Behavioral Avoidance Test for Obsessive Compulsive Disorder

BEHAVIORAL AVOIDANCE TEST FOR OBSESSIVE COMPULSIVE DISORDER

GAIL STEKETEE, ${ }^{1}$ DIANNE L. CHAMBLESS, ${ }^{2}$ GIAO Q. TRAN, ${ }^{2}$ HOPE WORDEN ${ }^{1}$ and MARTHA M. GILLIS ${ }^{2}$
${ }^{1}$ Boston University School of Social Work, 264 Bay State Road, Boston, MA 02215, U.S.A. and
${ }^{2}$ American University, Washington, DC 20016-8062, U.S.A.

(Received 1 December 1994)

Abstract

Summary-Few treatment outcome studies of Obsessive Compulsive Disorder (OCD) have employed Behavioral Avoidance Tests (BATs) to assess changes in symptomatology, probably because of the difficulty of constructing such tests for a disorder which has widely varying symptoms. The few studies that have examined the psychometric properties of BATs for OCD have found mixed evidence for validity but good treatment sensitivity. The present study presents psychometric findings for a multi-step/ multi-task BAT that assessed percentage of steps completed, subjective anxiety, global avoidance, and rituals. This measure was used with 50 clients diagnosed with OCD whose symptoms varied widely. The BAT demonstrated good convergent and divergent validity, as well as treatment sensitivity according to effect size calculations. A composite score combining steps, anxiety level, avoidance and rituals also performed well in psychometric tests. Strategies to reduce the complexity of scoring are presented, along with examples of several BAT tasks to enable researchers to employ this behavioral measure.

INTRODUCTION

The measurement of observable behavior has been a hallmark of assessment of treatment efficacy throughout the history of behavior therapy. The Behavioral Avoidance Test (BAT), a measure of observable avoidance behavior and associated self-reported anxiety levels, has been commonly used in assessing the effects of behavioral treatment for anxiety disorders, particularly for phobias and agoraphobia. However, the BAT has been used only rarely to assess outcome for Obsessive Compulsive Disorder (OCD), although a few investigators have made attempts (e.g. Foa, Steketee, Grayson, Turner & Latimer, 1984; Foa, Steketee & Milby, 1980; Philpott, 1975; Rachman, Hodgson & Marks, 1971). The reasons for its infrequent use for OCD most likely pertain to the difficulty of adapting BATs to this population. There is remarkable variability in the content of obsessive fears. Indeed, assessment of OCD has historically proved difficult for this very reason, and the validity of most measures of obsessive and compulsive symptoms has been challenged. Frustration with measurement difficulties had led only recently to the development of a new interview measure, the Yale-Brown Obsessive Compulsive Scale (YBOCS; Goodman, Price, Rasmussen, Mazure, Fleischmann, Hill, Heninger & Charney, 1989), which appears psychometrically sound (see Woody, Steketee & Chambless, 1995a). Several new self-report instruments also have been developed, such as the Padua Inventory (Sanavio, 1988; van Oppen, Hoekstra & Emmelkamp, 1995) and a self-report version of the YBOCS interview (Rosenfeld, Dar, Anderson, Kobak & Greist, 1992). In keeping with a multimethod approach to assessment (Michelson & Ascher, 1987), a valid measure is needed to assess observable behavior with regard to obsessive fear, avoidance and rituals in actual situations.

The methodology for conducting BATs before and after treatment has varied substantially from study to study. Most research on simple phobias has used a single task/multiple step method in which the researcher identifies a single feared object or situation and defines a series of steps to be taken in approaching it (e.g. Bernstein & Nietzel, 1973; Freund, 1986; Klieger, 1987; McGlynn & Barrios, 1978; McGlynn & Puhr, 1976; Trudel, 1978). More variation is evident in the types of behavior tests used in studies of agoraphobic $S \mathrm{~s}$. Multiple task/single step methods have required clients to perform a number of different feared tasks, each with only a single step (Barlow, 1988;

Craske, Rapee & Barlow, 1988; Mavissakalian, 1987). Others utilized multiple task/multiple step methods in which agoraphobics are asked to complete several different tasks, each with multiple steps (Chambless & Woody, 1990; deBeurs, Lange, VanDyck, Blonk & Koele, 1991; Mathews, Gelder & Johnston, 1981; Williams & Rappoport, 1983). Still others have combined the single task/multiple step and multiple task/single step methods in the same study (Mavissakalian & Hamann, 1986; Ost, 1990).

Findings from studies of clients with simple and agoraphobic disorders generally provide reasonably good evidence of test-retest reliability, and for the concurrent and criterion-related validity for BATs constructed by various methods (see Bernstein & Nietzel, 1973; deBeurs et al., 1991; Mavissakalian & Hamann, 1986; McGlynn, 1988; Trudel, 1978; Williams, 1985). The BAT appears to be a particularly useful measure of outcome because it can be tailored to individual patient symptoms and is highly sensitive to change (Barlow, 1988; Craske et al., 1988; McGlynn, 1988). Whether one of these BAT methods is more advantageous than another is difficult to determine because there is little evidence to argue for better reliability and validity of one over another.

In treatment research with OCD, a single task/multiple step method like that usually used for phobics was used by Foa and colleagues (Foa, Steketee & Grayson, 1985; Foa, Steketee, Grayson, Turner & Latimer, 1984) and by Rachman and associates (Rachman et al., 1971; Rachman, Marks & Hodgson, 1973). The latter investigators employed a BAT that assessed distance and duration of approach toward feared situations, and found nonsignificant correlations of this measure with self and assessor-rated anxiety and avoidance ( $r \mathrm{~s}=0.24-0.39$ ) and no relationship to scores on the Leyton Obsessional Inventory. Using data from Foa and colleagues’ research, Freund (1986) reported only on anxiety assessed via Subjective Units of Discomfort (SUDS) during steps of the BAT, rather than on actual avoidance behavior. She observed poor convergent validity of the BAT with target symptom ratings at pretest, with the exception of self-rated avoidance ( $r=0.53$ ), but good convergence with the same target ratings at posttest ( $r \mathrm{~s}=0.48-0.66$ ). These findings may be due to the severely restricted range for target symptoms at pretest, a problem that was not evident after treatment. Freund (1986) also reported mixed evidence for convergent validity with questionnaire measures of OCD symptoms that were much less restricted in range: Correlations with the Compulsive Activity Checklist and the Maudsley Obsessive Compulsive Inventory were 0.36 and 0.04 respectively, with SUDS levels from the BAT. Again, substantially higher correlations were evident at posttest ( $r \mathrm{~s}=0.62$ ). $T$ tests of BAT change at posttest and follow-up demonstrated treatment sensitivity (Freund, 1986), a finding also observed by Rachman et al. (1973), who noted that sensitivity to change of the BAT was comparable to other self-report and clinician measures.

Overall, then, the single task BAT showed modest concurrent validity and good evidence of treatment sensitivity. However, Freund reported only on subjective anxiety during tasks, rather than on actual avoidance behavior. Moreover, findings pertained only to OCD clients with contamination fears and washing rituals. This is because Foa and colleagues did not use the BAT for clients with other types of obsessions and rituals (checking, repeating), perhaps because of the difficulty of using the single task/multiple step method for more complex symptoms. In fact, Rachman et al. (1971) were unable to construct BATs for 2 of 10 patients in their study.

It may be that the complexity of the disorder should determine the BAT method, with disorders such as OCD and agoraphobia that have multiple manifestations requiring multiple tasks and multiple steps. The format employed should incorporate obsessive situations that tend to cluster together (e.g. obsessive fears of fire and associated checking of appliances, ashtrays, heaters, etc.; contamination fears and washing or cleaning rituals; magical ideas and associated repeating rituals; fears of losing things and related checking). Enough steps are needed within each task to accurately record the point at which $S \mathrm{~s}$ exhibit high levels of discomfort and avoid the task or resort to rituals. The present study provides evidence for the usefulness of a multiple task/multiple step method for constructing BATs applicable to OCD clients who exhibit a wide range of obsessions and compulsions. We first provide psychometric data regarding several measures derived from this method and then discuss strategies for streamlining these measures for research purposes, as well as some of the methodological complexities in designing such BATs.

METHOD

Participants

Data for the present study were collected at McLean Hospital in Belmont, MA, and at the American University in Washington, DC. Fifty individuals ( 34 women and 16 men) with a primary DSM-III-R diagnosis of OCD according to a Structured Clinical Interview for DSM (SCID-P; Spitzer, Williams, Gibbon & First, 1990) participated. All OCD clients engaged in overt rituals for at least 1 hr per day and were involved in an ongoing behavioral treatment study of 16 weeks duration in outpatient clinic settings. The majority were also receiving psychotropic medications. Clients were required to have been on a stable dose of any medications for 3 months before beginning treatment and to maintain that dosage throughout treatment. SCIDs were audiotaped and $22 \%$ of the interviews were blindly scored by a second rater. Kappa for OCD diagnosis was 1.0. Clients’ ethnic origin was $94 \%$ Caucasian, $4 \%$ African American, and $2 \%$ from other backgrounds. Mean age was 33.9 yr (range 17-62) and mean symptom duration was 20.3 yr (range $6-48)$.

Measures

A Behavioral Avoidance Test was administered by project therapists before treatment and by research assistants after treatment. BAT tasks were chosen after two sessions of information gathering to enable the therapist and client to identify all obsessive and compulsive symptoms that were troubling. The therapist and client jointly selected three tasks they deemed difficult or impossible for the client to do without significant anxiety or rituals, and for each task, they agreed upon up to 7 relevant steps. The number and type of steps were determined by the fear context and were intended to provoke steadily increasing levels of discomfort. Several examples are given in the Appendix. In some cases the nature of the task required fewer steps, but no less than 3 were permitted. Tasks could include periods of waiting without ritualizing, for example after leaving a setting without checking. Although BAT tasks were chosen from among the fear themes intended for treatment, whenever possible the steps were somewhat different than those used during the actual exposure treatment. Examples included approaching and touching the most contaminated object without washing, driving on progressively busier streets without checking, and walking by and eventually into a cemetery without repeating actions. Examples of several complete BATs are given in the Appendix and a manual for administration and scoring is available from the first author. To reduce demand characteristics of the test (Trudel, 1978), the therapist advised clients that the BATs were a test of their ability to approach feared obsessive situations as far as they could proceed comfortably without ritualizing, but that this was not a test of courage, and they were free to refuse all or any part of a task or to do steps only partially. Clients were told to inform the therapist when they wished to stop or to perform a ritual. In situations where the therapist could not be present to record information (e.g. bedtime rituals, work tasks), the client was to record the information as instructed and report back by phone. The following 5 BAT measures were calculated:

The percentage of steps (% Steps) completed at least partially was calculated (number of assigned steps completed divided by the total assigned steps, multiplied by 100). A composite % Steps score was generated by averaging across all three tasks. For the present study, steps of the tasks that involved periods of waiting without ritualizing were not included in this calculation, since they involved no active approach behaviors. Waiting steps were included in anxiety ratings below.

SUDS (Subjective Units of Disturbance) were determined by averaging the rated anxiety ( $0=$ none to $100=$ extreme) for all steps at least partially completed (including waiting periods) within a task. Scores were then averaged across all three tasks. SUDS were not included for steps that were avoided altogether. Note that although $S \mathrm{~s}$ were encouraged to approach only as far as they could proceed comfortably, many nonetheless attempted and completed steps while reporting high levels of anxiety (range $=3-100$ ). We note that this ability to engage in feared activities with high anxiety (typically followed by rituals to relieve the discomfort) is characteristic of many OCD sufferers.

Yoked SUDS were calculated for posttest BATs by determining average SUDS within a task only for those steps at posttest that had been attempted and rated for anxiety at pretest. That is, if a $S$ only completed 3 of 7 steps on a given task at pretest, only these 3 steps were used for calculation of yoked SUDS at posttest. Scores for each task were averaged to yield a total yoked SUDS score.

Avoidance for each task was rated by assigning an overall score as follows: $0=$ no avoidance; $1=$ partial avoidance/client failed some or part of some assigned steps; $2=$ complete avoidance of the entire task. Avoidance ratings were summed to yield a total avoidance score across all three tasks (range $0-6$ ). This measure was included to permit comparability with BATs for other populations such as agoraphobia. It was also intended to obtain an overall estimate of the clients’ reluctance to engage in the task. This measure was included in order to provide greater consistency across $S$ s compared to the $\%$ Steps measure. The latter contained substantial variability in: (1) the types of task assignments; (2) in the incremental difficulty of one step to the next; and (3) sometimes in number of steps assigned.

Rituals for each task were scored using a 3-point scale: $0=$ no rituals; $1=$ some but not extensive rituals; and $2=$ extensive rituals. If no steps were completed, ritualizing was not scored, since no rituals could be performed. Ritual scores were summed across the three tasks (range $0-6$ ), with missing data pro-rated from those tasks for which rituals could be rated. Because of changes in the procedure for conducting BATs during the study, only $30 S \mathrm{~s}$ had sufficient data on rituals to permit calculation of total ritual scores.

A composite BAT score was created by summing % Steps, SUDS, Avoidance and Rituals after dividing each variable by its standard deviation. When $S \mathrm{~s}$ were missing data on any one of these items, they were dropped from the analysis. This composite score captured in a single measure some of the various ways that performance might reflect OCD symptom severity. For example, at pretest, some clients chose to avoid most or all of the task, whereas others attempted most of the task steps but experienced very high anxiety and/or were unable to avoid ritualizing. In the composite, both kinds of difficulty would be expressed in one measure. Internal consistency of this measure was adequate at pretest (Cronbach’s $\alpha=0.64$ ), but somewhat lower at posttest ( $\alpha=0.49$ ). Note that high $\alpha \mathrm{s}$ are not to be expected for a 4 -item measure.

The Yale-Brown Obsessive Compulsive Scale (YBOCS; Goodman et al., 1989) was employed to assess the severity of obsessions and compulsions. This instrument has demonstrated generally good reliability and validity, with the exception of its lack of divergent validity with measures of depression (Goodman et al., 1989; Woody et al., 1995a). Interviewers recorded the presence of OC symptoms from a checklist and rated obsessions and compulsions separately on a 0 (none) to 4 (extreme) scale for time spent, interference, distress, resistance, and control. The scale was scored for Total score (range $0-40$ ), for Obsessions and Compulsions subscales (range $0-20$ ), and for Avoidance ( $0-4$ ). The latter is a new item not included in the total score. Interrater reliability for the YBOCS was 0.89 for Obsessions, 0.94 for Compulsions and 0.93 for the Total score (Woody et al., 1995a).

Target Symptoms reflecting the major idiosyncratic obsessions and compulsions were identified by the therapist and client during information-gathering sessions. Clients rated their fear and/or avoidance (Fear/Avoid) of up to three personally-relevant situations (e.g. situations that could be contaminated with HIV) and the frequency and/or duration of up to two rituals (Rituals) on 0 (no symptoms) to 8 (severe symptoms) scales. Scores for fear/avoidance and for rituals were calculated by averaging across situations. These scales have been widely used in behavioral research on OCD. Freund (1986) reported good interrater and test-retest reliability for target ratings of avoidance and rituals, with less reliability for fear. Test-retest reliability conducted over a 2 -week period for the present sample was mixed. The reliability of fear/avoidance was adequate ( $r=0.68$ ), but, unexpectedly, the reliability of the ritual rating was rather poor $(r=0.46)$.

The Maudsley Obsessional Compulsive Inventory (MOCI, Hodgson & Rachman, 1977) is a 30 -item true/false self-report questionnaire frequently used in treatment outcome research for OCD. This measure is internally consistent and has satisfactory test-retest reliability. It discriminated OCD patients well from those with other anxiety disorders and from anorexia, but not from those with depression (see Emmelkamp, 1988).

The depression subscale of the Symptom Checklist-90-Revised (SCL-90-R; Derogatis, 1977) was used to assess depressed mood in order to assess divergent validity. This subscale has shown good internal and test-retest reliability, and impressive evidence of convergent and criterion-related validity (Derogatis, 1983).

Because Obsessive Compulsive Personality Disorder (OCPD) has been commonly associated (or confused) with OCD, we included assessment of this Axis II diagnosis to examine the divergent validity of the BAT. The number of criteria met for OCPD from the SCID interview was included for further examination of divergent validity. Inter-rater reliability calculated for the present study was adequate ( $\rho=0.71, P<0.001, n=20$ ).

Procedure

Prior to treatment a trained research assistant interviewed all clients using the SCID interview and the YBOCS, and asked them to complete self-report measures that included the MOCI and SCL-90-R. Ratings of both target symptoms and the BAT were completed after two information-gathering sessions with the therapist. Over a 16 -week period, clients received 2 sessions of information gathering and 16 sessions of exposure and response prevention, tailored to clients’ OCD symptoms, followed by 4 sessions focused on maintenance of gains. After treatment, clients were again interviewed with the YBOCS and completed self-report and BAT measures.

RESULTS

To examine the interrelationship among BAT variables at pretest and posttest Spearman correlations were computed. Sample sizes vary because of missing data on individual variables, particularly for BAT rituals as noted earlier (and therefore for composite scores), and for posttest data since some $S \mathrm{~s}$ did not complete treatment. As Table 1 indicates, correlations among the variables were generally modest, suggesting that most of the BAT variables captured non-overlapping elements of behavioral performance. At pretest only % Steps was correlated substantially with Avoidance, but these variables were less strongly related at posttest. Some overlap between the two would be expected since both assess avoidance. The reverse was true for Rituals and Avoidance which were moderately correlated at pretest and strongly so after treatment. Note that since SUDS and Yoked SUDS are identical at pretest, the latter variable is not included in pretest calculations. Because these two variables were virtually identical at posttest, we omitted Yoked SUDS from the remainder of the analyses.

Convergent and divergent validity. Spearman correlations were used to examine convergent validity of pretreatment BAT variables with other pretreatment measures of OC symptoms (YBOCS Avoidance, Obsessions, Compulsions and Total scores; MOCI; Target Ratings of Fear/Avoidance and Rituals) and divergent validity with depressed mood (SCL-90-R) and other diagnostic criteria (OCPD). Table 2 presents these findings. Correlations of BAT variables with most measures of OC symptoms were significant and generally moderate in size (range from

Table 1. Spearman Correlations among BAT variables’ at pretest and posttest

${ }^{1} \%$ Steps $=$ percentage of task steps completed; SUDS $=$ Subjective Units of Discomfort $(0-100)$ averaged across task steps completed partially or fully; Yoked SUDS = average SUDS across task steps completed at pretest; Avoidance $=$ assessor’s rating of none ( 0 ), partial (1) or complete (2) avoidance of the task; Rituals $=$ assessor’s rating of no ( 0 ), some (1), or extensive (2) rituals occurring during the task.
Sample size is given in parentheses.
$+P<0.10 * P<0.05 * * P<0.01{ }^{* * *} P<0.001$.

Table 2. Spearman Correlations of pretreatment BAT variables with pretreatment measures of OCD symptoms, moodstate, and OC personality criteria

The range of sample sizes is given in parentheses for each BAT variable.
$\dagger P<0.10 * P<0.05 * * P<0.01$.
Composite $=\%$ Steps, +SUDS + Avoidance + Rituals. For each variable, scores were divided by the SD for that variable, and then summed to form the composite score.
approximately 0.30 to 0.40 ), with the YBOCS Avoidance item strongly associated with BAT SUDS and Rituals. A composite measure combining all 4 BAT variables correlated strongly with most YBOCS measures and the MOCI, but not with Target Ratings.

Because this low correlation could be a result of restriction of range of Target Rating scores to the upper half of the 9 -point range at pretest, we also examined the convergent validity of posttest BAT variables with posttest Target Ratings where scores spanned the range. Correlations were significant for % Steps and Target Rituals ( $r=0.37, P<0.04$ ) and for SUDS with Target Fear/Avoidance ( $r=0.54, P<0.001$ ); in addition, the correlation for BAT Avoidance and Target Rituals was marginally significant ( $r=0.32, P<0.08$ ). Other coefficients remained nonsignificant and lower than 0.30 . The composite BAT variable at posttest also correlated significantly with both Target Rating measures ( $r \mathrm{~s}=0.34-0.38, P<0.03$ ).

Relatively good divergent validity was evident for all BAT variables: Correlations with SCL-90 depression and criteria for OCPD were low and nonsignificant. Only the composite BAT variable correlated significantly with SCL-90 depression, and this correlation was lower than the composite’s concurrent validity coefficients.

Treatment sensitivity. $T$-tests were conducted comparing pre- and posttest scores for each of the BAT variables. Scores are presented in Table 3 and indicate the BAT is sensitive to the effects of treatment. Comparison of pre-post effect sizes from the BAT variables ( $d \mathrm{~s}=0.83-1.23$ ) with those from other measures of outcome indicated that the degree of change evident for the BAT was lower than that observed for the YBOCS Total (1.44) and for Target Ratings (2.19-2.32) but higher than the MOCI $(0.68)$.

To further study the relationship of change in BAT variables to change in other symptom measures, we calculated Spearman correlations for pretest/posttest residual gain scores for BAT variables and for YBOCS, MOCI, and Target Ratings. Residual gain scores control for the effects of initial severity by removing from change score calculations the correlation of pre- and posttest scores on the variable of interest (Manning & DuBois, 1962; Mintz, Luborsky & Christoph, 1979). Surprisingly, BAT residual gain scores for individual variables showed few significant correlations with residual gains scores on the interview and self-report variables; only 2 of 28 tests were clearly significant. However, residual gain for the BAT composite was consistently moderately correlated with residual gain for the YBOCS and questionnaire measures, $r$ s of $0.42-0.56$. All correlations were statistically significant except BAT Composite with the MOCI, for which a trend was observed, $P<0.08$.

Table 3. Comparison of scores for BAT variables at pre- and posttest

^[1]

DISCUSSION

Pretest individual and composite BAT variables exhibited generally good convergent validity with YBOCS symptom measures, which have demonstrated good to excellent psychometric properties in other research (for review see Woody et al., 1995a). Convergent validity with pretest MOCI and Target Ratings was less satisfactory, consistent with Freund’s (1986) findings in both cases. Given the unsatisfactory reliability of target ratings, high correlations were not expected. However, restriction of range in pretest scores probably contributed to the low correlations as well. Like Freund, we also found better convergence at posttest for the latter measure when range was not a problem. It is also possible that overlap between BAT steps and exposure tasks contributed to the higher concurrent validity at posttreatment. After exposure therapy, clients had considerably more experience in executing tasks similar to those posed in the BAT. Further, although we gave specific instructions that were designed to reduce the demand characteristics of the BAT task, undoubtedly these increased the performance of some clients, especially at pretest, thereby reducing correlations with pretreatment self-report measures of severity of OCD symptoms. It is noteworthy that, unlike individual BAT variables, the composite BAT showed reasonably good convergent validity with the MOCI. The strength of the correlations reported here (generally in the $0.30-0.50$ range) is typical of that observed in other monotrait/heteromethod psychometric research and suggests that the BAT is tapping similar underlying symptom constructs while also contributing unique variance.

Divergent validity was also quite good for BAT variables in relation to depressed mood and obsessive compulsive personality criteria: Only one correlation of the composite BAT and SCL-90 depression accounted for more than a small percentage of variance. This association of OCD symptom measures to depressed mood is quite common in the literature (e.g. Goodman et al., 1989; Wood, Steketee & Chambless, 1995a,b), presumably because of the frequent negative effects of severe anxiety symptoms on mood. However, arguing for its construct validity is the stronger association of the composite BAT score with other measures of OCD symptoms (except target ratings).

Substantial change representing a large effect size (Cohen, 1988) was evident following exposure and response prevention treatment for all BAT variables. The degree of change was noticeably less than for other variables (except for the MOCI), suggesting that the BAT provides a somewhat conservative measure of treatment effect. Surprisingly, residual gain scores on individual BAT measures were generally not related to residual gain on other outcome measures. Possibly some BAT components constituted a more severe test of change because of their focus on the most threatening tasks. Clients might show considerable improvement in obsessions and compulsions, but still have residual difficulty approaching hard tasks, and show strong anxiety in these contexts. This explanation also accords with the lower effect size of the BAT. Nonetheless, when BAT variables were combined into a composite, residual gain on this measure was moderately or strongly associated with gains on all but the MOCI. The different findings for the composite rather than the individual scores most probably results from the higher reliability generally observed for appropriately constructed composite variables and presents a strong argument for using the BAT composite score to assess outcome.

Based on the findings reported here, it appears that the measures collected during the BAT can be streamlined somewhat. The almost complete overlap of yoked SUDS to SUDS scored for all steps after treatment enabled us to eliminate this variable from further analyses. Practically, this represents a considerable saving of time for researchers, since calculations of yoked SUDS matched to only those steps completed at pretest requires substantial effort. We recommended that this variable be omitted in future use of the BAT. With regard to other BAT variables, the substantial overlap of the overall avoidance rating with $\%$ Steps especially at pretest suggests that scoring strategies for the BAT might be further reduced. When we compared the performance of a BAT composite with % Steps omitted (that is, SUDS + Avoidance + Rituals), convergent and divergent validity and internal consistency at pretest were very similar, the pre-post effect size changed minimally ( 0.80 vs 0.83 with the 4 -item composite), and the correlations of residual gain scores remained significant and of approximately the same strength ( $r \mathrm{~s}=0.44-0.56$ compared to $0.44-0.64$ ). Thus, it appears that a useful measure of behavioral avoidance could consist of ratings of avoidance, anxiety and rituals composited across steps and tasks. Given the difficulty in devising

comparable steps for a diverse group of clients, we recommend using steps to structure the task but not for scoring purposes.

The methodological complexities in designing BATs for clients with OCD symptoms are undeniable, but not insurmountable. We believe they are worth the effort to demonstrate behavioral change since they appear to capture some unique components of outcome. Moreover, BATs provide data that are not dependent solely on the clients’ self-report, as are both questionnaires and, ultimately, clinican-rated interview measures.* The Appendix provides several examples of how the therapists and clients in this study elected to construct BAT tasks and steps for different types of obsessions. Based on our experiences and findings from this study, we suggest that the activities selected for avoidance tests be ones that the client would be expected to engage in on a frequent basis and can therefore be reproduced on any given week after treatment. These should be characterized as active behaviors in which some exposure to fear-evoking situations is to be accomplished, rather than merely waiting tasks in which clients are asked not to do rituals. The latter are subject to distraction and perhaps other coping strategies which do not necessarily reflect the client’s typical behavior when confronted with a feared situation. Steps for the task should be arranged in the order expected to provoke increasing anxiety, although this is not always possible because the sequential nature of some activities includes elements that vary in discomfort level. For example, in the laundry task for Client B, the folding step was less disturbing than the handling of clean laundry from the dryer after touching dirty clothes. Practically, however, the folding step must follow the washing and drying ones. This variable rather than hierarchical progression in anxiety level for some tasks argues for the need to permit clients to omit some steps but attempt later ones. Likewise the need to average anxiety across all steps, rather than employ the highest anxiety level, seems essential to capture the degree of discomfort across the entire task.

At times we were unable to determine more than 3 steps for a task, as in the case of one woman for whom getting dressed in the morning provided avoidance of certain clothing and magical rituals when this was not possible. The BAT task for this case included: (1) picking out an outfit with green in it; (2) putting on the clothing; and (3) wearing the clothing. Once on, anxiety and urges to ritualize did not increase so further steps were not possible. This case also illustrates difficulty in repeating exactly the same task after treatment for some clients. Because her clothing obsessions were idiosyncratic to the thoughts that occurred to her on a given day (that is, the feared color changed, as did the type of clothing she wished to avoid), the BAT task had to remain flexible, but could be kept constant in that wearing some type of clothing was predictably difficult.

Clients vary considerably in the degree of generalization of their obsessions. For example, Task 3 for Case C in the Appendix illustrates a typically generalized checking situation in which the client checks a variety of situations prior to leaving home. In Case D, however, this client’s problem focused almost exclusively on checking the door lock, and therefore the BAT was constructed around this fear alone. Likewise, contamination fears and magical obsessions that require repeating rituals may be generalized across many contaminants or situations in which bad thoughts occur, or may be highly specific to certain contexts. For clients with a main theme (e.g. pesticide contamination, harm coming to husband) evident in multiple contexts (widely generalized), BATs steps will include a variety of situations pertinent to the theme. For those with less generalized fears, the steps will typically occur in the same situation.

Overall, our findings suggest that the use of a Behavioral Avoidance Test to assess the outcome of behaviorally or pharmacologically treated OCD clients is both feasible and useful. The 4 -item and 3 -item composite measures performed well psychometrically and contributed some unique variance to assessment of outcome.

Acknowledgements - We gratefully acknowledge assistance in this research from Sheila Woody, Cheryl Sheffler Rubenstein, Whitney Wykoff, Leslie Shapiro, and Elaine Williams. This research was supported by NIMH grant No. MH44190 awarded to the first two authors. Note that some of the findings from the present paper are also reported in Woody, Steketee, and

^[1]

Chambless (1995a, b) for a somewhat smaller sample; in those papers, findings are focused on the concurrent validity of the YBOCS and SCL-90-R with the BAT.

REFERENCES

Barlow, D. H. (1988). Anxiety and its disorders. New York: Guilford Press.
Bernstein, D. A. & Nietzel, M. T. (1973). Procedural avoidance in behavioral avoidance tests. Journal of Consulting and Clinical Psychology, 4, 165-174.
Chambless, D. L. & Woody, S. R. (1990). Is agoraphobia harder to treat? A comparison of agoraphobics and simple phobics and response treatment. Behaviour Research and Therapy, 28, 305-312.
Cohen, J. (1988). Statistical power analyses for the behavioral sciences. Hillsdale, NJ: LEA.
Craske, M. G., Rapee, R. M. & Barlow, D. H. (1988). The significance of panic-expectancy for individual patterns of avoidance. Behavior Therapy, 19, 577-592.
deBeurs, E., Lange, A., VanDyck, R., Blonk, R. & Koele, P. (1991). Behavioral assessment of avoidance in agoraphobics. Journal of Psychopathology and Behavioral Assessment, 13, 285-300.
Derogatis, L. R. (1977). SCL-90-R: Administration, scoring and procedures manual-I. Baltimore, MD: Clinical Psychometrics Research.
Derogatis, L. R. (1983). Description and bibliography for the SCL-90-R and other instruments of the psychopathology rating scale series. Baltimore, MD: Johns Hopkins Univ. Press.
Emmelkamp, P. M. G. (1988). Maudsley Obsessional Compulsive Inventory. In Hersen, M. & Bellack, A. S. (Eds), Dictionary of behavioral assessment techniques. Oxford: Pergamon Press.
Foa, E. B., Steketee, G. & Grayson, J. B. (1985). Imaginal and in vivo exposure: A comparison with obsessive-compulsive checkers. Behavior Therapy, 16, 292-302.
Foa, E. B., Steketee, G. & Milby, J. B. (1980). Differential effects of exposure and response prevention in obsessivecompulsive washers. Journal of Consulting and Clinical Psychology, 48, 71-79.
Foa, E. B., Steketee, G., Grayson, J. B., Turner, R. M. & Latimer, P. R. (1984). Deliberate exposure and blocking of obsessive-compulsive rituals: Immediate and long-term effects. Behavior Therapy, 15, 450-472.
Freund, B. (1986). Comparison of measures of obsessive-compulsive symptomatology. Doctoral Dissertation, Dissertation Information Service, Ann Arbor, MI.
Goodman, W. K., Price, L. H., Rasmussen, S. A., Mazure, C., Fleischmann, R. L., Hill, C. L., Heninger, G. R. & Charney, D. S. (1989). The Yale-Brown Obsessive Compulsive Scale-I. Development, use, and reliability. Archives of General Psychiatry, 46, 1006-1011.
Hodgson, R. J. & Rachman, S. J. (1977). Obsessional-compulsive complaints. Behaviour Research and Therapy, 15, 389-395.
Klieger, D. M. (1987). The snake anxiety questionnaire as a measure of ophidophobia. Educational and Psychological Measurement, 47, 449-459.
Manning, W. H. & DuBois, P. H. (1962). Correlational methods in research on human learning. Perceptual and Motor Skills, $15,287-321$.
Mathews, A. M., Gelder, M. G. & Johnston, D. W. (1981). Agoraphobia: Nature and treatment. New York: Guilford Press.
Mavissakalian, M. (1987). The relationship between panic, phobia, and anticipatory anxiety in agoraphobia. Behaviour Research and Therapy, 26, 235-240.
Mavissakalian, M. & Hamann, M. S. (1986). Assessment and significance of behavioral avoidance in agoraphobia. Journal of Psychopathology and Behavioral Assessment, 8, 317-327.
McGlynn, F. D. (1988). Behavioral avoidance tests. In Hersen, M. & Bellack, A. (Eds), Dictionary of behavioral assessment techniques. Oxford: Pergamon Press.
McGlynn, F. D. & Barrios, B. A. (1978). Psychophysiological response to presentation of a caged snake among behaviorally avoidant and non-avoidant college students. Journal of Clinical Psychology, 34, 313-319.
McGlynn, F. D. & Puhr, J. J. (1976). Heart rate responses to snakes among behaviorally avoidant and non-avoidant college students. Journal of Clinical Psychology, 32, 134-140.
Michelson, L. & Ascher, L. M. (1987). Anxiety and stress disorders: Cognitive-Behavioral assessment and treatment. New York: Guilford Press.
Mintz, J., Luborsky, L. & Crits-Cristoph, P. (1979). Measuring the outcomes of psychotherapy: Findings of the Penn Psychotherapy Project. Journal of Consulting and Clinical Psychology, 47, 319-334.
Öst, L. (1990). The Agoraphobia Scale: An evaluation of its reliability and validity. Behaviour Research and Therapy, 28, 323-329.
Philpott, R. (1975). Recent advances in the behavioural measurement of obsessional illness: Difficulties common to these and other measures. Scottish Medical Journal, 20, 33-46.
Rachman, S., Hodgson, R. & Marks, I. M. (1971). The treatment of chronic obsessive-compulsive neurosis. Behaviour Research and Therapy, 9, 237-247.
Rachman, S., Marks, I. M. & Hodgson, R. (1973). The treatment of obsessive-compulsive neurotics by modelling and flooding in vivo. Behaviour Research and Therapy, 11, 463.
Rosenfeld, R., Dar, R., Anderson, D., Kobak, K. A. & Greist, J. H. (1992). A computer-administered version of the Yale-Brown Obsessive Compulsive Scale. Psychological Assessment, 4, 329-332.
Sanavio, E. (1988). Obsessions and compulsions: The Padua Inventory. Behaviour Research and Therapy, 26, 169-177.
Spitzer, R., Williams, J., Gibbon, M. & First, M. (1990). Structured Clinical Interview for DSM-III-R-Patient Edition Version 1.0. Washington, DC: American Psychiatric Press.
Trudel, G. (1978). The effects of instructions, level of fear, duration of exposure and repeated measures on the Behavioral Avoidance Test. Behaviour Research and Therapy, 17, 113-118.
van Oppen, P., Hoekstra, R. J. & Emmelkamp, P. M. G. (1995). The structure of obsessive compulsive symptoms. Behaviour Research and Therapy, 33, 15-24.
Williams, S. L. (1985). On the nature and measurement of agoraphobia. In Hersen, M., Eisler, R. M. & Miller, P. M. Progress in behavior modification. New York: Academic Press.

Williams, S. L. & Rappoport, A. (1983). Cognitive treatment in the natural environment for agoraphobics. Behavior Therapy, 14, 299-313.
Woody, S., Steketee, G. & Chambless, D. L. (1995a). The reliability and validity of the Yale-Brown Obsessive Compulsive Scale. Behaviour Research and Therapy, 33, 597-605.
Woody, S., Steketee, G. & Chambless, D. L. (1995b). The usefulness of the Obsessive Compulsive Scale of the Symptom Checklist 90 -Revised. Behaviour Research and Therapy, 33, 607-611.

APPENDIX

Examples of Behavioral Avoidance Tests for OCD Clients

Case Example A illustrates three tasks for a client with multiple types of repeating rituals. These were focused on various situations in which she feared that her children or husband might come to harm because of a magical bad association with numbers, clothing, disastrous news events, and so forth.

Task 1: Media information

Step 1. Entering a store to buy the National Enquirer
Step 2. Picking up the Enquirer to buy
Step 3. Reading a bad article
Step 4. Looking for a news station on the radio
Step 5. Listening to news on the radio
Step 6. Watching a TV sitcom program for 10 minutes
Step 7. Watching a TV news program for 10 minutes

Task 2: Numbers

Step 1. Thinking a bad number ( 5 or 12 )
Step 2. Reading a card with the numbers 5 and 12 on it
Step 3. Writing down the numbers 5 and 12
Step 4. Saying the numbers 5 and 12
Step 5. Thinking about either child becoming age 12
Step 6. Saying “Susan is 12 years old”
Step 7. Saying the name “Donna” aloud (associated with a bad number)
Task 3: Items/possessions associated with death
Step 1. Think about looking at bad clothing
Step 2. Picking up a piece of bad clothing and putting it in a different spot
Step 3. Thinking of pants with a (particular store’s) label
Step 4. Reading the (specific store’s) label once only, not four times
Step 5. Not reading the label and putting pants back underneath other clothes
Step 6. Wearing the night shirt that was next to the purple shirt
Step 7. Putting purple shirt in drawer with other clothes and wearing one of them
Case Example B illustrates the BAT from a woman whose fears of contamination centered around a variety of “dirty” items at home and in public places, with the possibility of seriously becoming ill:

Task 1: Laundry

Step 1. Gathering the family’s dirty clothes together into a basket
Step 2. Carrying clothes to the basement
Step 3. Putting a load of underwear into the washer without shaking the clothes, looking for spots or wiping the washer
Step 4. Putting soap into the dispenser and leaving
Step 5. Transferring clothes to the dryer without checking clothes or wiping dryer
Step 6. Putting more dirty clothes into washer and immediately removing clean clothes from dryer without washing hands
Step 7. Folding clean clothes onto table without wiping table or checking clothes
Task 2: Money contamination
Step 1. Looking at money in wallet
Step 2. Touch new dollar bill
Step 3. Touch remote control
Step 4. Touch old dirty-looking dollar bill
Step 5. Touch remote control
Step 6. Touch telephone receiver
Step 7. Make a phone call while holding receiver
Task 3: Toilet in public restroom
Step 1. Drive car to Mobil gas station
Step 2. Get keys from attendant
Step 3. Handle doorknob and enter bathroom
Step 4. Touch toilet seat without using tissue or washing
Step 5. Sit on toilet seat without using seat cover
Step 6. Use toilet
Step 7. Flush toilet without kleenex or washing

Case Example C illustrates BAT situations selected for a man who suffered from multiple checking rituals, including fears that he would leave on faucets or appliances at home, causing a flood or fire, fears that he had hit a pedestrian while driving, and problems with hoarding mail and newspapers for fear of discarding important items.

Task 1: Driving past pedestrians at midday

Step 1. Enter car without circling around it
Step 2. Start motor and back car out of driveway by looking backward but not rechecking
Step 3. Drive down one street at speed limit without checking mirror or turning around
Step 4. Drive two blocks down Market Ave. at speed limit without checking
Step 5. Drive further through shopping area at speed limit without checking
Step 6. Turn right into parking lot across pedestrian walkway
Step 7. Back up in parking lot without extra checking and recross pedestrian walkway
Task 2: Throwing out mail and papers
Step 1. Place pile of unopened mail on table
Step 2. Open mail and separate items to be thrown out into pile
Step 3. Throw empty envelopes into kitchen trash
Step 4. Throw unwanted mail into kitchen trash
Step 5. Put weekly newspapers into recycle container
Step 6. Put kitchen trash bag into dumpster (unretrievable)
Step 7. Put recycle container out for curb pickup and leave house
Task 3: Leaving home in the morning without checking
Step 1. Make coffee and turn off coffee maker without checking
Step 2. Shower and turn off tub faucets without checking
Step 3. Brush teeth and turn off sink faucets without checking
Step 4. Using toilet without returning to check
Step 5. Boil water for eggs and turn off burner once only
Step 6. Exit the front door without checking the back door
Step 7. Turn key in lock without rechecking
Case Example D illustrates the use of a different hierarchy for a similar checking difficulty. This example is for a client for whom a main ritual concerned checking the apartment lock multiple times, often making her late to work. Accordingly, in this case the relevant BAT task included the steps leading to and from one situation, whereas for Case C’s Task 3, multiple situations were included as steps in the same task.

Task: Leaving apartment

Step 1. Assembling purse and items to take to work
Step 2. Shutting apartment door behind you
Step 3. Putting key in lock and locking door
Step 4. Walking down steps
Step 5. Getting into car
Step 6. Driving away
Step 7. Driving to work ( 15 min away)