The impact of emergency medicine residents on clinical productivity

Methods

The Summa Health System’s Akron City Hospital (ACH) is an academic medical center and Level 1 Trauma Center, stroke center, geriatric center, bariatric center, and STEMI center in Akron, Ohio. ED clinical care is contracted with US Acute Care Solutions (USACS). The Summa Health System sponsors residency training in multiple specialties across multiple institutions. The EM program was started on July 1, 2020 with a new class of eight EM R1s and a single EM R2 who transferred from another program. For the period of July 1, 2020 to June 30, 2021, we queried the USACS billing data for all ED encounters at ACH. We obtained individual physician billing data for every day of the year, as measured by RVUs, for all 43 attending EPs who worked at this facility during the study period.

Data were exported to Microsoft Excel, including the RVU for each attending EP on each day of the year. The clinical work schedule was reviewed to determine shift length in order to calculate RVU/h. The authors analyzed the medical students’ and residents’ schedules and end-of-shift evaluations to determine what learners (if any) were working with each EP on each shift. The data were recorded in five separate columns on the spreadsheet (no learners, medical students, EM R1, EM R2, and off-service resident rotators). When trainee shifts did not mirror attending shifts perfectly, we classified it to be a teaching shift only when the resident or student was present for 50 % or more of the attending EP’s shift.

We calculated the total number of RVUs per clinical hour worked by each attending physician overall, when working without learners and when working with each of the four learner categories. The primary outcome (determined a priori) was the difference in RVU/h for the attending EPs when they worked without learners compared to when they worked a majority of their shift with at least one learner. The secondary outcome (also determined a priori) was determining the influence of the learners of each type on EP RVU/h for the subgrouped shifts in which a learner was present for the majority of the shift. A sample size calculation was performed with an 80 % power to detect a 10 % difference in RVU/h between groups with an alpha set at 0.05, resulting in a required sample size of 787 shifts.

The Microsoft Excel database was imported into SPSSv25.0 software (IBM Corp., Armonk, NY). The generalized linear model (GLM) contained a factor to indicate whether a learner was present for most of the shift in order to determine the associated effect (if any) on the RVU/h metric and a factor for the interaction of the learner presence with provider. This interaction was included in the model to determine whether the presence of a learner for most of the shift was consistently affecting the providers or if there was some enhancement or dilution of the learner effect with certain providers. When the interaction was significant (p<0.05), a boxplot was determined to graphically indicate the per-physician influence of the associated effects of the student learner on the RVU/h metric. For the subgroup of those shifts with a student learner present for most of the shift, another GLM was determined with effects for the number of each learner level (medical student, resident) to determine the influence of these on the RVU/h metric.

The Summa Health System’s Institutional Review Board determined that this project did not meet the definition of human subject research and therefore their approval was not required.

Results

A total of 1761 ED encounters were analyzed (Table 1). There were 571 encounters in which no learners were working with the EP. These had a mean faculty productivity of 12.52 RVU/h. There were 1,190 encounters with a faculty EP paired with either a medical student or non-EM R1 resident learner. In those situations, the mean faculty productivity was 12.95 RVU/h, which was not significantly different than the baseline productivity when the EP worked without a learner (p=0.125). However, there was a significant difference between providers on the effect of having a learner present (p<0.001), indicating that the influence of the presence of learners was not uniform across the providers. Figure 1 displays a boxplot of the mean difference in productivity with learners present for each provider, identifying multiple faculty with both positive and negative effects on productivity when learners were present.

Figure 1:

Boxplot of the difference in RVU/h with learners present per provider.

There were 665 encounters in which the learner was an EM R1. When working with one EM R1, the faculty EP productivity was 15.19 RVU/h. When two EM R1 learners were present, the faculty EP productivity was 15.25 RVU/h. When three EM R1 learners were present, productivity was 24.75 RVU/h. There is a significant difference between the number of EM R1 learners present and productivity (p<0.001) for all three situations and between each group (Table 2).

There were 120 encounters where a single EM R2 was present. In these encounters, EP productivity was 17.96 RVU/h, significantly greater than when other learner types were present (16.84 RVU/h, p=0.001) (Table 3).

Discussion

Our study adds to the body of literature suggesting that the presence of medical students and residents does not negatively affect the productivity of EPs. Our single-center data show no significant difference in RVUs/h when physicians worked with medical students and residents rotating to the ED from other specialties in comparison to when working alone. However, there was a positive association between RVUs/h and working with an EM resident. The amount of increased RVUs/h was positively associated with the number of residents present (more EM R1s meant more RVUs) as well as resident level (a single EM R2 was associated with more productive shifts on average than shifts with one or even two EM R1s).

Whereas the presence of learners was associated with higher productivity for a majority of providers (22/29, 75.9 %), productivity was associated with lower productivity for others (7/29, 24.1 %) with a fairly wide range of outcomes (−1.47 RVU/h – +3.90 RVU/h). This has been observed previously. Utilizing the API in their large retrospective observational study, Robinson et al. [4] evaluated individual physician’s productivity and efficiency and found that EPs with baseline higher API had decreased performance when working with residents, while physicians with lower baseline API saw their productivity increase [4]. Whether other attending physician characteristics (academic rank, number of years in practice, nocturnist vs. primarily day shift, etc.) may contribute to this observed difference is a potentially interesting topic for future study.

This is the first time in the EM literature that resident seniority has been reported to have an effect on attending physician productivity. When Bhat et al. [10] analyzed the impact of learners on the number of PPH seen by EPs in which attending physicians were paired with only one learner at a time, they found that EPs saw more PPH when working with a resident than when working alone, but they found no difference if EPs were working with a medical student and no difference when the specialty or the seniority of the resident was taken into account [10].

Our study results were consistent with previous publications on the effect of acting interns (EM-bound fourth-year medical students) on clinical productivity. Hiller et al. [11] found that gross, procedural, and critical-care RVUs billed consistently did not differ significantly when attending physicians worked alone vs. with a fourth-year acting intern.

Our results are also consistent with those of Clinkscales et al. [9] In their retrospective observational analysis of attending physician productivity before and after the establishment of a new EM residency program, they observed a 70 % increase in attending physician productivity after the implementation of a new EM residency program that accompanied a decrease in physician and advanced practice provider staffing of over 10,000 h [9].

Although teaching physicians working with trainees on shift may spend less time at the bedside and more time teaching and supervising, learners may save them time by performing time-consuming tasks such as obtaining and documenting in-depth histories and physical examinations, talking to consultants and families, counseling patients, and documenting a complete record of the encounter. Such tasks make learners valuable team members. Given the variability of results in multiple single-center studies, continued large-scale investigations will be needed to establish a reproducible measure of how trainees effect the productivity and efficiency of attending physicians in the ED.

Limitations

This study has some important limitations. It was a retrospective nonrandomized observational study, as randomization is logistically difficult with scheduling boundaries and departmental flow. This type of study design can only measure associations between productivity with shift conditions and may be influenced by potential selection bias. For example, in a teaching environment, higher acuity presentations that would bill at a higher rate are more likely to have trainees involved, because a teaching attending would be more likely to make sure a trainee benefits from such a learning opportunity rather than manage the case on their own. It was also conducted in a single community ED, and therefore requires replication to ensure external validity. Although the newly established residency has the benefit of allowing EPs to serve as their own controls by evaluating their RVUs with and without learners, and with different types and number of learners in each shift, the small sample size of eight EM1 residents and one EM2 resident may have influenced our results. Although our study model limited the impact of many of the confounding variables, it is possible that improved productivity while working with EM residents was in part the result of residents being scheduled during times when ED volume was highest in order to maximize both their effect on department flow and their learning opportunities.

Conclusions

The presence of EM residents was positively associated with the clinical productivity of EM faculty as measured by RVU/h. There was also a positive association between productivity and the number of EM residents present as well as their training level. Non-EM residents and medical students had no effect on EP productivity.