Intensive Care Unit Rotations and Predictors of Career Choice in Pulmonary/Critical Care Medicine: A Survey of Internal Medicine Residency Directors

Background The United States (US) is experiencing a growing shortage of critical care medicine (CCM) trained physicians. Little is known about the exposures to CCM experienced by internal medicine (IM) residents or factors that may influence their decision to pursue a career in pulmonary/critical care medicine (PCCM). Methods We conducted a survey of US IM residency program directors (PDs) and then used multivariable logistic regression to identify factors that were predictive of residency programs with a higher percentage of graduates pursuing careers in PCCM. Results Of the 249 PDs contacted, 107 (43%) completed our survey. University-sponsored programs more commonly had large ICUs (62.3% versus 42.2%, p=0.05), primary medical ICUs (63.9% versus 41.3%, p=0.03), and closed staffing models (88.5% versus 41.3%, p < 0.001). Residents from university-sponsored programs were more likely to pursue specialty fellowship training (p < 0.001) overall but equally likely to pursue careers in PCCM as those from community-sponsored programs. Factors predictive of residencies with a higher percentage of graduates pursuing training in PCCM included larger ICUs (>20 beds), residents serving as code leaders, and greater proportion of graduates pursuing specialization. Conclusions While numerous differences exist between the ICU rotations at community- and university-sponsored IM residencies, the percentage of graduates specializing in PCCM was similar. Exposure to larger ICUs, serving as code leaders, and higher rates of specialization were predictive of a career choice in PCCM.


Introduction
Since their advent during the polio epidemic of the midtwentieth century, intensive care units (ICUs) and the specialty of critical care medicine (CCM) have become fundamental pillars in the care of critically ill patients [1].Specialists in CCM, also referred to as "intensivists," receive additional procedural and medical training and are ideally positioned to care for this increasingly complex population.
e United States (US) is currently experiencing a worrisome shortage of intensivists, and the inability to meet staffing demands will likely continue to worsen as our population ages [2,3].Innovative ways to increase the supply of CCM trained physicians are now being explored.
Historically, intensivists in the US have largely been internal medicine (IM) residency graduates with subsequent training in pulmonary/critical care medicine (PCCM), along with surgeons and anesthesiologists [4].More recently, trainees from other residencies (specifically emergency medicine and neurology), as well as IM graduates pursuing CCM fellowships without pulmonary training (IM-CCM), have begun entering the field in increasing numbers, with expanded board certification opportunities partially directed at increasing the supply of intensivists [3,5,6].e largest pool of potential intensivist trainees, however, continues to be PCCM [7].
e Accreditation Council for Graduate Medical Education (ACGME) requires all graduates of IM residency programs to experience three to six months of dedicated critical care rotations during the 36 months of training but does not provide further guidance on how to structure such rotations [8].A survey of university-affiliated PCCM fellowship directors about resident CCM education found that there was no standardized approach used by different programs [9].Other studies of CCM education have largely focused on specific changes related to duty-hours or educational interventions [10][11][12].
Even less is known about how differences in CCM rotations influence the choice of IM graduates to pursue further training in PCCM.A survey study from 2002 found that IM residents harbored serious reservations about PCCM as a field, with heavy workloads and perceived stress among faculty and fellows being the most dissuading factors [13].Conversely, residents cited the intellectual challenge, opportunity to manage critically ill patients, application of physiology, and ability to perform procedures as characteristics that increased their interest in the field.Other studies have investigated factors influencing career choice in a variety of specialties, with mentorship, intellectual challenge, and procedural opportunities being common themes [14][15][16][17].
In order to better inform future efforts to reform CCM education and increase interest in this specialty, we conducted a nationwide survey of US IM residency programs.Our primary objectives were to characterize the critical care rotations experienced by IM trainees at both communityand university-sponsored residency programs, as well as to identify aspects of the rotation that are predictive of IM graduates pursuing careers in PCCM.

Study Design and Population.
A cross-sectional survey of IM residency PDs in the United States was conducted from February to March 2017.Participants included all directors of IM residency programs listed in the Electronic Residency Application Service (ERAS) for whom contact information was available online.An initial invitation letter was sent by email to all participants that provided an overview of the study objectives and methods.One week later, a link to the electronic survey was distributed, with biweekly reminders for up to 6 weeks.e protocol for this study was approved by the institutional review board at the Massachusetts General Hospital.

Surveys and Data Collection.
Our survey instrument was modeled after the one developed by Almoosa et al. [9] and was subsequently modified with the author's permission. is revised survey primarily focused on the characteristics of medical ICU (MICU) rotations, CCM educational practices, ICU team structure, and fellowship choices.Prior to distribution, it was tested by multiple faculty members at the Massachusetts General Hospital for readability and appropriateness for the desired study population.e final survey (provided in the Supplementary Materials available here) was administered using Research Electronic Data Capture (REDCap) software.

Data Analysis.
e survey data were categorized according to sponsorship (university versus community) and described using summary statistics (proportions, medians and interquartile ranges (IQR), or means and standard deviations).Categorical variables were compared using Fisher's exact test, while continuous variables were compared using two-sided t-tests and the nonparametric Kruskal-Wallis test.In order to evaluate for predictors of career choice in PCCM, we first classified programs as having a "high" or "low" percentage of the class pursuing training in PCCM (cutoff defined as the median percentage of a class going into PCCM).Adjusted odds ratios and 95% confidence intervals were obtained from a parsimonious multivariable logistic regression model to predict programs with a "high" percentage of the class pursuing training in PCCM and included variables considered to potentially influence the career choice of IM residents.Two-sided p values of ≤0.05 were considered to be significant.All analyses were conducted in STATA version 14 (Statacorp 2015, College Station, TX).
Few respondents reported having more than one attending per ICU team (16.7% versus 17.4%, p � 1.00), while the presence of midlevel providers (50.0%versus 24.4%, p � 0.01) and residents from other specialties (83.3% versus 35.6%, p < 0.001) as part of the ICU rounding team were more common in university-sponsored programs.e base specialties of intensivist attendings are presented in Figure 1.

2
Critical Care Research and Practice

Relationship of Variables to Percentage of Class Matching in PCCM.
e career choices of graduating residents are shown in Figure 2. A higher median percentage of residents graduating from university-sponsored residencies go on to pursue any fellowship training than do those from community-sponsored programs (60 (50-75) versus 40 (30-60), p < 0.001), while the median percentage of the class pursuing fellowship training in PCCM was similar between the two groups (10 (10-15) versus 10 (10-20), p 0.64).
Table 2 shows the comparison of variables by whether the residency program had a "high" (above the median) or "low" (at or below the median) percentage of the graduating class pursuing fellowship training in PCCM.Programs with a high percentage of residents pursuing PCCM more frequently had ICUs with >20 beds (66.7% versus 44.1%, p 0.03) and lower percentages of residents pursuing any fellowship (17.2% versus 20.6%, p < 0.01).

Predictors of Programs with
Higher PCCM Interest.Table 3 shows predictors for a higher/lower percentage of a class pursuing fellowship in PCCM in the nal multivariable logistic regression model.Residents serving as hospital code leaders (OR 3.14 (1.01-9.82),p 0.05), the principle ICU having greater than 20 beds (3.32 (1.21-9.15),p 0.02), and the percentage of a class going into any fellowship (OR 1.07 (1.03-1.11),p < 0.001) predicted a higher percentage of the class pursuing training in PCCM after accounting for other variables.

Discussion
In this study, we surveyed US IM residency PDs, described the characteristics of the ICU rotation at both universityand community-sponsored residencies, and identi ed predictors of residency programs with higher percentages of graduates pursuing careers in PCCM.University-and community-sponsored residencies utilized similar educational methods and did not di er in the percentage of graduates specializing in PCCM.Programs with more graduates specializing in PCCM tended to have larger ICUs and lower rates of specialization.Residents serving as inhospital code leaders, larger ICUs, and percentage of the class pursuing specialty training were identi ed as predictors of higher PCCM participation in our nal multivariable logistic regression model.Little has been published on the di erences between community-and university-sponsored residency programs, and to our knowledge, this represents the rst study to describe CCM education in both settings.While community  Our study also provided interesting insight into the changing staffing patterns of CCM.Both community-and university-sponsored programs reported having non-PCCM intensivists on the ICU teaching service, reflecting the increasing multidisciplinary nature of CCM as a specialty [3].Notably, the presence of non-PCCM attending did not negatively influence residents' decision to specialize in PCCM, in contrast to prior studies that have emphasized the role of PCCM mentors in positively influencing resident perceptions of the specialty [13].
Finally, we have identified that exposure to larger ICUs, the opportunity to serve as in-hospital code leaders, and percentage of a class pursuing fellowship are predictors of resident career choice in PCCM.Exposure to larger ICUs may reflect a greater diversity of patient presentations, thereby increasing the intellectual challenge of the CCM rotation, a factor cited as one of the most attractive features of PCCM by IM graduates [13].It might also reflect getting to experience more complex and innovative technologies (e.g., extracorporeal membrane oxygenation (ECMO)) and offer an opportunity to work with more skilled nurses and providers.Serving as the in-hospital code leader is often viewed as a stressful experience, for which residents may feel underprepared [18,19].Consequently, mastery of this skill may be perceived as a significant accomplishment by trainees and increase their interest in the care of critically ill patients.Some authors have noted a decline in the opportunity for residents to participate in in-hospital "code blues" in the era of quality improvement and resident work hour restrictions, highlighting the importance of designing residency education to maximize these valuable learning experiences [20].Lastly, our regression model identified the percentage of a class that pursues specialization as predictive of higher PCCM interest when accounting for other covariates, but the magnitude of this association was weak (OR 1.07 (1.03-1.11),p < 0.001) (Table 2).
Our study had a few potential limitations.e response rate of 43%, while comparable to similar uncompensated studies of this medical education faculty [9,12], was relatively low and thereby raises the possibility of response bias.While information was unavailable regarding the characteristics of the programs that failed to complete the survey, we found there to be a similar distribution of university-and communitysponsored programs among nonrespondents.Another important limitation of this study was the fact that the percentage of a class pursuing training in PCCM was estimated by PDs and therefore may be inaccurate.Previous studies on career choice, however, report similar percentages of IM residents planning to pursue PCCM fellowship training [21,22].
Additionally, some have questioned whether expanding the number of PCCM fellowship positions is the best approach to address the shortage of intensivists, as PCCM physicians may spend a minority of their time practicing critical care [3,4].IM-CCM programs without a pulmonary component have drawn increased attention, since these practitioners spend a greater percentage of their clinical time in the ICU [4].Unfortunately, our survey did not assess resident interest in this training pathway or exposure to IM-CCM mentors.However, this is unlikely to dramatically alter our conclusions, since IM-CCM training positions are dwarfed by those in PCCM (200 versus 1630, resp., in 2015-2016), and there has been a decline in the number of accredited IM-CCM fellowships in recent years [23].
Despite these limitations, this study adds valuable knowledge to the field of CCM education.Beyond better informing PDs regarding the diversity of ICU rotations nationwide, we identified easily implementable experiences (i.e., serving as hospital code leaders) that may increase resident interest in CCM.Modifications and improvements to residency critical care rotations could help to increase the intensivist supply, addressing an important public health concern.

Table 1 :
Program and rotation characteristics.
ǂ p values are based on two-sided Fisher's exact tests for categorical variables and Kruskal-Wallis tests for continuous variables.gramsreporting the presence of intensivists from di erent specialties on the ICU teaching service by residency sponsorship.In both groups, the most common intensivist specialty was PCCM, with the other base specialties variably represented in universityand community-sponsored programs.erewas not a signi cant di erence in the presence of non-PCCM attendings between the two types of residencies.Critical Care Research and Practice

Table 2 :
Di erent ICU characteristic in programs with low or high % of PCCM applicants.High" and "low" refer to the percentage of the class entering PCCM with respect to the median percentage observed (10%); § p values are based on two-sided Fisher's exact tests for categorical variables and Students t-tests for continuous variables. "