Motivational Interviewing: Key Ingredients Associated with Taking a Step Toward Employment

Abstract

Objective:

This study sought to identify key ingredients of motivational interviewing (MI) associated with taking a step in the direction of competitive employment for unemployed veterans with serious mental illness (SMI).

Method:

Data were analyzed from 195 audiotaped MI sessions targeted to employment conducted with 39 veterans with SMI. Sessions were coded and analyzed to identify components of MI practice predictive of taking any step in the direction of competitive employment (for example, asking for a referral to supported employment or conducting a job search). Predictor variables were a) counselor MI talk behaviors and adherence to MI technical and relational principles; and b) client intensity and frequency of change talk and sustain talk. Covariates were age, gender, race, duration of unemployment, receipt of disability income, health status, work importance, work confidence, mental health diagnosis and session number. Generalized Estimating Equations were used to create multivariate models.

Results:

After controlling for session number, work importance, work confidence and duration of unemployment, variables significant in the adjusted multivariate model were intensity of client change talk and sustain talk and counselor adherence to MI technical principles of cultivating change talk and softening sustain talk.

Conclusions and Implications for Practice:

Findings suggest that change talk and sustain talk during counseling sessions are associated with taking a step toward employment and that counseling focused on cultivating change talk and softening sustain talk increases the likelihood that unemployed veterans with SMI will take steps toward becoming competitively employed.

There is considerable evidence that individuals with serious mental illness (SMI), including veterans, consider work to be an important aspect of life (Dunn, Wewiorski, & Rogers, 2008; McQuilken et al., 2003; Mueser et al., 2001; Drebing et al., 2012). Nearly half of veterans in the United States are unemployed (Greenberg & Rosenheck, 2009) and research findings suggest that 94% of veterans with mental health diagnoses and vocational problems consider their unemployment to be a problem (Drebing et al., 2012). Work motivation has been associated with job search behavior which predicted employment outcomes (Corbiere et al., 2011). Although the Veterans Health Administration (VHA) has devoted considerable resources to offering supported employment (SE), an established evidence-based practice designed specifically to help individuals with SMI obtain and keep competitive employment, research evidence suggests that only 1.2% of VHA users with a psychiatric diagnosis use SE in a single fiscal year (Abraham et al., 2014). Thus, there is considerable potential to significantly increase competitive employment among veterans with SMI.

Motivational interviewing (MI) is an evidence-based counseling intervention that uses a directive, client-centered approach to address a person’s ambivalence about behavior change (Miller & Rollnick, 2013). MI has demonstrated efficacy with a wide range of behavioral targets, including engagement in substance use, anxiety, and symptom management (e.g., Carroll et al., 2006; Westra & Dozois, 2006; Murphy, 2008), and with various populations, including individuals with SMI (e.g., Smeerdijk et al., 2012; Wong-Anuchit et al., 2019). These findings suggest MI as a promising counseling approach for unemployed veterans with SMI.

MI rating manuals (Moyers, Manuel, & Ernst, 2014; Miller, Moyers, Ernst, & Amrhein, n.d.) are useful for investigating the process by which the core ingredients of MI contribute to behavior change. Desired counselor talk behaviors include open questions, complex reflections, and affirmations. Client talk behaviors include Change Talk (in the direction of the targeted behavior change) and Sustain Talk (away from the targeted behavior change). MI proficient counselors adhere to MI technical principles by Cultivating Change Talk and Softening Sustain Talk and they adhere to MI relational principles by demonstrating Collaboration and Empathy.

MI researchers investigating key elements of MI have depicted technical and relational pathways to client change talk and behavior change (Miller & Rose, 2009). Research on the association between therapeutic alliance and client outcomes (Horvath & Symonds, 1991;Martin, Garske, & Davis, 2000) and on the relationship between MI therapist interpersonal skill (e.g., empathy, support of client autonomy) and client involvement (e.g., cooperation, disclosure; Moyers, Miller, & Hendrickson, 2005) supports the relational pathway. Positive correlations have been found between positive treatment outcomes and client Change Talk, whereas Sustain Talk has been correlated with negative outcomes or no change (Moyers, et al., 2007). The most recent meta-analysis of MI studies (Magill et al., 2018) suggests that counselor proficiency in MI talk behaviors (open questions, complex reflections, affirmations) is positively associated with Change Talk and Sustain Talk, that the proportion of Change Talk predicts outcomes, and that MI technical components, rather than relational components, predict outcomes. However, because this meta-analysis included studies primarily focused on addiction behaviors and used data from only one or two intervention sessions, its findings may not be generalizable to veterans with SMI or to MI interventions that unfold over multiple sessions.

A feature of high-fidelity SE programs is the extent to which mental health providers create “opportunities for people to consider whether they want to work” (Becker et al., 2015, p.72) by asking about their interest in and concerns about working. Qualitative research suggests that these discussions influence client thoughts about work and participation in SE (Gowdy, Carlson, & Rapp, 2003). Better understanding of the MI key ingredients associated with steps toward employment would provide important guidance for mental health providers and SE specialists to increase motivation and reduce ambivalence toward engaging in SE and employment.

This study is a secondary analysis of data from a randomized trial that examined the efficacy of MI in enhancing SE enrollment and employment outcomes in Veterans with SMI (Mueller, et al., n.d.). Our aim was to examine client and counselor MI ingredients associated with taking a step toward employment. We hypothesized that the frequency and intensity of the participant’s Change Talk and Sustain Talk during a session would predict the likelihood of taking a step toward competitive employment (such as, requesting a referral to SE or filing a job application) either during that session or before the next monthly session. We also hypothesized that counselor MI technical proficiency and adherence to MI technical and relational principles would predict taking a step.

Method

Participants

Participants came from an Institutional Review Board approved randomized trial (Mueller et al., n.d.) of 84 veterans from a VHA medical center in the Northeastern United States who met the following criteria: (a) chart-documented diagnosis of schizophrenia, schizoaffective disorder, psychosis nos, bipolar I disorder, or depression with psychotic features, (b) unemployed, and (c) not currently receiving SE services. The study had no requirement regarding interest in work. Participants were told during recruitment and informed consent that this was a study about Veterans’ decisions to enter and participate in psychosocial rehabilitation services. Of the 42 participants randomized to the MI condition, 39 (92.9%) had audiotaped MI sessions and were the focus of this study.

Procedure

Intervention

MI sessions were conducted by research team members who utilized a variety of motivational enhancement strategies including evocative questions, importance and confidence scales, collaborative problem solving, and planning (Miller & Rollnick, 2002; Glynn, et al., n.d.). Sessions were organized around (a) building motivation for change, (b) strengthening commitment to change, and (c) follow-through strategies (i.e. steps) (Miller, 1995). Thus, for participants who were not interested in competitive employment, sessions focused on building motivation for change. For individuals who expressed interest in employment, sessions focused on steps to strengthen commitment to change and strategies to create change, such as placing a phone call to clinical provider to request a referral to SE. For those interested in competitive employment but not SE, MI targeted steps related to obtaining employment. For individuals who entered SE, MI focused on staying engaged in SE services, such as maintaining contact with their SE specialist and completing hiring paperwork. For participants who obtained employment, with or without SE, MI focused on keeping the job, and/or ongoing job search (if dissatisfied in the current job). In this way, the MI sessions were aligned with the participant’s individual process and did not interfere with SE activities (rapid job search, job development, assertive engagement and outreach). For those participants who entered SE during the six months of MI intervention, MI sessions were conducted concurrently but separate from meetings with their SE provider. Participants each received up to 6 monthly counseling sessions (mean, s.d.=5.3, 1.2; range=1-6).

Research and Data Collection

Before randomization, participants had a baseline research interview to obtain demographic and background information. To ensure that lack of information was not the reason that participants with employment goals had not enrolled in SE, participants also received an information session about all vocational rehabilitation services offered by VHA (including SE), Veterans Benefits Administration, state rehabilitation agencies and unemployment offices.

Measures

Participant Variables (Covariates)

Covariates from the baseline interview included self-reported age, gender, race, years unemployed, a dichotomous measure of receipt of any Social Security or VA disability income, health status (Veterans RAND 12-Item Health Survey; Kazis, et al, 2004), and work importance and work confidence. The work importance and confidence scales were based on Miller & Rollnick (2002) interest/confidence scaling tool, and were each a single self-report item, rated from 1 (not at all important/confident) to 7 (extremely important/confident). Mental health diagnoses (including lifetime diagnoses of PTSD and substance abuse or dependence disorder) were obtained from the medical record.

Participant predictors (see Table 1) for each session included ratings on a) frequency and intensity of Change Talk, and b) frequency and intensity of Sustain Talk (see Miller, Moyers, Ernst & Amrhein, n.d.).

Table 1.

Examples of Motivational Interviewing Utterances

Variable	Examples
Counselor Talk
Open Question	How did it go with getting in touch with SE Specialist after we last met? In what ways has working been difficult for you?
Closed Question	Did you apply for any jobs this past month? How important is work to you on a 1 to 10 scale?
Simple Reflection	(client says) I’m bored with my life–> (counselor reflects) You feel like it’s the same thing every day
Complex reflection	(client says) I’m bored with my life–> (counselor reflects) You want some excitement in your life; You would like something new and interesting to do
Cultivating change talk	(client says) Might try (asking for a SE referral)–> (counselor reflects) You plan to; You will do soon
Softening change talk	(client says) Plan to; Will do soon (ask for a referral for SE)–> (counselor reflects) Might try
Cultivating sustain talk	(client says) Not sure, Not interested (in work) –> (counselor reflects) Can’t imagine working; Never would consider work
Softening sustain talk	(client says) can’t go back to work because got hospitalized when working in the past –> (counselor reflects) You want to make sure that when you go back to work it doesn’t affect your mental health and send you backwards in your recovery
Participant Talk
High intensity change talk	I really wish I was working. I keep looking but I haven’t been able to find anything.
Moderate intensity change talk	I think I should get moving about trying to find some kind of work.
Low intensity change talk	I’ll probably start thinking about work if things settle down; With the mental problems I got, I’ve really got to approach going back to work carefully.
High intensity sustain talk	I have a heart problem, I have an advanced case of emphysema and I have a back problem, So, I can’t even do a part-time job.
Moderate intensity sustain talk	Work, in the sense that I’ve always known it, isn’t really in the picture for me anymore; It’s taken me so long to even think about interviewing for a job, I’m afraid to even go talk to the guy in the career center.
Low intensity sustain talk	I might be getting too old to work; I get tired. So I don’t know if I’m gonna be able to work or not.

Counselor Variables

Counselor predictor variables (see Table 1) for each session were ratings on a) counselor MI talk behaviors (proportion of questions that are open-ended, proportion of reflections that are complex and ratio of reflections to questions); b) ratings on counselor adherence to MI relational principles (average of scores on Partnership and Empathy; see Moyers, Manuel, & Ernst, 2014); and c) counselor adherence to MI technical principles (average of scores on Cultivating Change Talk and Softening Sustain Talk; see Moyers, Manuel, & Ernst, 2014).

Dependent Variable: Taking a Step

The dependent variable for this study was defined as “taking a step” toward using SE services or toward competitive employment either within the counseling session (such as calling for an appointment) or during the period between that session and the next session. A “step” was operationalized as any action preliminary to SE enrollment, such as requesting a referral to or making an appointment for SE, or, once enrolled in SE, taking an action that demonstrated active use of SE services, such as working with an SE provider on job exploration or creating a resume. A “step” also was any action related to obtaining or maintaining competitive employment, such as having a job interview or performing work at a competitive job, regardless of SE program enrollment. Based on review of all counselor notes, transcripts and relevant medical record entries, we rated each session as either a “step session” (1) or a “no step session” (0). Whether a step occurred between sessions was determined by reviewing the transcript of the next session and by medical record review. For example, if the participant reported during session 3 that he had an SE intake since the last session, we would rate session 2 as “step session”.

Data Coding

A total of 195 transcribed sessions from the 39 participants with audio recordings of their monthly MI sessions served as the data source for this study. A coding team (Drs. Wewiorski, Mueller, Wang and Ms. Dreifuss) with experience conducting and coding MI sessions reviewed the accuracy of transcripts prior to coding. Then, in consultation with an expert MI trainer and researcher (Dr. Rose), the team created a study coding manual adapted from the Manual for the Motivational Interviewing Skill Code (MISC) Version 2.1 (Miller, Moyers, Ernst, & Amrhein, n.d,) and the Motivational Interviewing Treatment Integrity (MITI) Coding Manual Version 4.2 (Moyers, Manuel, & Ernst, 2014).

The study coding manual utilized the MISC codes for categorizing client talk with respect to the targeted behavior change as either Change Talk or Sustain Talk. Client speech intensity was coded as low, moderate, or high for each utterance of client Change Talk and Sustain Talk. “Moderate” intensity was operationalized as the default rating, and utterances clearly at the top and bottom of the range were identified by modifiers such as “very,” “really,” and “definitely” for high intensity, and “maybe,” “probably,” and “a little” for low intensity. The study manual used counselor global ratings from the MITI to rate the relational principles of Empathy and Partnership and the technical principles of Cultivating Change Talk and Softening Sustain Talk.

First, 24 sessions selected for variability on step vs no step, as well as variability on participant baseline factors such as age, gender, and length of unemployment, were each coded by at least 3 coders, who then met to review codes, make coding decisions and come to consensus. The coding team consulted with Dr. Rose about resolving coding disagreements, creating project-specific codes and adapting the standard MITI and MISC coding schemes. Next, the remaining 171 sessions were each coded by one member of the research team. To assess coding drift, 18 (9%) of these transcripts, selected with representation across different MI counselors, were coded by a second coder. Intraclass Correlation Coefficients (ICC) for these 18 sessions ranged from .86 to 1.0, indicating good to excellent interrater agreement (Cicchetti, 1994).

Using the operational definition of the dependent variable (taking a step), Drs. Wewiorski, Wang and Dreifuss divided up and reviewed session transcripts, counselor notes and participant medical records and came to consensus about assigning a step score of 1 or 0 to each session. If the session transcript indicated that the participant had taken a step during the session (such as calling for an appointment) or if the medical record, counselor notes or transcript for the next session indicated a step was taken prior to the start of the next session (such as having an SE intake meeting), that session was rated as a step session. Thus, each session was rated as either a step session or a non-step session.

Analysis Plan

Univariate Analysis.

We first computed descriptive statistics for continuous variables and frequency distributions for categorical variables.

Multivariate Analyses.

We examined whether predictor variables and covariates differed between “step sessions” and “no step sessions.” Because observations (i.e., sessions) were clustered within each participant, we made an adjustment in calculating whether, in each session, there was a significant difference in the predictor variables between step and non-step sessions. For regression models we used generalized estimating equations (GEE). The GEE logit estimates the same model as the standard logistic regression. However, unlike logistic regression, GEE logit allows for dependence within clusters (i.e., sessions within participants) in longitudinal studies. We assumed a binomial distribution for the dependent variable (taking a step) and a logit link. We used Huber-White/sandwich estimators of the variance-covariance structure because they are robust to misspecification of the working correlation matrix and produce consistent unbiased estimates of the parameters’ standard errors.

Unadjusted Models.

We first used GEE logit models to estimate the unadjusted odds of taking a step, with each predictor and covariate entered alone in the model. When there were two variables measuring the same or similar constructs (e.g. frequency and intensity of Change Talk), we estimated models with both variables entered in the model together. Thus, we estimated separate models for: a) frequency and intensity of Change Talk; b) frequency and intensity of Sustain Talk, c) the three measures of counselor talk behaviors, and d) counselor adherence to MI technical and relational principles.

Multivariate Models (Adjusted).

We first developed a multivariate model by selecting all the predictors and covariates with p<.25 in the unadjusted models (Hosmer & Lemeshow, 2013). We then considered variables that became significant after controlling for other variables. The preliminary final model was restricted primarily to predictors and covariates that were statistically significant at conventional levels (p<.05). We then compared the values of the estimated coefficients for variables remaining in the preliminary final model to their respective values in the larger model. If a coefficient for a variable remaining in the smaller model had changed by more than 20% when compared to its value in the larger model, this change suggested that the excluded variable provided a needed adjustment to the estimate of the effect of the included variable on the outcome. Thus, some variables were included in the adjusted model even if they themselves were not statistically significant.

Results

Univariate Analyses and Multivariate Analyses

For univariate analyses, descriptive statistics for our predictors and covariates are presented in Tables 2 and 3. For bivariate analyses, unadjusted GEE model outcomes are presented in Table 4. We estimated the bivariate impact of each predictor and covariate on the odds of taking a step, with each variable entered alone in the model. Below, we discuss model results when multiple measures of a single construct were included together in the model.

Table 2.

Participant Characteristics (N=39)

Variable	Mean	S.D.
Age (years)	53.3	9.5
Years since competitive employment	12.5	10.8
Veterans Rand-12 Physical Health Score	45.6	11.4
Veterans Rand-12 Mental Health Score	42.0	13.6
Work Importance1	4.0	1.9
Work Confidence1	3.7	2.2
Total number of MI sessions	5.3	1.2
	N	%
Male	35	89.7%
Race/ethnicity
White, Hispanic	2	5.1%
White, not Hispanic	26	66.7%
Black, not Hispanic	8	20.5%
Asian/Pacific Islander	1	2.6%
Multi-racial	2	5.1%
Marital status
Never married	18	46.2%
Married/with partner	5	12.8%
Separated/divorced/widowed	16	41.0%
Educational level
High school/GED or less	29	74.4%
Some college	3	7.7%
BA or more	7	17.9%
Serious mental illness diagnosis (current)
Bipolar I	10	25.6%
Depression with psychotic features	3	7.7%
Psychosis NOS	4	10.3%
Schizoaffective disorder	16	41.0%
Schizophrenia	6	15.4%
Other diagnosis (lifetime)
Posttraumatic Stress Disorder	15	38.5%
Substance abuse or dependence	28	71.8%
Any current disability benefits	32	82.1%

¹Scale: 1 to 7 with higher scores indicating greater levels of work importance/confidence

Table 3.

MI Predictor Variables Across All Sessions (N = 195)

Variable
	Mean	S.D.
Participant talk per session
Change Talk frequency	7.02	6.51
Change Talk intensity	4.60	4.43
Sustain Talk frequency	1.76	3.61
Sustain Talk intensity	1.21	2.78
Counselor talk behaviors per session
Proportion of questions that are open-ended	.52	.17
Proportion of reflections that are complex	.65	.14
Ratio of reflections to questions	1.60	.09
Counselor adherence to MI principles1
Relational (Collaboration, Empathy)	4.19	0.28
Technical (Cultivate Change Talk, Soften Sustain Talk)	4.08	0.31

¹Scale: 1 to 5 with higher scores indicating greater adherence

Table 4:

Unadjusted Odds Ratios (OR) of Taking a Step

Predictor Variables	OR	95% CI
		LL	UL
Participant Change Talk
Frequency	1.20****	1.09	1.31
Intensity	1.37****	1.18	1.59
Participant Sustain Talk
Frequency	.81***	.72	.91
Intensity	.72****	.60	.87
Counselor MI talk behaviors
Proportion of questions that are open-ended	.27	.03	2.71
Proportion of reflections that are complex	.16	.01	3.60
Ratio of reflections to questions	.80	.49	1.29
Counselor adherence to MI principles
Relational	1.83	.44	7.70
Technical	8.42**	1.86	38.14
Covariates
Session number	1.20*	1.03	1.42
Months between baseline and session	1.16*	1.01	1.34
Work Importance1	.86	.66	1.11
Work Confidence1	1.21	.96	1.52
Years since competitive employment	.96*	.92	1.00
Serious mental illness diagnosis (current)
Depression with psychotic features/Psychosis NOS	ref
Schizophrenia/Schizoaffective disorder	.88	.21	3.75
Bipolar I	.81	.16	4.28
Substance use disorder (lifetime)
Dependence	ref
Absent	.28*	.09	.91
Abuse	1.14	.17	7.71
Posttraumatic Stress Disorder (lifetime)	1.12	.39	3.23
Veterans Rand-12 Physical Health Score	1.03	.99	1.07
Veterans Rand-12 Mental Health Score	.97	.93	1.02
Any current disability benefits	1.31	.28	6.09
Age	.96	.91	1.01
Gender Male	.86	.20	3.75
Race/Ethnicity
White, non-Hispanic	ref
Black, non-Hispanic	6.54**	1.82	23.41
Other (Hispanic, Asian, Multi-Racial)	1.55	.24	9.82

¹Scale: 1 to 7 with higher scores indicating greater levels of work importance/confidence

^*p≤.05;

^**p≤.01;

^***p≤.001;

^****p≤.0001

Predictor Variables.

Both frequency (OR = 1.20, p<.0001) and intensity (OR = 1.37, p<.0001) of Change Talk, with each included alone in the model, led to greater odds of taking a step. When they were included in the model together, the impact of intensity of Change Talk tripled (adj OR = 3.82, p=.003), whereas frequency of Change Talk (adj OR = 0.51, p=.019) now led to a 50% decrease in the odds of taking a step. Thus, in the multivariate model, we used intensity of Change Talk rather than frequency of Change Talk as the measure of client Change Talk. Because the coefficient for intensity of Change Talk tripled when frequency of Change Talk was contained in the model, we also estimated the final model with frequency of Change Talk included.

Both frequency (OR =0.81, p=.001) and intensity (OR=0.72, p<.0001) of Sustain Talk reduced the odds of taking a step, with each included alone in the model. When they were included in the model together, neither was statistically significant. Thus, in the multivariate model, because of its greater effect size, we use intensity rather than frequency of Sustain Talk.

None of the three measures of counselor talk behavior had a statistically significant effect on the odds of taking a step, either when in the model alone or when all three were in the model together. Counselor adherence to MI relational principles (Empathy, Collaboration) did not affect the odds of taking a step. However, counselor adherence to MI technical principles (Cultivating Change Talk, Softening Sustain Talk) did impact the odds of taking a step (OR=8.42, p=.006). Results were similar when these two adherence measures were included in the model together. When counselor adherence to the relational principles was included in the model, a one-unit change in the rating of counselor adherence to the technical principles led to a thirteen-fold increase in the odds of taking a step (adj OR=13.56, p=.001).

Covariates.

When included in the model alone, unadjusted odds of taking a step increased over each additional session (OR=1.20, p=.023). An increase in number of months between baseline and the session also was associated with an increase in the odds of taking a step (OR=1.16, p=.033). However, when these variables were included in the model together, neither was significant. In this combined model, the effect of session number (adj OR=1.26, p=.549) was similar to its effect when included in the model alone, whereas the effect of months between baseline and the session was now negligible (adj OR = 0.96, p=.900). Thus, in the multivariate model, we included session number rather than months since baseline.

An increase in number of years since competitive employment was associated with a small decrease in the odds of taking a step (OR=0.96, p=.047). Neither Work Importance nor Work Confidence was statistically significant when included in the model alone. Nonetheless, when included in the model together, Work Confidence was associated with a 50% increase in the odds of taking a step (adj OR = 1.45, p=.012), and Work Importance was associated with a 30% decrease in the odds of taking a step (adj OR= 0.69, p= .020). Receipt of any disability benefit had no impact on the odds of taking a step.

Lifetime diagnosis of substance use disorder was the only aspect of participant mental health or physical health that had an impact on the odds of taking a step. Participants without a history of substance abuse or dependence were 70% less likely to take a step (OR=0.28, p=.034).

Finally, we examined whether the odds of taking a step varied with participant demographic characteristics. Age and gender were not significant when each was included alone in the model. However, participants who identified as Black and not Hispanic had more than six times the odds of taking a step compared to those who identified as White and not Hispanic (OR=6.54, p=004).

Multivariate Models (Adjusted).

Final multivariate models are in Table 5. All predictors and covariates that were statistically significant in the unadjusted models were considered for these models. Thus, they estimate the combined effects of client talk behaviors and counselor adherence to MI technical principles (Cultivating Change Talk, Softening Sustain Talk) on the odds of taking a step.

Table 5:

Multivariate Models: Adjusted Odds Ratio (OR) of Taking a Step

Predictor Variables	Model 5b			Model 5a
	OR	95% CI		OR	95% CI
		LL	UL		LL	UL
Participant Change Talk
Frequency	.68	.42	1.09
Intensity	2.82*	1.26	6.34	1.55****	1.27	1.90
Participant Sustain Talk
Intensity	.55****	.44	.68	.53****	.42	.66
Counselor adherence to MI principles
Technical	9.46*	2.04	43.73	10.98**	2.22	54.30
Covariates
Session Number	1.40**	1.10	1.80	1.38**	1.09	1.76
Work Importance2	.71*	.54	.94	.74*	.56	.98
Work Confidence2	1.33*	1.04	1.70	1.34*	1.05	1.72
Years Since CE1	.93****	.89	.97	.92****	.89	.96

¹CE: Competitive Employment

²Scale: 1 to 7 with higher scores indicating greater levels of work importance/confidence

^*p≤.05;

^**p≤.01;

^***p≤.001;

^****p≤.0001

Predictor Variables.

The magnitude and significance of the odds ratios in these models were quite similar to their magnitude and significance in the unadjusted models. Thus, participant Change Talk and Sustain Talk, and counselor adherence to the technical principle of Cultivating Change Talk and Softening Sustain Talk had significant effects on the odds of participants taking a step.

For reasons described above, we estimated two final models. Model 5a estimated impact of Change Talk and Sustain Talk by including only intensity of Change Talk (adj OR = 1.55, p<.0001) and intensity of Sustain Talk (adj OR = 0.53, p<.0001) in the model. Model 5b, discussed below, also included frequency of Change Talk in the model. While the impact of frequency of Change Talk was not statistically significant (adj OR=.68, p=.109), its inclusion in Model 5b doubled the impact of intensity of Change Talk (adj OR=2.82, p=.012) and left the impact of intensity of client Sustain Talk unchanged (adj OR=0.55, p<.0001).

Only one measure of counselor performance had a significant impact on the odds of a participant taking a step: counselor adherence to MI technical principles (Cultivating Change Talk, Softening Sustain Talk) and its impact was large in magnitude. A one-unit change in counselor adherence to MI technical principles led to an almost tenfold increase in the odds that a participant would take a step toward employment (adj OR=9.46, p=.004).

Covariates.

The final multivariate model controlled for session number, work importance, work confidence, and years since competitive employment. Once again, the magnitude and significance of the odds ratios for these variables were quite similar to their magnitude and significance in the unadjusted models. However, lifetime diagnosis of substance abuse/dependence and race/ethnicity were no longer statistically significant and were not included in the final multivariate models.

Each additional session increased the adjusted odds of a participant taking a step by 40% (adj OR = 1.40, p=.007). Work confidence was associated with a 30% increase in the odds of taking a step (adj OR = 1.33, p=.022), whereas work importance was again associated with a 30% decrease in the odds of taking a step (adj OR = 0.71, p=.015). Finally, years unemployed was associated with a 7% decrease in the odds of taking a step (adj OR = 0.93, p<.0001).

Discussion

A major finding of importance to the field of psychiatric rehabilitation is that, consistent with past research and the Miller & Rose (2009) MI model, we found that client Change Talk and Sustain Talk have respectively positive and negative associations with taking steps toward employment for unemployed participants with SMI.

Our straightforward coding scheme for intensity of client talk, with “moderate” operationalized as the default rating, made it easy to assign intensity ratings. It should be easy for counselors to assess Change Talk and Sustain Talk as high or low intensity within the context of a clinical encounter and to use this as a way to predict the likelihood that the consumer will make a behavioral change. Noting client Change and Sustain Talk may benefit psychiatric rehabilitation practitioners as they self-monitor their MI skills and may help them increase the likelihood of positive consumer change by improving their skill in cultivating change talk and softening sustain talk.

The lack of relationship between steps and counselor adherence to the relational principles of Collaboration and Empathy is consistent with past findings (Magill et al., 2019; Morgenstern et al., 2017). Similar to findings from the meta-analysis conducted by Magill and colleagues (2018), scores on the MI relational components had a restricted range from average to high, most likely because MI counselors receive close supervision in a research study. Studies of MI in clinical practice with a wider range of scores may be helpful for identifying the potential benefits of these relational components of MI (Magill et al, 2018).

Results were mixed with respect to counselor demonstration of MI-consistent talk behavior. Our findings did not support the hypothesis that higher levels of MI-consistent counselor talk behavior (indicated by proportion of questions that are open questions, proportion of reflections that are complex reflections, and ratio of reflections to questions) would be associated with taking steps toward employment. There are several potential explanations. One is that our study included each MI session that took place over the 6-month intervention period. This is different from studies that include only single-session interventions. Thus, it is possible that MI-consistent talk behavior has less impact over multiple sessions than in a single session. Another explanation could be that taking a step toward employment is different from curbing addiction behaviors, the most common MI target; different elements of MI may impact different types of target behaviors. A third explanation is that our study population differs enough from other MI studies that different elements of MI emerged as predictive of change. As noted by Martino (2007), the standard MI approach without some modifications may not be a “best practice” for clients who have psychotic disorders. Further investigations of MI for individuals with SMI may identify other MI modifications, help explicate how MI affects the quality of the client/counselor relationship, and provide guidance for deciding under what conditions the use of standard MI may be contraindicated.

Counselor adherence to MI technical principles (Softening Sustain Talk and Cultivating Change Talk) was associated with taking steps toward employment. Consideration of whether these principles independently contribute to behavior change through increased client Change Talk or reduced Sustain Talk or are a reflection of existing levels of client Change Talk or Sustain Talk, is to be determined. The finding that the odds ratio for adherence to MI technical principles is similar in the unadjusted and adjusted models suggests there may be unique variance captured by this counselor variable independent of the variance associated with the client Change Talk and Sustain Talk variables. MI theory posits a pathway model of change such that the evocation and selective reflection of Change Talk increases frequency of the client’s intention to change, which then predicts actual behavior change (Miller & Rose, 2009). Our findings support the association of these factors with behavior change, if in a less direct manner than suggested by MI theory.

The finding that the number of sessions contributed to taking a step toward employment is consistent with prior reports of a dose-response effect within MI (Burke, et al. 2003), suggesting that repeated conversations about work may be associated with increased likelihood of SE program entry and/or of seeking employment even without SE support.

The positive association between baseline work confidence and taking a step is consistent with self-efficacy literature (Bandura, 1994). But the impact of self-efficacy, once receiving SE, is mixed (Waynor, Gill, & Goa, 2016; Regenold, Sherman, & Fenzel, 1999), suggesting that it may be beneficial for clinicians to address work confidence to better facilitate attaining/retaining employment.

It is more difficult to explain the finding that higher levels of work importance were associated with a decreased likelihood of taking a step. The work importance scale was based on an MI clinical tool for assessing the importance of change rather than to be a reliable and valid measure of work importance. Thus, the finding may be a limitation of the scale to capture work importance. Of note however, a very similar single-item measure of “centrality of work” (Corbiere et al., 2011), has been positively associated with job search behavior. Another possible explanation is that participants were not required to have work goals to enroll in the study.

A limitation of this study is that we artificially chunked the impact of a session by defining a step as only related to that session, whereas taking a step could possibly have been impacted by any previous session. Another limitation is that the coders who rated the sessions for the MI variables also identified the steps and were not blind to the study hypotheses which could have influenced their ratings. Additionally, the coders also served as MI counselors; thus, they may have had some bias related to that role. Of note, the MI counselors were distinct from the participants’ mental health providers and SE specialists. This structure was created for the purposes of training and fidelity of the sessions. It is possible that the elements of MI related to taking a step toward employment could be different for MI sessions conducted by a person’s existing providers.

Other limitations include the impact of variables not included in the model as well as the way variables in the model were coded. For example, cognitive impairment has been associated with both entry into vocational rehabilitation programs and work outcomes (O’Connor et al., 2011; McGurk & Mueser, 2004) and could be impacting the relationships among the variables described here. Also, our coding of receipt of disability income simply as present/absent across multiple sources rather than as an amount of income or specific type of benefit may have limited its impact in the model. There is a significant literature demonstrating the negative relationship between disability benefits and competitive employment outcomes, interest in working, and motivation (e.g., Rosenheck et al., 2017). Additionally, although race was not a primary focus of the study and race did not remain statistically significant when included in the full model, it could be an important focus for future research. Finally, the rating of steps as a dichotomy rather than on an ordinal scale limited the precision of our findings and could be examined in a future study.

Overall, this study suggests that counselors should adopt a counseling approach that cultivates client Change Talk and softens Sustain Talk. It supports the recommended SE best practice regarding having ongoing conversations about work with individuals with SMI (Becker & Drake, 2003). For counselors who are learning MI, study findings suggest that learning how to cultivate change talk and soften sustain talk may be the most important MI counseling skills for facilitating their clients’ behavior change.

For psychiatric rehabilitation practitioners, it is noteworthy that the MI relational factors are quite consistent with SAMHSA’s principles of recovery-oriented services as operationalized by Ellison and colleagues (2018). Counseling sessions that are individualized and person-centered, and that acknowledge the consumer’s autonomy and strengths are consistent with MI relational elements. Thus, counselors who have a recovery-oriented approach to psychiatric rehabilitation and who relate in a collaborative empathic manner that encourages autonomy are providing counseling that fulfills the core relational foundations of MI.

In conclusion, we recommend further research aimed at understanding the MI process for individuals with SMI. It is particularly important to continue to study MI relational variables, given their strong theoretical rationale. A core concept in MI is that MI relational components provide the foundation upon which the technical components are implemented (Miller & Rose, 2009). Thus, the field of psychiatric rehabilitation, as well as MI counseling in general, could benefit from researchers conducting more naturalistic investigations focused on the relational components and pathways of MI (Magill & Hallgren, 2019).

Impact and Implications.

In Motivational Interviewing (MI) sessions with unemployed veterans who have serious mental illness, the MI components that had the greatest impact on taking a step toward employment were how intensively the client talked about moving toward employment and how much the counselor cultivated this change talk and softened any talk about staying unemployed. These findings provide guidance for counselors who are helping individuals with SMI to take steps toward employment.