The target population was all personnel who could work or interact with the anesthesiologist. This included medical staff (surgical or medical specialties (senior physicians or resident), midwives, paramedical staff [Certified Registered Nurse Anesthetist (CRNA), Operating Room Nurse (ORN), State Registered Nurse (SRN) in surgery service, in the Post-Anesthesia Care Unit (PACU) and, in Intensive Care Unit (ICU), Nursing Assistant (NA)] and the administrative and managerial staff. Staff not belonging to the above-mentioned professional categories were excluded.

As usually recommended, we followed a three steps procedure for the elaboration of the survey (12, 13): (1) investigation phase, performed using the existing literature and by analyzing the semi-directive interviews; (2) creation of the survey; (3) validation of the survey. Each one of these steps is crucial to obtain good psychometric qualities (sensitivity, fidelity, and validity). The overall methodology used to produce the survey was validated by a University professor in psychological and social sciences. The design of the study is summarized in Figure 1.

After a bibliographical research phase, semi-directive interviews were conducted with anesthesiologists from the different hospitals (interview duration = 45 min). As the data collection was qualitative and the population concerned only anesthesiologists, a limited number of semi-directive interviews (n = 10) was sufficient (13, 14). This first phase provided precise and objective information on the perception of the anesthesiologist's profession and the supposed perception of other professionals. The semi-directive interviews included the general research question and the list of themes addressed (Supplementary File 1). The conduct of these semi-directive interviews was validated by a professor of communication.

The analysis and synthesis of the answers from the interviews helped us to list the real or supposed positive and negative representations of the anesthesia profession in as exhaustive manner as possible (Supplementary File 2). We were thus able to design the final survey, which includes 38 items divided into six main themes). The first section aimed to characterize the profile of the respondent: age, gender, profession, number of years of practice, and work location. The second section consisted of 38 items aimed at assessing the importance given by the respondents to the various positive or negative, real or supposed, representations of the anesthesiologist profession. The six main themes were: the image communicated, the skills and knowledge, the practice of anesthesia, the place in patient care, the workload, and the initial formation of the anesthesiologist. Each item was scored using a 5 levels Likert scale ranging from “completely agree” (+2) to “completely disagree” (−2) (Supplementary File 3).

We asked a group of anesthesiologists working in our institution (n = 10) to respond to the survey, indicating which items could be improved and if the survey tool was relevant and coherent. This final step of validation of our survey resulted in a minor reformulation of three items (items 29, 30, 31).

The anonymized survey was distributed via a computerized version produced by Google Forms software. The weblink to the survey was distributed to all medical secretariats, paramedical supervisors, and administrative staff. According to French law, individual participant consent is not required for this study (15).

As this was an original study, we did not have the data to calculate the sample size. Arbitrarily, we expected to collect a minimum of 600 responses. With an estimated average response rate of 30% according to the literature on online studies, we, therefore, sent the survey to 2,000 persons (16). The use of the Likert scale comprising five levels of approval associated with a numerical value enabled us to obtain for each representation a mean evaluation (Me) of the entire interviewed population, allowing a hierarchization of its representations but also the comparison of the same representation in different groups of healthcare professionals. The calculation of the mean was acceptable because we were looking for general trends and we had a large sample (17–19). We then compared the different data collected between 4 groups for all survey items: a group including all physicians (group “Med”), a group including all paramedical staff (group “PM”), a group including midwives (group “MW”) and, a group including all administrative staff (group “Adm”). For Me analysis, the Gaussian distribution of data was assessed using the Kolmogorov–Smirnov test. We compared the data using a non-parametric Kruskal-Wallis test followed by Dunn's multiple comparison test to highlight significant differences (P < 0.05) between the four groups. For continuous measurements, data are presented as mean ± standard deviation (SD) for normal distribution and, median [IQR: 25th and 75th percentiles] for non-normal distribution. For qualitative parameters, data are presented as number of case (n), percentage (%). When a significant difference between the Med group and one of the other groups (PM, MW, or Adm group) was found in univariate analysis (P < 0.05), we then performed a multivariate analysis using a generalized linear model (GLM) to search for associated factors (age, sex, practice location, specialty, and the number of years of practice) positively or negatively influencing the change in mean evaluation for the physician group (Med group). The Me can be considered as a reliable indicator, indeed we analyzed the score data with a univariate test and a multivariate test using a generalized linear model. To test the validity of the GLM, three assumptions were checked on the residues: (1) no outliers: the minimum and maximum values of the standardized residual are within [– 3, + 3] values; (2) the data points must be independent using the Durbin–Watson test; and (3) the distribution of the standardized residuals should be normal, with mean = 0 and a constant variance not different from 1, and graphically by means of a histogram, scatterplot, Q–Q plot and normal probability plot (scatterplot of standardized residual vs. standardized predicted value or Q–Q plot), and finally by a scatterplot of observed vs. predicted value. Multicollinearity was also checked using collinearity statistics (variance inflation factor of −1) (20–23). Throughout the study, a value of P < 0.05 was considered significant. Statistics were performed using GraphPad Prism 8.0 and R version 3.1.4 software.

In univariate analysis, we observed a significantly higher Me in the PM group compared to the Med group (PM, +0.76 ± 0.82 vs. Med, +0.30 ±1.06, P < 0.0001) (Figure 3A). On Multivariate analysis, we showed that the gynecologist-obstetrician group is associated with a significant increase of the Me (gynecologist-obstetrician: +0.83 ±0.27, P = 0.003). A higher number of years of practice and practicing in public hospital are significantly associated with an increase of the Me (number of years of practice: +0.01 ± 0.003, P < 0.0001 and public hospital: +0.35 ± 0.12, P = 0.0063). Practicing in a University hospital significantly decreases the Me (University hospital: −0.25 ± 0.12, P = 0.039) (Table 3). Data by different categories of responders are available in Supplementary Figure 1.

In univariate analysis, we observed a significantly higher Me in the PM and Adm groups compared to the Med group, (respectively PM, −0.08 ± 1.06 vs. Med, −0.55 ± 1.60, P < 0.0001 and Adm, 0.00 ± 1.25 vs. Med, −0.5 ± 1.16, P < 0.01). Me is significantly higher in the PM and Adm groups compared to the MW groups (PM, −0.08 ±1.06 vs. MW, −0.63 ± 1.11, P < 0.05 and Adm, 0.00 ± 1.25 vs. MW, –.63 ± 1.11, P < 0.05) (Figure 3B). Multivariate analysis showed that visceral surgeons contribute to decrease the Me (visceral surgeons: −1.13 ±0.57, P < 0.049) while orthopedic surgeons contribute to increase the Me (orthopedic surgeons: +0.85 ± 0.28, P < 0.003) (Table 3). Data by different categories of responders are available in Supplementary Figure 1.

In the univariate analysis, we observed a significantly higher Me in the PM, MW, and Adm groups compared to the Med group (PM, +0.56 ± 1.04 vs. Med,−0.15 ± 0.99, P <0.0001; MW, +1.16 ± 0.98 vs. Med,−0.15 ± 0.99, p < 0.0001; Adm, +0.65 ± 1.06 vs. Med, −0.15 ±0.99, P < 0.0001). We also found a significantly higher Me in the MW group compared to the PM group (MW, +1.16 ± 0.98 vs. PM, +0.56 ± 1.04, P < 0.01) (Figure 3C). In multivariate analysis, we observed a dichotomy between the different specialties in favor of a decrease of Me for surgeons (surgeons: −0.40 ± 0.10, P < 0.0001) and an increase of Me for medical specialties (medical specialties: +0.31 ± 0.10, P < 0.01). There was a significant increase of Me in the general public hospitals' subgroup whereas the University hospital subgroup tends to significantly decrease Me (public hospital: +0.42 ± 0.12, P < 0.0001 and University hospital: −0.43 ± 0.11, P < 0.001) (Table 3). Data by different categories of responders are available in Supplementary Figure 1.

Univariate analysis found a significantly higher Me in the PM group compared to the Med group (PM, −1.00 ± 0.89 vs. Med, −1.37 ± 0.81, P < 0.0001) and a higher Me in the Adm group compare to the Med group (Adm, −1.04 ±0.93 vs. Med, −1.37 ±0.81, P < 0.05) (Figure 3D). In multivariate analysis, the orthopedic surgeon was associated with an increase of Me (orthopedic surgeons: +2.36 ± 0.80, P < 0.01) (Table 3). Data by different categories of responders are available in Supplementary Figure 1.

Univariate analysis showed a significant increase of the Me in the PM, MW, and Adm groups compared to the Med group (PM, +1.21 ± 0.85 vs. Med, +0.13 ± 1.24, P < 0.0001; MW, +1.00 ± 0.91 vs. Med, +0.13 ± 1.24, P < 0.0001 and Adm, +1.33 ± 0.77 vs. Med, +0.13 ±1.24, P < 0.0001) (Figure 3E).

Multivariate analysis showed that the surgeons' response is associated with a significant decrease of the Me (surgeon: −0.70 ± 0.14, P < 0.0001) with a significant decrease for visceral surgeons (visceral surgeon: −1.08 ± 0.35, P = 0.0026), for urological surgeons (urological surgeon: −1.50 ± 0.50, P < 0.001) and for ear nose and throat (ENT) surgeons (ENT surgeons: −0.79 ± 0.37, P < 0.05). The response of medical specialties significantly increases the Me (medical specialties: +0.60 ± 0.14, P < 0.0001) as well as working in a public hospital (public hospital: +1.1 ± 0.1, P < 0.0001). The fact of working in a University hospital is associated with a significant decrease of the Me (University hospital: −1.17 ± 0.13, P < 0.0001) (Table 3). Data by different categories of responders are available in Supplementary Figure 1.

We observed that the anesthesiologist conveys a positive image: he/she inspires calmness and serenity, has a high degree of responsibility, benefits from significant decision-making power (especially from the point of view of the paramedical staff and in particular for the CRNA) and he/she is responsible for the patient's safety and comfort. These last two elements have already been shown to be shared by the general public (24). Other points remain unclear in the eyes of those interviewed, notably the “service provider aspect” that can potentially result from the practice of anesthesiology, characterized by the punctual nature of its acts and by the possible interchangeability of the anesthesiologists between them. This vision of the “service provider” in the study by Peyrache et al. appeared to be the second element that may make medical students hesitate in their choice of anesthesiology at the end of the course (the activity with a high degree of responsibility being the first) (25). In our study, it seems that the paramedical staff most share this vision of provider. This can be explained in part by the point of view of the ORN. They only participate in intraoperative anesthetic management and are often not involved in the pre-and post-operative period or the intensive care activity. In our multivariate analysis, visceral surgeons, unlike orthopedic surgeons, were more opposed to this reductive image of the profession. This discrepancy can probably be explained by a better perception of the challenges of anesthetic management on the part of visceral surgeons. Indeed, visceral surgeons manage a greater proportion of patients with multiple comorbidities and major surgeries, often requiring postoperative monitoring in the intensive care unit.

The anesthesiologist seems to be more available to paramedical staff than to medical staff. The paramedical staff is often the main interlocutors of the anesthesiologist, especially in the wards, during pre-anesthetic or post-operative visits both in the surgery department or intensive care unit. Although the anesthesiologist has multiple knowledge and skills, he/she is not qualified as a “general practitioner in the surgical sector” by all health professionals (8). Several studies have highlighted the patient's lack of knowledge of the anesthesiology profession (26–28). Our study shows that it is the doctors, in front of the paramedical staff, who are the most aware of it, and in particular the surgeons (probably because they make the observation directly with the patient), unlike the medical specialties.

The diversity of the exercise of anesthesia and intensive care appears to be well-recognized. The anesthesiologist is identified as a doctor with a wide transversal scientific knowledge, able of treating all types of emergencies, and he/she is well-recognized for his dual competence in anesthesia and intensive care. However, it appears to be the paramedics who are most aware of this, and in particular the CNAR. This perception is surprisingly worse for the management of analgesia (Me = 0.94), even though this is a major challenge in anesthesia management. The paramedical staff again seem to be more aware of this than the doctors (except for obstetrician-gynecologists). Indeed, paramedics often directly request the anesthesiologist for the management of postoperative analgesia in surgical departments for example. It is the midwives who have the best perception of this competence and it is reasonable to think that this is partly linked to the specific role of the anesthesiologist in obstetric analgesia, in particular the performance of epidural anesthesia (29).

Anesthesia and intensive care are identified as stressful activities, with exhausting on-call work, exposing to the risk of burnout (30). This last problem is identified within the profession itself and is the subject of much attention (31). Indeed, anesthesia is certainly one of the most stressful specialties. The anesthesiologist is daily exposed to stressful situations such as the management of life-threatening scenarios or end life decisions. In addition, this specialty includes night shifts as well as festivities and weekends. It is therefore not surprising to find a high incidence of burnout in this specialty. The potentially conflicting relationships with surgeons and interventional physicians are also recognized by all health professionals. This almost “historical” conflictual relationship of conflict has even been the subject of recommendations issued by the high authority for health in 2015 in France (32, 33). This point of view is mainly expressed by physicians, especially surgeons, and not by paramedical or even administrative staff. This can be explained by a lack of knowledge of the activity of the anesthesiologist outside the operating room (post-operative visits, management of PACU and intensive care units) (8).

This survey mainly contained closed questions. This limitation of answers allowed us to achieve certain conciseness, but also to perform a quantitative analysis of the data. However, we were exposed to a loss of information and nuance compared to what could be obtained with open answers or interviews. Moreover, the order of the questions was arbitrarily defined and grouped into sub-groups, which exposes us to the phenomenon of anchoring (the answer to one question can influence the answer to another, the respondent wants to keep coherence even unconsciously), a bias that is difficult to avoid and which is difficult to evaluate in our study. In addition, the way some questions are formulated may induce a confirmation bias. In this case, the question could positively or negatively confirm a preconceived idea. Our population also includes an important part of young physicians and paramedics working mainly in public hospitals. Our population is, therefore, not wholly representative of a substantial proportion of medical and paramedical professionals. Finally, our study reflects the French vision of the anesthesiologist with its history, its current difficulties (such as the shortage), and its specificities (such as the dual competence in anesthesia and intensive care). This vision is probably different in other countries even if common points probably exist.