcome a problem), and maintenance (work-
Josie Geller, Sarah J. Cockell, and Danae L. Drab, St. Paul's Hospital Eating Disorders Clinic, and University of British Columbia. This research was supported by a grant from the British Columbia Health Research Foundation. Correspondence concerning this article should be addressed to Josie Geller, Eating Disorders Clinic, St. Paul's Hospital, 1081 Burrard Street, Vancouver, British Columbia, V6Z1Y6 Canada. Electronic mail may be sent to
[email protected].
189
190
GELLER, COCKELL, AND DRAB
Similar to findings in the addictions literature, individuals with eating disorders have been shown to report varying degrees of ambivalence about recovery and to use change strategies that correspond to their stage of change (Stanton, Rebert, & Zinn, 1986; Blake, Turnbull, & Treasure, 1997; Ward, Troop, Todd, & Treasure, 1996). For example, in one study using hierarchical multiple regressions to investigate the relationship between process and stage scores in 35 inpatients, self-reevaluation scores were negatively associated with precontemplation scores, accounting for 33% of the variance. In the same study, consciousness raising and self-reevaluation were positively associated with contemplation scores, accounting for 58% of the variance, and counterconditioning and stimulus control were positively associated with action scores, accounting for 40% of the variance (Ward et al., 1996). Despite these preliminary findings supporting the utility of the transtheoretical model of change in individuals with eating disorders, to date, global stage of change has only inconsistently predicted actual clinical variables and treatment outcome in this population. For instance, although in one study individuals who were categorized in the action stage made greater improvements to their binge eating symptoms than did individuals in contemplation (Treasure et al., 1999), in other research, stage of change failed to predict clinical outcome (Levy, Lucks, & Pike, 1998; Pike, 1998). A number of possible explanations for this inconsistency exist. First, the extent to which measures developed for single-symptom problems (e.g., drug or alcohol abuse) capture the complexity of the multiple symptom domains of an eating disorder is questionable. The eating disorders are distinct in that the core diagnostic symptoms consist of several behavioral (e.g., purging, bingeing, restricting) and cognitive-affective (e.g., feelings of fatness, fear of weight gain) features. Given that individuals may be more willing to make changes to some symptoms than to others, assessment tools that capture individuals' experiences of different symptom areas are needed. Second, the difficulty of clearly articulating complex experiences about change may make self-report questionnaires ill-suited to identifying and measuring these constructs. Related to this difficulty is a lack of clarity in our own definitions and conceptualizations of motivation for change in the eating disorders. In this article, we define readiness and motivation for change as related but distinct constructs. Whereas motivation is more commonly understood as referring to an individual's desire and drive for change, the term readiness emphasizes that change occurs as a result of an individual's capability for change (i.e., the client has the skills to change) and faith that change both is possible and will produce a positive outcome. Together, then, we view motivation and readiness for change to be the product of desire, drive, capacity, and beliefs about the outcome of change. For instance, readiness and motivation to gain weight may require wanting to gain weight, having the skills to tolerate the distress and disorganization that occurs as a result, and having the conviction that life will improve following weight gain. The Readiness and Motivation Interview (RMI; Geller & Drab, 1999) is a new collaborative interview in which interviewer and client work together to determine readiness and motivation status for each symptom of an eating disorder. In conducting the RMI, the assessor's stance is of central importance. The RMI assessor expresses curiosity and interest about any ambivalence the client
may have about her1 symptoms and recognizes the importance of exploring conflicting feelings about recovery. In contrast to other clinical situations, in which clients may feel pressured to express readiness in order to receive approval or treatment, RMI assessors inform clients that their responses will not be shared with clinical team decision makers. The RMI provides ratings of the extent to which participants are in precontemplation, contemplation, and action/maintenance and establishes the extent to which change is occurring for internal versus external reasons. To avoid confusion resulting from overlapping categories, readiness and motivation is defined in the RMI as a finite entity that is divided into three categories, such that the precontemplation, contemplation, and action/maintenance ratings add up to 100%. In this way, participants explore their experiences of precontemplation, contemplation, and action/maintenance relative to one another, in the context of all of their feelings about change. Inclusion of the intemality rating is based on research showing an association between an internal locus of control and a number of clinical outcome variables, including length of sobriety (Sandoz, 1991), smoking quit attempts (Stuart, Borland, & McMurray, 1994), and compliance with behavioral contracting (Trice, 1990). In this research, we examine the psychometric properties of the RMI and its clinical utility. The goals are (a) to determine interrater and internal consistency reliabilities, (b) to examine the degree to which RMI scores converge with established measures of motivation for change (i.e., Stages of Change Questionnaire, McConnaughy, Prochaska, & Velicer, 1983; Processes of Change Questionnaire, Ward et al., 1996) and diverge from constructs that are theoretically unrelated to readiness for change (e.g., age, body mass index [BMI], socioeconomic status [SES], social desirability, eating disorder symptoms, psychiatric symptoms), and (c) to examine the degree to which RMI scores predict measures of behavioral and clinical outcome. We use four measures of outcome: a questionnaire measure of anticipated difficulty of completing recovery activities, self-reported completion of recovery activities in the week following the research assessment, commitment to an intensive symptom-reduction treatment program, and dropout from the program. We also compared descriptive and predictive information provided by the RMI with that provided by a standard global measure of stage of change (i.e., the Stages of Change Questionnaire).
Method Participants Ninety-nine women participated in this research as part of their standard intake assessment. Participants were selected from 112 consecutive new referrals to a metropolitan Canadian eating disorder clinic. Diagnoses were made by a clinical psychologist and four clinical psychology graduate students using the Eating Disorders Examination (EDE; Cooper & Fairburn, 1987). A diagnosis of anorexia nervosa or bulimia nervosa was made when all of the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (American Psychiatric Association, 1994), criteria were satisfied for a period of 3 months prior to being assessed, Subthreshold
1 Given that the prevalence of eating disorders is markedly higher among women, we use the personal pronoun her throughout this article.
READINESS AND MOTIVATION INTERVIEW diagnoses were made if one or more of the criteria for either of the disorders was not met. Any difficulties or ambiguous cases were discussed with the primary investigator (Josie Geller) in regular supervision meetings. Twelve individuals were excluded because they did not have a current or subthreshold diagnosis of anorexia or bulimia nervosa. Of those included in the study, 15 were rated as anorexia nervosa binge-purge subtype, 32 as anorexia nervosa restricting subtype, 41 as subthreshold anorexia nervosa, 7 as bulimia nervosa, and 4 as subthreshold bulimia nervosa. We included subthreshold cases in an effort to maximize variability in RMI scores. The mean age of participants was 25.7 years (SD = 8.9); mean SES was 2.0 (SD = 1.0) on the Hollingshead Index (Hollingshead, 1979), indicating middle class; and mean BMI was 18.2 (SD = 3.9). The mean duration of eating disorder in this sample was 9.25 years (SD = 8.94).
Convergent and Divergent Validity Measures RMI. The RMI is a semistructured interview that elicits information on individuals' readiness and motivation to change their eating disorder symptoms. To ensure that all of the relevant dimensions of an eating disorder are adequately sampled, the RMI relies on the established validity of the EDE diagnostic questions as a valid sample of the facets of an eating disorder. All of the diagnostic items in the EDE and one additional item (restraint over eating) are included in the RMI. This latter item is included because food restriction is a central aspect of anorexia nervosa and a key marker of recovery. The RMI questions are used in conjunction with each of the diagnostic questions from the EDE, so that both diagnostic information and motivational status can be obtained for each symptom. The RMI interview (including EDE diagnostic questions) takes between 30 and 50 min to administer. The diagnostic questions assess fear of weight gain, feelings of fatness, restraint over eating, maintenance of low weight, menstruation, importance of shape, importance of weight, objective bulimic episodes, dietary restriction outside bulimic episodes, self-induced vomiting, laxative misuse, diuretic misuse, and exercise. These symptoms are grouped into four categories: cognitive, restriction, bingeing, and compensatory strategies (see Geller & Drab, 1999). For each symptom, the RMI assesses readiness and motivation status and the extent to which change, when it occurs, is for internal versus external reasons. The RMI provides motivational stage scores (precontemplation, contemplation, action/maintenance, and intemality) for each of four symptom domains (restriction, cognitive symptoms, bingeing, and compensatory strategies) as well as global precontemplation, contemplation, action/ maintenance, and intemality scores. The global scores are calculated as the mean of the readiness ratings across all symptoms. The RMI provides a standard format for asking the motivationalreadiness questions. For each symptom (e.g., fear of weight gain), the frequency and severity is first determined using the EDE. If the symptom either is currently occurring or has occurred in the past, the assessor then asks the motivational-readiness questions. Specifically, the interviewer rates the extent to which the individual (a) is taking action to change the symptom or to maintain a change that has already been made (action/ maintenance rating), (b) is seriously thinking about changing the symptom (contemplation rating), and (c) either does not see the symptom as a problem or does not wish the symptom to change (precontemplation rating). When participants report taking action to change a symptom, the extent to which they are doing so for themselves versus for someone or something else is also determined (intemality rating). The order and format in which the readiness questions are asked for each symptom are determined by participants' answers to previous questions and are outlined in a flow chart provided in the RMI interview manual (Geller & Drab, 1999). For each motivational stage rating (precontemplation, contemplation, and action/maintenance), higher scores indicate a greater identification with the stage for that symptom. All ratings range from 1 to 5. For each rating, the interviewer has the option of coding the individual as being 0, 25, 50, 75, or 100% in each of the three stages. The three ratings add up to
191
100%. The intemality rating captures the locus of control for each symptom in which action is occurring. A rating of 1 corresponds to complete externality (all efforts to change are aimed at meeting other people's desires for improvement, or 0% of change efforts are for the self), and a rating of 5 corresponds to complete intemality (100% of change efforts are for the self). RMI assessors in this research went through the following training process: First, they reviewed key articles on readiness and motivation for change in eating disorders and substance abuse, including Geller and Drab's (1999) article, which provides a detailed account of the RMI and its administration. Trainees also received didactic instruction, listened to recordings of the RMI conducted by trained RMI assessors, participated in role plays, observed trained assessors conducting interviews, and administered the interview under supervision. For individuals who already possess basic clinical skills in the eating disorders, it is estimated that the total time to become trained on the RMI is 25-30 hr. In this study, regular interviewer meetings were held throughout the research study to minimize interviewer drift and to ensure that reliability standards were consistently met. Stages of Change Questionnaire (SCQ). This 32-item questionnaire (McConnaughy et al., 1983) operationally defines the four theoretical stages of change presented in the transtheoretical model of change. These stages include precontemplation, contemplation, action, and maintenance. The questionnaire has a 5-point Likert format in which a score of 1 indicates strong disagreement and a score of 5 indicates strong agreement. The psychometric properties of this tool are well established in the smoking literature (McConnaughy et al., 1983). Reliability of the subscale scores in the original sample of smokers (N = 155) was high (Cronbach's a = .88), with item loadings of 0.6 or above on the principal components analysis (McConnaughy et al., 1983). Although application of the SCQ is relatively new to the eating disorders, this questionnaire has been used in at least two other studies (Treasure et al., 1999; Ward et al., 1996), but reliability and validity data were not presented in the results. In the current study, minor modifications were made to make the measure applicable for individuals with eating disorders. Coefficient alphas ranged from .73 to .90. Processes of Change Questionnaire (PCQ). The PCQ (Ward et al., 1996) operationally defines the cognitive-affective and behavioral processes of change described in the transtheoretical model of change (Prochaska et al., 1988). Modifications to the original questionnaire were made for use with bulimia nervosa (Stanton et al., 1986) and anorexia nervosa (Ward et al., 1996), yielding a 47-item instrument in which 8 of the original 10 processes were deemed relevant. The questionnaire has a 5-point Likert format, measuring the extent to which participants are engaging in various cognitive-affective (consciousness raising, dramatic relief, and self-reevaluation) and behavioral (self-liberation, counterconditioning, stimulus control, helping relationships, reinforcement management) change activities. A not applicable option is attached to purging questions, which, if this option is endorsed, are not scored. The psychometric properties of this tool are well established in the smoking literature (Prochaska et al., 1988), with intercorrelations between the 10 processes in the low .30 range (indicating little overlap between the subscales) and internal consistency reliabilities ranging between .69 and .92, the majority being .80 or above. Although the factor structure has not been assessed in the eating disorders, coefficient alphas of .71 to .93, with a mean of .86 (Stanton et al., 1986), are reported for the eight subscales in the bulimia nervosa version. In the current sample, coefficient alphas ranged from .77 to .91. Eating Disorders Inventory—2 (EDI-2). This is a 91-item self-report questionnaire (Garner, 1991) designed to measure attitudes, personality features, and eating disorder symptoms thought to be relevant to anorexia nervosa and bulimia nervosa. Participants are asked to rate each item on a 6-point scale ranging from never to always. A factor analysis in a nonclinical sample revealed that three of the scales (Drive for Thinness,
192
GELLER, COCKELL, AND DRAB
Bulimia, and Body Dissatisfaction) loaded on the same factor and appeared to tap a general concern with shape, weight, and eating, and this factor was recommended as a screening measure of eating disorder symptoms (Welch, Hall, & Walkey, 1988). Accordingly, in this study, the sum of the Drive for Thinness, Body Dissatisfaction, and Bulimia subscales was used to measure severity of eating disorder symptomatology. Scores on this composite scale can range from 0 to 69. Extensive psychometric support for this instrument is available in the treatment manual. Brief Symptom Inventory (BS1). The BSI (Derogatis & Spencer, 1982) is a 53-item inventory of psychiatric symptoms. Respondents indicate the extent to which they are distressed by various problems on 5-point scales. The BSI yields nine primary symptom scale scores and three global indices of distress, including the global severity index (GSI). For the purposes of this study, only the GSI, the average distress experienced across all nine symptom dimensions, was used. The BSI has demonstrated good internal consistency, test-retest reliability, and construct and criterion-related validity (Derogatis & Spencer, 1982). Balanced Inventory of Desirable Responding (BIDR). The BIDR (Paulhus, 1994) is a 40-item measure of the tendency to respond in a socially desirable way. There are two subscales, with 20 items each. One subscale, Self-Deceptive Enhancement (SDE), reflects cognitive overconfidence. This subscale correlates with high Extraversion and low Neuroticism, suggesting that this style is common among energetic individuals who have a positive orientation to life. The other subscale, Impression Management (IM), reflects exaggerated social conventionality. This subscale correlates with Agreeableness and Conscientiousness, suggesting that this style is common among socially conventional and cautious individuals. This measure has good internal consistency (.65 to .75 for SDE, and .75 to .80 for IM), test-retest reliability (.69 for SDE, and .77 for IM), and construct validity (Paulhus, 1994). Because women with anorexia nervosa are more inclined to be conscientious and agreeable, as opposed to extraverted, only the IM items were administered to assess discriminant validity of the RMI.
Criterion Validity Measures Anticipated difficulty of recovery activities. Participants were provided with a list of 27 recovery activities and were asked to rate, on 11-point scales ranging from 0 (not at all difficult) to 10 (extremely difficult), their perception of how difficult it would be for them to engage in each recovery activity. In constructing the recovery activities, we reviewed cognitivebehavioral treatment manuals and identified commonly prescribed recovery goals. Items capturing a range of difficulty were included, and items that produced a restricted range of responses in pilot testing were eliminated. Examples of recovery activities include eating a meal (i.e., breakfast) each day, weighing oneself only once during the week, reducing exercise by 50%, and delaying vomiting by 5 min. For behaviors that were not relevant for all participants (i.e., bingeing and compensatory strategy activities), the option of not applicable was provided. The Anticipated Difficulty of Recovery Activities Scale was pilot tested to ensure that the items elicited a wide range of responses. The total difficulty score was computed as the mean of all applicable difficulty ratings and could therefore range from 0 to 10. Internal consistency was calculated at .87 for the entire set of items and .88 when items with a not applicable option were removed from the analysis. Completion of recovery activities. Participants were asked to attempt three of the recovery activities during the week following their assessment, one that they rated as somewhat difficult (rating of 3), one that was moderately difficult (rating of 6), and one that was very difficult (rating of 9). We made a priori decisions regarding behavioral task selection in an effort to ensure consistency across participants. Specifically, whenever possible, we assigned different types of activities (e.g., food restriction vs. exercise vs. purging) for the easy, moderate, and difficult goals. As well, if more than one activity was rated 3, 6, or 9, participants were instructed to select whichever goal they preferred. If no items were rated 3, 6, or 9,
activities were adjusted so that they would reflect these difficulty ratings. For instance, if eating one starch was rated a 4, and there were no other items rated a 3, then we would adjust the goal to a portion of starch that the client rated as a 3, for example a three-quarters portion of starch. Participants were encouraged to attempt the three goals in the week following the assessment and were informed that they would be contacted by a research assistant 1 week later to inquire about their progress. Consistent with the frame of the entire research assessment, participants were reminded that the focus of the research was on understanding the clients' experience and that there were no expectations with regard to task completion. Clients reported the outcome in a follow-up phone call. Each activity was scored as 0 (not completed), 1 (partially completed), or 2 (fully completed), yielding a total behavioral task completion score that ranged from 0 to 6. It is of note that data on the behavioral outcome task were not available for 20 participants. Half of these were not assigned the recovery activities, because they were assessed immediately prior to beginning the clinic day treatment program. This program requires adherence to activities similar to those described in the Anticipated Difficulty of Recovery Activities list, and we considered this a confound to completion of recovery activities. The other half were assigned recovery activities but could not be reached for the follow-up call. Decision to enroll in treatment. Participants' treatment decisions were tracked over time following the research assessment. On the basis of a review of charts, we determined that 64 participants were considered appropriate for and offered a space in the clinic's intensive 12-15-week day treatment program (clients have the option of requesting a 3-week extension), which requires a commitment to weight gain and/or normalization of eating. For participants to be offered a spot, their symptoms must be considered severe enough by clinic assessors to warrant intensive day treatment, they must not have medical complications requiring immediate attention, and they must meet all day treatment program inclusion criteria (e.g., BMI > 15). Thirty-three of those offered this treatment chose to enroll. The remaining 31 chose less intensive treatment options that do not require reducing symptoms, including weekly or biweekly group or individual therapy, follow-up visits with their family physician, or no treatment. Commitment to treatment was therefore coded as a dichotomous variable. Dropout from day treatment program. We are currently collecting dropout data from the intensive day treatment program. To date, 48 individuals2 who entered into the day treatment program have been followed over the course of their 12-15-week admission, and 37 have completed the treatment successfully. Participants were coded as dropped either if they decided to leave the program or if they were asked to leave because they did not meet program guidelines (e.g., weight gains, symptom reduction, abstinence from alcohol, drugs, and self-harm behaviors). Dropout was coded as a dichotomous variable. Three individuals were excluded from the analyses. One committed suicide after completing 11 weeks of the program, and 2 individuals left the program after requesting an extension, maintaining a healthy weight, and completing at least 11 weeks of the program.
Procedure Initial contact with participants was made by the research assistant prior to the participants' scheduled clinic intake assessment. The research assistant described the purpose and nature of the project, answered questions, and addressed concerns about participation. The voluntary nature of the study was made explicit. Appointments were scheduled at the participants' convenience.
2 This sample included all of the individuals who decided to enroll in treatment (n = 33), plus 15 additional individuals who participated in a larger treatment outcome study.
READINESS AND MOTIVATION INTERVIEW In the research assessment, participants received a verbal and written description of the study and provided written informed consent. In an effort to minimize the tendency to respond in a socially desirable way, we assured participants that information shared in the research assessment would not be communicated with members of the clinic team and would not affect treatment recommendations or care in any way. We also emphasized that the focus of the research was on understanding ambivalence about recovery and that we expected that participants might have a range of feelings about readiness for change. This information was also reiterated prior to the participants' completion of the RMI. Participants completed a packet of questionnaires that included a general information sheet, the SCQ, the PCQ, the EDI-2, the BSI, the impression management items of the BIDR, and the Anticipated Difficulty of Recovery Activities. After completing the questionnaire packet, participants completed the RMI. The research assistant then assigned three recovery activity goals from those listed in the Anticipated Difficulty of Recovery Activities and asked participants to attempt these in the coming week. A telephone interview was arranged for 1 week following the assessment.3 In the follow-up phone call, participants were again reminded that the goal of the research was to learn about their experience and to understand what barriers, if any, came up in attempting the recovery activities. Each goal was read aloud by the research assistant, and participants were asked to describe whether they were able to complete the goal entirely, partially, or not at all. Participants were then debriefed and thanked for their participation.
193
Internal consistency. Given that the RMI is a symptomspecific measure, calculation of coefficient alphas for readinessmotivation ratings per symptom domain would have been optimal. However, because respondents endorsed only a subset of the RMI items (i.e., those that were relevant to their diagnosis) and mean readiness-motivation ratings calculated for each of the four symptom domains were based on a different number of items for each individual, this was not pragmatically possible. We considered calculating internal consistency per stage of change a reasonable alternative. That is, mean precontemplation, contemplation, action/ maintenance, and internality ratings for each of the four symptom domains, based on the items endorsed for each individual, were used to calculate internal consistency. Because we expected readiness to change to vary across symptoms, we anticipated that internal consistency would be moderate. In the current study, coefficient alphas were as follows: precontemplation = .73, contemplation = .75, action = .86, and internality = .85. Because the sample on which these coefficients were computed consisted only of participants who endorsed all four symptom domains, (n = 41), to maximize the number of individuals included in the analysis, we recalculated coefficients, dropping bingeing items, as this was the domain for which there was the lowest number of endorsements. The resulting sample was larger (n = 98) and produced coefficients of .71, .63, .84, .69, respectively).
Results Item Analysis For each symptom assessed by the RMI, we examined the range and standard deviation of responses for the precontemplation, contemplation, action, and internality ratings. Of the 52 items, the full range of responses (1 to 5) were endorsed, and standard deviations were acceptable (greater than 1.0) for all but 4 of the items (objective bulimic episode precontemplation, objective bulimic episode internality, diuretic misuse internality, and exercise contemplation). No items were deleted because of restricted range. To examine the relative predictive ability of each RMI item with the criterion indices, we also performed item-by-item correlations with the two continuous criterion measures (anticipated difficulty of recovery activities and completion of recovery activities) and item-by-item t tests with the commitment to treatment variable, which is dichotomous. These analyses revealed some items to be better predictors of the criterion indices than were others. However, because the least predictive items were those that were endorsed less frequently (i.e., binge episodes, diuretic misuse), we consider it premature to drop items on the basis of these results until more data are available.
Reliability Interrater reliability. Participants were interviewed with the RMI by one of four trained interviewers. All of the interviews were audiotaped, and 16 of these (20%) were randomly selected and coded independently by an interviewer who was unaware of the original interviewer ratings. Agreement was scored when two raters gave the same rating for a symptom and stage of change. According to this strategy, interrater reliability coefficients for the stages of change subscales were as follows: precontemplation = 97.4%, contemplation = 95.6%, action/maintenance = 96.9%, and internality = 96.4%.
Description of Sample Readiness Using the SCQ and the RMI SCQ. On the SCQ, total scores on the Precontemplation, Contemplation, Action, and Maintenance scales were 14.64 (SD = 4.70), 35.49 (SD = 3.69), 31.18 (SD = 5.28), and 29.39 (SD = 5.80), respectively. A repeated measures ANOVA, F(3, 264) = 358.78, p < .001, with follow-up Tukey's honestly significant difference (HSD) test revealed that all SCQ total scores differed significantly from one another (p < .001), except for action and maintenance. That is, the relative magnitude of the SCQ scales was Contemplation > Action = Maintenance > Precontemplation. RMI. RMI mean readiness scores for each symptom domain are provided in Figure 1. For each of the four symptom domains (e.g., compensatory), we performed an ANOVA to compare the relative magnitude of precontemplation, contemplation, and action scores. As shown in Figure 1, different patterns of readiness emerged for different symptom types. The ANOVAs for restriction, compensatory, and binge symptoms were significant, F(2, 192) = 11.76, p < .001; F(2, 176) = 12.54, p < .001; F(2, 82) = 4.73, p < .05, respectively. Follow-up Tukey's HSD tests revealed that for restriction and compensatory strategies, precontemplation scores were significantly higher than contemplation and action scores were. In contrast, for bingeing, action was significantly higher than precontemplation was. For cognitive, no significant differences were detected between the three stages of change. Similar analyses were computed for the RMI total scores (averaging across all symptom ratings). The means and standard 3
As part of a larger study, participants also completed three questionnaires and responded to a qualitative interview during the follow-up phone call.
194
GELLER, COCKELL, AND DRAB
• Precontemplation D Contemplation Q Action
Compensatory
Restriction
Cognitive
Binge
Figure 1. Mean readiness scores for the four Readiness and Motivation Interview subscales.
deviations for the precontemplation, contemplation, and action scores were 2.62 (SD = 0.95), 2.18 (SD = 0.75), and 2.14 (SD = 1.03), respectively. The ANOVA comparing these scores was significant, F(2, 194) = 5.41, p < .01, and follow-up Tukey's HSD revealed that the highest global mean stage score was precontemplation, which was significantly higher than both contemplation and action/maintenance. Therefore, in contrast to the SCQ, in which the lowest stage score was precontemplation, in the RMI, precontemplation was the highest stage score. The RMI therefore provides a more conservative readiness profile than the SCQ does.
Convergent Validity Correlations between RMI subscale scores and the SCQ and PCQ are reported in Table 1. To protect from Type I error, we used an alpha level of .01 in interpreting significance. As can be seen from Table 1, the diagonal of correlations representing relation-
ships between the RMI and SCQ precontemplation and action scores were in the anticipated directions. Off-diagonal correlations between SCQ and RMI scores were also in the anticipated directions. RMI contemplation and RMI internality scores did not correlate significantly with any of the SCQ scores, suggesting that these subscales tap different aspects of readiness than the SCQ scores do. With regard to the relationships between the RMI and processes of change, precontemplation scores were negatively correlated with all cognitive-affective and three behavioral PCQ scores, contemplation scores were negatively correlated with four of the behavioral PCQ scores, and action scores were positively correlated with two cognitive-affective and four behavioral PCQ scores. Although power was significantly less for RMI internality scores, internality was significantly associated with two cognitiveaffective PCQ scores and two behavioral PCQ scores. Overall,
Table 1 Correlations Between Readiness and Motivation Interview (RMI) Scores and Stages of Change Questionnaire (SCQ) and Processes of Change Questionnaire (PCQ) Scores SCQ subscales RMI subscales Pre (n = 95) Con (n = 95) Action (n = 95) Internality (n = 66)
Pre
Con
.43** .05 -.43** -.22
-.49** .14 .33** .22
Action
PCQ subscales Main
-.52** -.46** -.04 .10 .52** .34** -.00 -.01
CR
SR
DR
Lib
SC
-.39** -.12 .45** .37**
-.34** .17 .18 .29
-.36** .01 .35** .41**
-.34** -.21 .52** .40**
-.31** -.29** .53** .16
Help
CC
.02 -.40** -.31** -.36** .68** .25 .34** .23
RM .03 -.38** .28** .08
Note. Pre = Precontemplation; Con = Contemplation; Main = Maintenance; CR = Consciousness Raising; SR = Self-Reevaluation; DR = Dramatic Relief; Lib = Self-Liberation; SC = Stimulus Control; Help = Helping Relationships; CC = Counterconditioning; RM = Reinforcement Management. **p < .01.
195
READINESS AND MOTIVATION INTERVIEW
precontemplation and action scores were the strongest predictors of both cognitive-affective and behavioral processes of change. Discriminant Validity Correlations between RMI Precontemplation, Contemplation, Action, and Internality scores were performed with age, BMI, SES, and the BIDR Impression Management score. Correlations were also performed with the EDI-2 composite score and the GSI of the BSI. For these analyses, to balance the need to protect from Type I error and also ensure the detection of meaningful relationships, we set alpha at .01 in reporting significance. RMI scores were unrelated to age, SES, BMI, or Impression Management. RMI Action scores were negatively related to the BSI global severity index (r = -.33, p < .01) and the EDI composite (r = -.33, p < .01) score, indicating that individuals who were actively engaged in change experienced fewer eating disorders and general psychiatric symptoms. In addition, RMI internality was negatively related to the BSI severity score (r = -36,p < .01), suggesting that when individuals make changes for themselves (as opposed to for others), they experience fewer psychiatric symptoms.
Recovery activities. Correlations between RMI scores and the two continuous criterion measures (anticipated difficulty of recovery activities and completion of recovery activities) are shown in Table 2. As can be seen from the table, RMI precontemplation and action scores were significantly related to both measures. Although power was a limitation with RMI internality ratings (valid internality scores were only available for 68 individuals who provided difficulty ratings and 53 individuals who were asked to perform the behavioral task), internality was nevertheless related to anticipated difficulty of recovery activities in the anticipated direction. That is, actively working on change for oneself, as opposed to for others, was associated with perceiving recovery activities as less difficult. We conducted a series of regression analyses comparing the RMI with the SCQ in accounting for variance in the two recovery activity scores. Separate regressions were performed for precontemplation, contemplation, and action scores, respectively. In each instance, the SCQ score was entered into the regression equation first, followed by the RMI score. The results of these analyses are reported in Table 3. With regard to anticipated difficulty of recovery activities, a consistent pattern emerged. For each stage of
Table 2 Correlations Between Readiness and Motivation Interview (RMI) Subscales and Criterion Measure Scores Anticipated difficulty of recovery activities
Precontemplation Contemplation Action Internality
* p < .01. ***p < .001.
.39***
.16 -.50*** -.43***
Completion of recovery activities
-.35* -.07 .32*
.21
F change
Variable
Anticipated difficulty of recovery activities Precontemplation SCQ .10 9.79** RMI .22 12.49*** Contemplation SCQ .09 8.23** RMI .13 4.14* Action SCQ .09 8.15** RMI .25 18.10***
.19 .37 -.33
.21 -.06 -.46
Completion of recovery activities Precontemplation
SCQ RMI
.09 .18
7.46** 8.15**
.06 .07
4.53* 0.51
-.08
.03 .11
2.14 6.22*
-.02 .33
-.19
-.32
Contemplation
SCQ RMI
.25
Action
SCQ RMI
Criterion Validity
RMI subscales
Table 3 Hierarchical Regression Analyses With Stages of Change (SCQ) and Readiness and Motivation Interview (RMI) Scores as Predictors of Criterion Measures
*p