Edited by: Kylie Rochford, Case Western Reserve University, USA
Reviewed by: Anita Howard, Case Western Reserve University, USA; Deborah Anne O'Neil, Bowling Green State University, USA
*Correspondence: Joann F. Quinn, Office of Educational Affairs, USF Health, Morsani College of Medicine, University of South Florida, 12901 Bruce B. Downs Blvd. MDC 54, Tampa, FL 33612-4799, USA
This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
The career path for many professionals is often into a leadership role, yet many professionals do not have the competencies or inclination to lead. This study explores physician leaders as a representative group of professionals. While there have been many efforts at understanding the characteristics of effective physician leaders, a greater understanding is needed on the nature of physician leadership. The largest healthcare organization for physician leaders in the United States was surveyed to gain a greater understanding of the nature of leadership. Partial Lease Squares (PLS) was used to analyze results from 677 online surveys to understand the causal relationship of role conflict and role endorsement to participation. The findings reveal the mediating influence that positivity exerts upon participation, and offers health care leaders an opportunity to increase understanding of the social identification process that leads a higher level of professional participation, which may increase effectiveness for physicians in leadership.
Physicians have long held leadership roles within hospitals and other healthcare organizations (Reinertsen,
The differentiation in identity from a clinician to a leader offers a unique perspective of study that could greatly impact how we understand physician leadership. The challenge for healthcare organizations concerned with improving physician leadership goes beyond selection and development (Stoller,
To better understand the nature of physician leadership, this study seeks an understanding of the impact of role conflict, as well as role endorsement, upon physician leader participation. It offers a model, which theorizes that two aspects of positive affect (compassion and vision) mediate the relationship of role conflict and role endorsement upon participation and seeks to validate these hypotheses based upon responses from 677 physician members of the American College of Physician Executives.
The understanding of the nature of physician leadership is an important topic to explore as it has implications at both the practitioner and theoretical level. While many theories of leadership exist, none specifically have sought to understand leadership from the perspective of an at times unwitting physician leader who is thrust into the role, which may be temporary and even part time.
This paper begins with a review of the literature on physician leadership, identity and role, positive affect and organizational participation. Building upon existing theory, a theoretical framework is developed, along with an associated set of hypotheses. The research methodology and sample are then presented, along with the analysis of results, as well as a discussion of findings. Finally, the paper concludes with a discussion of the implications to research and practice.
In this section, literature discussing physician leadership, the impact of positive relationships, identity and role will be reviewed.
Like many other professionals, physicians often assume part or full time leadership roles as department chairs, committee members, directors, and other administrative roles in hospitals and healthcare organizations. These leadership roles are often held within clinical departments or specific functions that operate somewhat separately from the larger organization (Lobas,
Montgomery suggests that the intra-professional divisions between clinicians into areas of functional expertise may not be as relevant as the division of physicians into clinician and manager due to the changing structure of healthcare (Montgometry,
How then, can physician leaders be developed into successful organizational leaders? Physicians are educated and professionalized to value autonomy (Stoller,
The increasingly complex environment in healthcare also requires more inter-departmental collaboration VanVactor (
The field of positive psychology has aided in our understanding of what leads to individuals thriving emotionally at the individual, community and societal level (Seligman and Csikszentmihalyi,
Boyatzis has argued that Positive Emotional Attractors (PEA) and Negative Emotional Attractors (NEA) are critical in affecting behavior, influencing one on a cognitive, emotional, social and physiological basis (2008). Positive and negative emotional attractors are described by Boyatzis (
The PEA is related to a shared vision, compassion and overall positive mood (Boyatzis and McKee,
Compassion, another sub-dimension of Boyatzis' P/NEA measure (Boyatzis,
Therefore, the following hypotheses are proposed:
Stets and Burke (
Individuals define their identity through membership within various groups, such as work groups, organizations (Tajfel and Turner,
According to Larson, the focus upon the uniqueness and specialization of the role exaggerates the “dignity” of the profession (Larson,
This social classification process begins in medical school. Physicians are not only gaining technical expertise, but are also being socialized into a profession and assuming their identity as a physician (Hall,
Social identity theory explains that the process of self-categorization accentuates the similarities of those belonging to the same category and the differences of those in different categories (Turner,
Not only are physicians then confined to a professional group that excludes others, but there is reluctance to become subordinate to those outside of their group (Bate,
Social identity theory aids our understanding of the nature of physician leadership with respect to the multiple roles they must assimilate, which does not specifically address “roles,” but does set out “to explain individuals' role-related behaviors” (Hogg et al.,
The role of physician, or leader, then creates a norm for behavior as an incumbent of that role, and in turn “the self as a structure of role-identities… operate[s] as a social force, affecting the structure of society by affecting behavior in important ways” (Callero,
DeRue and Ashford (
Moreover, identity theory informs us that that role-identities are hierarchically positioned, thus having differing effects upon behavior (Callero,
“According to the chain-of-command principle, organizations set up on the basis of hierarchical relationships with a clear and single flow of authority from the top to the bottom should be more satisfying to members and should result in more effective economic performance and goal achievement than organizations set up without such an authority flow” (Rizzo et al.,
Increasing the level of participation and improving performance is in effect dependent upon a greater understanding of role acceptance and conflict. While many scholars have studied role conflict (Kahn et al.,
Negative emotions, such as those associated with role conflict, are correlated with a lower likelihood of cooperation (Cremer and Hiel,
To explore the nature of physician leadership, a survey-based study was conducted to validate the hypotheses. A psychometric survey methodology was used that maps individual responses to the underlying concepts within the model. In an effort to capture representative data on physician in leadership, the membership of the American College of Physician Executives (ACPE) was surveyed.
To ascertain and measure the relevant dimensions of the model, this process proceeded in four stages: development of the survey instrument, development of measurement scales, pretesting to assess validities of the survey instrument and data collection from a sample of physicians with membership in the American College of Physician Executives (ACPE), the largest health care organization for physician executives in the US.
Where possible, construct items were based upon previously validated measures; otherwise, indigenous items were developed based on a review of pertinent literature and using a procedure consistent with prior studies (Churchill,
Although most scale items were adapted from those in the existing literature with slight modifications to reflect the focus of this study, a new scale was developed to measure role endorsement.
Role conflict items were adapted from the work of Rizzo et al. (
Role endorsement was informed by the author's earlier work on physician leadership and adapted from DeRue and Ashford (
A multi-item construct of organizational citizenship behavior was adapted using participation as a major component. The five measures of participation were adopted from the original 54 items in Van Dyne et al.'s measure for organizational citizenship behavior Van Dyne et al. (
To measure positivity, Boyatzis' PNEA scale (2008) was used, which includes three subscales, vision, compassion and overall positive mood. All items were measured on a 5-point scale with “strongly agree” at the extreme positive end and “strongly disagree” at the opposite end of each scale.
Several controls were also included, including role, tenure in role and in organization, age and gender. Rousseau and McLean Parks (
The multi-items for each of the constructs are summarized in Appendix A and the relationships of the model represented in Figure
The population sampled was the membership of the American College of Physician Executives (ACPE)*, which is the largest organization for physician executives in the nation. The ACPE is accredited by the Accreditation Council for Continuing Medical Education and has greater than 9000 members from the United States and 45 other countries, holding roles including chief medical officer, chief executive officer, vice president of medical affairs, directorships, as well as others (ACPE Website).
The survey was delivered online, which was emailed out to members of the American College of Physician Executives (ACPE) by the management of the organization. The ACPE has over 9000 registered members who are self-selecting into the organization, with the requirement of full members being allopathic (MD) and osteopathic (DO) physicians; dentists (DDS or DMS); and podiatrists (DPM).
Individual respondents were provided a URL to the survey, which was deployed through Qualtrics, a popular online survey research tool. Of the 9083 contacts that received the email with the survey link, 8672 emails were delivered, 2148 were tracked as opened, 1030 clicked on the link for the survey, and 936 physicians started the survey. The sample was then reviewed for missing values resulted in a final sample size of 677.
The data was collected beginning in July 2011, with 547 males and 128 females responding (81 and 19%, respectively). Of the respondents, 420 stated their leadership role as part time and 222 as full time, with the remaining responding with “not applicable.” 308 (46%) of respondents stated their age as 55 or older and of that age group, 240 reported their role as full time. The American Medical Association (AMA) delegates reported 79.4% as male and 20.6% as female and 77.3% as over age 50 as of December, 2010 (AMA,
In developing the survey instrument, a list of itemized questions was sent to 10 respondents, including several physician leaders, and asked them to comment on the flow, clarity, timing, and the respondents' interest through completion rate. Two of the items were modified to ensure that exact meaning was conveyed and understood. The pre-test was then followed by asking three individuals to read the questions aloud and answer them in order to assess cognitive difficulties presented by the survey items (Bolton,
Next, a pilot survey was conducted with 65 physicians working in four hospitals within a single healthcare organization to perform an exploratory factor analysis (EFA) for each hypothesized construct within the model. The pilot survey was carried out online. The items were found to be acceptable for factoring within each construct and no adjustments were made following this step.
Prior to analysis, missing values were removed related to the latent constructs. The data was screened for linearity, normality, multicollinearity, skewness, and outliers and found the data adequate for analysis. 260 data points were dropped due to missing values. There were no significant outliers, as the survey contained primarily Likert scales.
The research model was tested using Partial Lease Squares (PLS-Graph, v3.0, Build 1060, Chin and Frye,
An EFA (exploratory factor analysis) was performed using principal axis factoring and PROMAX rotation. Sample size was adequate with 677 usable responses across 28 items. The Kaiser-Meyer-Olkin (KMO) value was 0.932 and the Barlett's Test of Sphericity was significant (
The pattern matrix for initial convergent and discriminant validity. Criterion was employed as designated by Hair et al. (
After eliminating one item (role endorsement item 1), 27 items measured five factors—four reflective and one formative. Table
Role conflict | 4 | −0.736, −0.652, −0.591, −0.541 | 0.728 |
Role endorsement | 5 | 0.866, 0.854, 0.721, 0.695, 0.464 | 0.872 |
Vision | 8 | 0.876, 0.853, 0.818, 0.805, 0.689, 0.604, 0.590, 0.586 | 0.912 |
Compassion | 6 | 0.773, 0.767, 0.656, 0.620, 0.518, 0.437 | 0.841 |
Role conflict | 1 | ||||
Role endorsement | −0.414 | 1 | |||
Vision | −0.433 | 0.644 | 1 | ||
Compassion | −0.389 | 0.568 | 0.642 | 1 | |
Participation | −0.111 | 0.357 | 0.289 | 0.291 | 1 |
Partial Least Squares (PLS), a structural equation modeling (SEM) technique, was used for testing the research model. PLS approach was superior to other SEM approaches for this study because of its flexibility on distributional assumptions, its small sample size requirements, and its strength on complex predictive models (Chin and Newsted,
To assess the psychometric properties of the latent constructs, a PLS measurement model was created. To assess convergent validity, the internal consistency reliability (ICR), the average variance extracted (AVE), and the item factor loadings for the reflective constructs were assessed.
The survey employed multi-item scales to measure the reflective first-order factors. The measurement properties for the reflective constructs were examined by conducting confirmatory factor analyses using PLS. To assess the internal consistency of the reflective factors, AVEs, coefficient alpha and composite reliability measures were assessed. For participation, it was not possible to assess validity and reliability, since the very nature of formative measurement renders irrelevant traditional assessments of convergent validity and item reliability.
Accordingly, as seen in
Tests were conducted to evaluate the convergent and discriminant validity and the reliability of reflective measures. Convergent validity of the constructs is assessed by examining the constructs factor loadings, composite scale reliability and average variance extracted (Fornell and Larcker,
The square root of AVEs ranged from 0.553 to 0.663 for reflective constructs. For a second test of discriminant validity, individual items may be assumed to possess sufficient discriminant validity if they load higher on their own respective construct than on any other latent variable (Gefen et al.,
It was expected that items belonging to the same scale would have factor loadings exceeding 0.70 on this common factor. As indicated by the results in
As a result of the construction of a formative variable, “conventional procedures used to assess the validity and reliability of scales composed of reflective indicators (e.g., factor analysis and assessment of internal consistency) are not appropriate for composite variables (i.e., indexes) with formative indicators” (Diamantopoulos and Winklhofer,
A test for common method bias was performed, as survey item responses were all self-reported. In order to test for common method bias, Harman's one-factor test was applied, including all items in the model in a principle components factor analysis (Podsakoff et al.,
The test of the structural model includes estimating the path coefficients and the
H7: Role conflict → Vision | 0.448 | 6.498 | 0.0598 | Small effect |
H4: Role endorsement → Vision | 0.448 | 19.0748 | 0.4710 | Large effect |
H7: Role conflict → Compassion | 0.348 | 4.7257 | 0.0414 | Small effect |
H5: Role endorsement → Compassion | 0.348 | 14.4921 | 0.3037 | Medium effect |
H1: Vision → Participation | 0.107 | 2.2769 | 0.0134 | No effect |
H2: Compassion → Participation | 0.107 | 3.444 | 0.0269 | Small effect |
A series of tests were run to investigate the predictive power of the structural model (Chin and Frye,
R2 represents the amount of variance in the construct that is explained by the model. Cohen (
H4: RE → P | 0.364 |
8.2065 | 0.0452 | Partial mediation |
RE → VI | 0.572 |
19.2890 | 0.0297 | |
VI → P | 0.149 |
2.2789 | 0.0654 | |
H5: RE → P | 0.364 |
8.2065 | 0.0452 | Partial mediation |
RE → COMP | 0.495 |
14.3085 | 0.0346 | |
COMP → P | 0.210 |
3.5177 | 0.0597 | |
H7: RC → P | 0.033 |
0.2773 | 0.0433 | Full mediation |
RC → VI | −0.186 |
6.2113 | 0.0382 | |
VI → P | 0.149 |
2.2789 | 0.0654 | |
H8: RC → P | 0.033 |
0.2773 | 0.0433 | Full mediation |
RC → COMP | −0.177 |
4.6366 | 0.0382 | |
COMP → P | 0.210 |
3.5177 | 0.0597 |
See Table
H1: Supported | |
H2: Supported | |
H3: Not Supported | |
H4: Supported | |
H5: Supported | |
H6: Not Supported | |
H7: Vision fully mediates the relationship between role conflict and participation | |
H8: Compassion fully mediates the relationship between role conflict and participation | |
H9: Not Supported |
Role conflict | 0.831 | 0.553 | |||
RC1 | 0.8103 | 0.0238 | 15.4711 | ||
RC2 | 0.8082 | 0.0276 | 14.451 | ||
RC3 | 0.6396 | 0.0302 | 7.979 | ||
RC4 | 0.7028 | 0.0282 | 11.35 | ||
Role endorsement | 0.907 | 0.663 | |||
RE2 | 0.6799 | 0.015 | 13.871 | ||
RE3 | 0.8816 | 0.0089 | 29.3311 | ||
RE4 | 0.7718 | 0.0103 | 20.4912 | ||
RE5 | 0.8897 | 0.0090 | 31.4706 | ||
RE6 | 0.8295 | 0.0096 | 26.5433 | ||
Vision | 0.929 | 0.621 | |||
VIS1 | 0.7027 | 0.0093 | 15.5335 | ||
VIS2 | 0.7028 | 0.0082 | 16.5069 | ||
VIS3 | 0.7931 | 0.0073 | 24.6386 | ||
VIS4 | 0.8114 | 0.0068 | 23.5194 | ||
VIS5 | 0.7949 | 0.0084 | 19.6612 | ||
VIS6 | 0.8413 | 0.0064 | 26.8095 | ||
VIS7 | 0.8111 | 0.0081 | 19.0003 | ||
VIS8 | 0.8346 | 0.0075 | 21.0442 | ||
Compassion | 0.884 | 0.562 | |||
COMP1 | 0.7927 | 0.0124 | 18.4857 | ||
COMP2 | 0.7182 | 0.0153 | 13.8153 | ||
COMP3 | 0.7049 | 0.0139 | 15.7524 | ||
COMP4 | 0.7980 | 0.0131 | 17.8802 | ||
COMP4 | 0.6577 | 0.0190 | 9.0571 | ||
COMP6 | 0.8127 | 0.0125 | 20.8252 | ||
Participation | |||||
PART1 | 0.5394 | 0.1217 | 4.4338 | ||
PART2 | 0.6406 | 0.1264 | 5.0676 | ||
PART3 | 0.0786 | 0.1233 | 0.6375 | ||
PART4 | 0.0713 | 0.1094 | 0.6518 |
The final model, shown in Figure
The results structural model testing provide evidence to support H1 and H2, vision (β = 0.149*) and compassion (β = 0.210***), have a significant and positive relationship with participation. Vision plays a mediating role in the relationship of role conflict and role endorsement to participation. The direct effect of role endorsement to participation was significant, as was the indirect effect via both vision and compassion. Moreover, role endorsement showed a significant direct effect with the vision (β = 0.572***) and compassion (β = 0.495***) mediators, and both vision (β = 0.149*) and compassion (β = 0.149*) had significant relationships with participation. These findings are therefore consistent with the hypotheses of partially mediated effects; therefore H4 and H5 are both supported.
It was hypothesized that the role conflict on participation would be partially mediated by both vision and compassion. Instead, it was found that the there was not a significant relationship from role conflict to participation, and role conflict showed a significant negative direct effect with the vision (β = −0.186***) and compassion (β = −0.177***) mediators, and as previously stated, both vision (β = 0.149*) and compassion (β = 0.149*) had significant relationships with participation. In summary, the results of indicate that role conflict has a direct effect on the mediators of vision and compassion, and both of these mediators has a significant relationship with participation, and that the direct effect of role conflict to participation is no longer significant. These results are consistent with the hypothesis of a full mediation for H7 and H8.
H4, H6, and H9 were not supported, as compassion, a sub-construct of the P/NEA Scale was not found to be significant.
Insights were applied from positive psychology, social psychology and management literature to demonstrate that role factors, such as role conflict and role endorsement, are an important consideration in participation by physician leaders. Specifically, this study found support that role conflict negatively affects participation; while role endorsement has a positive relationship with participation.
These results also show support for the argument that positivity, in this instance vision and compassion, mediate the relationship of role factors and participation. The largest effect found in this model was the relationship of role endorsement to vision, which may speak to the importance of an individual being endorsed in their role by both their peers and organization, in addition to their own certainty in their authority. This is inline with DeRue and Ashford's (
The findings also confirm the importance of the leader-member relationship, as it relates to the importance of the endorsement of the leadership role by peers. Specifically, the mediators of vision and compassion partially mediate the relationship of role endorsement to participation.
In testing the mediated relationship of role conflict and participation, surprisingly, full mediation was found in that that the association was completely accounted for by vision and compassion (James et al.,
In the case of role conflict, the findings demonstrate the significant role that vision and compassion play within the model, as these mediating factors fully explain the relationship to participation. These results illuminate the importance of positivity in buffering role conflict, with the hope of increasing participation and potentially effectiveness. It was unexpected, however, that no significant differences were found in the model when testing for part time physician leaders vs. full time physician leaders. It was anticipated that there may be a difference in these results involving role conflict, as a previous inquiry by this author suggested that there was a distinct difference between how part and full time leaders viewed their role as it pertains to conflict. Role endorsement was also found to be an important factor in this study. DeRue and Ashford (
Finally, it is curious that mood was not found to be significant in our testing, while the other two sub-scales of the P/NEA were found to be significant. It may be that mood is something physicians do not have the luxury of allowing to impact their work in life and death situations, therefore they have conditioned themselves not to allow affect from mood. With enough conditioning and time, this mood may not have a significant impact upon on a physician's behavior, whether they are acting in a clinical capacity or not.
Further research is needed to explore additional mediating factors, which may explain the relationship of role factors to participation.
A potential limitation to this study is one that may actually strengthen the results- the fact that the sample is comprised of both part and full time physician leaders who have
While great strides were taken to protect the results from common methods bias, no statistical test can guarantee such bias does not exist within these results (Podsakoff et al.,
As the first study to empirically examine the impact of relational and organizational endorsement of role, this study offers previously undiscovered insight as to the impact role perception to healthcare leaders concerned with physician leadership.
A practical implication of these findings is the understanding of the factors that influence the acceptance of a leadership identity for physicians by healthcare administrators, so that they may positively influence the interpreted psychological climate by physician leaders. If healthcare leaders know the factors that influence physician leaders to fully accept and engage in their role, they will be better prepared to assist in the development of physician leaders. Pratt et al. (
While psychological climate is an understanding of meaning by the individual, there are several ways in which an organization can influence that perception. At the organizational level, healthcare leaders may be informed by the impact of vision and compassion upon a physician leader's engagement and increase awareness surrounding these concepts. Leadership development workshops and programs can aid in an individual's self- awareness and an understanding of the factors that enable them to participate at a higher level and potentially become a better leader.
Physicians entering into leadership roles may also be informed by these findings. If physicians are aware of the factors that may limit or enhance how they enact their role as a leader, they may be better prepared to deal with the challenges. The basic realization that they may struggle with the acceptance of the secondary identity as a leader may alone be enough to encourage them to explore options to overcome the limitations to acceptance of that role.
Finally, these results should also inform medical school administrators and faculty members of the importance of including leadership skills and specifically emotional and social competencies into the curriculum. Chaudry et al. note “because leadership skill sets are not emphasized during training and practice, physicians, whose education is rooted in quantitative science, tend to address most problems with technical solutions” (Chaudry et al.,
Stoller et al. (
Future research should continue to examine the impact of role endorsement upon not only organizational participation, but also effectiveness. As well, although no significant differences were found between those in part and full time leadership roles within this study, this may be an aspect for further examination, especially with regard to individuals from a single organization.
It was anticipated going into this exploration that there may be a difference in the results involving role conflict for part and full time physician leaders; however, there were no significant differences found related to part or full time status. Therefore, a more detailed exploration of how part or full time status may be impacted by organizational climate may be beneficial.
This model only examined the linear relationships associated with the intervening effects. However, moderator relationships could be incorporated into future explorations involving these constructs.
The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
The Supplementary Material for this article can be found online at: