Have you ever taken a quiz to see which Harry Potter character you are or which Marvel villain you would marry? I built something like that for emergency medicine jobs, though with a few key differences.
This nine-question Likert scale questionnaire (1 was “strongly disagree,” 5 was “strongly agree”) asked respondents about themselves and their views on emergency medicine, and attempted to predict their employment structure. (Table 1.) The computer’s goal was to predict the user’s job, and my goal was to use the data to gain a better understanding of employment models in emergency medicine. (http://bit.ly/BelangerJobQuiz.)
The application first guessed if the respondent was an owner or employee, verified the correct answer, and then guessed the employment structure from seven categories: democratic group, hospital employee, contract management group, independent or locums, academic, resident, or government. I let the user pick between two machine-learning algorithms for prediction (k-nearest neighbors or random forest).
I picked questions about the physician’s life and career but not about the employer. The predictive algorithm was also used to provide an incentive for accurate answers, which I hoped would reduce bias in the responses.
You are Hermione!
A total of 573 people participated over about four days in February. Participants were 49.7 percent male and 50.3 percent female; 23 percent said they had administrative duties (director or higher). Male and female respondents had relatively similar answers. The data were further broken down by employment structure and ownership status. (Table 2.)
I limited my analysis to subgroups with more than 10 participants. Each question represented a positive value, so a higher response value indicated a more positive result, and I transformed the data so that responses on the “disagree” side of the scale were negative and on the “agree” side were positive. (Table 3.) Democratic group owners were the most positive respondents in almost every respect.
Employees of contract management groups (CMGs) tended to be the most negative respondents, followed closely by the respondents who identified as owners of CMGs. In fact, CMG owners and employees disagreed with all but two of the positive survey statements, and they were the only subgroups that disagreed in six of these statements. (Table 4.)
Most respondents said emergency medicine was a good career and that they were confident in their worth. They were pessimistic, however, about the future of the specialty. The greatest divergence in the responses between subgroups appeared to be in feeling valued as an individual and in believing one’s voice was important. (Tables 5 and 6.)
These answer profiles were striking. Less than 25 percent of CMG employees said they felt their voices were important or that they were valued as individuals, and more than 50 percent of these physicians believed or strongly believed their voice was unimportant. CMG owners seemed to closely parallel employees, which may be surprising.
A one-way ANOVA test demonstrated significance (p<0.001) in the differences between employment structures in both of these questions, and pair-wise groupings (which are beyond the scope of this article) demonstrated significance in all but a few pairings. Notably, employees and owners of CMGs did not demonstrate a statistically significant pair-wise difference, but employees of CMGs demonstrated significant pair-wise differences between themselves and all other groups.
I also analyzed the data split by administrative role rather than ownership status, which produced more pronounced differences. The graph shows the difference between the average of all administrators and non-administrators for each question. (Table 7.)
Administrators answered more positively than non-administrators, but the difference between the two groups was most pronounced for physicians working for CMGs. Responses from 21 CMG administrators and 66 non-administrators were striking: Answers by administrators were almost opposite those of non-administrators. (Table 8.)
I also broke each question down into the individual responses and compared administrator responses with non-administrator responses within CMGs. (Table 9.) CMG administrators answered much more positively than their charges in every respect. In fact, almost all of the questions in the administrative group were answered positively by about 50 percent of the respondents, but few of the questions in the non-administrative group were responded to positively by more than 25 percent of the respondents.
Participants by Employment Structure
|Employee, hospital employee||167||29.14|
|Owner, democratic group||78||13.61|
|Employee, democratic group||52||9.08|
|Employee, independent or locums||10||1.75|
|Owner, independent or locums||9||1.57|
|Owner, hospital employee||1||0.17|
Room for Improvement
Emergency medicine has rapidly come to an inflection point of sorts. Forces have converged to produce a vastly different world for our specialty. Unfortunately, traditional data sources may lag behind change that moves too quickly, forcing us to act blindly or delay acting. Neither is a satisfying solution, and that demands unconventional data sources.
In acknowledging the limitations of this work, we need to recognize that Likert scales and surveys do have some inherent problems. These results may be biased; respondents may have completed the survey multiple times or may have lied. And the survey is only as generalizable as its respondents are representative of the entire population. The respondents appear to know the difference between a democratic group and a contract management group, but these terms are poorly defined. I suspect framing the survey in the form of a predictive quiz improved the quality of the answers, but this is conjecture.
Limitations aside, a good possibility exists that the employment structure of emergency physicians may have real effects on their lives and careers. It is also clear that a large segment of emergency physicians may be in potentially detrimental employment structures.
Respondents identifying as being owners or employees of contract management groups, nebulous as the term may be, gave significantly more negative responses to the questionnaire and represented a very large portion of the total responses. Trends within this subgroup point to a large disparity between administrators and non-administrators.
Better understanding of these problems should be an area of focus for the leaders of our field, and these should also interest administrators in contract management groups because the data demonstrated real areas for improvement, generating differential value.
The survey dataset is available to all athttp://bit.ly/BelangerJobQuiz. This study was also released in the ACEP Workforce Section Newsletter.
Dr. Belangeris the secretary of the American College of Emergency Physicians Locum Tenens section and an emergency physician in McKinney, TX. Read his past articles athttp://bit.ly/EMN-numbERs.
What EM Needs
Share how you feel about EM and what it needs most right now in Dr. Belanger’s word cloud at http://bit.ly/EMneeds. Results in August!