Therapist First Impression of Treatment Motivation Moderates the Relationship between Client-Rated Therapeutic Alliance and Drinking Outcomes During Treatment

Abstract

Introduction:

First impressions have been demonstrated to influence interpersonal relationships for extended periods, with negative first impressions leading to more negative judgments and behaviors between individuals months after initially meeting. Although common factors such as therapeutic alliance (TA) are well studied, less is known of the potential influence of a therapist’s first impression of their client’s motivation on TA and drinking outcomes. Based on data from a prospective study of clients’ perceptions of the TA receiving cognitive behavioral treatment (CBT), this study examined how therapists’ first impressions may moderate the relationship between client-rated TA and drinking outcomes during treatment.

Methods:

A total of 154 adults participated in a 12-week CBT and completed measures of TA and drinking behaviors following each treatment session. Additionally, therapists completed a measure of their first impression of their client’s motivation for treatment following the first session.

Results:

Time-lagged multilevel modeling revealed a significant within-person TA by therapist first impression interaction predicting percent days abstinent (PDA). Specifically, among participants rated as lower on first impressions of treatment motivation, higher within-person TA predicted greater percent days abstinent (PDA) in the interval prior to the next treatment session. Within-person WA was not associated with PDA among those rated higher on first impressions of treatment motivation who demonstrated higher PDA throughout treatment. Furthermore, significant between-person TA by first impressions interactions were found for both PDA and drinks per drinking day (DDD), such that among those with lower treatment motivation, TA positively predicted PDA and negatively predicted DDD.

Conclusion:

Although therapists’ first impressions of a client’s treatment motivation are positively associated with treatment outcomes, clients’ perception of the TA may mitigate the impact of poor first impressions. These findings further highlight the need for additional nuanced examinations of the relationship between TA and treatment outcomes, emphasizing the contextual factors that influence this relationship.

Introduction

Alcohol use disorder (AUD) is a debilitating psychological disorder that has a lifetime prevalence of 28-31% in the United States (Grant et al., 2015). In addition to significant negative behavioral and mental impact, AUD has been associated with higher rates of numerous negative health outcomes, highlighting the need for treatments aimed at reducing drinking and related negative outcomes (Verplaetse et al., 2021; Weil et al. 2018). Although treatments effective in reducing drinking have been identified for AUD (Magill & Ray, 2009), factors influencing the effectiveness of AUD treatments are still being examined. Previous research has suggested common factors of psychotherapy, such as client motivation and client-rated therapeutic alliance, are among the most salient predictors of positive outcomes in AUD treatment (Miller & Moyers, 2015), and the initial interaction between a client and therapist may have lingering effects on treatment success (Coleman, 2006). First impressions formed during initial social interactions have been demonstrated to influence behaviors and judgements of individuals long after their precipitation (Gunaydin, et al., 2017), though their influence on the therapeutic alliance and treatment success in clients with AUD has not yet been thoroughly examined. Investigating therapists’ first impressions is important for furthering the field’s understanding of the contextual factors that influence the therapeutic process and treatment outcomes in clients with AUD. The purpose of this study was to examine the influence of the therapist’s initial impression of their client’s motivation for treatment on the relationship between the client’s therapeutic alliance and treatment outcomes in clients seeking treatment for alcohol use disorder.

Common Factors in AUD Treatment

Psychotherapy has long been used as an effective treatment option for clients suffering from AUD (Magill & Ray, 2009; Lo Coco et al., 2019). Although numerous approaches and practices exist across different theoretical orientations, several mechanisms of treatment-related change are common to multiple treatment approaches (Weinberger, 1993). These “common factors” remain one of the most consistent predictors of positive outcomes across different AUD treatments and have been shown to be more predictive of positive treatment outcomes than specific treatment approaches (Imel et al., 2008; Wampold, 2015). For example, therapeutic alliance, a common factor summating the goal consensus, agreement on treatment plan, and interpersonal bond between a therapist and client (Bordin, 1979), is thought to enhance client engagement and retention in treatment, and serve as a model for improved relationships outside therapy, thereby improving treatment outcomes (see Miller & Moyers, 2015 for review). Indeed, a previous study by Cook and colleagues (2015) found that higher therapeutic alliance predicted higher percent days abstinent (PDA) over the course of treatment within clients seeking treatment for alcohol use disorder. Similarly, time-lagged regression analysis showed that higher session therapeutic alliance scores predicted lower numbers of drinking days and heavy drinking days over the next week before the next session (Connors et al., 2016), especially among clients who had lower rates of pre-treatment change. The overwhelming consensus on therapeutic alliance’s positive association with treatment outcomes highlights the importance of interpersonal factors and a therapist’s ability to adapt their treatment plans to fit their client’s needs.

Therapist characteristics, though examined less thoroughly compared to client-centered common factors, have also demonstrated a role in treatment success. Specifically, several therapist characteristics have been identified as influencing treatment outcomes, including expectancies, therapist allegiance to specific treatment approaches, interpersonal skills, and treatment fidelity (see Miller & Moyers, 2015 for review). For example, therapist empathy has been found consistently to be inversely associated with number of drinking days in clients with AUD (Miller & Moyers, 2015; Moyers et al., 2016), and the use of techniques such as open question asking and summarizing client statements have also been linked with improving treatment efficacy (Moyers et al., 2009). Importantly, therapist and client characteristics are expressed and likely interact at different timepoints throughout the treatment process, with therapeutic alliance often used as a metric for the quality of therapist-client interactions (Miller & Moyers, 2015). Thus, therapist characteristics may influence the client’s perception of the therapeutic alliance and contribute to the development and maintenance of a therapist’s relationship with their client. One factor that may be associated with therapeutic alliance and treatment outcomes that has received little attention in the psychotherapy literature is therapists’ first impressions of their client.

First Impressions Bias in Psychotherapy

First impressions, which begin forming within 100ms of seeing a person’s face for the first time, can be defined as the trait inferences someone makes about a person after receiving any information about them (Willis & Todorov, 2006). These initial impressions, which may derive from either an experienced interaction with a person or information received about that person from a third party, can bias further judgements and behaviors towards a person months after initially meeting. For example, Gunaydin and colleagues found that participants who “liked” a photographed person displayed increased warmth towards that person after meeting them for the first time, a full month after initially viewing their photo (Gunaydin, et al., 2017). The salience of first impression bias is additionally present in its observed effect on performance outcomes, known as the Pygmalion Effect (Jenner, 1990; Rosenthal, 1973). In a study of academic expectations and growth in elementary school students, a teacher’s expectations of a student at the beginning of the year modestly predicted a student’s grades in their class, regardless of the student’s actual ability as measured by a standardized exam (Jussim, 1989). These results suggest that a negative first impression may be detrimental to future interactions between individuals and may decrease the likelihood of performance success.

Though the research on a therapist’s first impressions of their client is sparse, prior evidence suggests that such impressions may be related to treatment success. An early study by Leake and King (1977) found that clients whose therapists were told had a high likelihood of recovery from alcohol use, regardless of symptom severity, were more likely to show signs of symptom improvement and were rated as more motivated for treatment by their therapist; conversely, clients not described as having a high likelihood of recovery were much less likely to show symptom improvement and were rated as far less motivated for treatment by their therapists. Further, Wild et al. (1998) found that therapists who perceived a client as being autonomously motivated for AUD treatment had better treatment outcomes than those who were not perceived as being autonomously motivated and had fewer negative attributions made about their treatment motives. Taken together, these findings suggest that a therapist’s first impression of their client may affect their attitudes toward them, which in turn may be related to aspects of the therapeutic alliance, including treatment collaboration effectiveness and the strength of the interpersonal bond.

Current Study

To date, no study has examined therapists’ first impression of their client’s motivation for treatment as a moderating factor on the relationship between therapeutic alliance and drinking outcomes in clients with AUD. The current study examined this relationship in therapist-client dyads completing a 12-week cognitive behavioral treatment for AUD. We hypothesized (Hypothesis 1) that both client-rated therapeutic alliance and percent days abstinent (PDA) are higher, and drinks per drinking day (DDD) lower, at the outset of treatment and change at a faster rate during the course of treatment among clients whose therapists rated them as having a higher motivation for treatment compared to those whose therapists rated them lower in motivation. Further, consistent with prior research suggesting that convergence between therapists and clients in therapeutic factors such as the therapeutic alliance is associated with better treatment outcomes (Laws et al., 2017; Zilcha-Mano et al., 2015), we hypothesized (Hypothesis 2) that a client whose therapist initially rated them as having a higher motivation for treatment will report higher PDA and lower DDD if they also have a higher client-rated therapeutic alliance score compared to therapist-client dyads whose client-rated therapeutic alliance score and therapist-rated initial motivation were discrepant, or when both scores were lower.

Materials and Methods

Participants

Participants were 175 individuals who were seeking outpatient treatment for AUD and were recruited using local newspaper and radio advertisements. Participants were recruited as part of a study examining the effect of providing therapists with client feedback on the course of therapeutic alliance and alcohol use outcomes in adult outpatients during a 12-session, individual CBT treatment for AUD (see Maisto et al., 2020). Inclusion criteria were as follows: (i) were between 18 and 65 years of age at the time of recruitment, (ii) met DSM-IV-TR (American Psychiatric Association, 2000) criteria for a diagnosis of alcohol dependence, (iii) resided within commuting distance of the program site, (iv) demonstrated basic reading skills sufficient to complete assessment materials, and (v) signed a consent form stating a willingness to participate in and complete all facets of the study. Exclusion criteria were as follows: (i) met criteria for a current psychotic disorder, (ii) presented with gross neurocognitive impairment, or (iii) had received alcohol or other drug treatment during the previous 12 months. The present analyses are based on the 154 participants who attended at least 2 treatment sessions (see Figure 1 for participant flow chart). Table 1 is a summary of participant demographic and descriptive information.

Figure 1.

CONSORT participant flow chart.

Table 1.

Descriptive Statistics for Overall Sample (n = 154)

Mean Age (SD)	48.67 (8.78)
Sex (% female)	32.5% (n=50)
Race
White/Caucasian	94.2% (n=145)
Black/African American	2.6% (n=4)
More than One Race or unknown	3.2% (n=5)
Ethnicity
Hispanic	1.9% (n=3)
Non-Hispanic	98.1% (n=145)
Marital Status
Single/Never Married	18.2% (n=28)
Married	46.8% (n=72)
Cohabiting/Living Together	5.2% (n=8)
Separated	8.4% (n=13)
Divorced	20.8% (n=32)
Widowed	0.6% (n=1)
Employment
Full-Time	55.2% (n=85)
Part-Time	11.7% (n=18)
Not Employed/Looking	13.6% (n=21)
Not Employed/Not Looking	6.5% (n=10)
Disabled	7.8% (n=12)
Retired	5.2% (n=8)
Household Income
$0-$4,999	9.7% (n=15)
$5,000-$9,999	3.9% (n=6)
$10,000-$14,999	4.5% (n=7)
$15,000-$24,999	11.7% (n=18)
$25,000-$34,999	16.9% (n=26)
$35,000-$49,999	14.9% (n=23)
$50,000-$74,999	18.8% (n=29)
$75,000-$99,999	10.4% (n=16)
$100,000 OR MORE	9.1% (n=14)
Baseline Mean Percent Days Abstinent (SD)	26.2% (26.7)
Baseline Mean Drinks Per Drinking Day (SD)	7.91 (4.24)

Measures

Participants completed assessments before treatment and following each treatment session. Before beginning treatment, participants filled out questionnaires assessing demographics in addition to the Timeline Follow-back (Sobell and Sobell, 1992), which measured drinking behaviors during the year prior to enrollment. Following each treatment session, participants filled out a series of questionnaires that included measures of client-perceived therapeutic alliance and a Client Session Report (Project Match Research Group, 1997). Therapists filled out the Therapist Rating Form, which measured the therapist’s initial impression of their client’s motivation for treatment. Table 2 provides the means, correlations, and standard deviations of the measures used.

Table 2.

Means, Standard Deviations, and Correlations for Level 2 Variables.

Level 2 Variables
	Mean	SD	1	2	3	4	5	6	7	8	9	10
1. Treatment Condition	0.33	0.47	--	.107	−.061	−.043	.096	−.074	−.006	−.125	−.022	−.026
2. No. Sessions	9.82	3.35		--	−.036	−.135	−.014	−.073	.077	−.013	.021	−.075
3. ADS	0.69	0.28			--	−.195*	−.190*	.185*	.085	.216**	.009	.444***
4. PreTx Change (PDA)	−0.43	0.53				--	.525***	−.093	.008	−.166*	.034	−.190*
5. PreTx Change (DDD)	−0.32	0.94					--	.045	.089	−.167*	.136	−.194*
6. URICA-Readiness	10.60	1.55						--	.355***	.313***	.223**	.099
7. Between-Participant WAIa	0.004	0.604							--	.282**	.026	.069
8. TRF	16.12	2.59								--	.007	.122
9. PDA (baseline)	0.26	0.27									--	−.005
10. DDD (baseline)	7.91	4.24										--

^*p<.05,

^**p<.01,

^***p<.001

^aBetween-Participant WAI scores are deviation scores calculated following detrending procedures. See data analytic strategy for more details.

Timeline Follow-back.

The Timeline Follow-back (TLFB; Sobbell & Sobbell, 1992) is a calendar-based, retrospective recall of daily alcohol use. The TLFB was administered prior to treatment to assess daily alcohol consumption over the 6-month period prior to the baseline pre-treatment assessment. Participant responses were coded for each day as number of standard drinks (.6 oz ethanol) consumed, from which multiple measures of alcohol consumption over a previous time period can be derived. In this study, the main outcome was percentage of total days in which no alcohol was consumed (percentage of days abstinent from alcohol). Additionally, the TLFB was administered following each treatment session to assess drinking behaviors and frequency over the week prior to that session, and a weekly PDA was generated for each participant. There is considerable support for the reliability and accuracy of the TLFB interview when used with this (and multiple other) population (e.g., Ehrman and Robbins, 1994; Sobell and Sobell, 1992; Sobell et al., 1996).

Working alliance Inventory- Short Version.

Measures of client-rated therapeutic alliance were obtained following every treatment session using the Working Alliance Inventory- Short Version (WAI-S; Horvath & Greenberg, 1989; Tracey & Kokotovic, 1989). The 12-item inventory consists of questions assessing the participant’s perception of three components of the therapeutic alliance: goal congruence between the client and therapist, agreement with the tasks of therapy, and the interpersonal bond between the therapist and client. The WAI was used to calculate within-participant, between-participant, and between-therapist effects of the WAI (see data analytic strategy for more details). Internal consistency at both the between- and within-levels were estimated for each subscale following procedures outlined by Geldhof et al. (2014). Specifically, internal consistency (Cronbach’s alpha) in the current sample for the within-participant, between-participant, and between-therapist were .796, .966, and .983 respectively, indicating good to excellent reliability.

Therapist Rating Form.

The Therapist Rating Form (TRF) is a measure created by our research group and was designed to assess therapist first impressions of their clients motivation for treatment. The TRF was completed by therapists only after their first session with a client, and consists of seven, 4-point Likert Scale questions assessing the therapist’s confidence in their client’s motivation to (i) remain abstinent, (ii) complete treatment, and (iii) be an active participant in future sessions. Responses from the items were summed to create a total score. Internal consistency for this measure was good with a Cronbach’s alpha of .80. The complete list of questions and possible responses is showcased in Appendix A.

Alcohol Dependence Scale.

To assess baseline alcohol dependence severity, the Alcohol Dependence Scale (ADS; Skinner & Horn, 1984) was used. The ADS is a measure of severity of the participant’s dependence on alcohol. The measure contains 29 items regarding symptoms and occurrences associated with dependence on alcohol. A total score for the measure is yielded by adding across items, with higher scores indicating more severe dependence. Internal consistency for this measure was good with a Cronbach’s alpha of .772.

University of Rhode Island Change Assessment Scale.

To assess overall motivation the change, the University of Rhode Island Change Assessment Scale (URICA; McConnaughty, E. A., Prochaska, J. O., & Velicer, W. F., 1983) was used. The URICA contains 32-items rated that form four subscales related to stages of change (i.e., precontemplation, contemplation, action, maintenance) and an overall readiness to change score (summation of the contemplation, action and maintenance scores minus the precontemplation score). Within the current sample, Cronbach’s Alphas for the precontemplation, contemplation, action and maintenance subscales were .742, .740, .817, and .825, respectively, indicating good to excellent internal consistency.

Procedure

All potential participants who responded to newspaper advertisements or were referred to the treatment site were scheduled for an initial screening interview. After determining eligibility, eligible participants completed a baseline assessment and were randomly assigned to experienced therapists who either received or did not receive therapeutic alliance feedback throughout treatment. All participants received 12 weeks of cognitive–behavioral therapy (CBT; Kadden et al., 1992) tailored to treat AUD, which was conducted in an outpatient research clinic. Each participant was assigned to one of six different therapists who conducted the AUD-tailored CBT treatment. All six therapists were white females with training and experience either as mental health therapists, social workers, or counselors. Specifically, each therapist had at least 5 years’ experience in AUD treatment and was credentialed (e.g., certified addictions counselors). Each session lasted 60 minutes, aside from the first session which lasted 90 minutes, and measures of client-rated therapeutic alliance and prior week drinking behaviors were recorded following each session. Treatment gains and termination were discussed during the final session, occurring in the twelfth week of treatment. Participants completed assessments at baseline prior to treatment, throughout treatment, and for 12 months posttreatment. For more details on the procedures, see Maisto et al. (2020).

Data Analytic Strategy

The original study was designed to examine the effect of therapeutic alliance feedback (given to the therapist) on drinking outcomes during the course of treatment. As reported in Maisto et al. (2020), therapeutic alliance feedback had no effect on drinking outcomes or therapeutic alliance scores during the course of treatment, nor long-term drinking outcomes (i.e., at 12-month follow-up). Therefore, data across the two feedback conditions were combined and were analyzed as a single arm trial in the current study. Further, the original Maisto et al. (2020) study utilized the California Psychotherapy Alliance Scale (CALPAS), which is an alternative measure of therapeutic alliance and was used to provide feedback to therapists for participants assigned to the feedback condition. The current study focuses on the more commonly used Working Alliance Inventory, which was chosen to limit any potential confounds with treatment condition. Finally, due to our interest in the relations among first impressions, client-rated therapeutic alliance, and drinking behaviors, only data from participants who attended at least 2 sessions (n= 155) were examined. One participant was missing data on therapist first impression (TRF), thus the final sample size analyzed was 154 (see Figure 1 for participant flow).

Analyses examining the effect of therapist first impressions on the relationships between therapeutic alliance and drinking outcomes were conducted using a similar strategy outlined in the main outcome study, Maisto et al. (2020). Specifically, multilevel modeling (MLM; HLM 7.03, Raudenbush et al., 2017) was used to examine changes in both drinking and therapeutic alliance during treatment, with therapist first impressions of the client’s motivation for treatment entered as a level 2 moderator. The data were structured on 3 levels: Measurements were nested within participants, and participants were nested within therapists. To account for the correlation between measurements within-participant and measurements of participants with the same therapist, random intercept of participants nested within therapist and the random intercept of therapist were modeled, and all analyses were conducted as a 3-level hierarchical nested model to protect against potentially biased estimates regardless of significance (see Crits-Christoph and Mintz, 1991; Wampold and Serlin, 2000). Further, data obtained at the end of each session were used for the treatment analyses, with percentage of days abstinent and average drinks per drinking day representing the interval between sessions (i.e., treatment session 1 represents drinking between session 1 and 2).

Therapeutic alliance scores were divided into three levels: within-participant effect, between-participant effect, and between-therapist effect. To disaggregate the three effects of alliance, similar to Zilcha-Mano & Errazuriz (2015), we followed recommendations by Bauer and Curran (2011) by applying linear regression of alliance against the best fitting growth model (with time centered at the mean session). Specifically, following the identification of the best fitting growth model (i.e., quadratic random effect, see results below), residuals and intercepts from these regression models were used to represent within-participant, between-participant, and between-therapist effects. The within-participant effect was measured by the difference between the observed measurement and its expected value given growth over time (i.e., residuals). The estimated intercepts at both level 2 and level 3 were used to obtain the between effects. The difference between the participant’s intercept (level 2) and the intercept of the therapist (level 3) was used as an estimate of between-participant effect. The intercept for each therapist was used as an estimate of the between-therapist effect. Further, because the within-participant and between-participant effects represented deviation scores, these effects were entered into models without further centering. Specifically, the within-participant effect was entered as a level-1 time-varying covariate and modeled as random, and the between-participant effect was entered as a cross-level interaction (level-2) on the intercept. Time and the between-therapist effect were both grand mean-centered, with time entered as a time-varying covariate on level-1, and the between-therapist effect entered on the intercept at level-3.

To examine the effect of therapist first impression of client’s motivation on both the within- and between-participant effects of WAI, first impression score was entered as a cross-level moderator on the within-participant alliance effect and a between-participant WAI X TRF term was added to the level 2 intercept, with significant interactions probed at low (15^th percentile) and high (85^th percentile) values of first impression scores, and regions of significance (Johnson-Neyman technique; Preacher, Curran, & Bauer, 2006). Finally, we controlled for the overall alcohol dependence severity, changes in drinking behavior prior to starting treatment¹, condition (feedback versus no feedback), readiness to change and number of sessions attended (on the level 2 intercept) due to their potential to affect therapist first impressions of their clients and drinking outcomes (all grand mean centered). Therefore, the full model examined was:

$$\begin{gathered} \text{Level-1 Model} \\ \text{Outcome}{(\text{PDA or DDD})}_{\text{tij}}={\pi }_{\text{0ij}}+{\pi }_{\text{1ij}}*({\text{TIME}}_{\text{tij}})+{\pi }_{\text{2ij}}*(\text{Within-participant}{\text{WAI}}_{\text{tij}})+{\text{e}}_{\text{tij}} \end{gathered}$$

$$\begin{gathered} \text{Level-2 Model} \\ {\text{π}}_{\text{0ij}}={\beta }_{\text{00j}}+{\beta }_{\text{01j}}*({\text{Condition}}_{\text{ij}})+{\beta }_{\text{02j}}*({\text{No. Sessions Attended}}_{\text{ij}})+{\beta }_{\text{03j}}*({\text{Dependence Severity}}_{\text{ij}})+{\beta }_{\text{04j}}*({\text{Pretreatment Change}}_{\text{ij}})+{\beta }_{\text{05j}}*({\text{URICA-Readiness}}_{\text{ij}})+{\beta }_{\text{06j}}*({\text{Between-participant WAI}}_{\text{ij}}\text{)}+{\beta }_{\text{07j}}*({\text{Between-participant TRF}}_{\text{ij}})+{\beta }_{\text{08j}}*(\text{Between-participant WAI X}\text{Between-participant}{\text{TRF}}_{\text{ij}})+{\text{r}}_{\text{0ij}} \\ {\text{π}}_{\text{1ij}}={\beta }_{\text{10j}}+{\text{r}}_{\text{1ij}} \\ {\text{π}}_{\text{2ij}}={\beta }_{\text{20j}}+{\beta }_{\text{21j}}*(\text{Between-participant}{\text{TRF}}_{\text{ij}}{)+\text{r}}_{\text{2ij}} \end{gathered}$$

$$\begin{gathered} \text{Level-3 Model} \\ {\text{β}}_{\text{00j}}={\gamma }_{\text{000}}+{\gamma }_{\text{001}}(\text{Between-therapist}{\text{WAI}}_{\text{j}})+{\text{u}}_{\text{00j}} \\ {\text{β}}_{\text{01j}}={\gamma }_{\text{010}} \\ {\text{β}}_{\text{02j}}={\gamma }_{\text{020}} \\ {\text{β}}_{\text{03j}}={\gamma }_{\text{030}} \\ {\text{β}}_{\text{04j}}={\gamma }_{\text{040}} \\ {\text{β}}_{\text{05j}}={\gamma }_{\text{050}} \\ {\text{β}}_{\text{06j}}={\gamma }_{\text{060}} \\ {\text{β}}_{\text{07j}}={\gamma }_{\text{070}} \\ {\text{β}}_{\text{08j}}={\gamma }_{\text{080}} \\ {\text{β}}_{\text{10j}}={\gamma }_{\text{100}} \\ {\text{β}}_{\text{20j}}={\gamma }_{\text{200}} \\ {\text{β}}_{\text{21j}}={\gamma }_{\text{210}} \end{gathered}$$

Missing Data.

Data at level 2 (between-participant) were complete for 154 participants. Of the potential 1848 data points at level 1 (154 participants X 12 sessions each), complete data was available for 1511 data points (82%) and used in the analyses (i.e., runtime deletion at level 1). Because data was collected at the end of each treatment session, missing data were related to the number of treatment sessions completed. On average, participants completed on average 9.82 (SD = 3.35; mdn = 12), with 58% of the sample completing 12 sessions (n=90). To further explore missingness of data, correlations between number of sessions attended and demographic and baseline measures were examined. Results indicated that the number of sessions attended was not related to numerous baseline indices including gender, race, household income, alcohol dependence severity, baseline drinking rates, readiness to change, nor pretreatment change (see Table 2 for summary of correlations). Furthermore, number of sessions was not associated with the two primary constructs of interest at the between-participant levels, therapist first impressions and therapeutic alliance. Taken together, these results suggest that total number of sessions, and thereby missingness, was not systematically associated with common covariates in treatment outcome research, nor the primary constructs of interest.

Results

Change in WAI scores During Treatment: Moderated by First Impressions (Hypothesis 1)

Results from model comparisons indicated that the random quadratic model best fit the data (deviance test of quadratic versus quadratic with random slope; χ2 ₍₃₎ = 44.32, p < .001). A summary of results is presented at the top half of Table 3. There was a significant increase in WAI score during treatment (i.e., significant Time fixed effect), with such increases slowing towards the end of treatment (i.e., significant Time² fixed effect).

Table 3.

Growth Curve Models During Treatment (H1)

	WAI (n=154)			PDA (n=154)			DDD (n=154)
	b	SE	p-value	b	SE	p-value	b	SE	p-value
Intercept	6.137	.062	<.001	.717	.025	<.001	3.061	.271	<.001
Time (linear fixed effect)	.059	.010	<.001	.041	.006	<.001	−.357	.071	<.001
Time2 (quadratic fixed effect)	−.003	.0008	.001	−.002	.0005	<.001	.014	.006	.014
Random Effects Variances	Var	χ 2	p-value	Var	χ 2	p-value	Var	χ 2	p-value
Intercept (level 3)	.012	13.03	.023	.00001	4.56	>.500	.00026	4.398	>.500
Intercept (level 2)	.396	1599.68	<.001	.095	2005.91	<.001	9.021	622.034	<.001
Time (linear random effect)	.010	352.44	<.001	.004	601.59	<.001	.278	239.59	<.001
Time2 (quadratic random effect)	.00005	271.71	<.001	.00002	401.83	<.001	.0007	189.09	.008
Level-1, e	.059	---	---	.0116	---	---	4.048	---	---
Moderation by Therapist First Impression
	b	SE	p-value	b	SE	p-value	b	SE	p-value
Intercept	6.137	.069	<.001	.717	.020	<.001	3.060	.266	<.001
X Between- participant TRF	.073	.014	<.001	.026	.009	.005	−.143	.055	.011
Time (linear fixed effect)	.059	.009	<.001	.041	.006	<.001	−.355	.071	<.001
X Between-participant TRF	−.002	.001	.040	ns	ns	ns	ns	ns	ns
Time2 (Quadratic fixed effect)	−.003	.0008	.001	−.002	.0005	<.001	.013	.006	.016
X Between-participant TRF	ns	ns	ns	ns	ns	ns	ns	ns	ns
Random Effects Variances	Var	χ 2	p-value	Var	χ 2	p-value	Var	χ 2	p-value
Intercept (level 3)	.021	18.65	.003	.00001	2.85	>.500	.00013	2.947	>.500
Intercept (level 2)	.360	1477.32	<.001	.088	1947.68	<.001	8.612	604.91	<.001
Time (linear random effect)	.009	351.82	<.001	.004	601.63	<.001	.279	239.66	<.001
Time2 (quadratic random effect)	.00004	271.56	<.001	.00002	401.81	<.001	.0007	189.14	.008
Level-1, e	.059	---	---	.012	---	---	4.046	---	---

Note: WAI = Therapeutic alliance Inventory; PDA = Percent Days Abstinent; PHD= Percent Heavy Drinking; Feedback = 0 (no feedback), 1 (feedback); b = unstandardized estimate; SE = standard error; Var = Variance estimate; ns = non-significant and trimmed from model.

The effect of therapist first impressions on changes in WAI scores during treatment was examined next. Results indicated that the interaction between first impressions and the fixed effect of Time² was nonsignificant (b = .0002, SE = .0003, p = .542). Therefore, to further examine its potential relation to the fixed effect of Time, therapist first impression was removed as a cross-level interaction on Time² and the model was estimated again. Results indicated that therapist first impressions significantly moderated the fixed effect of Time on WAI scores (b = −.0023, SE = .001, p = .040). Follow-up analyses indicated the fixed effect of Time was significant for both those with lower rated therapist first impressions (b = .066, SE = .010, p < .001) and for those rated higher (b = .052, SE = .010, p < .001), with those rated lower increasing at a slightly faster rate (see Figure 2). Further, therapist first impressions were also significantly associated with WAI at the beginning of treatment (b = .073, SE = .014, p < .001), such that higher initial first impressions were associated with higher WAI scores.

Figure 2.

Growth Curve for WAI during Treatment

Change in Drinking Behaviors: Moderated by First Impressions (Hypothesis 1)

Percent Days Abstinent (PDA).

Similar to Maisto et al. (2020), results from model comparisons indicated that the random quadratic model best fit the data (deviance test; χ2 ₍₃₎ = 131.51, p < .001). There was a significant increase in PDA during treatment (i.e., significant Time fixed effect), with such increases slowing towards the end of treatment (i.e., significant Time² fixed effect; see top half of Table 3 for summary of results).

Next, therapist’s first impressions of their client’s motivation was entered as a cross-level moderator. Results indicated that the interaction between first impressions and the fixed effect of Time² was nonsignificant (b = −.00001, SE = .0001, p = .914). To further examine its potential effect on Time, therapist first impression was removed as a cross-level interaction on Time². Therapist first impressions (i.e., TRF) also did not moderate the fixed effect of Time on PDA during treatment (b = −.0008, SE = .0008, p = .317), and thus was subsequently removed. Although therapist first impressions did not moderate change during treatment, it was associated with drinking behavior at the beginning of treatment (b = .026, SE = .009, p = .005), such that higher first impressions of the clients’ motivation were associated with greater percent days abstinent (see bottom half of Table 3 for summary of results).

Drinks per Drinking Day (DDD).

Similar to PDA, results the random quadratic model was the best fitting model (deviance test; χ2 ₍₃₎ = 19.11, p < .001). There was a significant decrease in DDD during treatment (fixed effect of Time), with such decreases slowing towards the end of treatment (fixed effect of Time²); see top half of Table 3 for summary of results].

The effect of therapist’s first impressions of their client’s motivation on changes in DDD scores during treatment was examined next. Results indicated that the interaction between first impressions and the quadratic growth term was nonsignificant (b = −.0014, SE = .002, p = .498). To further examine its potential effect on initial change (i.e., Time effect), therapist first impression was removed as a cross-level interaction on the fixed effect of Time². Therapist first impressions (i.e., TRF) also did not moderate the fixed effect of time on DDD during treatment (linear effect; b = .016, SE = .010, p = .134), and thus was subsequently removed from all growth paraments in the model. Although therapist first impressions did not moderate change during treatment, it was associated with drinking behavior at the beginning of treatment (b = −.143, SE = .055, p = .011), such that higher first impressions of the clients’ motivation were associated with lower average number of drinks per drinking day (see bottom half of Table 3 for summary of results).

Therapeutic Alliance Effects on Drinking Moderated by First Impression (Hypothesis 2)

Percent Days Abstinent (PDA).

Results indicated a significant between-participant WAI X TRF interaction (b = −.022, SE = .010, p = .038; see Table 4 and Figure 3 for summary of results). Follow-up analyses indicated that among participants with lower rated therapist first impressions, between-participant WAI predicted PDA (b = .108, SE = .042, p = .012), such that higher overall WAI during treatment predicted higher percent days abstinent. In contrast, among participants with higher rated therapist first impressions, the effect of the between-participant WAI was nonsignificant (b = −.0242, SE = .044, p = .581), with participants demonstrating relatively high percent days abstinent regardless of WAI scores. Furthermore, results also revealed that therapist first impressions moderated the within-participant effect of alliance on PDA (b = −.010, SE = .004, p = .030) (see Figure 4). Follow-up analyses indicated that among participants with lower rated therapist first impressions, within-participant WAI predicted drinking during the upcoming week (b = .050, SE = .017, p = .005), such that increases in WAI predicted higher percent days abstinent. In contrast, among participants with higher rated therapist first impressions, the effect of the within-participant WAI was nonsignificant (b = −.008, SE = .023, p = .719), with participants demonstrating relatively high percent days abstinent regardless of changes in WAI scores.

Table 4.

Effect of WAI and Therapist First Impressions on Drinking Outcomes During Treatment (all parameters) (H2)

	PDA (n=154)			DDD (n=154)
	b	SE	p-value	b	SE	p-value
Intercept	.857	.017	<.001	1.628	.154	<.001
Between-Therapist WAI	−.214	.156	.243	1.561	1.352	.313
Treatment Condition	−.061	.034	.077	.432	.293	.143
No. Sessions Attended	.010	.006	.066	−.056	.061	.360
Between-Participant ADS	−.056	.060	.352	1.276	.512	.014
Between-Participant Pretreatment Change	−.115	.032	<.001	.180	.142	.209
Between-Participant URICA-Readiness	.025	.011	.031	−.153	.100	.127
Between- Participant WAI	.039	.029	.186	−.359	.257	.164
Between- Participant TRF	.016	.007	.022	−.100	.058	.089
Between- Participant WAI X TRF	−.022	.010	.038	.206	.091	.025
Time	.018	.002	<.001	−.203	.030	<0.001
Within-Person WAI	.020	.015	.204	−.278	.315	.378
X Between- Participant TRF	−.010	.004	.030	1.628	.154	<0.001
Random Effects Variances	Var	χ 2	p-value	Var	χ 2	p-value
Intercept (level 3)	<.00001	3.42	>.500	.0006	2.74	>.500
Intercept (level 2)	.035	3213.07	<.001	2.471	939.61	<.001
Time (linear random effect)	.001	730.89	<.001	.083	459.62	<.001
Within-person WAI (random effect)	.001	204.46	<.001	2.715	210.80	<.001
Level-1, e	.014	---	---	4.048	---	---

Note: WAI = Therapeutic alliance Inventory; Treatment Condition (0 = no feedback, 1=feedback); TRF = Therapist Report Form; ADS = Alcohol Dependence Scale.

The moderating effect of treatment condition on the within-participant WAI X TRF interaction and between-participant WAI X TRF interactions on PDA and DDD were also examined, with results indicating that the 3-way interactions at both the between (PDA, b = −.037, SE = .023, p = .113; DDD, b = .162, SE = .197, p = .414) and within levels were non-significant (PDA, b = .008, SE = .010, p = .436; DDD, b = −.131, SE = .229, p = .567). Therefore, condition was removed as a moderating effect and controlled for on the intercept only.

Figure 3.

Between-Participant WAI X TRF Interaction Predicting Percent Days Abstinent. Simple Slopes (left panel) and Region of Significance (right panel); Shaded area indicates significant WAI simple slopes at varying TRF values, TRF ≤ −0.9365).

Figure 4.

Within-Participant WAI X TRF Interaction Predicting Percent Days Abstinent. Simple Slopes (left panel) and Region of Significance (right panel; shaded area indicates significant WAI simple slopes at varying TRF values, TRF ≤ −0.9681).

Drinks per Drinking Day (DDD).

Results indicated a significant between-participant WAI X TRF interaction (b =.206, SE = .091, p = .025; see Table 4 and Figure 5 for summary of results). Similar to PDA, follow-up analyses indicated that among participants with lower rated therapist first impressions, between-participant WAI was associated with DDD (b = −1.002, SE = .370, p = .008), such that higher overall WAI during treatment predicted lower average drinks per drinking day. In contrast, among participants with higher rated therapist first impressions, the effect of the between-participant WAI was nonsignificant (b = .232, SE = .379, p = .541), with participants demonstrating relatively low average drinks per drinking day regardless of WAI scores. The within-participant WAI X TRF was non-significant (b = .064, SE = .114, p = .575).

Figure 5.

Between-Participant WAI X TRF Interaction Predicting Average Drinks per Drinking Day. Simple Slopes (left panel) and Region of Significance (right panel; shaded area indicates significant WAI simple slopes at varying TRF values, TRF ≤ −0.7699).

Discussion

The purpose of the current study was to examine the influence of therapists’ first impression of their clients’ motivation for treatment on client-rated therapeutic alliance and drinking outcomes during treatment. We predicted that therapist first impressions would be positively associated with therapeutic alliance scores and drinking outcomes at the beginning of treatment, and that higher therapist first impressions would be associated with greater increases in therapeutic alliance and percent days abstinent, and greater decreases for drinks per drinking day (Hypothesis 1). This study’s results partially supported our hypothesis. We found that higher therapist first impressions of their clients’ motivation were associated with higher client-rated therapeutic alliance scores at the beginning of treatment and remained higher throughout, with therapeutic alliance scores increasing at a slightly faster rate for those lower on therapist first impressions. This finding is consistent with prior literature demonstrating that psychotherapists who hold more positive expectations for their clients’ treatment success may develop a stronger therapeutic alliance at the onset of treatment (Connor & Callahan, 2015; Long, 2004), suggesting that therapists’ perceptions of their client’s motivation may play a role in both the formation and maintenance of the therapeutic alliance. Further, consistent with previous literature indicating that clients of therapists with higher impressions of their likelihood of recovery are more likely to remain abstinent (Leake & King, 1977), results indicated that higher therapist first impressions were associated with greater percent days abstinent and fewer drinks per drinking day throughout the course of treatment. Therapist first impressions, however, were not associated with changes in drinking behaviors during treatment.

This study was also the first to examine how therapist first impressions of their clients’ motivation for treatment affected the relationship between therapeutic alliance and drinking outcomes. We predicted that, among clients whose therapist rated them as having a higher motivation for treatment, higher therapeutic alliance scores (within- and between-participant effects) would be associated with greater percent days abstinent and fewer drinks per drinking day compared to those who rated lower on motivation for treatment (Hypothesis 2). Results indicated that therapist first impressions moderated the relationship between average client-rated therapeutic alliance during treatment (between-participant effect) and drinking outcomes; however, inconsistent with our hypothesis, individuals whose therapists rated them as being highly motivated for treatment after the first session reported greater precent days abstinent and fewer drinks per drinking day regardless of average therapeutic alliance scores during treatment. In contrast, among those with lower therapist initial impressions, higher therapeutic alliance scores were associated with greater precent days abstinent and fewer drinks per drinking day. A similar pattern was observed for changes in therapeutic alliance (higher scores relative to their mean) predicting percent days abstinent between treatment sessions (i.e., within-participant effect). Specifically, among those with lower therapist first impressions, increases in therapeutic alliance (higher scores relative to their mean) predicted higher percent days abstinent between treatment sessions among those whose therapists rated them lower on treatment motivation. In contrast, those rated higher on treatment motivation reported greater percent days abstinent regardless of changes in therapeutic alliance. Taken together, these findings have several theoretical and clinical implications.

Although much of the first impression and clinical judgment literature has focused on negative impact, i.e., poor impressions resulting in poorer outcomes, it is possible that therapist perceptions of their client’s motivation are an accurate reflection of how motivated their client is for treatment and their readiness to change. Indeed, ratings of therapist first impressions in the current study were associated with the client’s overall readiness to change and levels of pretreatment change, suggesting that therapists may have accurately assessed their client’s motivation. Consistent with this notion are findings that client-rated readiness to change is one of the stronger predictors of both therapist and client rated therapeutic alliance scores (e.g., Connors et al., 2000). To further explore this possibility, we re-ran our growth curve analysis controlling for both readiness to change and pretreatment changes in drinking, with results indicating that therapist first impressions continued to predict alliance scores. Thus, therapist first impressions of their client’s motivation appear be capturing more than the client’s actual readiness to change and may have significant effects on the therapeutic alliance and treatment outcomes.

The finding that higher overall therapeutic alliance and increases in therapeutic alliance (relative to one’s mean) predicted higher percent days abstinent between treatment sessions only when therapist first impressions of their client’s motivation was low contradicts studies demonstrating that client-therapist dyads with discordant therapeutic alliance ratings typically have worse treatment outcomes than those in dyads whose alliance ratings are both high and in agreement (Laws et al., 2017; Zilcha-Mano et al., 2015). Once thought to remain consistent throughout a relationship, research suggests that an individual’s current impressions about a person are susceptible to change from their initial impression when exposed to new information that contradicts their original impression (Wyer, 2010). Changes in therapeutic alliance may therefore serve as a reflection of updated information on a client’s motivation for treatment and help to mitigate the negative consequences of initial poor impressions. Alternately, it may be that when a therapist perceives their client to have low treatment motivation, they engage in greater efforts to build a strong therapeutic alliance with the goal of improving the client’s motivation and engagement in treatment. In this regard, in a review of the therapeutic alliance literature, Meir and colleagues (2005) concluded that early therapeutic alliance was a consistent predictor for treatment engagement and retention. Taken together, results from this study continue to highlight the need for research examining the contextual factors around common factors, and the mechanisms through which these contextual factors, such as therapist first impressions, influence treatment outcomes.

Although future research is needed to determine whether therapist first impressions create a bias that can influence treatment outcomes or simply reflect good clinical judgment, the current findings have several clinical implications. First, higher perceived therapeutic alliance by the client may serve to mitigate the effect of lower perceptions of treatment motivation by the therapist, whether accurate or not. Within the AUD field, treatments often emphasize the importance of common factors, encouraging the use of non-specific therapeutic skills such as reflective listening to understand a client’s perspective, thoughts, and feelings without judgement or criticism, in efforts to bolster the therapeutic alliance and improve client motivation for treatment (Epstein & McCrady, 2009; Rotgers et al., 2002;). Thus, it is plausible that therapists adept in these skills would be perceived by their clients as being more approachable and empathetic, thereby improving the therapeutic alliance and mitigating the potential negative outcomes associated with poor first impressions or lower treatment motivation. Conversely, clients with therapists who are unable to overcome the potential biases associated with poor first impressions may perceive the therapeutic alliance more poorly and thereby be less motivated over the course of treatment. Indeed, prior research has suggested that AUD treatments are less efficacious when clients viewed their therapists as being driven by controlled motivation, which is distinguished by behaviors that reflect non-genuine interest (Wild et al., 1998). These theoretical explanations for differences in treatment outcomes are in concordance with Miller and Moyers (2015), who suggested that therapist-specific factors may serve as important predictors and moderators of treatment outcomes. Therefore, findings from this study continue to highlight the importance of common factors in AUD treatments not only because they are consistently associated with improved drinking outcomes (Miller & Moyers, 2015; Moyers et al., 2009; Moyers et al., 2016), but because they may mitigate the influence of therapist perceptions of treatment motivation (e.g., Ilgen et al., 2006). Therefore, continued efforts to improve or refresh non-specific skills among those working with AUD populations have the potential to reduce the effects of any negative first impressions of their clients and improve clinical outcomes.

Limitations

This study has several limitations that should be considered in interpreting its findings. First, the study was a secondary analysis of data collected in a clinical trial investigating the influence of providing therapists with feedback on client-rated therapeutic alliance and treatment outcomes during a 12-week CBT treatment for AUD, which warrants several methodological considerations when interpreting results and drawing conclusions (see Maisto et al., 2020). Second, self-report was used to gather information on alcohol consumption and therapeutic alliance; however, these methods (e.g., timeline follow-back) have been used extensively in the AUD research and shown to be both reliable and valid in AUD treatment contexts (e.g., Robinson et al., 2014). Third, the study utilized a convenience sample of individuals presenting for treatment at an outpatient clinic, and as such, may not be representative of all individuals seeking treatment. Fourth, there were only 6 therapists that participated in the study protocol, which limits the ability to examine possible individual differences in therapist variables that may influence treatment outcomes. However, prior research has demonstrated that therapist characteristics such as racial congruence and gender tend not to be associated with treatment outcomes nor therapeutic alliance (Beutler et al., 1994; Hersoug et al., 2001; Miller & Moyers, 2015).

Therapist characteristics may, however, play a role in the first impression formation process. First impression formation has been demonstrated as being associated with demographic variables such as race (Blair et al., 2002) and gender (Friedman & Zebrowitz, 1992), and these impressions have been shown to be associated with social outcomes such as hireability (Cortez et al., 2017) and judicial decisions (Zebrowitz & McDonald, 1991). Further, stereotypes of therapists held by clients before therapy, such as features that an ideal therapist should have, have been previously associated with therapy satisfaction (Russel et al., 2020), indicating a possible positive feedback loop between expectations and therapy efficacy. Future research should therefore look to examine the possible influence of client and therapist characteristics on first impression formations, and the role stereotyping may play in both therapeutic first impression formation and therapeutic alliance.

Conclusion

The current study sought to examine the relation between therapists’ first impression of their clients’ motivation for treatment and the relationship between client-rated therapeutic alliance and drinking behaviors in a 12-week CBT treatment for AUD. Consistent with prior findings, therapists’ initial impressions of their clients’ motivation for treatment serve as a moderator in the relationship between client-rated therapeutic alliance and treatment outcomes. This effect remained consistent even with therapists who rated their clients as having lower motivation for treatment, suggesting therapeutic alliance may mitigate the effect of negative therapist first impressions. These results highlight the importance of therapeutic strategies such as empathy expression and non-judgment in fostering positive, productive therapy dyads and improving treatment efficacy. Future research should examine how different contextual factors of therapy, as well as therapist characteristics (e.g., Miller & Moyers, 2015), influence first impression formation in therapeutic contexts and treatment efficacy.

Appendix A. Therapist Rating Form

	Not involved	Somewhat involved	Fairly involved	Very involved
(1) How involved was the client in this session?	1	2	3	4
	Not Motivated	Somewhat Motivated	Fairly Motivated	Very Motivated
(2) How motivated did the client appear this session?	1	2	3	4
	No, definitely will not	No, I don’t think so	Yes, I think so	Yes, Definitely will
(3) How confident are you that the client will complete treatment?	1	2	3	4
(4) How confident are you that the client will be motivated to abstain from drinking?	1	2	3	4
(5) How confident are you that the client will abstain from drinking during treatment?	1	2	3	4
(6) How confident are you that the client will become involved/ interact in individual sessions?	1	2	3	4
(7) How confident are you that the client will be motivated by individual sessions?	1	2	3	4