Should I Stay, or Should I Go? Early Phase Instrument Development of Workforce Movement—A Pilot Study with Breast Radiologists

Abstract

Objective:

The goal of this study was to develop a psychometrically valid survey on workplace satisfaction and examine predictors of workforce movement among breast radiologists.

Methods:

Actively practicing members of the Society of Breast Imaging were invited to complete a survey on workplace satisfaction. Radiologists also indicated whether they had recently left their practice or were thinking of leaving their practice.

Results:

In total, 228 breast radiologists provided valid responses (8.7% response rate); 45% were thinking of leaving or had left their practice. Factor analysis yielded five factors, and discriminant function analysis found six main aspects associated with workforce movement in breast radiologists: (1) not enough work–life balance; (2) salary too low; (3) not feeling valued; (4) wanting a different challenge and/or more growth opportunity; (5) safety concerns; and (6) not feeling respected by physician leadership.

Conclusions:

Pending further validation in larger and different cohorts, the survey created here can be administered by radiology practices to predict when breast radiologists are vulnerable to quitting. Atlhough this measure was designed for breast radiologists specifically, it could be adapted for other subspecialties.

Graphical Abstract

A growing body of work suggests that burnout is of concern for radiologists [1–3]. The World Health Organization defines burnout as an occupational syndrome resulting from chronic workplace stress that has not been successfully managed, characterized by three dimensions: exhaustion, depersonalization, and ineffectiveness [4]. The exhaustion dimension refers to the physical and emotional response to stress from the job, with people feeling overextended by work demands and depleted of emotional and physical resources, e, feeling drained and used up [5]. The depersonalization dimension refers to the interpersonal context of burnout—namely, a negative, callous, or excessively detached response to the job. The ineffectiveness dimension refers to the self-evaluation component of burnout, with associated feelings of incompetence and lack of achievement and productivity at work [5].

In a survey of 367 radiology practice leaders, 77% indicated that burnout was a “significant” or “very significant” problem [6]. Burnout is likely to exacerbate current radiology worker shortages. The Association of American Medical Colleges estimates that a shortage of 54,100 to 139,000 physicians will occur by 2033 [7], and radiology is one vulnerable specialty. In the United States and Ontario, rates of medical imaging increased for adults aged 18–64 years, and for adults aged 65+ years, between 2000 and 2016 [8]. The number of radiology residents joining the workforce has not matched this increase in demand [9]. Therefore, workforce retention is more important than ever; at the same time, radiologist burnout has been associated with intention to leave [10]. According to Merritt Hawkins, a physician search firm, “radiologists were the third most requested search by healthcare employers in 2021” [11]. When radiologists change jobs, this can cause disruption to patient care and to the practice, and it results in other radiologists having to work additional hours, which can exacerbate burnout. Job changes also come at a considerable monetary cost, as the price to replace a physician in the United States is estimated at $500,000 to $1,000,000 [12]. Therefore, a critical need exists for practices to address the related issues of burnout and workforce movement.

For radiology practice leaders to address workforce movement, they need to know what factors cause radiologists to leave their jobs. Although the burnout literature is informative, burnout is a broad construct with the three aforementioned underlying dimensions [5]. Understanding specific predictors of workforce movement can enable employers to intervene on the most relevant factors. The goal of the present study is to develop a psychometrically valid survey on workplace satisfaction, to examine which items are predictive of whether radiologists left or were considering leaving their practice.

METHODS

This study was exempt from ongoing evaluation by the Rhode Island Hospital Institutional Review Board. We piloted the study in breast radiologists, as this is one of the radiology subspecialties currently in greatest demand. Actively practicing members of the Society of Breast Imaging (SBI) received an e-mail newsletter from the SBI inviting them to voluntarily complete the survey. The SBI is the largest organization of practicing breast radiologists in the United States. E-mail invitations were sent between January 31, 2022 and February 7, 2022. The survey was closed on March 15, 2022. Data were collected using research electronic data capture (REDCap) [13]. The survey link used was public, and therefore, responses were anonymous. Survey questions were developed by consensus of the study authors. The survey is available in Appendix 1.

Psychometric Analysis

All analyses were conducted using SAS software, version 9.4 (SAS Inc. Cary, NC). As an early phase of instrument development, exploratory factor analysis (EFA) was used with maximum likelihood estimation and Promax rotation with the FACTOR procedure to evaluate construct validity. Factor solutions were evaluated using a scree plot and a parallel test. Scale and subscale internal consistency was evaluated using Cronbach’s alpha using the CORR procedure. Concurrent validity was examined using discriminant function analysis (DFA). Missing data patterns were assessed using the MI procedure. Statistical analyses include comparisons of individual items (questions) among those who had left their job within the preceding 2 years, those who were thinking of leaving, and those who are not thinking of leaving. These analyses were performed using medians and the interquartile range, along with the omnibus Kruskal–Wallis H test and the Dwass–Steel–Critchlow–Fligner multiple-comparison procedure for pairwise comparisons, with the NPAR1-WAY procedure. In addition, workforce movement was examined by age and gender identification using a χ² test with the FREQ procedure. No adjustments were made for the number of these comparisons. Alpha was set a priori at the .05 level, and all interval estimates were calculated for 95% confidence.

RESULTS

Response and Demographics

Of the 2,612 delivered e-mails, 1,670 opened the e-mail, and 266 opened the survey. No skip patterns (missing data) were observed for any particular item (all 1.1%−3.3%). Of the 1,670 physician members who opened the e-mail, 230 responded, or 14%. Of those, 2 responded that they had retired over 2 years ago, leaving 228, of which 23.2% (53 of 228) responded that they had left a practice within the last 2 years, 21.9% (50 of 228) responded that they were thinking of leaving their current practice, and 54.8% (125 of 228) responded that they were not thinking of leaving their current practice. Overall, 45% of respondents either had thought of leaving or had left their practice within the preceding 2 years.

Of the 228 respondents included in the analysis, 82% (182 of 222) identified as White, 3.6% (8 of 222) as Black or African American, 12.6% (28 of 222) as Asian, 1.8% (4 of 222) as Other, and 6 did not answer; 91% (198 of 217) reported being not Hispanic, Latino, or Spanish, with 77% (173 of 225) identifying as female. Regarding age, 5% (11 of 226) were aged 35 years or younger, 33% (75 of 226) were aged 36 to 45 years, 29% (66 of 226) were aged 46 to 55 years, 25% (57 of 226) were aged 56 to 65 years, 7.5% (17 of 226) were aged 65 years or older, and two did not answer. The majority (73.6%; 167 of 227) reported 76% to 100% of their practice was focused on breast imaging, with 15% (34 of 227) being focused 51%−75% on breast imaging. In total, 69% (155 of 225) reported being full time, 42% (96 of 226) reported being primarily private practice, 24% (54 of 226) reported being academic, and 19.5% (44 of 226) reported being primarily hospital-based. A total of 75% of respondents (171 of 228) reported being fellowship-trained.

Of those who left within the preceding 2 years, 84% (42 of 50) started a new position elsewhere in radiology, 12% (6 of 50) left the workforce altogether, and 4% (2 of 50) were terminated. Of those who started a new position in radiology, 64% (27 of 42) sought out the position, and 36% (15 of 42) had been actively recruited. Of those who left the workforce, one (17%; 1 of 6) left owing to dependent care (children, grandchildren, parent, spouse), one (17%; 1 of 6) left owing to personal health issues, one (17%; 1 of 6) retired (1 year sooner than planned), and three (50%; 3 of 6) left for other reasons.

Survey Results

Of those who were 35 years of age or younger, 36% (4 of 11) were thinking of leaving, 0 (0%) had left, and 64% (7 of 11) were staying. Of those who were 36 to 45 years of age, 17% (13 of 75) were thinking of leaving, 31% (23 of 75) had left, and 52% (39 of 75) were staying. Of those who were 46 to 55 years of age, 20% (13 of 66) were thinking of leaving, 18% (12 of 66) had left, and 62% (41 of 66) were staying. Of those who were 56 to 65 years of age or younger, 23% (13 of 57) were thinking of leaving, 26% (15 of 57) had left, and 51% (29 of 57) were staying. Of those who were 65 years of age or more, 41% (7 of 17) were thinking of leaving, 18% (3 of 17) had left, and 41% (7 of 17) were staying. These differences were not statistically significant, P = .16. Of those identifying as male, 23% (12 of 52) were thinking of leaving, 12% (6 of 52) had left, and 65% (34 of 52) were staying. Of those identifying as female, 22% (38 of 173) were thinking of leaving, 27% (47 of 173) had left, and 51 (88 of 173) were staying. Although over twice as many women left, relative to the number of men (27% versus 12%), this difference only approached statistical significance, P = .06.

In addition, differences among and between the three groups, as counts/percentages (Supplemental Table 1), medians/ interquartile ranges (Supplemental Table 2), and boxplots (Supplemental Figure 1) are provided as supplemental material. As seen in Supplemental Table 2, in general, those thinking of leaving and those who had left agreed more that work–life balance is an issue and that the work schedule is too inflexible, relative to those staying (quality-of life-factors, see next section), P < .05. Those thinking of leaving and those who had left agreed more with four of the five questions pertaining to physician sociocultural factors, relative to those staying, P < .05. Those thinking of leaving and those who had left agreed more with two of the three questions pertaining to nonphysician sociocultural factors, relative to those staying, P < .05. Those thinking of leaving and those who had left agreed more with only one (want a different challenge and/or more growth opportunity) of the four questions pertaining to career factors, relative to those staying, P < .05. Finally, those thinking of leaving and those who had left agreed more with only one (safety) of the 11 questions pertaining to workplace factors, relative to those staying, P < .05.

Psychometric Analysis

Construct validity: factors of workforce movement.

All three groups (those who had left, those thinking of leaving, and those staying) were asked about their attitudes toward the same 31 aspects of their current job or the job they left. Exploratory factor analysis was used to identify overall factors of workplace movement. The factor analysis shows how the 30 questions cluster together, as these clusters of questions have a common theme called a “factor.” How well each question relates to each of the five factors can be interpreted like a Pearson’s correlation coefficient or “r,” with values closer to 1 or −1 denoting a stronger correlation with the factor, and + or − denoting the direction of the correlation. As can be seen in Supplemental Table 3 and Supplemental Figure 2, the 30 questions can be categorized into five latent variables, as follows: quality-of-life factors, physician sociocultural factors, workplace and/or position conditions factors, nonphysician sociocultural factors, and career factors (the cluster of questions is denoted with pink, and the strong correlations are shown in boldface). One question was removed because it did not load with any construct (geography). The five-factor solution was selected relative to other models based on interpretation, the scree plot, and the parallel test (Supplemental Figs. 3 and 4).

Internal consistency.

Cronbach’s alpha for the scale was .89. Cronbach’s alpha was .82 for the quality-of-life category, .82 for the physician sociocultural category, .76 for the “workplace and/or position conditions category, .82 for the nonphysician sociocultural category, and .57 for the career category, all reflecting adequate levels of internal consistency, except for the career category (although this could be due, in part, to the fact that this construct had only four items).

Concurrent validity: responses and group membership.

Discriminant function analysis was used to examine how well the 30 questions discriminated among those who had left, those who were thinking of leaving, and those staying. Briefly, the discriminant function predicts group membership and whether the respondent had left, was planning to leave, or was staying. How well each of the 30 questions relates to each of the two discriminant functions can be interpreted as a Pearson’s correlation coefficient or r, with values closer to 1 or −1 denoting a stronger correlation with the factor, and + or − denoting the direction of the correlation. For example, “not enough work–life balance” correlates positively (agrees) with the first discriminant function at the .63 level, whereas “not enough vacation” correlates negatively (disagrees) with the first discriminant function at the .52 level. As seen in Figure 1, the first discriminant function correlates positively with having “left” and correlates negatively with having “stayed.”

Figure 1.

Illustration of discriminant functions 1 and 2 by radiologists who are staying, have left, or are thinking of leaving. Clearly, function 1 (x-axis) discriminates who “left” (blue) and who is “staying” (red) with those thinking of leaving (“leaving”) being in the middle (green). That is, function 1 shows “agreement” for those who “left” regarding: (1) not enough work–life balance; (2) salary too low; (3) I don’t feel valued; (4) want a different challenge and/or more growth opportunity; (5) safety concerns; (6) I don’t respect physician leadership and “disagreement” for those who are “staying.” Function 2 does not discriminate between these groups.

As seen in Table 1, the first discriminant function’s canonical correlation was .70, and this was significant (P < .0001) and was correlated negatively and positively with certain issues. The second discriminant function’s canonical correlation was .42, which was not significant (P = .22). If focusing on the larger magnitude loadings (| > .29|), then 9 of the 30 items discriminated among the three groups, as follows: (1) not enough work–life balance; (2) too low a salary; (3) not feeling valued; (4) wanting a different challenge and/or more growth opportunity; (5) safety concerns; and (6) not respecting physician leadership all correlated positively with the first discriminant function; by contrast, (7) not enough vacation; (8) job required new skills; and (9) technology and/or PACS and/or electronic medical record (EMR) challenges all correlated negatively with the first discriminant function.

Table 1.

Discriminant functions

Questions/Issues	DF(1)	DF(2)
Salary and/or financial benefits too low	0.31	0.17
Not enough work–life balance	0.63	−0.14
Work schedule too inflexible	−0.10	0.20
Daily clinical workload too high	−0.15	−0.09
Too much on-call and/or weekend work	−0.15	0.02
Don’t feel valued	0.29	0.05
Only want to do breast imaging	−0.16	0.23
Want to do more general radiology	−0.11	−0.08
Want a different challenge and/or more growth opportunity	0.49	0.09
Diversity and/or equity and/or inclusion concerns	−0.13	0.27
Safety concerns	0.51	−0.18
Interpersonal conflicts at work	−0.01	0.05
Too much administrative work	−0.13	−0.12
Not enough nonclinical time for the amount of administrative and/or academic and/or leadership work	−0.17	0.20
Technology and/or PACS and/or EMR challenges	−0.32	0.32
Malpractice issues	0.03	0.39
Not promoted and/or made a partner	0.11	−0.23
Poor quality of technical staff and/or imaging and/or patient care	0.14	−0.18
Change in leadership	0.01	0.23
Job required new skills I didn’t have	−0.32	0.24
Not enough vacation	−0.52	0.18
Practice downsized and/or lost contract	−0.10	−0.26
Practice acquisition and/or consolidation	−0.14	−0.11
Nonphysician leadership doesn’t respect me	−0.19	−0.88
I don’t respect nonphysician leadership	0.14	1.12
Physician leadership doesn’t respect me	−0.03	−0.04
I don’t respect physician leadership	0.45	−0.12
I want a more academic position	−0.05	0.40
I want a more leadership and/or administrative position	−0.22	−0.07
I want a more clinical position	0.05	−0.34
Canonical functions	0.70	0.42
P-value	P < .001	.2189

Values denote correlation (between |1.0 and 0.0|) with the DF. For example, “not enough work–life balance” correlates positively with the first DF (agrees) at the .63 level, and “not enough vacation” correlates negatively with the first DF at the .52 level (disagrees). DF = discriminant function; EMR = electronic medical record.

Visualized on the x-axis in Figure 1, staying (red) correlated negatively with the first discriminant function (mean value −0.77), and such respondents largely disagreed with “not enough work–life balance,” “salary too low,” “I don’t feel valued,” “want a different challenge/more growth opportunity,” “safety concerns,” and “I don’t respect physician leadership,” but they largely agreed with “not enough vacation,” “job required new skills I didn’t have,” and “technology/PACS/EMR challenges.”

Conversely, having left (blue) correlated positively with the first discriminant function (mean value 1.59) and consistent with this correlation, such respondentslargely agreed with “not enough work–life balance,” “salary too low,” “I don’t feel valued,” “want a different challenge/more growth opportunity,” “safety concerns,” and “I don’t respect physician leadership,” but they disagreed with “not enough vacation,” “job required new skills I didn’t have,” and “technology/PACS/EMR challenges.” Finally, those thinking of leaving (green) had a correlation between the two (mean value 0.21).

Responses to the open-ended question are available in Appendix 2.

DISCUSSION

Although validated surveys exist to measure physician burnout, to our knowledge, no such measure can assess radiologists who have left or are seriously considering leaving their practice. To address this need, we developed a 30-item questionnaire across five factors (quality of life, physician sociocultural, workplace conditions, nonphysician sociocultural, and career) relating to radiologist workplace satisfaction. The survey demonstrates internal consistency and concurrent validity. In particular, the main factors associated with workforce movement in breast radiologists are as follows: (1) not enough work–life balance; (2) too low a salary; (3) not feeling valued; (4) wanting a different challenge and/or more growth opportunity; (5) safety concerns; and (6) not feeling respected by physician leadership. Interesting to note is that not enough vacation, not having new job-related skills, and technology and/or PACS and/or EMR challenges were not problematic enough to provoke respondents to leave.

One challenge for practice leaders is that improving work–life balance often results in lower income, and we found these two opposing forces to be common reasons for workforce movement. However, individuals place different values on these factors, and practice leaders could endeavor to address both issues by increasing opportunities that allow radiologists to prioritize either income or work–life balance. Effective leadership is inversely related to burnout [14], and radiology leaders need to engage with physicians and understand what motivates them on an individual level. Such engagement may also address the feeling of not being respected by physician leadership, which was one driver of radiologists’ decision to leave. Consistent with earlier recommendations [15], practice leaders can hold career development conversations with radiologists to reduce burnout and identify new growth opportunities.

An interesting finding is that although work–life balance was the strongest reason given by breast radiologists for leaving or seriously considering leaving a practice (+.63), lack of vacation was not (−.53). We hypothesize that this result means that radiologists generally receive sufficient vacation time and that vacation does not compensate for the daily lack of work–life balance. We did not find that workforce movement among breast radiologists was related to a desire for more or less subspecialization or a change in skill set. Participants may have negotiated the quantity of breast imaging they would be required to review, before accepting a position. Alternatively, practices may be adequately accommodating the skill set or degree of subspecialization desired by breast radiologists once they are hired. Also notable is that although technology issues may cause irritation or stress in radiologists, these are not a driver of workforce movement for breast radiologists (eg, they may perceive that changing practices will not provide a remedy).

Safety concerns causing workforce movement in breast radiologists need more investigation. Results regarding this factor may relate to the COVID-19 pandemic and fears of exposure to patients and other health care workers. Reduced psychological safety has been reported to be associated with burnout [16]. Future studies should evaluate both physical and psychological safety issues further, to elucidate what radiologists’ concerns are and how they might be addressed.

One noteworthy finding is that 45% of respondents indicated that they either had left their practice within the preceding 2 years or were thinking of leaving their practice. Santavicca and colleagues [17] examined workplace changes among more than 25,000 radiologists. Between 2014 and 2018, 41% separated from at least one practice, and the annual rate of separation increased, from 13.8% in 2014 to 2015, to 19.2% in 2017 to 2018.

The disruption caused to practices by radiologists leaving their jobs can be substantial and comes at a high monetary cost. Han and colleagues estimated that the cost of physician turnover attributable to burnout in the United States is several trillion, in 2015 dollars [18]. In a study of more than 5,000 physicians in the United States, those who reported burnout had a 2.43 times higher odds of intending to leave their practice relative to those who were not burned out [19]. An advisable approach, therefore, is for practices to assess workplace satisfaction so they can determine whether their workforce is in jeopardy of a high level of turnover, and if so, to intervene to improve job satisfaction. Indeed, just as some practices are measuring burnout on a yearly basis, consistent with recommendations from the AMA [20], we propose that they also should regularly measure the related yet different construct of workplace satisfaction. The measure developed here provides an important tool to be used toward this goal. Unlike other workplace satisfaction surveys that may be currently used in hospitals or large multispecialty groups, the survey we developed is specific to radiology, with only one item focusing on breast imaging specifically—“I only want to do breast imaging.” This item could easily be reworded for use with other subspecialties or general radiologists.

Limitations and Future Directions

In the present study, we focused on only breast radiologists, which is one specialty at high risk of stress and burnout [21,22]. However, burnout is a concern for radiologists in general [6,23–25]. All or nearly all of the items from our survey are applicable to general radiologists and to radiologists from other specialties. As noted earlier, future research should test or adapt the survey we developed for use among nonbreast radiologists. If validated, the 30-item questionnaire we developed could be widely used by radiology practices. Also, due to the cross-sectional nature of our data, we could examine only concurrent validity. Likewise, these results need to be validated with a larger and independent dataset. The initial sample is small because the population of interest (breast radiologists) is correspondingly small. Additionally, a larger cohort also potentially could better represent the diversity of breast radiology in the United States, thereby strengthening the validity of the survey. Longitudinal research could validate the predictive validity of the questionnaire. The survey was administered during the COVID-19 omicron wave, so how results can be generalized to an era in which pandemic concern is lower is not clear. For instance, work–life balance may be more important for parents facing regular school closures, but less important now that schools have generally opened following the COVID-19 omicron wave. Finally, as survey development is an ongoing process, and given that this was a pilot study, future surveys may benefit from rewording or the addition of items.

TAKE-HOME POINTS.

• We developed and validated a questionnaire on workplace satisfaction among breast radiologists, although the survey may be valid among radiologists in general.

• The main factors associated with workforce movement in breast radiologists appear to be as follows: (1) not enough work–life balance; (2) a salary that is too low; (3) not feeling valued; (4) wanting a different challenge and/or more growth opportunity; (5) safety concerns; and (6) not feeling respected by physician leadership.

• Radiologists’ responses to these six specific items correlated with whether they had recently left, were thinking of leaving, or were not thinking of leaving their current job.

• Once this survey is validated among larger and more diverse cohorts, radiology practices may consider administering it regularly, to examine and address radiologist workforce turnover risk.