Working alliance and empathy as mediators of brief telephone counseling for cigarette smokers who are not ready to quit

Abstract

Working alliance and empathy are believed to be important components of counseling, though few studies have empirically tested this. We recently conducted a randomized controlled trial in which brief motivational and reduction counseling failed to increase the number of participants who made a quit attempt (QA) in comparison to usual care (i.e., brief advice to quit). Our negative findings could have been due to non-specific factors. This secondary analysis used a subset of participants (n=347) to test a) whether, in comparison to usual care, brief telephone-based motivational or reduction counseling predicted greater working alliance or empathy, b) whether changes in these non-specific factors predicted an increased probability of a QA at a 6-month follow-up, and c) whether counseling affected the probability of a QA via working alliance or empathy (i.e., mediation). Findings were similar for both active counseling conditions (motivational and reduction) vs usual care. In comparison to usual care, active counseling predicted greater working alliance (p<.001) and empathy (p<.05). Greater working alliance predicted a greater probability of a QA (p<.001) but, surprisingly, greater empathy predicted a decreased probability of a QA (p<.05) at the 6-month follow-up. Both working alliance (p<.001) and empathy (p<.05) mediated the active counseling's effects on the probability of a QA. One explanation for our motivational and reduction interventions' failure to influence QAs in comparison to usual care is that working alliance and empathy had opposing effects on quitting. Our analyses illustrate how testing non-specific factors as mediators can help explain why a treatment failed.

Most adult smokers are not ready to make a quit attempt (QA) in the near future (PROPEL Centre for Population Health Impact, 2014). Making a QA predicts future cessation (Hughes et al., 2014). Thus, increasing QAs in those not ready to quit is one strategy to decrease the prevalence of smoking.

Both motivational interviewing and nicotine replacement therapy (NRT)-aided interventions to reduce cigarettes per day (CPD) increase QAs among smokers who are not ready to quit (Lindson-Hawley, Thompson, & Begh, 2015; Wu, Sun, He, & Zeng, 2015). Our prior randomized controlled trial (RCT) found that brief telephone-based motivational counseling increased the odds of making a QA five-fold and reduction counseling with NRT increased the odds of making a QA four-fold (Carpenter et al., 2004).

We recently conducted a second RCT to replicate our initial findings of the efficacy of brief motivational counseling and to test whether brief reduction counseling without NRT is effective. However, neither motivational nor reduction (without NRT) interventions increased the odds of making a QA more than usual care (i.e., brief advice to quit) at a 6-month follow-up (Klemperer, Hughes, Solomon, Callas, & Fingar, 2016).

Mediation analysis is one method to determine why interventions in trials are ineffective (McCarthy et al., 2010). Specifically mediation analyses can test whether negative results are due to the intervention not influencing the mediator variables or due to the mediators not influencing the outcomes, or both (Hayes, 2013). We examined participants' working alliance and empathy with their counselors as mediators of counseling because these non-specific factors appear to be important components of counseling (Elliott, Bohart, Watson, & Greenberg, 2011; Horvath, Del Re, Fluckiger, & Symonds, 2011) and substance abuse interventions (Meier, Barrowclough, & Donmall, 2005; Miller & Moyers, 2015), and could be important components of smoking cessation interventions (Michie, Hyder, Walia, & West, 2011).

Working alliance is the quality of relationship between client and counselor that develops around the goals, cooperation, mutual trust, confidence in, and liking of one another (Horvath & Greenberg, 1989). Greater working alliance predicts early engagement and positive outcomes in many (Coco, Gullo, Prestano, & Gelso, 2011; Meier et al., 2005) but not all (Feldstein & Forcehimes, 2007) studies on brief counseling for substance use disorders. The only study that examined working alliance as a predictor of cessation among smokers found that it did not predict cessation in an 8-week intensive mindfulness program (Goldberg, Davis, & Hoyt, 2013).

Empathy is the extent to which a counselor takes their patient's perspective and understands their frame of reference (Burns & Nolen-Hoeksema, 1992; Duan & Hill, 1996). Empathy is considered a core component of motivational interviewing (Miller & Moyers, 2015; Smedslund et al., 2011) and a strong predictor of outcomes for a variety of behavior therapies (Elliott et al., 2011). However, empathy is understudied in brief interventions for smokers. One unpublished trial found that greater empathy from counselors predicted a greater probability of cessation (Tutty, McAfee, Mahoney, Wassum, & Roberts, 2010).

In a taxonomy of behavior change techniques (BCT) for stop smoking services, Michie and colleagues identified a group of non-specific factors associated with interactions between counselor and patient (Michie et al., 2011). One of these BCTs, “build general rapport,” encompasses both working alliance and empathy (Michie et al., 2011). However, few smoking cessation treatment manuals (<25%) include “build general rapport” as an explicit goal, and the manuals that do include this BCT are not more successful than those that do not (Bartlett, Sheeran, & Hawley, 2014; Lorencatto, West, & Michie, 2012; West, Walia, Hyder, Shahab, & Michie, 2010). Importantly, we know of no study that has examined patients' report of this BCT as a predictor of outcome in counseling for smoking cessation.

This secondary analysis of our recently completed RCT (Klemperer et al., 2016) uses mediation analysis to examine whether, in comparison to usual care, our motivational and reduction interventions' failure was due to 1) counselors' failures to establish a relationship with participants (i.e., develop working alliance and express empathy); or 2) a failure of these non-specific factors' to influence the probability of a QA.

Method

Participants

The RCT was approved by the University of Vermont's Institutional Review Board and enrolled 560 adult smokers of ≥ 10 CPD who stated that they were not ready to quit in the next 30 days. Participants were recruited via email invitations to the Nielsen (www.nielsen.com) consumer panel of over 350,000 participants who use the Internet and elected to receive invitations for surveys in return for online store credit. The email invitations did not refer to smoking cigarettes. We added the measures for this secondary analysis after the first 83 participants were enrolled. Among the 477 that enrolled after this, 130 did not complete their initial counseling call. Thus, data from 347 participants who had been randomly assigned to receive usual care (n=117), motivational counseling (n=118), or reduction based counseling (n=112) were included in these analyses.

Participants were predominantly middle-aged (mean age=52), Caucasian (88%), and female (67%). They smoked a mean of 20 CPD at baseline, and were moderately dependent (mean Fagerstrom Test for Cigarette Dependence [FTCD]=5.4 (Fagerstrom, 2012); Table 1). There were no significant differences among study groups.

Table 1.

Demographic information and smoking history.

Demographics	Usual Care (n=117)	Motivational Intervention (n=118)	Reduction Intervention (n=112)	Total (n=347)	National Health Interview Survey: Current daily smokers (N=3,362)‡
Mean Age (SD)	51 (10)	52 (10)	53 (11)	52 (10)	42
% Women	64	70	66	67	46
% White, non-Hispanic	86	92	87	88	78
% Some college or more	68	67	69	68	39
% Employed full time	39	32	35	35	-
% Married	42	48	55	48	-
Smoking History
Mean CPD (SD)	20 (8.2)	20 (8.8)	19 (8.3)	20 (8.4)	16
Mean QA in life (SD)	3.6 (5.8)	3.9 (8.6)	3.7 (4.1)	3.7 (6.4)	-
Mean age started smoking (SD)	18 (8.6)	19 (7.8)	18 (6.5)	18 (7.7)	17
Mean FTCD (SD)	5.3 (2.1)	5.6 (2.0)	5.2 (2.1)	5.4 (2.1)	4.3–4.6†
Mean Intention to Quit (SD)	3 (2.3)	2.9 (2.4)	2.8 (2.5)	2.9 (2.4)	-
Median longest QA in Days	21	30	46	30	-

^†(Fagerstrom & Furgerg, 2008);

^‡(Hughes & Callas, 2010);

CPD=Cigarettes per day; FTCD=Fagerstrom Test for Cigarette Dependence (1=lowest dependence, 10=highest dependence); QA=Quit attempts; Intention to quit in the next month (0=very definitely no, 10=very definitely yes); SD=Standard deviation.

Interventions

We conducted all interventions by telephone and did not provide any medication. The motivational and reduction conditions received three 10–15 minute counseling calls at baseline, week 2, and week 4. Counselors concluded with advice to quit smoking at the last call. The usual care condition consisted of a single 5-minute call at baseline.

Motivational intervention

Brief motivational counseling was based on the United States Public Health Service 5Rs guidelines (Fiore, 2008) and was a replication of the motivational intervention in our initial study (Carpenter et al., 2004). The intervention included certain MI strategies (e.g., develop discrepancy and support self-efficacy) but not others (e.g., roll with resistance). The intervention focused on participants' a) relevant reasons for quitting, b) risks of smoking of concern to the smoker, c) rewards of smoking cessation, d) roadblocks to quitting, and e) repetition of the topics (Fiore, 2008).

Reduction intervention

The reduction counseling was an update of the treatment found to be effective in our initial RCT of NRT-aided reduction (Carpenter et al., 2004), except no NRT was used to aid reduction. Counselors encouraged participants to set their own goals for reducing CPD and helped them choose one of two strategies: a) scheduled reduction; i.e., smoking on a schedule and increasing time between cigarettes, or b) hierarchical reduction; i.e., eliminating certain cigarettes beginning with those that are the easiest to give up (Cinciripini, Lapitsky, Seay, Wallfisch, & Kitchens, 1995; Riggs, Hughes, & Pillitteri, 2001).

Usual care

The usual care condition was based on a prior description of usual care (Flocke & Stange, 2004). Counselors asked questions about the participants' smoking, provided brief advice to quit, and offered treatment information.

Counselors

The four counselors had or were pursuing graduate degrees in social work or clinical psychology and completed 13 hours of training guided by treatment manuals for each treatment condition. One author (LS) provided the training and observed counseling calls approximately five times per month to ensure fidelity to the interventions. Each counselor provided motivational, reduction, and usual care counseling. Once assigned to a counselor, most participants (86%) received all their calls from the same counselor to maintain continuity of care. Participants' assigned counselor did not significantly predict working alliance, empathy, or the probability of making a QA.

Assessments

Assessments of working alliance and empathy occurred approximately 24 hours after each counseling call. Assessments of QAs occurred weekly during the first 4 weeks and monthly for 6 months.

Non-specific factors

Participants completed a modified 12 item version of the Short Revised Working Alliance Inventory (WAI-SR) (Hatcher & Gillaspy, 2006). The WAI-SR is widely used and has good reliability (α>0.8) and convergent validity (r>0.6) (Hatcher & Gillaspy, 2006; Horvath et al., 2011; Munder, Wilmers, Leonhart, Linster, & Barth, 2010). The WAI-SR consists of Task, Goal, and Bond oriented subscales. We excluded all items on the Bond subscale because they appeared inappropriate to assess the relatively brief and structured interventions used in our trial. We also made minor wording changes to make the survey more applicable to our telephone counseling. Possible scores ranged from 1 (lowest alliance) to 5 (highest alliance).

Participants completed a modified 10-item version of the Empathy Scale (ES) (Persons & Burns, 1985). The ES has good internal consistency (α=0.7) and has been used in a prior study to measure smokers' perception of telephone counseling (Persons & Burns, 1985; Tutty et al., 2010). We made minor wording changes to make the questions more applicable to our telephone counseling. Possible scores ranged from −15 (lowest empathy) to 15 (highest empathy).

Outcome

Our outcome was whether or not participants made a QA that lasted ≥24-hours between baseline and the 6-month follow-up. We did not use biochemical verification because the Society for Research on Nicotine and Tobacco (SRNT) stated that verification was usually not necessary when treatment contact was minimal (SRNT Subcommittee on Biochemical Verification, 2002).

Data Analysis

We conducted all analyses using SPSS (IBM SPSS Statistics for Windows, Version 23.0. Armonk, NY: IBM Corp) and PROCESS, a statistical macro for conducting path and mediation analyses (Hayes, 2013). Despite prior recommendations (Baron & Kenny, 1986), mediation analysis in the absence of a significant main effect can provide useful information (Hayes, 2013).

We tested WAI-SR and ES scores as parallel mediators of the motivational vs usual care and reduction vs usual care interventions' influence on the probability of a QA (i.e., both WAI-SR and ES were included in each mediation model). We also conducted sensitivity analyses where WAI-SR and ES were analyzed independently as simple mediators. We determined a priori not to compare the motivational to the reduction intervention because a sample size large enough to detect the difference that would be expected between two accepted interventions with a dichotomous outcome was not feasible. We used ordinary least squares (OLS) regression to test the interventions' influence on WAI-SR and ES scores and maximum likelihood logistic regression to test WAI-SR and ES scores' influence on QAs. We also calculated indirect effects to test mediation; i.e., whether the interventions affected QAs via WAI-SR or ES scores. Bias-corrected bootstrap analyses (10,000 resamples) were used to estimate the 95% confidence intervals (CI) for the standardized beta coefficients (Hayes, 2013).

Results

Preliminary Analyses

At the 6-month follow-up, 39% of participants in the usual care, 44% in the motivational, and 38% in the reduction condition made a QA that lasted ≥ 24 hours. Twenty-two percent of participants missed a follow-up survey. Missing outcome data were not related to treatment condition. We assumed that those who missed a follow-up did not make a QA since the last follow-up (i.e., we treated missing outcome data as continued smoking). We conducted sensitivity analyses where we excluded missing data and there were no substantial differences in findings compared to the findings reported below.

Mean WAI-SR and ES scores are reported in Table 2. Both the WAI-SR (α=0.94) and the ES (α=0.85) had good reliability. Correlations between all included variables are reported in Supplemental Table 1 [Insert link to Supplemental Material here]. WAI-SR and ES scores from participants' last completed counseling call were used for the primary analyses to account for any development of working alliance or empathy across calls. We were unable to contact 25 participants (11% distributed evenly between the motivational and reduction conditions) for their third and final call, and thus, WAI-SR and ES data from their last counseling call were used. Whether or not a participant missed a counseling call was not associated with treatment condition or the probability of making a QA. We conducted a sensitivity analysis using mean WAI-SR and ES across the three calls for the motivational and reduction conditions, and there were no substantial differences in findings from those reported below.

Table 2.

Mean (SD) values for the non-specific factors reported after participants' final counseling call.

	Usual Care (n=117)	Motivational Intervention (n=118)	Reduction Intervention (n=112)
Working Alliance Scale: 1 (least) to 5 (most)	3.5 (1.0)	4.4 (0.7)	4.7 (0.4)
Empathy Scale: −15 (least) to 15 (most)	10.9 (4.9)	13.3 (2.6)	13.3 (3.2)

SD=Standard deviation.

Interventions' effects on non-specific factors

In comparison to usual care, both motivational and reduction interventions predicted greater WAI-SR and ES (Table 3).

Table 3.

Interventions as predictors of non-specific factors.

Outcomes (non-specific factors)	Motivational vs Usual Care Interventions: Standardized Beta (95% CI)	Reduction vs Usual Care Interventions: Standardized Beta (95% CI)
Working Alliance	1.03 (0.80 to 1.26)***	1.22 (1.02 to 1.43)***
Empathy	0.57 (0.33 to 0.81)***	0.55 (0.30 to 0.81)***

^***p<.001;

CI=Confidence interval.

Non-specific factors' effects on outcome

Greater WAI-SR predicted a greater probability of making a QA in both models (i.e., motivational vs usual care and reduction vs usual care). Surprisingly, greater ES predicted a decreased, not increased, probability of making a QA in both models (Table 4).

Table 4.

Non-specific factors as predictors of outcome.

Predictors (non-specific factors)	Probability of a Quit Attempt in the Motivational vs Usual Care Model Standardized Beta (95% CI)	Probability of a Quit Attempt in the Reduction vs Usual Care Model Standardized Beta (95% CI)
Working Alliance	0.90 (0.49 to 1.31)***	1.02 (0.53 to 1.51)***
Empathy	−0.40 (−0.75 to −0.06)*	−0.45 (−0.80 to −0.10)*

^*p<.05;

^***p<.001;

CI=Confidence interval.

Mediation: indirect effects of interventions on outcomes

In a parallel mediator model, both WAI-SR and ES mediated the motivational and reduction interventions' effects on the probability of making a QA in comparison to usual care. Both interventions predicted greater WAI-SR, which increased the probability of making a QA. Both interventions also increased ES in comparison to usual care, but greater ES predicted a decreased probability of making a QA (Table 5). When WAI-SR and ES were examined in separate models as independent mediators, findings for WAI-SR were similar to that reported above. ES, however, did not significantly mediate the motivational or reduction interventions' effects on the probability of a QA without WAI-SR in the model.

Table 5.

Indirect effects of the interventions on the outcome via working alliance and empathy when both were tested as parallel mediators.

Intervention	Mediator (non-specific factors)	Outcome	Indirect Effect: Standardized Beta (95% CI)
Motivational vs Usual Care	Working Alliance	Quit Attempt	0.93 (0.52 to 1.49)***
Motivational vs Usual Care	Empathy	Quit Attempt	−0.23 (−0.49 to −0.04)*
Reduction vs Usual Care	Working Alliance	Quit Attempt	1.25 (0.65 to 2.00)***
Reduction vs Usual Care	Empathy	Quit Attempt	−0.25 (−0.53 to −0.07)*

^*p<.05;

^***p<.001;

CI=Confidence interval.

Discussion

The primary findings were similar for motivational vs usual care and reduction vs usual care comparisons (Figure 1). The motivational and reduction counseling's influence on working alliance and empathy suggests that our interventions' failure to increase the probability of a QA in comparison to usual care was not due to our counselors' inability to establish a relationship with participants. These findings could be due to the fact that motivational and reduction counseling were longer and more collaborative (three 15-minute calls to identify the pros and cons of smoking or reduce CPD) than usual care (one 5-minute call with brief advice to quit).

Figure 1.

Summary of mediation findings for both motivational vs usual care and reduction vs usual care comparisons.

Greater working alliance predicted an increased probability of making a QA. This finding is consistent with a large body of literature that suggests working alliance is crucial to counseling for substance use disorders (Meier et al., 2005; Miller & Moyers, 2015; Norcross & Wampold, 2011). Our findings provide empirical support for the importance of establishing working alliance in brief interventions for smokers; a strategy that has previously been proposed as an “active ingredient” of behavior change techniques (Michie et al., 2011).

Surprisingly, greater empathy from counselors predicted a decrease, not increase, in the probability of making a QA. However, empathy only mediated the interventions' influence on the probability of a QA when the mediator model accounted for working alliance. Nonetheless, empathy's negative influence on QAs could be one reason why our motivational and reduction interventions did not have a significant total effect on the probability of making a QA in comparison to usual care. This finding, however, contradicts prior research that suggests empathy is an important predictor of outcome in psychotherapy (Elliott et al., 2011; Norcross & Wampold, 2011) and a core component of substance abuse and smoking cessation interventions (Miller & Moyers, 2015; Smedslund et al., 2011; Tutty et al., 2010). Thus, future research is needed to replicate our unexpected findings regarding empathy's negative influence on making a QA. One explanation for this finding is that participants could have interpreted counselors' empathy as support for participants' perceived inability to quit, which may have enabled them to postpone change (i.e., make a QA).

One potential limitation is that participants' responses could have been subject to a desire to please their counselor. However, data were collected approximately 24 hours after each call in order to minimize demand characteristics. Another limitation is that we did not include a measure to test counselors' fidelity to the treatment manuals. For practical reasons, counselors (non-medical professionals) delivered the interventions via telephone which could have resulted in less working alliance or empathy than in-person interventions from professional clinicians. Comparisons between our findings for smokers not ready to quit and previous literature are limited because prior research on non-specific factors often recruited participants motivated for treatment (Elliott et al., 201; Horvath et al., 2011). Further, the exclusion of the “Bond” scale from the WAI-SR limits our ability to interpret findings.

Conclusions

Non-specific factors may be important components of counseling for smoking cessation (Michie et al., 2011) and could explain our interventions' failure to increase the probability of a QA. In comparison to usual care, brief motivational and reduction counseling predicted greater working alliance and empathy between counselors and participants. Greater working alliance predicted an increased, but greater empathy predicted a decreased probability of making a QA. Thus, one explanation for our interventions' failure could be that working alliance and empathy had opposing effects on quitting. Future research is needed to replicate our finding that empathy decreased the probability of making a QA. Our analyses illustrate how testing non-specific factors as mediators can help explain why a treatment failed.