Measures, Methods, and Models of Doctor Satisfaction: Future Research Challenges

Burnout

Burnout has its origins in field work with human service professionals and hence has special relevance to physicians, especially generalists and psychiatrists. Burnout has been associated with impaired job performance and poor health, including headaches, sleep disturbances, irritability, marital difficulties, fatigue, hypertension, anxiety, depression, myocardial infarction, and may contribute to alcoholism and drug addiction (Spickard, Gabbe, & Christensen, 2002). An extensive body of literature documents burnout in physician samples across a wide variety of specialties and populations. Three components of burnout include emotional exhaustion (EE: the depletion of emotional resources), depersonalization (DP: the development of a negative, callous and cynical attitude towards service recipients) and reduced personal accomplishment (PA: the tendency to evaluate one's work negatively, feelings of insufficiency and poor professional self-esteem) (Maslach & Jackson, 2013). Burnout develops over time (Houkes, Winants, Twellaar, & Verdonk, 2011), and some components (e.g., DP) may not be immediately cognitively accessible thus not reliably measured by survey self-report. Yet large physician surveys (Shanafelt et al., 2012) have used single items drawn from larger scales to represent EE and DP components (West, Dyrbye, Sloan, & Shanafelt, 2009). A single 5 statement item developed to represent the complex burnout concept (Schmoldt, Freeborn, & Klevit, 1994) has been validated (Rohland, Kruse, & Rohrer, 2004) and used in physician surveys; this index most closely correlates with the EE burnout subscale. A recent narrative review identified correlates of doctor burnout that elevate risk (Amoafo, Hanbali, Patel, & Singh, 2015) as demographic factors (i.e., female gender, younger age), workplace factors (i.e., heavy workloads and job dissatisfaction) and life situations (poor social support).

The US Tradition

The current framework for measurement of work satisfaction used in many US physician studies can be traced back to Herzberg’s mid 20^th century studies of US workers which found distinct sources of work satisfaction and dissatisfaction (Herzberg, 1966). On the positive side, satisfaction resulted from intrinsically rewarding and intellectually stimulating “work itself,” and environments that granted recognition, fostered achievement, and supported advancement. Conversely, dissatisfaction was associated with “extrinsic” or “hygienic” factors such as poor supervision, restrictive company policies, suboptimal working conditions, contentious interpersonal relationships with colleagues and/or customers, and inadequate compensation. Herzberg asserted that minimal levels of these “hygienic” factors were required to prevent dissatisfaction with work, but removing these barriers did not necessarily increase satisfaction unless more satisfying elements were “intrinsic” to the job. This conceptual framework undergirded a widely used measurement tool called the Job Description Index (JDI) (Kinicki, McKee-Ryan, Schriesheim, & Carson, 2002).

This approach was applied to nursing studies (Stamps, Keuter, Byrne, Voell, & Larson, 2000), and physician specific items were developed for various components of job satisfaction (Stamps & Cruz, 1994). Stamps worked in hospital settings, while Lichtenstein (1984) focused on prisons, so the physician specific items developed for these tools concretely described work arrangements of doctors in those workplaces involving interdependent relationships with administrators, peers, and ancillary personnel. Unfortunately, these early scales failed to include items representing normative elements of professionalism discussed by medical sociologists, for example altruism, a service orientation, or putting patients’ interest first, that might be distinctive sources of gratification affecting physicians career or practice setting choices (Hafferty, 2006).

The UK tradition

In contrast to the JDI paradigm, most studies of physician job satisfaction in the UK and many other countries developed from a parallel industrial psychology tradition and measurement model (Warr, Cook, & Wall, 1979). These scales features multiple distinct intrinsic and extrinsic satisfaction facets, but subsequent adaptations for physicians simply omitted items that seemed not to make sense for this population. The most widely used summary scale uses 10 of the original 15 Warr items (Cooper, Rout, & Faragher, 1989; Rout & Rout, 1997). The five single item intrinsic factors are: freedom to choose your own method of working; amount of variety in your work; opportunities to use your abilities; recognition you get for good work; and amount of responsibility you are given. The four extrinsic factors are: your hours of work; your remuneration; physical working conditions; your colleagues and fellow workers. There is also a general satisfaction item: Taking everything into consideration, how do you feel about your work? Few studies analyze items separately. Some report the general satisfaction item alone, but most sum the 10 items to represent a global construct of physician job satisfaction, using Likert-type response options with a neutral midpoint anchored by "very dissatisfied" and "very satisfied." Varying numbers of response options, scoring methods, and satisfaction cutoff points have made it difficult to compare studies across settings or over time.

Despite its widespread use the last two decades, only one validation study of this scale has been conducted recently on a large population of Australian doctors (Hills, Joyce, & Humphreys, 2012), which found high level of internal consistency —but no support for differentiating intrinsic and extrinsic facets. Hills et al. (2012) suggest including other profession specific job-specific attributes in the scale, especially for physicians working in organizational settings. They also caution that a more inclusive, multi-factor instrument might be needed for in-depth, diagnostic approaches to job satisfaction, and concede that the scale’s brevity may have omitted important facets of job satisfaction in clinical medicine, limiting its capacity to discriminate between different levels of job satisfaction. The authors cautioned that single-item measures of global job satisfaction be may be biased because respondents may overstate levels of summary job satisfaction when not given the choice to answer specific items that would have required the choice of lower values.

It is interesting to note that the original Warr items omitted from physician versions of the scale describe elements of the work environment inconsistent with the notion of the solo practitioner (e.g., immediate supervisor, promotion opportunities, input into organizational policy, effectiveness of local management of the firm, and relationships between management and workers). As more physicians take on employee status and/or work in more tightly regulated environments (McKinlay & Marceau, 2002), these aspects of doctors’ work have come back into research designs either through the pathway of environmental descriptors or as additional facets of job satisfaction measures.

The Physician Worklife Study

One of the most extensive efforts to measure physician job satisfaction using the multi-faceted approach was the Physician Worklife Study (PWS) conducted in the US in the late 1990s by a multi-disciplinary, multi-institutional research team (Konrad et al., 1999; Linzer et al., 2000; McMurray et al., 1997; Williams et al., 1999). Information was gathered by examining open-ended responses from a 1988 survey of large group practice physicians (McMurray et al., 1997). Other items were gleaned from analyzing transcripts of six physician focus groups including ethnic minority and female physicians and those working with challenging populations. This source data confirmed the importance of relationships with patients, subordinates and physician peers, and of day-to-day practice environments, and identified distinctive issues for subgroups of physicians by gender, ethnicity, and practice type.

From this information an extensive item bank was developed and winnowed down in national surveys of US physicians (Konrad et al., 1999; Williams et al., 1999). Application of psychometric techniques yielded 10 facets; three dealt with resources: pay, clinical resources, and personal time; another three dealt with restrictions: autonomy, administration, and patient care issues; four facets dealt with relationships: physician colleagues, other clinical and administrative staff, patients, and the larger community. The survey also featured three 5-item global satisfaction scales measuring job, career, and specialty satisfaction. PWS adaptations and replications internationally have been successful with generally consistent pattern of positive relationships between facets and global measures, although somewhat different facets emerge in different countries. The Japanese adaptation added a prestige facet (Ozaki, Bito, & Matsumura, 2008); while a Swiss multi-faceted version identified five facets: patient care, burden, income-prestige, personal rewards, and professional relations and includes a single item global satisfaction measure (Bovier & Perneger, 2003).

Analyses of facet specific patterns of association with global measures suggest that there are multiple pathways to overall job satisfaction. The pattern of relationships between the various facets and global job satisfaction can differentiate different segments of the medical workforce. For example, global job satisfaction is more strongly associated with satisfaction with relationships with patients among primary care physicians that is the case for specialists, while medical specialists’ global job satisfaction is more closely correlated with satisfaction with other physician colleagues than is the case for primary care physicians (Linzer et al., 2000).

The MEMO study

The MEMO study (Linzer et al., 2009) initiated and conducted by this study team of investigators, examined working conditions, physician reactions to those conditions, and health care quality at 119 ambulatory care clinics in the USA. The logic model is displayed in Figure 1.

Primary data collection employed site visits, direct observation in waiting rooms, interviews with clinic managers, physician surveys, lengthy abstractions of systematically selected patient records and patient surveys. Despite interesting descriptive findings, data analyses only partially supported the hypothesized model. There were strong associations between adverse working conditions and negative physician reactions (dissatisfaction, stress, burnout, and intention to leave practice) with some associations between adverse working conditions and health care quality and errors. Workflow (time pressure and a chaotic work pace), job characteristics (lack of work control), and poor organizational culture (e.g., lack of values alignment between staff and leadership) were strongly associated with adverse physician reactions. A few working conditions were inconsistently associated with poorer patient care, but adverse physician reactions, such as dissatisfaction, stress, and burnout, were not associated with poorer patient care.

One plausible interpretation of this array of findings is that physicians act as buffers between adverse work conditions and patient care—situations affect them strongly, but their reactions do not translate into lower quality care for their patients. The implications of this for physician health, mental health and well-being are worth contemplating. On the other hand, because adverse work conditions were strongly associated with intent to leave the practice, care could eventually suffer from interruptions brought on by the departure of physicians. Because turnover could not be measured within the observation period, it was impossible to assess its impact on continuity or quality in this sample, although there is other evidence that discontinuous care does in fact adversely affect patient outcomes over time (Konrad, Howard, Edwards, Ivanova, & Carey, 2005). The investigators also concluded that their findings may explain difficulties in recruiting and retaining generalists but leave open the question of whether adverse work conditions influence health care quality and safety.

The Healthy Workplace Study

The next phase of the CSSG research team’s activity, the Healthy Workplace Study (HWP) (Linzer et al., 2015), was an intervention aimed at making physician work more satisfying while improving patient care. With this dual aim in mind, the investigators chose to setup and implement a cluster randomized controlled trial with clinicians in 34 different sites in three different regions of the US. When they volunteered to enroll, clinicians provided a variety of baseline survey data collected from their staffs and quality indicators gleaned from the organization’s own medical records. The investigators hypothesized that changes in the workplace, initiated in response to feedback about clinician perceptions and performance, would result in decreased negative outcomes for clinicians and better care for their patients. The specific research question was: "Does receipt of baseline data on work conditions and targeted clinical interventions improve satisfaction, stress, burnout, and intent to leave practice?"

The intriguing aspect of this trial was that the “intervention” employed with the aim of increasing physician satisfaction and reducing burnout, was not prescribed by the investigators, but designed by the clinician subjects themselves and not primarily directed at individual physicians, but at entire primary care teams. Because the specific content of the intervention was not pre-specified by the investigators, the issue of treatment fidelity, that is, heterogeneity in the major treatment variable and limited control over the “dosage” and pace of the implementation, would seem problematic. The practice leadership in all “intervention” clinics received structured feedback in a two-page summary which profiled issues that practice team members themselves had identified as problems. Clinic staffs then met to design and put in place some kind of innovation that they thought would be effective in improving care for patients while making their own work lives more satisfying or at least less chaotic.

This approach makes sense, given that so many of the complaints of physicians, particularly in the US, are negative reactions to a perceived loss of autonomy due to external directives—clinical guidelines, performance standards, or burdensome paperwork, as noted by Casalino & Crosson (2015) in this special issue. In the Healthy Work Place study, investigators functioned like management consultants, suggesting options with some underlying evidence; but participating practitioners ultimately decided which elements they would incorporate in their own plans and took responsibility for implementation.

The study found that while stress was not affected by the intervention, sizeable reductions in clinician burnout and increases in job satisfaction were observed overall. Drilling down further, the research team concluded that efforts to improve communication or workflow were effective in lowering burnout and improving job satisfaction, and may have decreased intention to leave. The implementation of quality improvement projects driven by clinician concerns also reduced burnout.

This research demonstrates that engaging medical staffs in an effort to empower themselves to alter workplace factors and clinical regimens of care can yield positive results. In sociological terms, this means identifying structural barriers in the work place and activating collective agency to overcome them. Although the impacts of these interventions on patient perceptions and clinical care are still being examined (Linzer, personal communication, April 2, 2015), it is clear that the Healthy Work Place approach has promise. It also suggests that using a sociological lens and developing a teamwork perspective may be an important and underappreciated factor in the study of physician job satisfaction and the improvement of care in real world settings.

Lessons from the CSSG experience

The three projects reported by this team illustrate what may be required to advance efforts in this field. First, the focus of the research is on a question that major segments of the medical profession and medical educators are concerned about—indeed, the question is about their survival as a vital segment of the medical profession. Secondly, a variety of disciplines—physicians, psychometricians, sociologists, epidemiologists, management—guided the research. Finally, active engagement and buy-in of the subjects in the design and implementation of the research was critical to the success of the enterprise.

Cross national comparative study

So where do we go from here? It is not too soon for an international collaborative study linking physicians’ work experiences and attitudes to the structure and culture of their workplaces. Such a project would examine how such factors might affect the outcomes experienced by their patients and perhaps even launch efforts to improve work lives of physicians. This undertaking would require careful planning, involving a variety of stakeholders by convincing them to make substantial long range commitments to the project as sponsors and/or subjects. Such a project would require identifying many stakeholders in different countries, fully engaging them and offering them options to participate in different components of the study phased in over a lengthy time period. To secure buy-in from key stakeholders, the potential benefits of carefully collected data-based cross-national studies should be clearly defined and made obvious to all including not only key physician groups and medical educators, but political figures and health policy makers in different countries. Both the strengths and limitations of proactive and deliberately initiated data collection should be well defined. The interests of various stakeholders in obtaining strategic but objective information about the physician workforce should be articulated. Such a process might serve as an effective counterweight to superficial comparative studies of health policy that have led to facile generalizations (Marmor, Freeman, & Okma, 2005).

The ideal cross-national comparative study would involve enough countries and systems to broadly represent major variations in the design and financing of health systems as well important variations in employment situations of physicians within those systems (e.g., small vs. large settings, self-employment vs. employee status, inpatient vs. outpatient focus, various patient populations, public vs. private sector, etc.). It might be best to initially target generalists to best address the questions raised in this volume. The overall study design would require linking clinic and community level characteristics to physicians working in each setting as well as to patients being served there. Ideally, a panel design would involve clinics as the primary sampling unit rather than individual practitioners. If individual patient-physician linkage is possible, some consistent, credible, readily implementable rules would be required linking a set of patients that is large enough to generalize to each individual physician assumed to be their doctor. Failing this, patients should be linked to a small stable team of practitioners virtually all of whom should be study participants. Ideally the same set of “process indicators” or “tracer conditions” would be used to assess “quality” outcomes across nations. But this might not inevitably be the case, assuming a statistical distribution of patient outcomes could generate enough interval or ordinal scale data that would be sensitive enough to detect “clinically important” differences as benchmarked against national averages or clinical standards. Patient perception and/or satisfaction data would also be useful. Based on the experience with the MEMO study and the HWP in the US, this type of study design is feasible, but challenging even in one country.

Multifaceted measurement of job satisfaction is recommended not only for its scientific validity, but also because it is likely to build and sustain commitment across countries as well as among different types of stakeholders. This is so because their distinctive concerns are likely to be addressed by more nuanced descriptions inherently possible with multifaceted measurements. Existing measures aimed at the clinic staff as well as the physicians should probably be the starting point for instrument development. Much of the instrumentation could include validated survey instruments such as the recent RAND/AMA study (Crosson and Casalino, 2015) but some primary scale development could be done. Physician surveys should use uniform wording for core clinical tasks, patient physician relationships, demographics, educational, and health history information. Yet, to make surveys interpretable by physician participants in different systems, country-specific modules should be developed using locally meaningful terminology to describe financing, payment, and organizational issues physicians deal with routinely and can meaningfully evaluate from their own perspective. However, prior to the conduct of the study itself, investigators should agree about how to set up and describe any statistical comparisons involving these system-related variables so valid comparative conclusions can be drawn, thus minimizing ad hoc assumptions or post hoc comparisons.

Given this broad outline, some strategic decisions would need to be made. Probably not every participating country needs to be engaged in both the clinic/physician and the patient care outcomes study phases; relatively few might opt for involvement in any intervention phase. Obviously the second and third phases are likely to be more time-consuming, difficult and risky for participants and relatively difficult to implement without extensive on the ground engagement of a study staff. Perhaps existing national or regional practice networks (Soler et al., 2008; Soler, Yaman, & Esteva, 2007) could be employed in such an enterprise. Creative data collection strategies such as the overlapping questionnaire design approach so helpful in longitudinal panel studies of Norwegian doctors (Aasland, Olff, Falkum, Schweder, & Ursin, 1997; Rosta & Aasland, 2014) could be applied in cross sectional and longitudinal design.