The Moderating Effect of Job Satisfaction on Physicians’ Motivation to Adhere to Financially Incentivized Clinical Practice Guidelines

Abstract

We examined moderating effects of professional satisfaction on physicians’ motivation to adhere to diabetes guidelines associated with pay-for-performance incentives. We merged cross-sectional survey data on attitudes, from 156 primary physicians, with prospective medical record-sourced data on guideline adherence and census data on ambulatory-care population characteristics. We examined moderating effects by testing theory-driven models for satisfied versus discontented physicians, using partial least squares structural equation modeling. Results show that attitudes motivated, while norms suppressed, adherence to guidelines among discontented physicians. Separate models for satisfied versus discontented physicians revealed motivational differences. Satisfied physicians disregarded intrinsic and extrinsic influences and biases. Discontented physicians, alienated by social pressure, favored personal inclinations. To improve adherence to guidelines among discontented physicians, incentives should align with personal attitudes and incorporate promotional campaigns countering resentment of peer and organizational pressure.

Keywords

primary care physicians job satisfaction value-based pay for performance health care quality adherence to clinical guidelines motivation theory

Introduction

Public report cards of the quality of physician services, citing metrics of their adherence to medical practice standards, are increasingly common and important (Kaplan, Griffith, Price, Pawlson, & Greenfield, 2009). Different stakeholders judge physician performance on the basis of an expanding array of measures, including, for example, adherence to clinical guidelines (American Association of Clinical Endocrinologists, 2013), achieving Patient-Centered Medical Home status (Friedberg, Schneider, Rosenthal, Volpp, & Werner, 2014; Grumbach, 2013), and Healthcare Effectiveness Data and Information Set performance measures (National Committee for Quality Assurance, 2015). These changes have heightened levels of practitioner discontent, anxiety, frustration, and burnout (Friedberg et al., 2013; Hoff, 2010).

In particular, concerns persist about the contribution of practice guidelines to eroding clinical autonomy (Emmanuel & Pearson, 2012) and the ability of guidelines to support appropriate care, for example, for elderly patients or those with multiple morbidities (Boyd et al., 2005). Institute of Medicine criteria for clinical guidelines aimed to alleviate such concerns (Graham, Mancher, Wolman, Greenfield, & Steinberg, 2011). Yet results from a recent nationwide survey of physicians in group practices with Medicare quality reporting show that 84% of respondents thought clinical guidelines did not improve quality and 85% felt they hindered productivity (Medical Group Management Association, 2014).

Physicians’ perceptions of clinical guidelines as enhancing or threatening their well-being could moderate their motivation to adhere to guidelines. The extent to which the job satisfaction of physicians affects their attitudinal responses to guidelines has not been examined sufficiently and merits further inquiry because these attitudes could represent a key opportunity for intervention. This study examines the extent to which job satisfaction moderates the effect of workplace attitudes on physicians’ adherence to diabetes care guidelines in a quality improvement program that included pay-for-performance (P4P) incentives.

We focus on diabetes because it was likely the first chronic illness to have guidelines promulgated (O’Connor et al., 2011), but its prevalence keeps rising (Menke, Casagrande, Geiss, & Cowie, 2015). The number of Americans living with diabetes rose from 12.1 million in 2002 (Hogan, Dall, & Nikolov, 2003) to 22.3 million in 2012 (American Diabetes Association, 2013), and it is estimated to climb to one in three adults by 2050 if current trends persist (Boyle, Thompson, Gregg, Barker, & Williamson, 2010). Economic costs of diabetes in the United States rose from $174 billion in 2007 to $245 billion in 2012 (American Diabetes Association, 2013). Provider adherence to guidelines can delay or prevent diabetes complications (American Diabetes Association, 2014; Handelsman et al., 2011). Yet half of adult Americans with diabetes still do not receive guideline-adherent care (Ali et al., 2013), and 70,000 Americans die from diabetes complications every year (National Committee for Quality Assurance, 2013).

New Contribution

Our primary aim is to examine the indirect, moderating effect of physicians’ satisfaction with their practice on their adherence to specific diabetes clinical guidelines. Our approach differs from prior studies that have examined the direct effects of physician job satisfaction on their performance (see Hickam et al., 2003, for a review) because we examine physician satisfaction as a moderator.

Few studies have examined job satisfaction as a moderator of the motivation to perform professional tasks. The literature on moderating variables suggests that identifying moderation effects can have a powerful role in developing interventions that are effective for different subgroups of a target population during early phases of an intervention program (Hopwood, 2007). Knowing the significant moderators of a predictor variable helps establish its boundary conditions on the outcome, that is, specific circumstances, stimuli, or persons for which its predictive effect is large versus small, present versus absent, or positive versus negative.

The prospect of motivation for clinical tasks differing by job satisfaction among physicians remains untested. In contrast to prior studies that test for a direct effect of job satisfaction on performance, the specific new contribution of this study is to pose the question: Are motivational factors differentially related to clinical guideline adherence among satisfied physicians as compared to discontented physicians?

To address this question, we draw on diverse theories to hypothesize that motivational drivers of task performance among physicians vary according to their levels of job satisfaction. To assess job satisfaction as a moderator, we use partial least squares structural equation modeling (PLS-SEM), a variance-based path analysis method for testing the predictive power of diverse pathways between constructs and indicators.

Conceptual Model

Figure 1 illustrates our conceptual model that adapts the principal constructs of the theory of planned behavior (TPB) and augments them with corresponding elements from other theories. The TPB proposes three predictors of forming an intention to engage in a behavior and actually performing the behavior: personal attitudes, subjective norms, and perceived behavioral control. Personal attitudes reflect how a behavior is evaluated by someone contemplating action, for example, whether a physician favors adherence to a guideline. Subjective norms refer to the perceived social appropriateness of the behavior, for example, how much physicians feel pressured by peers, organizations, and influential observers into adherent behavior. Perceived behavioral control is the assessed personal ability to control the environment sufficiently in order to act, for example, the extent to which a physician feels able to measure glycohemoglobin twice yearly or persuade patients to increase their exercise. For reviews of the TPB, see Azjen and Fishbein (Ajzen, 1991, 2014; Fishbein, 2008; Fishbein & Ajzen, 2010).

TPB has been tested widely in health care (Godin & Kok, 1996), including other studies of physicians (Godin, Naccache, & Fortin, 1998; Legare et al., 2005; Liabsuetrakul, Chongsuvivatwong, Lumbiganon, & Lindmark, 2003; Montano, Phillips, & Kasprzyk, 2000; Sable, Schwartz, Kelly, Lisbon, & Hall, 2006; Walker, Grimshaw, & Armstrong, 2001). It explains 28% to 31% of variance in clinical behavior (Godin & Kok, 1996; Sheeran, 2002). The TPB, despite being open to additional constructs (Head & Noar, 2014), does not include job satisfaction.

We now build conceptual bridges between the TPB and the theory of work adjustment (TWA), which initially proposed job satisfaction as a moderator of work performance (Dawis & Lofquist, 1984). The TPB’s personal attitudes are similar to personal values/motives in the TWA, while the TPB’s perceived behavioral control is analogous to the TWA’s concept of “ability requirements.” We argue that the TPB and TWA are compatible in their views of attitudes/values and behavioral control (or self-efficacy)/work abilities (or skills). The TWA posits that when satisfaction is high, the correspondence between abilities of workers and the requirements of a job correlates better with job performance; and the more closely a role’s rewards correspond with personal values/motives that an individual seeks to gratify/express at work, the more satisfied they will be (Dawis, 1994).

We extend the TWA’s intuition by arguing that job satisfaction moderates the association of guideline adherence with the personal attitudes of physicians and their perceived behavioral control. We hypothesize that both attitudes and behavioral control predict adherence among satisfied, but not discontented, physicians.

The TPB portrays goal-directed behavior as largely the result of deliberate intention. In the case of frequently repeated tasks, however, such deliberation or explicit intention formulation reduces to a minimum, and the work becomes routine (Bagozzi & Yi, 1989; Sutton, 1994). Once tasks become routine, little-to-no prior cognitive effort is required to keep executing them (Ouellette & Wood, 1998). We view clinical behaviors contained in diabetes guidelines as routine, repetitive tasks. We, thus, did not model behavioral intentions. This idea is supported by a study of physician adherence to mental health care guidelines (Rebergen et al., 2006), which found that the decision to adhere did not involve significant cognitive effort.

In place of intentions, we included the construct of past behavior to assess the extent to which guideline tasks were a “routinized” behavior that practitioners engage in with scarcely any deliberative effort, but by dint of sheer repetition. When past behavior is included in studies testing the TPB model, it explains an average 13% extra variance in subsequent behavior (Conner & Armitage, 1998). Past behavior often has a more significant effect than the TPB constructs (Conner & Armitage, 1998; Ouellette & Wood, 1998). Past behavior is a stronger predictor of future behavior in frequently versus infrequently performed tasks (Ouellette & Wood, 1998). We, thus, hypothesized that prior adherence behavior would predict latter behavior.

We now build conceptual bridges between the TPB and social identity theories. Within the TPB model, social norms are meant to capture the fact that individuals do not act in isolation but within a social context, subject to social influence from others. Yet TPB studies often find this construct to have the weakest effect (Armitage & Conner, 2001). Its influence is inconsistent (Askelson et al., 2010; Chau & Hu, 2002). Our model proposes that the satisfaction status of physicians influences their reaction to social pressure by acting as a perceptual filter through which they judge how workplace norms are likely to threaten or benefit their personal interests.

The social identity/self-categorization theories (Turner & Reynolds, 2010, 2011) view behavior as motivated by the “social identity” group(s) to which one “self-categorizes.” Those who feel organizational change threatens their professional identity or group norms (Crane & Platow, 2010) resist reforms (Hotho, 2008). Physicians identifying strongly with an institution (but weakly with their profession) are open to social influence from managers and readily adopt new work behaviors, but those who strongly identify with the profession (and weakly with the organization) resist such influence (Hekman, Steensma, Bigley, & Hereford, 2009). Those with a positive image of their organization incorporate it into their social identity (Dukerich, Golden, & Shortell, 2002). In the social identity/self-categorization models, everyone’s worldview consists of “us” versus “them” social identities (Terry & Hogg 2001; Terry, Hogg, & McKimmie, 2000).

We argue that the expressed satisfaction of a physician confers a unique group identity on them. It becomes a proxy for social identity, making a priori notions of professional identity more salient by priming feelings of belongingness to an in-group and nonbelongingness to or alienation from an out-group. Physicians weigh how much their adherence choice requires them to craft a new group identity or to reconfirm an existing one. We suggest that job satisfaction enters that calculus as a moderator.

We now build conceptual bridges between the TPB and self-determination theory (SDT). SDT views goal-directed behavior as driven by three intrinsic needs: autonomy, competence, and relatedness; and as divisible into four levels of self-determination: external regulation, introjection, identification, and intrinsic motivation (Deci & Ryan, 2011). Rewards/praise drive externally regulated behavior; internal pressure (e.g. shame/guilt) motivates introjected behavior; values or goals that one “identifies” with drive identified behavior; pure enjoyment drives intrinsically motivated behavior.

We argue that SDT’s intrinsic motivation and identification are analogous to personal attitudes in the TPB and external regulation is akin to social norms (Hagger & Chatzisarantis, 2009; Hagger, Chatzisarantis, & Harris, 2006), while both internal and external regulation underlie perceived behavioral control (Fishbein & Ajzen, 2010). We hypothesized that satisfied physicians view tasks involved in adhering to guidelines as internally/intrinsically regulated, while discontented peers experience them as externally regulated.

We also build conceptual links between the TPB and cognitive dissonance theory, as well as psychological reactance theory. The cognitive dissonance theory argues that social pressure suppresses the motivation to engage in a behavior when it surpasses a critical threshold (Festinger & Carlsmith, 1959). According to the psychological reactance theory, such “pressuring forms” of social influence arouse a rejection of the target behavior (Brehm & Brehm, 1981). We hypothesized that discontented physicians would perceive social pressure to adhere to guidelines as threatening their needs for autonomy, and exhibit “psychological reactance” by contradicting adherence-oriented norms. As a result, autonomy-suppressing social norms will demotivate adherence to guidelines among discontented physicians.

We hypothesized that satisfied physicians would be more open to P4P-based guidelines, and their adherence would correspond more with social norms; but discontented peers will resist guidelines, their adherence being negatively influenced by organizational or group norms. Unlike the TPB’s portrayal of attitudes and social norms as independent factors, we hypothesized that the two are correlated because individuals decide whether to comply with group norms based on how much those norms concur with their own personal attitudes. We further hypothesized that the attitudes–norms correlation is positive among satisfied physicians and negative among discontented counterparts.

In sum, we hypothesized that (a) personal attitudes predict guideline adherence among satisfied, but not discontented, physicians; (b) perceived behavioral control positively influences adherence for satisfied physicians, but negatively affects outcomes among discontented counterparts; (c) prior adherence predicts end point adherence among both satisfied and discontented doctors; (d) subjective norms positively affect guideline adherence for satisfied physicians, but negatively influence outcomes for discontented peers; and (e) personal attitudes are positively associated with social norms among satisfied physicians, but negatively associated with norms among discontented ones.

Method

Study Design and Setting

This is a retrospective cohort study, nested within a quasi-experiment, testing physician attitudes during a transition from volume-based reimbursement to value-based purchasing of physician services. The setting for the study was an Independent Practice Association (IPA), which functioned as a not-for-profit physician organization that negotiated both fee-for-service and capitated contracts for medical services provided by its members. The IPA represented about 3,400 individual practitioners: 900 primary care providers (PCPs) and 2,500 specialists from >20 specialties. From 1999 through 2005, its primary contract was with a nonprofit, nationwide, managed care plan headquartered in Rochester, New York. The contract was to provide physician services to the 250,000 to 450,000 health maintenance organization (HMO) enrollees in the 9-county region of greater Rochester.

After experimenting with P4P from 1999-2002, the IPA rolled out a physician-profiling program called the Value of Care Plan (VOCP) in 2002. Because of the high penetration of the specialized HMO product it offered in the Rochester area, the IPA was able to profile and financially incentivize individual physicians. Physicians received individual performance reports three times a year with the final report determining the physician’s incentive payment. The managed care organization contributed information technology and data analysis expertise to the VOCP. Three metrics were assessed and weighted: patient satisfaction (20%); quality of care: adherence to recommended care (40%); and cost efficiency (40%). Reports included registry data on patient adherence to preventive services and chronic and acute disease guideline-adherence measures. Cost of care metrics were based on comparative costs for risk-adjusted episodes of care. Physicians were incentivized to have costs below panel averages, guideline adherence and satisfaction above panel averages. Peer comparison was provided in reports throughout the year that included specialty averages. Target rates were set using community/national benchmarks. Data accuracy was validated at 92% to 95%. The VOCP adapted community-wide guidelines from the Rochester Health Commission, a nonprofit representative of local insurers, physician organizations, employers, and hospitals (Institute of Medicine, 2006).

From 2002 to 2006, the VOCP distributed $12 to $15 million in annual financial bonuses to physicians. The payments were funded from both risk withholds and the results of shared savings projects negotiated with the managed care organization, focusing on costs outside the IPA’s professional services agreement. A busy internist, on average, received performance-based pay bonuses totaling $5,000 to $15,000 annually. Administering the VOC program cost the IPA and managed care organization an estimated $1.2 million annually, including staff costs (Institute of Medicine, 2006). Additional expenses were accrued for time spent correcting patient data, improving current measures or developing/implementing new measures and other program refinements. “Shared Withhold” models withhold a portion of physician fees (i.e., place provider reimbursement partly “at risk”) or delay paying increases and release them contingent on a physician’s performance. From 2002 to 2005, the VOCP participated in the Robert Wood Johnson Foundation’s Rewarding Results evaluation of U.S.-based P4P initiatives to learn from best practices as well as failures (Young, Burgess, & White, 2007).

Study Population

From late 2002 to December 2003, IPA members from family medicine, internal medicine, and medicine-pediatrics for whom claims data were available for at least 50 deidentified patients were included in a mailed survey on Quality Targets and Incentives as part of the Rewarding Results evaluation. Researchers from Boston University School of Public Health administered the survey. Of the 574 eligible PCPs, 290 responded (a 51% response rate). From the 290 survey respondents, we excluded from the present study those (n = 73) who were ineligible to receive a financial bonus for adhering to a diabetes-care guideline because they had too few diabetic patients. We included PCPs who reported diabetes as the most typical illness of their patient panel, or whom VOCP was profiling on adherence to diabetes guidelines. Of 157 such PCPs, 94 were internists (who included physicians that were board certified in both Internal Medicine and Pediatrics), 62 family physicians, and 1 a pediatrician. We limited the sample for this analysis to internists and family practitioners (n = 156).

Data Collection

Data on physicians’ general attitudes to merit pay, specific attitudes to VOCP preventive medicine guidelines, and perceptions of resultant infringements on clinical autonomy were obtained from responses to the “Quality Targets and Incentives” survey conducted in 2002-2003. The data collection instrument was a 65-item prepiloted and prevalidated self-administered questionnaire that assessed general attitudes to incentive-based guidelines and experience-based attitudes to one patient-panel-typical context-specific clinical care guideline. The questionnaire also inquired about practice characteristics, and perceptions of autonomy support from the IPA and health plan. The survey method has been described extensively in previous reports (Meterko et al., 2006; Waddimba, Burgess, Young, Beckman, & Meterko, 2013; Waddimba, Meterko, Beckman, Young, & Burgess, 2010; Young, Beckman, & Baker, 2012; Young, Meterko, et al., 2007). Data on adherence to diabetes care guidelines were abstracted from the insurance plan’s claims database and aggregated over an entire year before they were linked to the individual’s survey responses. The survey was administered only once whereas prospective, annualized data on guideline adherence are available for multiple years. We obtained data on individual provider demographics, characteristics of each practice, and guideline-related clinical behaviors from the IPA. To risk-adjust for attributes of the catchment population, we obtained sociodemographics of the regional ambulatory care population from the U.S. Census data and linked them to zip codes for the practice locations.

Variables and Measures

Adherence to disease surveillance guidelines for diabetes at the conclusion of the final year of the parent study (2004) was the principal outcome. Specifically, we assessed provider compliance to four Healthcare Effectiveness Data and Information Set process measures for diabetes care. Within a 12-month period, a practitioner was judged to have completely adhered to practice standards if all eligible patients in their panel underwent the following: one urine microalbumin (urinalysis) screen, two glycohemoglobin (HbA_1c) tests, three annual low-density lipoprotein (LDL) cholesterol assay, and four annual dilated eye (fundoscopy) exam. Adherence in the final year was then compared to that in the baseline year of this study (2001).

The constructs or indicators that we measured were inspired mostly by the TPB, although they are supported in numerous other theoretical models, too. We assessed five indicators of personal attitudes: (a) the general attitude scales of effectiveness¹ and utility², and (b) the specific attitude scales for awareness³, financial salience⁴, and impact⁵. We examined three variables as indicators of subjective norms: (a) scales for clinical relevance⁶ and peer/staff cooperation⁷, and (b) a single item on perceived fairness/equity⁸ of financial bonus distribution. Indicators of perceived behavioral control were the following: (a) single items for achievability⁹, difficulty¹⁰ in complying with guidelines, and their lack of hindrance¹¹ to patient care, (b) the scale for locus of control¹², and (c) scales for perceived autonomy support¹³ from the HMO and the IPA. Each single item was rated on a 5-level bipolar Likert-style response format ranging from 1 = strongly disagree to 5 = strongly agree. Detailed descriptions and evaluations of these scales and items have been reported previously (Meterko et al., 2006; Waddimba et al., 2010).

Global job satisfaction was assessed by a single survey item asking PCPs to rate “overall” satisfaction on a 7-point bipolar Likert-type scale from “completely dissatisfied” to “completely satisfied.” Unlike facet satisfaction, global satisfaction is unidimensional and can be measured reliably by a single item (Wanous, Reichers, & Hudy, 1997). We then contrasted PCPs who were “satisfied” (ratings 5-7) with those who were “not satisfied” (ratings 1-4), whom we regarded as the discontented group.

Statistical Analysis

We adopted Anderson’s 2-step approach to SEM (Anderson & Gerbing, 1988) by specifying a theory-driven measurement (outer) model then estimating a parsimonious structural (inner) model that fit the data best. We used PLS-SEM, a variance-based method (Rigdon, 2013), to achieve our study goals. The advantages over covariance-based SEM, for our purposes, are that it makes a small number of distributional assumptions, is robust to small sample sizes, and permits both formative and reflective constructs in a single model (Chin & Newsted, 1999; Hair, Hult, Ringle, & Sarstedt, 2014). Reflective models are those where the latent construct causes change in its indicators, the latter are interchangeable and highly correlated; in formative models, it is indicators that cause change in the construct, they are not interchangeable and not correlated (Diamantopoulos & Siguaw, 2006). We specified adherence to diabetes care guidelines as a composite variable with formative indicators, with the attitudinal scales as reflective indicators. We implemented the PLS-SEM analyses using SmartPLS software Version 2.0 M3 (Ringle, Wende, & Will, 2005).

We used established criteria for validity and reliability of PLS-SEM models (Hair et al., 2014). Indicator loadings ≥0.7, composite reliability >.7, Cronbach’s alpha ≥.6, and average variance extracted >0.50 established the convergent validity/reliability of reflective constructs. Factor loadings <0.4 led to elimination of reflective indicators. Discriminant validity was supported by average variances extracted exceeding the squared correlations between a reflective construct and others, plus loadings that exceed cross-loadings.

We evaluated formative constructs based on multicollinearity (if variance inflation factor > 10, the indicator was dropped), indicator weights, and external validity (whether the formative model accounted for >50% of construct’s variability). We assessed the structural model on collinearity, direction, and significance of path (R²) coefficients (t tests derived from bootstrapping with 5,000 resamples), effect of exogenous on endogenous constructs (Cohen’s “effect size” f² of 0.02, 0.15, or 0.35 indicated weak, moderate, or substantial influence), and predictive validity (Stone-Geisser test (Q²) > 0).

The sample size precluded a split into evaluation and validation subsamples. The classic way to test moderator effects for a binary variable is to specify the model separately for each group, then assess differences in how well it fits each group (Baron & Kenny, 1986). We adapted the multigroup analysis strategy of Sarstedt and colleagues (Sarstedt, Henseler, & Ringle, 2011) to test for significant moderation effects. We also tested interactions of job satisfaction (as a continuous, nonnormal score) with latent variable (LV) scores for predictor variables (baseline adherence and the three TPB constructs) within a regression of LV scores for the outcome on LV scores for the exogenous constructs. Because tests for multiplicative interactions tend to have low power (Baron & Kenny, 1986), the significance criterion for interaction effects was p < .10.

Results

Of the 156 respondents, 77% were male; 60% internists and 40% family practitioners; 64% were affiliated with the University of Rochester Medical Center; 36% were late-career professionals (>20 years postresidency); 79% served in practices of ≤5 physicians; 86% in single-, not multispecialty, practices; 38% had panels of >2,500 patients; 67% served in populated areas, with >1,900 people/mile²; 82% served in zip codes where >15% of people lived below poverty; and 86% where >20% of the ambulatory care population belonged to a minority race. 15% practiced in rural (nonmetropolitan) areas.

From the survey, mean job satisfaction of the 156 respondents (on a scale of 1 to 7) was 4.9 (95% CI = 4.7-5.1). 71% of respondents reported being satisfied (ratings ≥ 5); 29% were discontented (ratings ≤ 4). At baseline, the mean (95% CI) adherence rates for the four diabetes quality measures were the following: fundoscopy 46% (CI = 43-48), gycohemoglobin assay 61% (CI = 58-64), LDL cholesterol 67% (CI = 64-71), and urinalysis 65% (CI = 61-68). Mean end point adherence rates for the study population were the following: fundoscopy 55% (CI = 52-57), glycohemoglobin 63% (CI = 60-66), LDL cholesterol 78% (CI = 75-81), urinalysis 69% (CI = 66-72). Differences by satisfaction status in adherence to urinalysis testing at the end point were nearly significant (Kruskal-Wallis χ² = 3.2159, p = .0729); adherence to other measures did not differ significantly (p > .15).

Measures for the TPB constructs (personal attitudes, subjective norms, and behavioral control) met the reliability criteria. Items for effectiveness and utility were not reliable indicators of personal attitudes, and they were dropped from further analysis. We likewise omitted three measures (difficulty of adherence tasks, and their perception as hindering patient care) that had poor reliability as indicators of behavioral control. Deletion of unreliable indicators did not affect the reliability or content validity of underlying constructs.

Table 1 shows characteristics of latent constructs and their indicators. Table 2 shows reliability/validity indices of reflective constructs. They were internally consistent and met convergent validity criteria. Diagonal rows in Table 2 show that reflective constructs satisfied Fornell-Lacker discriminant validity criteria (each construct shared more variance with its indicators than other constructs). The interconstruct correlation matrix shows that, of all latent constructs, baseline adherence correlated the most with end point adherence for satisfied and discontented physicians. It also shows that whereas the correlation between subjective norms and end point adherence was positive and nonsignificant among satisfied physicians, it was negative and significant among not-satisfied physicians. Attitudes correlated more strongly with end point adherence among discontented than among satisfied physicians. Correlations between baseline adherence and TPB constructs were positive among satisfied physicians but negative among those who were not.

Table 1.

Descriptive Statistics, Outer Weights, and Factor Loadings of the Indicator Variables for the Latent Constructs.

Latent construct	Type of indicator	M (SD)	t test	Outer weight	Factor loading	Squared loading
Rate of adherence to diabetes care guidelines at end point
● Annual eye exam	Formative	0.5453 (0.1550)	42.95***	0.2906*	0.5938**	0.3526
● Annual LDL cholesterol	Formative	0.7815 (0.1640)	58.17***	0.3886	0.7757***	0.6017
● Annual urinalysis test	Formative	0.6906 (0.2003)	42.09***	0.7124***	0.9067***	0.8221
● Biannual HbA1c test	Formative	0.6311 (0.1916)	40.21***	−0.2810	0.4321	0.1867
Rate of adherence to diabetes care guidelines at baseline
● Annual eye exam	Formative	0.4565 (0.1600)	34.59***	0.0411	0.2818	0.0794
● Annual LDL cholesterol	Formative	0.6742 (0.1935)	42.24***	0.7788***	0.8769***	0.7690
● Annual urinalysis test	Formative	0.6458 (0.1919)	40.81***	0.5289*	0.7251***	0.5258
● Biannual HbA1c test	Formative	0.6121 (0.1935)	38.35***	−0.1759	0.4496	0.2021
Personal attitudes
● Awareness of guideline	Reflective	2.8465 (0.7559)	46.88***	NA	0.8329***	0.6937
● Financial salience of bonus	Reflective	2.2043 (0.6936)	39.57***		0.8146***	0.6636
● Perceived impact of program	Reflective	2.5811 (0.6867)	46.79***		0.8161***	0.6660
Subjective norms
● Clinical relevance of guideline	Reflective	3.5581 (0.7486)	59.17***	NA	0.7243***	0.5246
● Cooperation from peers/staff	Reflective	2.9513 (0.6878)	53.25***		0.7329***	0.5371
● Fairness/equity of P4P bonus	Reflective	2.8684 (0.9256)	38.21***		0.8445***	0.7132
Perceived behavioral control
● Autonomy support from HMO	Reflective	2.9081 (0.7100)	50.99***	NA	0.8404***	0.7063
● Autonomy support from IPA	Reflective	3.2991 (0.7692)	53.57***		0.9000***	0.8100
● Perceived control	Reflective	2.6269 (0.8413)	38.87***		0.7612***	0.5794

Note. LDL = low-density lipoprotein; M = mean; SD = standard deviation; P4P = pay-for-performance; HMO = health maintenance organization; IPA = independent practice association; NA = not applicable to reflective constructs.

p < .05. **p < .01. ***p < .001.

Table 2.

Goodness of Fit Indices and Correlation Matrix of the Measurement Model.

Latent construct	Communality	Cronbach’s alpha (α)	Composite reliability (ρ_c)	Average variance explained	Correlations between latent constructs by satisfaction status
Latent construct	Communality	Cronbach’s alpha (α)	Composite reliability (ρ_c)	Average variance explained	End point adherence	Baseline adherence	Personal attitudes	Subjective norms	Perceived behavioral control
End point adherence	0.4710	Formative construct	Formative construct	Formative construct	Formative construct	0.5061	0.1087	0.0805	0.0712
Baseline adherence	0.3842	Formative construct	Formative construct	Formative construct	0.6051	Formative construct	0.2222	0.2164	0.2041
Personal attitudes	0.6744	.7604	.8614	0.6744	0.1503	−0.1264	0.8212	0.6398	0.6769
Subjective norms	0.5916	.6538	.8122	0.5916	−0.2430	−0.4099	0.5968	0.7894	0.7011
Perceived behavioral control	0.6986	.7817	.8737	0.6986	−0.1205	−0.2312	0.6999	0.6059	0.8227

Note. Correlations among satisfied physicians are above the diagonal; those among unsatisfied physicians are below the diagonal. Boldfaced numbers in the diagonal rows for reflective constructs are square roots of the average variance explained. These values exceed the interconstruct correlations, which fulfills the Fornell-Lacker criterion for discriminant validity.

Formative measurement models did not show significant collinearity (variance inflation factor = 1.03 to 1.55 for baseline adherence; 1.19 to 2.02 for end point adherence) and met convergent validity criteria (formative constructs account for >90% of variability in baseline/end point adherence). Indicator weights for fundoscopy and urinalysis at baseline and for LDL cholesterol assay and urinalysis at the study end point reached significance (see Table 1). The indicator weight for baseline glycohemoglobin tests was not significant, but its outer loading was >0.708, supporting its importance to the baseline adherence construct, even though it was less significant than fundoscopy or urinalysis. Outer loadings for baseline/end point glycohemoglobin rates were <0.50 but >.40, and were kept in the model. For conceptual validity (Hair et al., 2014), we retained baseline fundoscopy, despite a nonsignificant loading.

Table 3 compares the measurement model between physicians who were satisfied and those who were not. In both groups, baseline adherence to quality measures was the strongest predictor of end point adherence (β coefficients = .5114 and .5443, respectively; p < .0001), even after accounting for personal attitudes, subjective norms, and perceived behavioral control. Among physicians who were discontented, personal attitudes (awareness of guidelines, financial salience of P4P bonus, and their perceived impact) independently predicted end point adherence (β = .4779, p = .0028). We detected no similar independent effect for personal attitudes among satisfied physicians (β = .0405, p = .6577).

Table 3.

Measurement Model Path Coefficients and Hypotheses Test Results.

Group	Structural relationship	Path coefficient (standard error)	Bootstrap t test (p value)	Effect size (f²)
Satisfied physicians	Baseline adherence End point adherence	0.5114 (0.0529)	9.6717 (<.0001)	0.2785
	Baseline adherence Behavioral control	0.0213 (0.0538)	0.3949 (.6932)	0.0007
	Baseline adherence Subjective norms	0.0781 (0.1010)	0.7731 (.4400)	0.0106
	Baseline adherence Personal attitudes	0.2222 (0.0915)	2.4273 (.0157)	0.0520
	Behavioral control End point adherence	−0.0419 (0.0797)	0.5256 (.5995)	0.0012
	Subjective norm End point adherence	−0.0267 (0.0992)	0.2691 (.7880)	0.0016
	Personal attitudes End point adherence	0.0405 (0.0913)	0.4435 (.6577)	0.0031
	Subjective norms Behavioral control	0.4511 (0.0441)	10.2280 (<.0001)	0.2851
	Personal attitudes Behavioral control	0.3836 (0.0444)	8.6410 (<.0001)	0.1910
	Personal attitudes Subjective norms	0.6225 (0.0344)	18.0825 (<.0001)	0.6313
Physicians not satisfied with practice	Baseline adherence End point adherence	0.5443 (0.1232)	4.4208 (<.0001)	0.2566
	Baseline adherence Behavioral control	−0.0553 (0.0547)	1.0113 (.3126)	0.0035
	Baseline adherence Subjective norms	−0.3399 (0.0487)	6.9753 (<.0001)	0.1180
	Baseline adherence Personal attitudes	−0.1264 (0.0665)	1.9016 (.0581)	0.0163
	Behavioral control End point adherence	−0.2274 (0.1745)	1.3028 (.1935)	−0.0002
	Subjective norms End point adherence	−0.1673 (0.0738)	2.2675 (.0240)	0.0058
	Personal attitudes End point adherence	0.4779 (0.1589)	3.0079 (.0028)	0.0132
	Subjective norms Behavioral control	0.2642 (0.0781)	3.3821 (.0008)	0.3421
	Personal attitudes Behavioral control	0.5345 (0.0667)	8.012 (<.0001)	0.0708
	Personal attitudes Subjective norms	0.5539 (0.0491)	11.2892 (<.0001)	0.5788

Note. The boldfaced structural relationships are statistically significant at the 0.05 probability level.

Among physicians who were discontented, there was a small but significant negative effect of subjective norms (clinical relevance of guidelines, peer/staff cooperation in adherence efforts, and fairness/equity of P4P bonuses) on end point adherence (β = −.1673, p = .0240), independent of baseline adherence. No similar independent effect for norms was detected among satisfied physicians (β = −.267, p = .7880). Among both satisfied and discontented physicians, perceived behavioral control did not independently predict end point adherence (β = −0.0419, p = .5995; and β = −.2274, p = .1935, respectively).

As seen in Table 3, the intermediate relationships in the measurement model also differed by satisfaction status. Among satisfied physicians, baseline-year performance significantly predicted personal attitudes (β = .2222, p = .0157). For discontented physicians, this association was both negative and marginally nonsignificant (β = −.1264, p = .0581). The reverse was true for the baseline-performance-norms association, which was not significant for satisfied physicians (β = .0781, p = .4400), but significant and negative for discontented ones (β = −.3399, p < .0001). Among all physicians, personal attitudes predicted subjective norms and behavioral control perceptions, and subjective norms predicted behavioral control perceptions. Appendix A shows more details of the path analysis for the outer (measurement) model.

The factors that influenced provider adherence to guidelines differed by job satisfaction. As Appendix B shows, the linear regression of LV scores for end point adherence on LV scores for exogenous constructs and their interactions with satisfaction was significant (p < .10) for the interaction between personal attitudes and job satisfaction, supporting our hypothesis (Sarstedt et al., 2011) that satisfaction moderates the attitude–performance link. Job satisfaction, therefore, moderated the effect of awareness/knowledge of guidelines, sufficiency of financial bonuses, and their perceived impact on preventive measure performance. Appendix C shows further details of moderating effects. Predictive validity criteria were satisfied (Stone-Geisser Q² coefficients > 0).

Figure 2A and B illustrates the distilled “inner” models after nonsignificant paths were dropped from the measurement “outer” models. For discontented physicians, attitudes (awareness of guidelines, financial salience of P4P bonuses, and their perceived impact) had a significant, positive predictive association with adherence to guidelines. For this group, perceived social pressure from significant others (the latter’s views on relevance of guidelines, likely peer/staff cooperation, and fairness of P4P bonuses) had a significant negative association. Past adherence behavior, personal attitudes, and subjective norms explained 48% of variability in end point adherence, with prior behavior having the most significant influence. For satisfied physicians, prior adherence was also the strongest influence on current performance (accounting for 26.4% of variability by itself). Attitudes, social norms, and behavioral control did not significantly predict guideline performance for satisfied physicians. In summary, adherence of satisfied physicians was not influenced by behavioral control, subjective norms, or personal attitudes, while the latter two predicted adherence by discontented physicians.

Discussion

The primary purpose of this study was to test the degree to which job satisfaction moderates the three TPB-based predictors of adherence to ambulatory care guidelines within a P4P incentive program. We found that provider satisfaction had a significant indirect effect on performance. Physicians who are discontented with their practice were only motivated to adhere to guidelines that aligned with their personal attitudes. Discontented physicians were less likely to adhere to guidelines when they sensed social pressure to comply—pressure that they resented as coercive. For satisfied physicians, neither personal attitudes, nor subjective norms, nor perceived behavioral control significantly affected behavior at the study end point.

The results justify further studies of physician satisfaction as a moderator of their adherence to clinical guidelines. The results provide support for the prospect that monitoring job satisfaction among physicians could aid in identifying and crafting tailored interventions to promote their use of clinical guidelines.

Further research can also investigate whether quality improvement interventions that are tailored to differentially satisfied audiences, based on treating job satisfaction as a moderator, are effective in changing adherence behavior. Past physician behavior had a significant, positive association with attitudes, suggesting that satisfied physicians were more likely to favor adherence at the end point based on clinical habits, specific to these diabetes process measures, which arose from positive prior experiences. Past successes likely gave satisfied physicians a sense of accomplishment and confidence in achieving future success.

Our hypothesis that adherence to clinical guidelines would be positively associated with social norms among satisfied physicians was not supported. For physicians who are discontented, both personal attitudes and subjective norms influenced guideline adherence albeit in opposite ways (see Figure 2B). Strong awareness of the quality measures, believing that they were effective, and believing in the adequacy and value of the financial incentives, increased the likelihood of adherence among discontented, but not satisfied, physicians.

Beliefs that patients, practice managers, or professional peers considered the guidelines relevant; the distribution of P4P bonuses fair; that they would recommend adherence, demotivated the discontented physicians, but had no effect among their satisfied colleagues. Perceived social pressure from the opinions of influential observers in their work setting suppressed adherence among physicians who were discontented and who wanted to follow their own personal evaluations of the guidelines. In this group, higher baseline adherence was linked to more negative perceptions of pressure from others, suggesting that negative past experiences among physicians who are discontented could underlie their skepticism of the social pressure to adhere.

Viewed from the perspective of social identity theories, our findings could imply that discontented physicians resented the normative pressure from satisfied peers, patients, and/or managers, from whom they felt partially alienated (McKinlay & Marceau, 2011) and with whom they did not share a mutual social identity (Dukerich et al., 2002). Workplace norms tended to be perceived by alienated physicians as not being in their best interests.

Conscientious objections can motivate physicians to deviate from clinical guidelines (Mottur-Pilson, Snow, & Bartlett, 2001). The Minimizing Error, Maximizing Outcome (MEMO) study found higher levels of guideline-adherent care for diabetes when personal values of physicians aligned with those of organization leaders (Linzer et al., 2009). Results from a qualitative study of medical decision making (Ginsburg, Bernabeo, & Holmboe, 2014) show that physicians often do what is perceived as wrong, by their patients, peers, or organization, when asked to do something against their personal values or beliefs. Physicians may ignore social pressure to adhere to guidelines that they disagree with, and substitute them with actions “tailored” to specific patients (Vashitz et al., 2011), rationalizing these as done for the right motives. There is clearly potential for the expectations of important others, such as patients, to conflict with what physicians view as appropriate medical decisions (Regan, Ferris, & Campbell, 2010).

In our study, tension between personal attitudes and social norms manifested more among physicians who were discontented. Contrary to our hypotheses, perceived behavioral control did not reach statistical significance in either group. It is likely that among both types of physicians, personal control of the drivers of guideline adherence was perceived as unrealistic in the context of P4P incentives that included a punitive financial “withhold” for nonadherence.

Implications

In this study, neither attitudinal nor social pressures significantly influenced satisfied physicians. Subject to confirmation by future studies, this suggests that professional satisfaction can have a stabilizing effect on physician clinical behavior. Physician satisfaction, as a moderator, is thus a valuable component of practice management that ought to be monitored regularly.

While it is difficult to extrapolate from this limited context, physician satisfaction undoubtedly varies across organizations and settings. Tailored strategies to improve physician satisfaction and well-being should be studied in greater depth (Regehr, Glancy, Pitts, & LeBlanc, 2014). Recently, increasing attention is being paid to professional satisfaction, resilience and well-being as indicators of the quality of systems of health care delivery. Next to health status, patient satisfaction, and cost, provider satisfaction ought to be one of the four key outcomes of care (Grembowski et al., 2003). It has been dubbed “the missing quality indicator” (Spinelli, 2013; Wallace, Lemaire, & Ghali, 2009). Bodenheimer and Sinsky (2014) advocate expanding the “triple aim” approach to healthcare reform into a “quadruple aim,” to incorporate practitioner well-being as the fourth leg.

Incentive-based guidelines are more likely to motivate adherence among discontented physicians when they align with their personal attitudes. They are also more likely to improve adherence by discontented practitioners if programs engage in social marketing campaigns targeting negative reactions to approval of guidelines by their peers or supervisors/managers. Such programs can minimize “autonomy suppressing” forms of social influence that lead to psychological reactance and, instead, maximize “autonomy supporting” forms of social influence that enhance identification. One such strategy might be to use peer champions in every department who are early adopters of guidelines, providing them with leadership training, and empowering them to break social identity/self-categorization silos in the health care organization. This calls for contextualized guidelines that are tailored to the satisfaction profiles of individual physicians or groups.

Limitations

This study has a number of notable limitations. The available sample was confined to a particular IPA population. The subsample of physicians that were discontented is relatively small. Third, the survey method could have introduced a selection bias. We had no survey data on nonrespondents and thus could not test whether their attitudes significantly differed from those of respondents. Minor differences in specialty mix and practice size variation existed between survey respondents and nonrespondents (Meterko et al., 2006; Waddimba et al., 2010).

The study excluded pediatricians, higher proportions of whom are female, so the analyzed sample comprised a higher proportion of males (77.12%) than did respondents who were excluded from this analysis (56.15%). At the conclusion of the survey, 77.8% of U.S. physicians were male (Hing & Burt, 2007), which compares favorably with the proportions in the sample that we analyzed for this study. Survey data were cross sectional in nature. It would be valuable to examine these relationships over a longer time period. Finally, the financial incentive model employed was unique to the IPA studied. The findings from the present study must be viewed as preliminary and interpreted with caution. They need to be confirmed by similar studies conducted in diverse settings and contexts, among different populations, with larger sample sizes, across multiple time points and with clinical outcome (not just process) measures included.

The study also had important strengths. Since self-report measures are reported to overestimate adherence (Maue, Segal, Kimberlin, & Lipowski, 2004; O’Boyle, Henly, & Larson, 2001), our recourse to objective/observed measures of guideline adherence was a key strength. Use of real-world data to empirically test a conceptual model drawn from existing scientific theory (Reyna, 2008) was also a strength. Nonparametric analysis that is robust to small samples, nonnormal distributions, and inclusion of both reflective and formative constructs in a single model (Hair et al., 2014) was a further strength.

Conclusion

Influences on clinical guideline behavior among satisfied practitioners appear to be different from those who are discontented, illustrating a moderation effect that can be exploited in future research, policy and practice. Physicians might act quite independently of seemingly important influences, if they are satisfied with their practice. Those who are discontented, by contrast, seem to be influenced positively by personal attitudes and negatively influenced by perceived social pressures from influential others. Incentive-based quality improvement programs, based on interventions incorporating this view of job satisfaction as a moderator, have two basic on-the-ground choices: either to focus on boosting overall provider satisfaction in order to minimize extraneous influences on provider acceptance of clinical guidelines, or to determine who is discontented and deliver focused interventions to that group. Further research is required to sharpen these implications. However, routine monitoring of physician job satisfaction would appear to be a reasonable approach for practices, health systems, accountable care organizations, and medical groups interested in understanding and improving their clinical performance.

Footnotes

Appendix

Appendix B

Assessing the Moderator Effect of Job Satisfaction via Linear Regression of the PLS-SEM Latent Variable Scores^a.

Variable	Parameter estimate (standard error)	90% confidence interval	t statistic (p value)
Intercept	−0.0333 (0.0720)	−0.1524, 0.0859	−0.46 (.6448)
Latent variable score for baseline adherence to guideline	0.3287 (0.2596)	−0.1011, 0.7585	1.27 (.2075)
Latent variable score for the personal attitudes construct	0.8104 (0.4288)	0.1006, 1.5203	1.89 (.0608)
Latent variable score for the subjective norms construct	−0.5659 (0.3977)	−1.2243, 0.0925	−1.42 (.1569)
Latent variable score for the perceived behavioral control construct	−0.4540 (0.3984)	−1.1134, 0.2055	−1.14 (.2564)
Interaction between satisfaction and baseline adherence	0.0436 (0.0540)	−0.0458, 0.1330	0.81 (.4206)
Interaction between satisfaction and personal attitudes	−0.1413 (0.0836)	−0.2798, −0.0029	−1.69 (.0931)
Interaction between satisfaction and subjective norms	0.1067 (0.0819)	−0.0289, 0.2422	1.30 (.1947)
Interaction between satisfaction and perceived behavioral control	0.0687 (0.0766)	−0.0581, 0.1955	0.90 (.3711)

Note. PLS-SEM = partial least squares structural equation modeling.

These are results of an ordinary least squares regression of the latent variable score for the study outcome (adherence to diabetes care guidelines at the end point) on the latent variable scores for the exogenous constructs, including interaction terms between those latent variable scores and job satisfaction (the moderator variable). A statistically significant interaction effect (p < .10) is evidence of moderation by the variable tested. Boldfaced values are statistically significant at the 0.10 probability level.

Acknowledgements

The authors sincerely thank all the primary care physicians from the Rochester Independent Practice Association (RIPA) who generously responded to the Quality Targets and Incentives Survey. We are grateful to our fellow investigators on the national evaluation of the Rewarding Results Demonstration Projects, especially Gary J Young, JD, PhD; Robert Greene, MD; Mark Meterko, PhD; Dan Berlowitz, MD; and Bert White, DMin, MBA. Special thanks go to Greg Partridge for his immense help in organizing the administrative data that were employed in this study. Mr. Matt Guldin and Ms. Karen Sautter provided research assistance at various stages of the parent project. The editors and reviewers of the Medical Care Research and Review are commended for their numerous helpful insights and suggestions. We further acknowledge Joshua Hammonds, PhD, of the State University of New York in Oneonta, for his valuable comments on an earlier draft of this article.

Authors’ Note

The findings reported in this article are solely those of the authors and do not represent the official views of an institution to which any or all of the authors are individually or jointly affiliated.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Ajzen

(1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179-211.

Ajzen

(2014). The theory of planned behaviour is alive and well, and not ready to retire: A commentary on Sniehotta, Presseau, and Araújo-Soares. Health Psychology Review, 9, 131-137.

Ali

M. K.

Bullard

K. M.

Saaddine

J. B.

Cowie

C. C.

Imperatore

Gregg

E. W.

(2013). Achievement of goals in U.S. Diabetes Care, 1999-2010. New England Journal of Medicine, 368, 1613-1624.

American Association of Clinical Endocrinologists. (2013). AACE comprehesive diabetes algorithm. Endocrine Practice, 19, 327-336.

American Diabetes Association. (2013). Economic costs of diabetes in the U.S. in 2012. Diabetes Care, 36, 1033-1046.

American Diabetes Association. (2014). Standards of medical care in diabetes—2014. Diabetes Care, 37(Suppl. 1), S14-S80.

Anderson

J. C.

Gerbing

D. W.

(1988). Structural equation modeling in practice: A review and recommended two-step approach. Psychological Bulletin, 103, 411-423.

Armitage

C. J.

Conner

(2001). Efficacy of the theory of planned behaviour: A meta-analytic review. British Journal of Social Psychology, 40(Pt. 4), 471-499.

Askelson

N. M.

Campo

Lowe

J. B.

Dennis

L. K.

Smith

Andsager

(2010). Factors related to physicians’ willingness to vaccinate girls against HPV: The importance of subjective norms and perceived behavioral control. Women & Health, 50, 144-158.

10.

Bagozzi

R. P.

(1989). The degree of intention formation as a moderator of the attitude-behavior relationship. Social Psychology Quarterly, 52, 266-279.

11.

Baron

R. M.

Kenny

D. A.

(1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182.

12.

Bodenheimer

Sinsky

(2014). From triple to quadruple aim: Care of the patient requires care of the provider. Annals of Family Medicine, 12, 573-576.

13.

Boyd

C. M.

Darer

Boult

Fried

L. P.

Boult

A. W.

(2005). Clinical practice guidelines and quality of care for older patients with multiple comorbid diseases: Implications for pay for performance. JAMA Journal of the American Medical Association, 294, 716-724.

14.

Boyle

J. P.

Thompson

T. J.

Gregg

E. W.

Barker

L. E.

Williamson

D. F.

(2010). Projection of the year 2050 burden of diabetes in the U.S. adult population: Dynamic modeling of incidence, mortality, and prediabetes prevalence. Population Health Metrics, 8(29). doi:10.1186/1478-7954-8-29

15.

Brehm

S. S.

Brehm

J. W.

(1981). Psychological reactance: A theory of freedom and control. New York, NY: Academic Press.

16.

Chau

P. Y. K.

P. J.-H.

(2002). Investigating healthcare professionals’ decisions to accept telemedicine technology: An empirical test of competing theories. Information and Management, 39, 297-311.

17.

Chin

W. W.

Newsted

P. R.

(1999). Structural equation modeling analysis with small samples using partial least squares. In Hoyle

R. H.

(Ed.), Statistical strategies for small sample research (pp. 307-342). Thousand Oaks, CA: Sage.

18.

Conner

Armitage

(1998). Extending the theory of planned behavior: A review and avenues for further research. Journal of Applied Social Psychology, 28, 1429-1464.

19.

Crane

M. F.

Platow

M. J.

(2010). Deviance as adherence to injunctive group norms: The overlooked role of social identification in deviance. British Journal of Social Psychology, 49(Pt. 4), 827-847.

20.

Dawis

R. V.

(1994). The theory of work adjustment as a convergent theory. In Savickas

M. L.

Lent

R. W.

(Eds.), Convergence in career development theories (pp. 33-44). Palo Alto, CA: Consulting Psychologists Press.

21.

Dawis

R. V.

Lofquist

L. H.

(1984). A psychological theory of work adjustment: An individual-differences model and its applications. Minneapolis: University of Minnesota Press.

22.

Deci

E. L.

Ryan

R. M.

(2011). Self-determination theory. In Van Lange

P. A. M.

Kruglanski

A. W.

Higgins

E. T.

(Eds.), Handbook of theories of social psychology (pp. 416-437). Thousand Oaks, CA: Sage.

23.

Diamantopoulos

Siguaw

J. A.

(2006). Formative versus reflective indicators in organizational measure development: A comparison and empirical illustration. British Journal of Management, 17, 263-282.

24.

Dukerich

J. M.

Golden

B. R.

Shortell

S. M.

(2002). Beauty is in the eye of the beholder: The impact of organizational identification, identity, and image on the cooperative behaviors of physicians. Administrative Science Quarterly, 47, 507-533.

25.

Emmanuel

E. J.

Pearson

S. D.

(2012). Physician autonomy and health care reform. JAMA Journal of the American Medical Association, 307, 367-368.

26.

Festinger

Carlsmith

J. M.

(1959). Cognitive consequences of forced compliance. Journal of Abnormal Psychology, 58, 203-210.

27.

Fishbein

(2008). A reasoned action approach to health promotion. Medical Decision Making, 28, 834-844.

28.

Fishbein

Ajzen

(2010). Predicting and changing behavior: The reasoned action approach. New York, NY: Psychology Press.

29.

Friedberg

M. W.

Chen

P. G.

Van Busum

K. R.

Aunon

Pham

Caloyeras

. . . Tutty

(2013). Factors affecting physician professional satisfaction and their implications for patient care, health systems, and health policy. Santa Monica, CA: Rand Corporation.

30.

Friedberg

M. W.

Schneider

E. C.

Rosenthal

M. B.

Volpp

K. G.

Werner

R. M.

(2014). Association between participation in a multipayer medical home intervention and changes in quality, utilization, and costs of care. JAMA Journal of the American Medical Association, 311, 815-825.

31.

Ginsburg

Bernabeo

Holmboe

(2014). Doing what might be “Wrong”: Understanding internists’ responses to professional challenges. Academic Medicine, 89, 664-670.

32.

Godin

Kok

(1996). The theory of planned behavior: A review of its applications to health-related behaviors. American Journal of Health Promotion, 11, 87-98.

33.

Godin

Naccache

Fortin

(1998). Understanding physician intention to use a simple infection control measure: Wearing gloves. American Journal of Infection Control, 26, 413-417.

34.

Graham

Mancher

Wolman

D. M.

Greenfield

Steinberg

(2011). Clinical practice guidelines we can trust. Washington, DC: National Academies Press.

35.

Grembowski

Ulrich

C. M.

Paschane

Diehr

Katon

Martin

. . . Velicer

(2003). Managed care and primary physician satisfaction. Journal of the American Board of Family Medicine, 16, 383-393.

36.

Grumbach

(2013). The patient-centered medical home is not a pill: Implications for evaluating primary care reforms. JAMA Internal Medicine, 173, 1913-1914.

37.

Hagger

M. S.

Chatzisarantis

N. L.

(2009). Integrating the theory of planned behaviour and self-determination theory in health behaviour: A meta-analysis. British Journal of Health Psychology, 14(Pt. 2), 275-302.

38.

Hagger

M. S.

Chatzisarantis

N. L.

Harris

(2006). From psychological need satisfaction to intentional behavior: Testing a motivational sequence in two behavioral contexts. Personality and Social Psychology Bulletin, 32, 131-148.

39.

Hair

J. F.

Hult

G. T. M.

Ringle

C. M.

Sarstedt

(2014). A primer on partial least squares structural equation modeling (PLS-SEM). Thousand Oaks, CA: Sage.

40.

Handelsman

Mechanick

J. I.

Blonde

Grunberger

Bloomgarden

Z. T.

Bray

G. A.

. . . Wyne

K. L.

(2011). American Association of Clinical Endocrinologists Medical Guidelines for clinical practice for developing a diabetes mellitus comprehensive care plan: Executive summary. Endocrine Practice, 17, 287-302.

41.

Head

K. J.

Noar

S. M.

(2014). Facilitating progress in health behaviour theory development and modification: The reasoned action approach as a case study. Health Psychology Review, 8, 34-52.

42.

Hekman

D. R.

Steensma

H. K.

Bigley

G. A.

Hereford

J. F.

(2009). Effects of organizational and professional identification on the relationship between administrators’ social influence and professional employees’ adoption of new work behavior. Journal of Applied Psychology, 94, 1325-1335.

43.

Hickam

D. H.

Severance

Feldstein

Ray

Gorman

Schuldheis

. . . Helfand

(2003). The effect of health care working conditions on patient safety: AHRQ evidence report summary. Rockville, MD: Agency for Healthcare Research and Quality.

44.

Hing

Burt

C. W.

(2007). Characteristics of office-based physicians and their practices: United States, 2003-04 (Vital and Health Statistics, Series 13, No. 164). Hyattsville, MD: Centers for Disease Control and Prevention.

45.

Hoff

(2010). Practice under pressure: Primary care physicians and their medicine in the twenty-first century. New Brunswick, NJ: Rutgers University Press.

46.

Hogan

Dall

Nikolov

(2003). Economic costs of diabetes in the U.S. in 2002. Diabetes Care, 26, 917-932.

47.

Hopwood

C. J.

(2007). Moderation and mediation in structural equation modeling: Applications for early intervention research. Journal of Early Intervention, 29, 262-272.

48.

Hotho

(2008). Professional identity: Product of structure, product of choice. Journal of Organizational Change Management, 21, 721-742.

49.

Institute of Medicine. (2006). Case study 4: Rochester Individual Practice Association. In Performance measurement: Accelerating improvement (Pathways to quality health care series) (pp. 158-159). Washington, DC: National Academies Press.

50.

Kaplan

S. H.

Griffith

J. L.

Price

L. L.

Pawlson

L. G.

Greenfield

(2009). Improving the reliability of physician performance assessment: Identifying the “Physician Effect” on quality and creating composite measures. Medical Care, 47, 378-387.

51.

Legare

Godin

Ringa

Dodin

Turcot

Norton

(2005). Variation in psychosocial determinants of intention to prescribe hormone therapy prior to the release of the women’s health initiative trial: A survey of general practitioners in France and Quebec. BMC Medical Informatics and Decision Making, 5, 31. doi:10.1186/1472-6947-5-31

52.

Liabsuetrakul

Chongsuvivatwong

Lumbiganon

Lindmark

(2003). Obstretricians’ attitudes, subjective norms, perceived controls, and intentions on antibiotic prophylaxis in cesarean section. Social Science & Medicine, 57, 1665-1674.

53.

Linzer

Manwell

L. B.

Williams

E. S.

Bobula

J. A.

Brown

R. L.

Varkey

A. B.

. . . Schwartz

M. D.

(2009). Working conditions in primary care: Physician reactions and care quality. Annals of Internal Medicine, 151, 28-36, w26-29.

54.

Maue

S. K.

Segal

Kimberlin

C. L.

Lipowski

E. E.

(2004). Predicting physician guideline compliance: An assessment of motivators and perceived barriers. American Journal of Managed Care, 10, 383-391.

55.

McKinlay

J. B.

Marceau

(2011). New wine in an old bottle: Does alienation provide an explanation of the origins of physician discontent? International Journal of Health Services, 41, 301-335.

56.

Medical Group Management Association. (2014). MGMA physician practice assessment: Medicare quality reporting programs survey report. Washington, DC: Author.

57.

Menke

Casagrande

Geiss

Cowie

C. C.

(2015). Prevalence of and trends in diabetes among adults in the United States, 1988-2012. Journal of the American Medical Association, 314, 1021-1029.

58.

Meterko

Young

G. J.

White

Bokhour

B. G.

Burgess

J. F.

Berlowitz

. . . Nealon Seibert

(2006). Provider attitudes toward pay-for-performance programs: Development and validation of a measurement instrument. Health Services Research, 41, 1959-1978.

59.

Montano

D. E.

Phillips

W. R.

Kasprzyk

(2000). Explaining physician rates of providing flexible sigmoidoscopy. Cancer Epidemiology, Biomarkers & Prevention, 9, 665-669.

60.

Mottur-Pilson

Snow

Bartlett

(2001). Physician explanations for failing to comply with “Best Practices.” Effective Clinical Practice, 4, 207-213.

61.

National Committee for Quality Assurance. (2013). Improving quality and patient experience: The state of health care quality 2013. Retrieved from http://www.statecoverage.org/node/5079

62.

National Committee for Quality Assurance. (2015). HEDIS 2015. Retrieved from http://www.ncqa.org/HEDISQualityMeasurement/HEDISMeasures/HEDIS2015.aspx

63.

O’Boyle

C. A.

Henly

S. J.

Larson

(2001). Understanding adherence to hand hygiene recommendations: The theory of planned behavior. American Journal of Infection Control, 29, 352-360.

64.

O’Connor

P. J.

Bodkin

N. L.

Fradkin

Glasgow

R. E.

Greenfield

Gregg

. . . Wysham

C. H.

(2011). Diabetes performance measures: Current status and future directions. Diabetes Care, 34, 1651-1659.

65.

Ouellette

J. A.

Wood

(1998). Habit and intention in everyday life: The multiple processes by which past behavior predicts future behavior. Psychological Bulletin, 124, 54-74.

66.

Rebergen

Hoenen

Heinemans

Bruinvels

Bakker

van Mechelen

(2006). Adherence to mental health guidelines by Dutch occupational physicians. Occupational Medicine, 56, 461-468.

67.

Regan

Ferris

T. G.

Campbell

E. G.

(2010). Physician attitudes toward personal relationships with patients. Medical Care, 48, 547-552.

68.

Regehr

Glancy

Pitts

LeBlanc

V. R.

(2014). Interventions to reduce the consequences of stress in physicians: A review and meta-analysis. Journal of Nervous and Mental Disease, 202, 353-359.

69.

Reyna

V. F.

(2008). Theories of medical decision making and health: An evidence-based approach. Medical Decision Making, 28, 829-833.

70.

Rigdon

E. E.

(2013). Partial least squares path modeling. In Hancock

G. R.

Mueller

R. O.

(Eds.), Structural equation modeling: A second course (2nd ed., pp. 81-116). Charlotte, NC: Information Age.

71.

Ringle

C. M.

Wende

Will

(2005). SmartPLS 2.0. Retrieved from www.smartpls.de

72.

Sable

M. R.

Schwartz

L. R.

Kelly

P. J.

Lisbon

Hall

M. A.

(2006). Using the theory of reasoned action to explain physician intention to prescribe emergency contraception. Perspectives on Sexual and Reproductive Health, 38, 20-27.

73.

Sarstedt

Henseler

Ringle

(2011). Multi-group analysis in partial least squares (PLS) path modeling: Alternative methods and empirical results. Advances in International Marketing, 22, 195-218.

74.

Sheeran

(2002). Intention-behavior relations: A conceptual and empirical review. European Review of Social Psychology, 12, 1-36.

75.

Spinelli

W. M.

(2013). The phantom limb of the triple aim. Mayo Clinic Proceedings, 88, 1356-1357.

76.

Sutton

(1994). The past predicts the future: Interpreting behaviour-behaviour relationships in social psychological models of health behaviour. In Rutter

D. R.

Quine

(Eds.), Social psychology and health: European perspectives (pp. 71-88). Aldershot, UK: Avebury.

77.

Terry

D. J.

Hogg

M. A.

(2001). Attitudes, behavior, and social context: The role of norms and group membership in social influence processes. In Forgas

J. P.

Williams

K. D.

(Eds.), Social influence: Direct and indirect processes (pp. 253-270). Philadelphia, PA: Psychology Press.

78.

Terry

D. J.

Hogg

M. A.

McKimmie

B. M.

(2000). Attitude-behaviour relations: The role of in-group norms and mode of behavioural decision-making. British Journal of Social Psychology, 39(Pt 3), 337-361.

79.

Turner

J. C.

Reynolds

K. J.

(2010). The story of social identity. In Postmes

Branscombe

N. R.

(Ed.), Rediscovering social identity: Key readings (pp. 13-32). New York, NY: Psychology Press.

80.

Turner

J. C.

Reynolds

K. J.

(2011). Self-categorization theory. In Van Lange

P. A. M.

Kruglanski

A. W.

Higgins

E. T.

(Eds.), Handbook of theories of social psychology (Vol. 2, pp. 399-417). Thousand Oaks, CA: Sage.

81.

Vashitz

Meyer

Parmet

Henkin

Peleg

Liebermann

Gilutz

(2011). Adherence by primary care physicians to guidelines for the clinical management of dyslipidemia. Israel Medical Association Journal, 13, 657-662.

82.

Waddimba

A. C.

Burgess

J. F.

Jr. Young

G. J.

Beckman

H. B.

Meterko

(2013). Motivators and hygiene factors among physicians responding to explicit incentives to improve the value of care. Quality Management in Healthcare, 22, 285-301.

83.

Waddimba

A. C.

Meterko

Beckman

H. B.

Young

G. J.

Burgess

J. F.

Jr. (2010). Provider attitudes associated with adherence to evidence-based clinical guidelines in a managed care setting. Medical Care Research and Review, 67, 93-116.

84.

Walker

A. E.

Grimshaw

J. M.

Armstrong

E. M.

(2001). Salient beliefs and intentions to prescribe antibiotics for patients with a sore throat. British Journal of Health Psychology, 6, 347-360.

85.

Wallace

J. E.

Lemaire

J. B.

Ghali

W. A.

(2009). Physician wellness: A missing quality indicator. Lancet, 374(9702), 1714-1721.

86.

Wanous

J. P.

Reichers

A. E.

Hudy

M. J.

(1997). Overall job satisfaction: How good are single-item measures? Journal of Applied Psychology, 82, 247-252.

87.

Young

G. J.

Beckman

Baker

(2012). Financial incentives, professional values and performance: A study of pay-for-performance in a professional organization. Journal of Organizational Behavior, 33, 964-983.

88.

Young

G. J.

Burgess

J. F.

Jr. White

(2007). Pioneering pay-for-quality: Lessons from the rewarding results demonstrations. Health Care Financing Review, 29, 59-70.

89.

Young

G. J.

Meterko

Beckman

Baker

White

Sautter

K. M.

. . . Burgess

J. F.

Jr. (2007). Effects of paying physicians based on their relative performance for quality. Journal of General Internal Medicine, 22, 872-876.