David and Sara have established and managed their family business, a medical device company, for over 30 years. David is the

https://doi.org/10.1177/1359105316676330

Journal of Health Psychology 2018, Vol. 23(5) 743 –753 © The Author(s) 2016 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/1359105316676330 journals.sagepub.com/home/hpq

For the modern clinician, it is not enough to provide competent medical care; clinicians must also motivate their patients through inter- personal exchanges to achieve optimal levels of patient satisfaction and health (DiMatteo et al., 2012). For example, clinicians’ nonverbal behavior such as tone of voice, posture, and use of humor all play a role in patients’ satisfaction, trust, and adherence (Wrench and Booth- Butterfield, 2003). Unfortunately, the list of cli- nicians’ behaviors that researchers have identified as motivational is long and unfo- cused, such that it would be nearly impossible to incorporate all of these idiosyncratic behav- iors into a cohesive behavior modification strat- egy to improve patient care. To counter this challenge, we propose a theoretical approach to instead identify underlying mechanisms that

render each behavior effective and then cluster those mechanisms according to their similari- ties (Huynh and Sweeny, 2014). We suggest that these clusters of behaviors, which we refer to as patient care styles (PCSs), can be organ- ized through the lens of transformational lead- ership theory (Bass and Riggio, 2006). In this article, we examine the utility of this approach for predicting patient outcomes.

Transformational leadership in primary care: Clinicians’ patterned approaches to care predict patient satisfaction and health expectations

Ho Phi Huynh1, Kate Sweeny2 and Tricia Miller2

Abstract Clinicians face the complex challenge of motivating their patients to achieve optimal health while also ensuring their satisfaction. Inspired by transformational leadership theory, we proposed that clinicians’ motivational behaviors can be organized into three patient care styles (transformational, transactional, and passive-avoidant) and that these styles differentially predict patient health outcomes. In two studies using patient-reported data and observer ratings, we found that transformational patient care style positively predicted patients’ satisfaction and health expectations above and beyond transactional and passive-avoidant patient care style. These findings provide initial support for the patient care style approach and suggest novel directions for the study of clinicians’ motivational behaviors.

Keywords health education, health expectations, health psychology, patient satisfaction, transformational leadership

1Armstrong State University, USA 2University of California, Riverside, USA

Corresponding author: Ho Phi Huynh, Department of Psychology, Armstrong State University, 11935 Abercorn Avenue, Savannah, GA 31419, USA. Email: [email protected]

676330HPQ0010.1177/1359105316676330Journal of Health PsychologyHuynh et al. research-article2016

Article

744 Journal of Health Psychology 23(5)

Transformational leadership theory stands as the most widely researched approach to under- standing motivational leadership behaviors (Bass and Riggio, 2006). Its popularity stems from the fact that leaders’ effectiveness is not judged by individual behaviors but rather by styles, or clusters of behaviors grouped together based on their effectiveness for motivating fol- lowers. Although the study of leadership typi- cally concerns leaders and work-group members in business settings, we extend the scope of leadership research to include clinician–patient relationships. Considering the similar dynamics of leader–follower and clinician–patient rela- tionships (e.g. repeated interactions and uneven power status; French and Raven, 1959), and given the comparable motivational goals of the two parties (i.e. one party motivating the other party toward the completion of a task), the inte- gration of leadership and health research can yield valuable insights for the process and delivery of quality patient care (Gabel, 2012).

The PCSs framework does not supersede pre- vious models of patient care, such as patient-cen- tered care (Epstein and Street, 2011; Stewart et al., 2000) or patient empowerment (Anderson et al., 1995; Wallerstein, 1992). Rather, the model incor- porates and organizes factors such as goal setting, patient autonomy, and motivation into a frame- work that not only emphasizes the role of the patient but also the important dynamics between the clinician and patient. In addition to serving as an organizing framework, transformational lead- ership theory allows for specific predictions about the effectiveness of each style, which have been consistently and significantly supported through cross-sectional and longitudinal studies and experiments with random assignment (Judge and Piccolo, 2004). Finally, transformational leader- ship research empowers institutions to systemati- cally train leaders to be more effective at motivating followers in a variety of settings (Collins and Holton, 2004). Therefore, using transformational leadership as the theoretical framework, we can make predictions about which PCS will likely be most effective, and these insights can ultimately aid clinicians in becoming more effective motivators.

PCSs

Essential to the transformational leadership framework is the idea that leaders can shift through the full range of styles, but the style they display most frequently represents their primary style (Bass and Riggio, 2006). An updated version of the traditional leadership model contains three primary styles: transfor- mational, transactional, and passive-avoidant leadership. Each style comprises several com- ponents, as described in greater detail below.

Just as there are three distinct leadership styles, we propose that there are three distinct PCSs. Similarly, we propose that clinicians can shift through a range of styles, but the style they display most frequently represents their “habit” and thus their primary PCS. Evidence suggests that clinicians do indeed have habits. For exam- ple, clinicians can practice the four habits of medical interviewing, which include eliciting the patient’s perspective, demonstrating empa- thy, and investing in the beginning and the end of the interaction, to increase the flow of medi- cal visits (Frankel and Stein, 1999). In the fol- lowing sections, we describe each PCS and its components.

Transformational PCS

Transformational PCS characterizes clinicians who not only create health plans for patients and monitor their progress but also inspire and motivate patients to achieve to those goals. The transformational PCS includes four compo- nents (“the four I’s”): idealized influence, inspi- rational motivation, intellectual stimulation, and individualized consideration (Bass and Avolio, 1991). These components are not mutu- ally exclusive; clinicians can engage in any or all of these components in order maximize their effectiveness as care providers.

Clinicians display idealized influence when they serve as role models for patients (Anderson et al., 1987). They display good health habits, which are often apparent to patients (e.g. non- smoking and maintenance of healthy weight; Harsha et al., 1996). In displaying inspirational

Huynh et al. 745

motivation, clinicians create a compelling vision for their patients’ health and clearly com- municate that vision to the patient. During the process of creating a health plan, they enthusi- astically engage patients in the information exchange process and achieve patient consen- sus (Charles et al., 1997), which conveys that the clinician and patient are working together as a team toward a common goal (Charles et al., 1997). With intellectual stimulation, clinicians engage patients to view their health concerns in new ways and to come up with innovative solu- tions to those issues. They facilitate an environ- ment in which patients can reflect on their concerns and do not force or coerce patients to accept a unilaterally determined solution (Rollnick and Miller, 1995). Finally, clinicians display individualized consideration when they treat each patient as a unique individual. They prioritize the process of creating warm interper- sonal relationships (Beach and Inui, 2006) and adjust their style to allow for differences in patient autonomy (Deber et al., 1996).

Transactional PCS

This PCS characterizes clinicians who set health goals for patients and provide them with instruc- tions, feedback, and reinforcement to pursue those goals. For example, a clinician–patient “transaction” may begin with a description of a health plan during an initial visit (e.g. an exercise regimen and medications) and end when the cli- nician provides feedback regarding the patient’s success (or failure) in executing the plan during a follow-up appointment. Transactional PCS has two components: problem-focused active care and contingent reward care.

Clinicians who display problem-focused active care monitor patients to anticipate adher- ence failures or deviations from patients’ stand- ard level of health. They focus on avoiding or preventing major health problems before they occur. Clinicians who display contingent reward care articulate the goals of treatment and the outcomes patients can expect when they follow through with the health plan. These cli- nicians provide extensive feedback and offer

reinforcements to patients during the course of their care (e.g. Seaburn et al., 2005). The pri- mary distinction between problem-focused active care and contingent reward care is that problem-focused active care focuses only on preventing potential negative outcomes, whereas contingent reward care focuses on goal setting and reinforcement.

Passive-avoidant

Clinicians who engage in a passive-avoidant style do not maximize their capacity to care for patients. At worst, they are unsympathetic to patients’ needs and leave health concerns to patients to sort out for themselves (laissez- faire). For example, they may order many unnecessary tests or refer patients to specialists, with the sole intention of avoiding decisions about the patients’ health (Axt-Adam et al., 1993). At best, these clinicians only concentrate on corrective actions, such as addressing symp- toms only after they have occurred (problem- focused passive). The laissez-faire component appears similar to the default style in Roter and Hall’s classic model of patient–physician rela- tionships (Roter, 2000); however, our model only focuses on clinician behaviors rather than the broader patient–physician relationship. Although we suspect that this PCS is rare, at least in its most extreme form, we assess it in the current studies to provide a thorough test of our theoretical approach.

It is important to note that the transforma- tional and transactional styles are not necessar- ily independent of each other; instead, the transformational style enhances the effects of the transactional style. In other words, transfor- mational clinicians are effective because they transcend transactional behaviors, not because they neglect transactional behaviors. Therefore, we anticipate that the transformational PCS engenders positive patient outcomes above and beyond the contribution of the transactional PCS (i.e. an augmentation effect; Den Hartog et al., 1997), not that the effect of one style depends on the effect of the other style (i.e. an interaction effect).

746 Journal of Health Psychology 23(5)

Patient health outcomes associated with PCSs

In addition to proposing the structure of PCSs, we also suggest that PCSs may contribute to patient satisfaction and health expectations, which are important markers of effective patient care.

Patient satisfaction

Patient satisfaction refers to personal evaluations of the healthcare process. Patient satisfaction can be used to assess the quality of care, to highlight areas in need of improvement, and to assess patient loyalty and commitment (Sitzia and Wood, 1997). Additionally, patient satisfaction is associated with better adherence to treatment recommendations and health outcomes (e.g. DiMatteo, 2004). Because research indicates that transformational leadership produces more satis- fied work-group members than transactional and passive-avoidant leadership (Bass and Riggio, 2006), we believe transformational clinicians will have more satisfied patients than clinicians characterized by either of the other two styles.

Health expectations following visit (Study 2 only)

Patients who are more optimistic about their future health status may be more likely to pur- sue or continue their care (Mann, 2001). Furthermore, creating health regimens that align with patients’ expectations can increase the likelihood that patients will adhere to the recommended treatment (Horne and Weinman, 1999). We propose that when transformational clinicians display the inspirational motivation component, they show optimism about their patients’ health, and when they display the indi- vidualized consideration component, they tailor their care to each patient. This combination may lead to the creation of health plans that patients believe are effective for improving their health in the future. Thus, we anticipate that patients’ expectations about their future health status following the medical visit will be associated with clinicians’ PCSs.

Overview and hypotheses

The primary goal of this article is to examine each PCS’ association with patients’ satisfac- tion and health expectations. In Study 1, we examined PCSs using patient-reported ques- tionnaire data. In Study 2, we examined PCSs using third-party observer ratings of clinician– patient interactions to overcome potential bias in the patient-reported questionnaires from Study 1. Study 1 assessed only patient satisfac- tion, whereas Study 2 examined satisfaction and health expectations. Specifically, we tested the following hypotheses in two studies:

•• Hypothesis 1. Scores on the passive- avoidant PCS dimension will negatively predict (or not predict) patients’ satisfac- tion and health expectations.

•• Hypothesis 2. Higher scores on the trans- actional PCS dimension will positively predict patients’ satisfaction and health expectations.

•• Hypothesis 3. Higher scores on the trans- formational PCS dimension will posi- tively predict patients’ satisfaction and health expectations, above and beyond the predictive power of scores on the transactional dimension.

Moreover, many demographic factors pre- dict patient satisfaction and health expectations (DiMatteo, 2004). For example, patients with lower educational attainment and lower incomes tend to be more satisfied with their care (Huynh et al., 2014). To address the unique predictive relationship between PCSs and our outcomes, we controlled for these demographic variables in our analyses.

Study 1

Method

Participants. Healthcare recipients (N = 164; 62% female; 36% 18–25 years old, 40% 26–55, 24% > 56; 49% White, 13% Hispanic, 20% Asian, 10% African American, and 8% did not state) completed an online questionnaire about

Huynh et al. 747

their most recent medical visit and were awarded US$5 to an online retailer for their participation. Participants had most recently visited their clini- cian for a preventive care issue (e.g. physical; 49%) or to address an acute illness (e.g. to care for a cold, flu, or physical injury; 50%) or chronic illness (e g. to care for diabetes or can- cer; 14%). Participants could select more than one option regarding the reason for their visit. In all, 54 percent completed our survey within 3 months of seeing their clinician, 31 percent between 3 and 12 months, and 14 percent the visits occurred more than a year prior to com- pleting our survey. In total, 51 percent of the participants had been with the clinician for more than a year, 17 percent between 1 and 11 months, and for 32 percent of participants, this was their first visit with the clinician.

Procedures. Trained undergraduate research assistants recruited participants from their social networks to complete an online question- naire. The research assistants did not know the study’s hypotheses and were simply instructed to ask participants to complete a questionnaire about their most recent medical visit. The research assistants were instructed to oversam- ple non-undergraduate students, and the age distributions in our sample suggest that we were successful in recruiting an adult sample that was not dominated by undergraduate students.

Measures. Participants answered questions about their most recent medical visit. Patient satisfaction with the clinician was assessed using a one-item measure (“I was satisfied with the care I received from my doctor”; the scale included all numbers from 1 to 7 with the fol- lowing labels: 1 = strongly disagree, 3 = some- what disagree, 5 = somewhat agree, 7 = strongly agree; Robbins et al., 1993; M = 5.81, standard deviation (SD) = 1.44). Participants also responded to a set of items designed to measure their clinician’s PCSs (27 items total). Items from the Multifactor Leadership Questionnaire (Bass and Avolio, 1991) were selected and mod- ified to reflect clinician–patient relationships in the medical context instead of leader–follower

relationships. The items were presented in a ran- dom order so that items pertaining to each style were not grouped together.

Sample items for each component are pre- sented below according to their PCS. Participants responded to the stem prompt: “My clinician …”: (1) passive-avoidant, 7 items, α = .80; for example, “Fails to interfere until my health issues become serious,” “Avoids making deci- sions with regard to my health”; M = 1.97, SD = 1.14; (2) transactional, 7 items, α = .83; for example, “Makes clear what I can expect to receive when health goals are achieved,” “Directs my attention toward failures to meet standards set for my health”; M = 2.22, SD = 1.00; and (3) transformational, 13 items, α = .93; “Gets me to look at my health problems from many different angles,” “Expresses confidence that my health goals will be achieved”; M = .54, SD = 1.00. Because the transformational style has more components, this style required more scale items than the other two styles.

Data analysis

We first created a composite score for each style by averaging items across components within each of the three styles. We then tested our hypotheses using hierarchical multiple regres- sion. We used hierarchical multiple regression because we were interested in each style’s unique contribution to variance in patient satis- faction. In particular, we wanted to partition variance of the outcomes by each PCS. Because transactional PCS may represent an adequate model of care in some situations but may be insufficient for others, we specifically exam- ined the effects of transformational PCS above and beyond the effects of transactional PCS (i.e. an augmentation effect; Seltzer and Bass, 1990). We entered the predictor variables in four steps: (1) control variables (age, income, education, race, sex, elapsed time from medical visit to survey completion, and length of relationship with clinician),1 (2) passive-avoidant PCS, (3) transactional PCS, and (4) transformational PCS. Because of the high correlation between transactional and transformational PCS, we

748 Journal of Health Psychology 23(5)

examined their collinearity by regressing trans- actional PCS onto transformational PCS. The resulting tolerance was higher than .20, which suggests that the estimated coefficients are reli- able in the following regression models (Miles and Shevlin, 2001).

Results

Results from step 1 showed that only the length of relationship with clinician and elapsed time from medical visit to survey completion were significant predictors of patient satisfaction with clinician, β = .17, p = .04 and β = −.19, p = .02, respectively; all other βs < .10, ps > .10; R2 = .14, F(10, 148) = 2.12, p = .03. In step 2, the passive- avoidant style did not account for additional variance, β = .11, ΔR2 = .01, F(1,147) = 2.09, p = .20. In step 3, the transactional style signifi- cantly and positively accounted for additional variance in satisfaction, β = .85, ΔR2 = .26, F(1,146) = 62.13, p < .001. In step 4, transforma- tional PCS significantly and positively explained yet additional variance in patient satisfaction, β = .35, ΔR2 = .03, F(1, 145) = 8.07, p < .01; total R2 = .44.

Study 2

Study 1 relied exclusively on patient-reported evaluations of their clinicians’ PCSs, which may reflect biased perceptions of their clinicians. In contrast, Study 2 used audio recordings of clini- cian–patient interactions to more objectively evaluate clinicians’ PCSs. Additionally, Study 1 asked participants to rate their most recent medi- cal encounter, whereas Study 2 involved medi- cal interactions with only primary care providers. Despite many apparent differences between the two studies, the fundamental question (i.e. do PCSs predict patient outcomes?) is the same.

Method

Description of data set. We used data from the Clinician–Patient Communication to Enhance Outcomes (CPC) program conducted by the Insti- tute for Healthcare Communication (formerly the

Bayer Institute for Health Care Communication) in 1994–1998. Clinician–patient interactions were audio recorded with patient and clinician consent, and both parties completed post-visit question- naires. For confidentiality purposes, each audio recording was assigned a unique number identify- ing the clinician, patient, and practice site. Because these recordings are from primary care visits, the visits vary greatly in length, with some interac- tions lasting well over 30 minutes.

Sample characteristics. The sample consisted of 297 patients (55% women; 58% White/Cauca- sian, 21% Asian, 7% Hispanic, 6% African American, and 8% did not state) and 100 clini- cians (39 female; 47 White/Caucasian, 44 as Asian American, 7 Hispanic, and 2 African American) with an average age of 37.59 years (SD = 9.63).

Procedures and measures. Four raters coded each audio recording and each rater coded approxi- mately 100 audio recordings. Recordings were provided to raters in a counterbalanced order to reduce fatigue effects (Haskard et al., 2008). Raters listened to the full audio recording of each interaction and then evaluated the clinician on items representing the components of PCSs. Raters received two waves of training. In the first wave, each item on the rating scale was explained in great detail to ensure there was consensus on the meaning of each item. All raters then listened to the same audio recording (not included in the sample) and provided initial practice ratings. Then the raters, along with the first author, discussed each item at length with regard to the target audio recording. The goal of this exercise was not necessarily to gain consen- sus for the ratings themselves but for the raters to fully comprehend the meaning and intent of each item. In the second wave of training, raters listened to and coded two additional audio recordings and discussed them in a group with the first author until consensus was reached about the meaning of each item. Inter-rater reli- ability was satisfactory (intraclass correlation coefficient (ICC) = .71; Shrout and Fleiss, 1979), which enabled us to appropriately average the

Huynh et al. 749

raters’ scores across each item if there was disa- greement between raters during data analysis.

Rating measures. Items describing PCSs were reworded and presented from an observer per- spective. Each item was rated on a 1–7 scale with two anchors, 1 = not at all, 7 = a great deal. For example, “The clinician avoids responding to urgent questions” (passive-avoidant, M = 1.37, SD = .38); “The clinician sets clear goals for patient’s health and specifies the benefits of achieving these goals” (transactional, M = 4.62, SD = 1.15); “The clinician expresses confidence in the patient’s ability to become or stay healthy” (transformational, M = 4.22, SD = .87).

Patient questionnaire from CPC program. Post- visit patient-reported questionnaires collected in the original study were marked by the same identification number as the audio recordings to permit linkage between the measures. The questionnaire included an assessment of the patient’s satisfaction with the clinician (“How would you rate the overall care you received from the doctor who treated you today?”; 1 = poor, 5 = excellent; Robbins et al., 1993; M = 4.47, SD = .80) and their expectation for their future health status (“I expect my health to get worse”; 1 = definitely true, 5 = definitely false (reverse coded); Ware and Sherbourne, 1992; M = 3.73, SD = 1.17).

Data analysis

Due to the clustered nature of the data (patients nested within clinicians), which can violate the independence assumption in regression analyses, we began using multilevel modeling (MLM) to explore the data. MLM is advanta- geous because it allowed us to partition the error variance at the appropriate level of anal- ysis (either at the patient or clinician level; Raudenbush and Bryk, 2002). For each of the outcomes (satisfaction and patient’s health expectations), we tested an unconditional analysis of variance (ANOVA) model in HLM 7 Student Edition (Raudenbush and Bryk, 2014). We examined the ICC to determine

whether the proportion of variance in each out- come was due to differences between clini- cians rather than differences between patients treated by the same clinician. A large ICC indi- cates that patients seeing the same clinician are very similar and/or that there are great dif- ferences across clinicians (Adelson and Owen, 2012). MLM should only be used when ICCs are greater than .10 (Lee, 2000). If ICCs are less than .10, we would examine our hypothe- ses using hierarchical multiple regression, similar to Study 1. For both outcomes, neither of the ICCs were greater than .06 (test of U0 is not significant at the alpha < .05 level, all χ2s(98) = 96.71, ps > .13). These results indi- cated that patients are quite different within clinician groups and/or there are not great dif- ferences across clinicians, and more impor- tantly, we would be unlikely to violate the independence assumption when conducting regression analyses. Therefore, we continued to use hierarchical multiple regression to examine our hypotheses.

Similar to Study 1, for each of the outcomes, we entered the predictor variables in four steps: (1) control variables (patient sex, income, edu- cation, and race), (2) passive-avoidant PCS, (3) transactional PCS, and (4) transformational PCS. Tolerance values were all higher than .20, which indicated that collinearity was not an issue (Miles and Shevlin, 2001).

Results

Patient satisfaction. Results from step 1 showed that the control variables did not account for significant variance in patient satisfaction, all βs < .10, ps > .10, R2 = .02, F(8, 282) = .65, p = .73. In step 2, passive-avoidant style did not account for additional variance, β = −.07, ΔR2 = .01, F(1, 281) = 1.46, p = .23. In step 3, the transactional style also did not account for additional variance in satisfaction, β = .03, ΔR2 < .01, F(1,280) = .23, p = .63. However, in step 4, transformational PCS significantly explained additional variance in patient satis- faction, β = .22, ΔR2 = .02, F(1, 279) = 5.22, p = .02; total R2 = .04.

750 Journal of Health Psychology 23(5)

Patients’ health expectations following visit. Results from step 1 showed that only patients’ sex pre- dicted expectation of health status, such that males expected poorer health after the visit, β = −.21, p < .001; all other βs < .10, ps > .10; R2 = .07, F(8, 282) = 2.62, p < .01. In step 2, pas- sive-avoidant PCS negatively predicted patients’ health expectations following visit, β = −.16, ΔR2 = .02, F(1, 281) = 7.54, p < .01. In step 3, transactional PCS did not account for additional variance in expectations, β = .01, ΔR2 < .01, F(1,280) = .04, p = .85. In step 4, transforma- tional PCS significantly and positively explained additional variance in patients’ health expecta- tions following visit, β = .22, ΔR2 = .02, F(1,179) = 5.47, p = .02; total R2 = .11.

Discussion

In two studies, we examined the predictive power of clinicians’ styles of patient care. …