Physician associate/assistant (PA) staffing in US EDs is expanding. According to some estimates, up to 10% of the PA workforce is employed in the ED.1,2 The National Hospital Ambulatory Medical Care Survey of 2016 revealed that one in six ED visits involve a PA, compared with one in 12 in 2008.3-5 PAs may be part of the solution for throughput delays; with skills and scope of practice well suited for the ED, the trend of increased PA staffing in EDs is anticipated to continue.4 Despite recent increases in the number of physician emergency medicine residency programs, a shortage of board-certified emergency physicians persists in the United States.6 Efficient throughput, defined as the flow of patients from triage to disposition, requires sufficient clinicians to treat the large volumes of patients presenting to the ED.7 Inadequate staff can lead to overcrowding, which leads to prolonged length of stay (LOS). Prolonged LOS is associated with delays in treatment, adverse outcomes, and reduced patient satisfaction; extended wait times also can have negative health consequences for patients who choose to leave without treatment.8-10PA staffing may mitigate this problem by increasing the number of patients who can be seen in a given period.

Studies in general EDs in the United States suggest that PAs have lower productivity, measured in relative value units per hour and patients per hour, than emergency physicians.4,11 Conversely, a recent study in Canada demonstrated a statistically significant difference in ED LOS that favored treatment by a single PA, compared with treatment by family physicians.12 Limited data compare PA with pediatrician efficiency in treating low-acuity children. We performed this study to compare the efficiency of PAs with pediatricians in treating children with low-acuity complaints in the ED. We considered LOS as the primary metric of efficient care because of the broad implications of prolonged LOS: increased LOS for one group of patients in the ED directly worsens the experience for that group of patients and in turn leads to prolonged LOS for other patients in the ED.7

METHODS Study design and setting

We performed a retrospective paired analysis to compare ED patient LOS between concurrent PA and pediatrician shifts. The setting was a large, urban, tertiary, academic pediatric ED and level 1 pediatric trauma center with about 90,000 annual visits. The pediatric ED is a two-track system with a high-acuity area staffed by fellowship-trained pediatric emergency medicine faculty, fellows, and residents. The study population was low-acuity patients, defined as Emergency Severity Index (ESI) triage level 4 or 5. These patients predominantly are treated in a separate, Fast Track unit staffed by pediatricians as well as NPs and PAs who operate independently with indirect supervision available by phone, if needed. At our institution, PAs are required to have at least 1 year of clinical experience in pediatrics before they are able to treat patients independently and work in the Fast Track unit of our ED. Before treating low-acuity patients independently, PAs must present every case to a supervising clinician. Occasionally, pediatric emergency medicine fellows and attendings work extra shifts in the low-acuity area of our ED for supplemental income. Low-acuity patients are seen in order of arrival by the next available clinician.

The Children's National Hospital institutional review board made a determination that this project was exempt from ongoing review, because it was secondary research using existing electronic health record data.

Selection of shifts

Between July 1, 2017, and February 29, 2020, we identified all paired shifts during which the low-acuity area in our ED was staffed with a single PA working concurrently with a single pediatrician. We did not analyze shifts with two pediatricians, which occur frequently because there are many days when PAs do not staff the low-acuity area in our ED. Shifts with potential PA-pediatrician overlap typically start in the morning (7 a.m. or 8 a.m.) or afternoon (3 p.m., 4 p.m., or 5 p.m.) and last for 8 or 9 hours. Overnight shifts do not regularly have dual staffing. To ensure overlap throughout the entire shift, we defined morning shifts as 8 a.m. to 2 p.m. and evening shifts as 5 p.m. to 11 p.m. We chose paired shifts to ensure both groups were exposed to an equivalent population of low-acuity patients.

Data collection and measurements

All clinician, timestamp, and patient visit data were extracted from the electronic health record (EHR) and the ED tracking system. We obtained data retrospectively for all visits during paired shifts and excluded patients who were admitted to the hospital or who left without treatment. We then excluded outlier patients whose LOS was greater than the 99th percentile, in accordance with our routine quality improvement data cleaning procedures, to address implausible data caused by computer entry errors. LOS for each visit was defined as the time from patient check-in with registration, which occurs on arrival, to patient discharge from the ED. We calculated the mean LOS for each PA shift and each pediatrician shift from the average for every patient whose clinician evaluation began between 8 a.m. and 2 p.m. (for overlapping morning shifts), or between 5 p.m. and 11 p.m. (for evening shifts).

Outcomes and Analysis

Our primary outcome was the difference in mean shift LOS between PA and pediatrician shifts. For each paired shift, we subtracted mean LOS for the pediatrician shift from mean LOS for the PA shift to determine the difference in LOS for each pair. These data were assessed to determine if they were normally distributed before applying a paired t-test to evaluate for a statistically significant difference between PA and pediatrician LOS.

We used a threshold difference of +/- 15 minutes for our power calculation. A priori, we chose 15 minutes, roughly 10% of the mean LOS for low-acuity patients in our ED (154.6 minutes, SD: 71.1), as a clinically important difference, based on guidance from staff who work regularly with low-acuity patients, discussions with parents, and the director of the low-acuity Fast Track unit in our ED.

To describe the variability between PAs, we reported the mean LOS for patients seen by clinicians performing at the 25th percentile, median, and 75th percentile. We also reported median and interquartile range of the mean LOS for patients seen by clinicians in our pediatrician group.

Sensitivity analysis

To ensure our primary findings were not related to subtle differences in the population of patients seen by PAs and pediatricians in our Fast Track, we repeated our primary analysis, including only patients who were diagnosed with the most common categories of diagnoses in our sample. Shifts were only included if both the PA and pediatrician diagnosed at least one patient with the relevant category of diagnoses. The mean shift LOS was calculated from the ED LOS of patients that fell within that category.

RESULTS Characteristics of shifts and patients

We identified 267 shifts during which the low-acuity area in our ED was staffed with a single PA working concurrently with a single pediatrician. Each shift was shared between one of 15 PAs and one of 22 pediatricians. A total 5,585 patients were evaluated during these shifts. After excluding patients who were hospitalized, had left without being seen, or had LOS greater than 7 hours (greater than the 99th percentile for LOS), we included the individual ED LOS for 5,460 patients in the calculation of mean shift LOS for the 267 shifts. PAs evaluated 2,597 patients (1.62 per hour) and pediatricians evaluated 2,863 patients (1.79 per hour). The most common diagnoses for patients were related to fever and upper respiratory infections. Gastrointestinal problems, injuries, and dermatologic problems were the second, third, and fourth most common groups of diagnoses (Table 1).

TABLE 1. - Descriptive characteristics of patients seen by PAs and pediatricians PA N (%) Pediatrician N (%) Total∗ 2,597 (100) 2,863 (100) Primary diagnosis: Upper respiratory problem 815 (31.4) 1,103 (38.5) Lower respiratory problem 57 (2.2) 74 (2.6) Ear problem 184 (7.1) 243 (8.5) Eye problem 72 (2.8) 79 (2.8) Mouth problem 49 (1.9) 44 (1.5) Gastrointestinal problem 395 (15.2) 374 (13.1) Genitourinary problem 68 (2.6) 56 (2) Dermatologic problem 264 (10.2) 298 (10.4) Injury 395 (15.2) 301 (10.5) Other 298 (11.5) 291 (10.2) Acuity level (ESI)∗∗: 2-3 (emergent, urgent) 14 (0.5) 4 (0.1) 4 (less urgent) 1,962 (75.5) 2,121 (74.1) 5 (nonurgent) 621 (23.9) 737 (25.7) Female 1,212 (46.7) 1,399 (48.9) Laboratory testing performed 471 (18.1) 503 (17.6) Radiograph obtained 353 (13.6) 322 (11.2) Language: English 1,704 (65.6) 1,860 (65) Spanish 669 (25.8) 748 (26.1) Other/not listed 224 (8.6) 255 (8.9) Antibiotics 334 (12.9) 466 (16.3) Return visit within 7 days 113 (4.4) 126 (4.4) LOS <90 minutes 441 (17) 559 (19.5)

∗Median patient age (4 years, IQR: 4 to 5 years) was the same for PAs and pediatricians.

∗∗Some patients started as fast track but subsequently had their ESI increased after initial evaluation. No patients were upgraded to ESI 1, the highest acuity level.

Main results

Mean shift LOS for children seen by PAs was 160.1 minutes (SD: 48.6) and mean shift LOS for children seen by pediatricians was 150 minutes (SD: 47.3). Mean shift LOS for children seen by PAs was 10.1 minutes longer (95% CI: 6.1, 14.1) than mean shift LOS for children seen by pediatricians. The 95% CI did not cross our a priori definition of a clinically important difference.

Assuming PAs and pediatricians were both required to work 32 clinical hours each week and expected to see 1.79 patients per hour, a single clinician might see 57 patients over the course of a week. A 10.1-minute difference in LOS for each patient could contribute an additional 9.5 patient-hours to the ED census over the course of each week.

The mean LOS for all patients treated by the median-performing PA was 166 minutes (IQR: 162.7-203.6). The mean LOS for all patients treated by the median-performing pediatrician was 153.8 minutes (IQR: 135-170.3). Patients treated by the most efficient PA and the most efficient pediatrician had a mean LOS of 139.1 and 125.2 minutes, respectively.

Sensitivity analysis

Both the PA and pediatrician saw at least one patient with an upper respiratory problem on 248 of 267 shifts. Mean shift LOS was 7.9 minutes longer (95% CI: 1.1, 14.7) for PAs. Again, the 95% CI did not cross our a priori definition of clinically important difference.

DISCUSSION

In this paired analysis of PAs and pediatricians, we found a statistically significant, but not clinically important, difference between LOS for patients seen by PAs and pediatricians. This finding was robust to sensitivity testing with a repeat analysis including only patients with upper respiratory problems.

Of note, we used a tighter threshold for considering possible clinically relevant differences in ED LOS than has been used in prior literature. When considering the effect of system changes on LOS for ED patients of every acuity level, +/- 30 minutes may be more appropriate as a threshold for clinically important changes to LOS.13,14 In conversations with our study team and clinicians who regularly work in our fast-track area, we chose a tighter clinically important difference threshold of 15 minutes because efficient care is especially important for low-acuity patients, given the relatively uncommon need for hospital admission, specialty consultation, or extensive diagnostic workups. Our group has previously used 15 minutes, about 10% of our mean fast-track LOS, as a clinically important threshold to evaluate the effect of system changes on LOS for urgent care patients.15 For healthcare administration stakeholders, especially ED leadership involved in hiring, even a 10-minute difference might be important, because such a difference contributes nearly 10 patient-hours to the ED census over the course of each week.

One possible explanation for the 10-minute difference is individual clinician-level action. Paired analysis can control for scheduling confounders but cannot systematically control individual clinician-level action. A 10-minute difference in LOS may not necessarily reflect on the category of clinician, but instead may reflect characteristics across clinicians, such as years of experience or attention to careful history-taking. Separately evaluating the variability across PAs and pediatricians, the difference between first and third quartile clinicians was greater than 30 minutes. The variability between clinicians of the same category was greater than the difference in LOS between PAs and pediatricians generally. In fact, the most efficient PA in our study had a mean LOS, across all the patients they treated, shorter than the median-performing pediatrician. This finding supports the results of a recent Canadian study that showed staffing an ED with a single PA, compared with control days with an assortment of physicians, can improve ED efficiency across a variety of throughput metrics.12 Efforts to reduce practice variability across all clinicians treating low-acuity patients in a pediatric ED may have substantial benefits for patient LOS.15 Given the variability in LOS across PAs who treat low-acuity patients in our pediatric ED, efforts to reduce clinician variability might be especially important for this group. Finally, a 10-minute difference in the amount of time PAs and pediatricians take to manage low-acuity care visits in a large, academic, urban ED may be overshadowed by other advantages PAs bring to workforce staffing.

To our knowledge, this is the first paired clinician analysis comparing LOS for care rendered by PAs relative to physicians focused specifically on the low-acuity pediatric ED population. Our findings extend previous studies forecasting the role of PAs in pediatric EDs and extend the demonstrated overall positive throughput effects of PAs on ED performance into the pediatric ED realm.4,5,12,16 Given the large proportion of low-acuity patients in pediatric EDs and the expanding use of PAs in the ED, this is relevant from a healthcare administration perspective.17

LIMITATIONS

The most important limitation of our study is the risk of selection bias. In our experience, PAs and pediatricians work collegially in the low-acuity area of our ED and reliably select their next patient based only on how long that patient has been waiting to be seen. In the busy clinical environment, patient selection is not strictly controlled. It is possible that PAs, for example, might be more likely to select patients with specific chief complaints, ages, language requirements, or other patient characteristics that could be associated with increased LOS. Alternatively, pediatricians might be motivated to preferentially select patients with simple complaints who can quickly be evaluated and discharged. However, the overall distribution of age, language, sex, and ESI level were similar between patients seen by PAs and pediatricians (Table 1). Although PAs see a higher percentage of patients with injury in our Fast Track, our sensitivity analysis, including only patients with upper respiratory problems in the calculation of shift LOS, makes clear that the small difference in efficiency between PAs and pediatricians is not driven exclusively by selection of patients with specific chief complaints.

We sought to determine if patients seen by PAs or pediatricians could expect a significant difference in LOS, an important factor in the experience of patients and families in adult and pediatric EDs.18,19 Other administrative and throughput factors merit further study. Pediatricians saw a greater number of patients than PAs across shifts included in our study, with an equal number of shift hours. Further studies could evaluate other differences between PAs and pediatricians, including patient satisfaction or a formal cost-benefit analysis evaluating productivity, use of diagnostic testing, scheduling availability, and salary and benefits. Other opportunities for future work might include study of educational efforts to decrease variability in LOS between clinicians.

CONCLUSIONS

In a large, academic, urban ED, no clinically important difference in LOS existed for low-acuity pediatric patients treated by PAs compared with those treated by pediatricians in the Fast Track unit. This study suggests that PAs can be effectively integrated in a pediatric ED low-acuity staffing model without a negative effect on ED patient LOS.

REFERENCES 1. Ginde AA, Espinola JA, Sullivan AF, et al. Use of midlevel providers in US EDs, 1993 to 2005: implications for the workforce. Am J Emerg Med. 2010;28(1):90–94. 2. Pavlik D, Sacchetti A, Seymour A, Blass B. Physician assistant management of pediatric patients in a general community emergency department: a real-world analysis. Pediatr Emerg Care. 2017;33(1):26–30. 3. National Hospital Ambulatory Medical Care Survey: 2016 emergency department summary tables. www.cdc.gov/nchs/data/nhamcs/web_tables/2016_ed_web_tables.pdf. Accessed January 20, 2023. 4. Pines JM, Zocchi MS, Ritsema T, et al. The impact of advanced practice provider staffing on emergency department care: productivity, flow, safety, and experience. Acad Emerg Med. 2020;27(11):1089–1099. 5. Doan Q, Hall W, Shechter S, et al. Forecasting the effect of physician assistants in a pediatric ED. JAAPA. 2014;27(8):35–41. 6. Reiter M, Wen LS, Allen BW. The emergency medicine workforce: profile and projections. J Emerg Med. 2016;50(4):690–693. 7. Kenny JF, Chang BC, Hemmert KC. Factors affecting emergency department crowding. Emerg Med Clin North Am. 2020;38(3):573–587. 8. Mentzoni I, Bogstrand ST, Faiz KW. Emergency department crowding and length of stay before and after an increased catchment area. BMC Health Serv Res. 2019;19(1):506. 9. Gaucher N, Bailey B, Gravel J. Who are the children leaving the emergency department without being seen by a physician. Acad Emerg Med. 2011;18(2):152–157. 10. Gravel J, Gouin S, Carrière B, et al. Unfavourable outcome for children leaving the emergency department without being seen by a physician. Can J Emerg Med. 2013;15(5):289–299. 11. Brook C, Chomut A, Jeanmonod RK. Physician assistants contribution to emergency department productivity. West J Emerg Med. 2012;13(2):181–185. 12. de la Roche MR, Dyer N, Froats M, et al. Effect of a physician assistant on quality and efficiency metrics in an emergency department: population cohort study. Can Fam Physician. 2021;67(2):e61–e67. 13. Muller K, Chee Z, Doan Q. Using nurse practitioners to optimize patient flow in a pediatric emergency department. Pediatr Emerg Care. 2018;34(6):396–399. 14. Burt H, Doan Q, Landry T, et al. The impact of universal mental health screening on pediatric emergency department flow. Acad Pediatr. 2022;22(2):210–216. 15. McKinley KW, Chamberlain JM, Doan Q, Berkowitz D. Reducing pediatric ED length of stay by reducing diagnostic testing: a discrete event simulation model. Pediatr Qual Saf. 2021;6(2):e396. 16. Merdler I, Hochstadt A, Sheffy A, et al. The Israeli physician assistant in a tertiary medical center emergency department. Isr Med Assoc J. 2020;22(7):409–414. 17. Li J, Ramgopal S, Marin JR. Resource utilization during low-acuity pediatric emergency department visits. Pediatr Emerg Care. 2022;38(2):e983–e987. 18. Chang AM, Lin A, Fu R, et al. Associations of emergency department length of stay with publicly reported quality-of-care measures. Acad Emerg Med. 2017;24(2):246–250. 19. Byczkowski TL, Fitzgerald M, Kennebeck S, et al. A comprehensive view of parental satisfaction with pediatric emergency department visits. Ann Emerg Med. 2013;62(4):340–350.