a b s t r a c t

Objectives: Assess the impact of scribes on an academic emergency department’s (ED) patient-specific throughput. Methods: Study design, setting, participants: A prospective cohort design comparED throughput metrics of patients managed when scribes were and were not a part of the treatment team during pre-defined study hours in a tertiary academic ED with both an adult and pediatric ED.

Intervention: Eight scribes were hired and trained on-site by a physician with experience in scribe implementation. Scribes provided 1-to-1 support for a provider’s work shift. An alternating-day pattern in months 2 to 5 post implemen- tation ensured balance between the scribe and non-scribe groups in time of day, day of week, and patient complexity. Results: Adult: Overall length of stay (LOS) was significantly longer for scribed patients (265 vs. 255 minutes, P = .028). The remaining throughput measures analyzed (door to provider, provider to disposition, and patient duration in treat- ment room) had higher summary values, but were not significant. Subgroup analysis revealed that patients seen by postgraduate year (PGY) 3 residents had significantly shorter LOS when seen with a scribe (244 vs. 262 minutes, P = .021).

Pediatric: Overall LOS (163 vs. 151 minutes, P = .011), door to provider (21 vs. 16 minutes, P b .001), and treatment room duration (130 vs. 123 minutes, P = .020) were significantly longer when the treatment team had a scribe.

Conclusions: Scribes failed to improve patient-specific throughput metrics in the first few months post implementation.

Future work is needed to understand whether throughput efficiencies may eventually be gained after scribe implementation.

(C) 2016

Introduction

Increasing documentation requirements and growing patient work- load have led some healthcare settings to explore the use of medical scribes [1-4]. Scribes are non-licensed health care team members who document patient encounters on behalf of a provider contemporane- ously with the patient visit. They can keep track of laboratory findings, radiological studies and record other pertinent information to improve physician productivity and patient care [5]. They do not act indepen- dently, but rather they assist with documentation, retrieve test results, and support workflow [6]. Although the use of scribes and the numbers of scribe programs are growing nationally, especially in emergency

? Presented at: abstract accepted for presentation at the American College of Emergen- cy Physicians’ Scientific Assembly, October 2016

?? Grant support: None

? Conflicts of Interest: None

* Corresponding author at: Department of Emergency Medicine, Mayo Clinic, 200 First St. SW, Rochester, MN, 55906. Tel.: +1 507 255 4399; fax: +1 507 255 4399.

E-mail address: [email protected] (H.A. Heaton).

departments (EDs), presently only five peer-reviewed studies on the ef- fectiveness of scribes exist to guide evidence-basED decision making re- garding implementation [7-11].

Understanding how scribes influence the ED practice is important. ED operations leaders are navigating a complex healthcare environment with rising patient volumes and acuity, an increasingly complicated re- imbursement and technologic environment, and increasing visibility of throughput metrics. In addition, poor ED throughput has been linked to increased morbidity and mortality, increased financial risk, higher mal- practice risk, and lower patient satisfaction [7]. Scribes have been pro- posed as a potential ED operations adjunct to improve provider efficiency and, thus, ED operational performance in many editorials [11,12,13]; however, the paucity of peer-reviewed literature on the topic makes it difficult to understand what to expect when EDs hire scribes, with only 3 of the 5 peer-reviewed studies addressing the im- pact of scribes in an academic ED and none addressing the impact in an academic pediatric ED [7,8,10].

In early 2015, our department added scribes to our ED workforce to assist our provider group. The scribes were hired, trained, and managed by an attending physician with several years of experience in

http://dx.doi.org/10.1016/j.ajem.2016.07.011

0735-6757/(C) 2016

H.A. Heaton et al. / American Journal of Emergency Medicine 34 (2016) 1982-1985 1983

implementing and managing scribe programs nationally. To evaluate scribes’ operational effect on our ED, we designed a prospective cohort study with a primary aim to assess whether scribes have an immediate impact on patient-specific throughput in our adult and pediatric aca- demic ED settings.

Methods

Study design and setting

The study was conducted at the Mayo Clinic (Saint Marys campus) ED, an academic ED with a separate embedded pediatric ED, and which hosts an EM residency training program. Our ED manages 73 000 patient visits annually, 82% of whom are adults (age N17 years). Thirty-five percent of adult patients and 16% of pediatric patients are admitted. Our study utilized a prospective cohort design. The study was deemed exempt by our institutional review board.

Selection of participants

We evaluated all patients roomed during predefined study times from July 1, 2015 through September 30, 2015, with the following predefined exceptions: behavioral health patients (due to excessive boarding and predictably long LOS in our population, potentially skewing throughput data), resuscitation patients (due to a complex and unique cross-coverage model that challenges comparison to the general ED population), patients who Left without being seen, and nurse-only visits (since these visits are not staffed by licensed providers).

We evaluated two groups: (1) patients managed by a “traditional” care team (without a medical scribe), where providers used personal preference to construct their own documentation in the medical record through transcription, voice recognition software, or self-entry in the electronic health record ; and (2) patients managed by a team that included a medical scribe. Scribes were assigned to senior residents in high-Acuity areas of the adult ED from 6:00 AM until 11:00 pm or 7:00 am until midnight, and 12:00 pm until 9:00 pm with NPs and PAs (who work both as independent providers and supervised by attendings), or attendings in both adult and pediatric EDs. Staffing followed an alloca- tion pattern which was developed separately from the providers’ sched- ules with no preference given based on specific providers. The pattern ensured balance between the scribe and non-scribed groups in the times and days of the week, with an equal number of scribe days and non-scribed days in this study through alternating “A” and “B? days.

Intervention

Scribes were recruited and trained through an “in-house” program with a defined curriculum developed by a physician with prior experi- ence implementing scribe programs [14]. The scribes were largely pre- health students hired as temporary employees for 1- to 2-year periods. May 2015 marked the completion of scribe training. Each scribe provid- ed 1-to-1 support to a predetermined provider (attending, resident, NP/ PA) for the entirety of the provider’s shift. Some providers had brief ex- posure to the use of scribes at other institutions, but most had no scribe experience prior to the beginning of the study period. A preferred scribe workflow was demonstrated several times to providers in meetings prior to clinical shifts with the scribes.

We developed an allocation scheme to allow for accurate compari- son between intervention (scribe) and control (non-scribed) groups. The scheme was followed without deviation throughout the study peri- od. Patients seen during shift A with a scribe were compared to patients seen during shift A without a scribe, and so on.

Methods and Measurements

Investigators extracted patient demographics, patient-specific timestamps and type of provider (attending, resident, NP/PA) from the EHR. We evaluated patient-specific throughput metrics for each visit during the study period. The overall study period included patients seen from July 1, 2015, through September 30, 2015 (2-5 months after scribe training was complete). Discrete fields in the EHR allowed us to identify all patients for whom a scribe was part of the care team.

Patients were further categorized by provider type to identify if a resident, NP or PA participated in his or her care. Patients with an atyp- ical provider for location (such as a patient cared for by a NP or PA in a high acuity hallway during resident conference time) or those seen by post-graduate year (PGY) 1 residents, were excluded from this particu- lar subgroup analysis, since they did not have adequate comparison pa- tients in this head-to-head design.

Analysis

Continuous features were summarized with medians. Categorical features were summarized with frequency counts and percentages. Comparisons between the scribe group and non-scribed group were evaluated using Wilcoxon rank sum or ?² tests. Statistical analyses were performed using version 9.3 of the SAS software package (SAS In- stitute Inc, Cary, NC). Any P b .05 was considered statistically significant.

Measures analyzed for both the scribe and non-scribed groups

included:

Length of stay (minutes) (LOS): arrival time until departure from the department

Door to provider (minutes): arrival time until first seen by a provider

Treatment room time (minutes): total time spent in the treatment room (as opposed to the waiting room)

Provider to disposition (minutes): time first seen by a provider until the Disposition decision was made and entered in the EHR

Patients per hour: For shifts where scribes were assigned to an at- tending provider responsible for an entire hallway, we calculated the number of patients seen per hour. We were did not calculate pa- tients per hour for shifts where scribes were assigned to residents or NP/PAs, as these providers work as pairs (along with an attending). For example, a shift may have an attending and two residents. If the scribe was assigned to one of these residents, that patients per hour number would be too confounded by the Experience level of the other resident, which patients each resident would manage, etc. Therefore, we only calculated patients per hour for shifts where the scribe was assigned to an attending physician, since the attend- ing physician is ultimately responsible for patients per shift in the en- tire treatment hallway.

Results

Characteristics of study subjects

During the study period, 8015 adult patients were seen, of whom 2091 (26%) were managed by a provider with a scribe. During the same study period, 1921 pediatric patients were seen, of whom 1003 (52%) were managed by a provider with a scribe. Table 1 describes pa- tient demographics. The scribed and non-scribed groups were similar in their baseline characteristics.

Main results

When evaluating adult patient-specific throughput metrics shortly after implementation of a scribe program, the scribe group had higher summary values on most measures, though only reached statistical signif- icance for LOS (Table 2). Median LOS was 10 minutes longer in the scribed

1984 H.A. Heaton et al. / American Journal of Emergency Medicine 34 (2016) 1982-1985

Table 1

Participant Demographics (adults and pediatrics)

Adults Pediatrics

Table 3

Pediatric ED

Non-scribed N = 918 Scribed N = 1003 P value

Non-scribed N = 5924

Scribed

N = 2091

P

value

Non-scribed N = 918

Scribed

N = 1003

P

value

Median

Length of stay (minutes) 151 163 0.011

Median or N (%) Median or N (%)

Age	59	58	0.62	8	7	0.20
Male	2925 (49)	985 (47)	0.074	497 (54)	550 (55)	0.76
ESI 1	6 (b1)	3 (b1)	0.62	5 [1]	2 (b1)	0.35
ESI 2	1562 (26)	556 (27)		66 [7]	87 [9]
ESI 3	3816 (64)	1352 (65)		583 (64)	593 (59)
ESI 4	519 [9]	176 [8]		250 (27)	312 (31)
ESI 5	17 (b1)	3 (b1)		11 [1]	7 [1]
Disposition:	2665 (45)	924 (44)	0.53	111 [12]	113 [11]	0.57

Admitted

ESI: Emergency Severity Index.

vs. the control group (265 minutes vs. 255 minutes (P = .028) and pro- vider to disposition time was 153 minutes with scribes vs. 149 minutes in the control group (P = .15) (Table 2). Patients per hour for residents and NP/PAs were not calculated, given these providers did not see all pa- tients in a given hallway. For attending physicians scheduled to work di- rectly with a scribe, no benefit in patients per hour was demonstrated.

For pediatric patients, LOS and treatment room time were significantly

longer in the scribe group at 163 minutes vs 151 in the control group (P = .011) and 130 minutes vs. 123 minutes (P = .020), respectively (Table 3). For these patients, there were 1003 patients seen during 381 hours of scribe coverage (2.61 patients per hour) and 918 patients seen during the comparison 381 non-scribed hours (2.57 patients per hour).

Post hoc subgroup analysis of the providers revealed that patients seen by 3 PGY-3 residents had an 18 minute shorter LOS when managed with a scribe, compared to patients managed by PGY-3 residents without a scribe (244 minutes vs. 262 minutes (P = .021), but no other provider type’s op- erational metrics improved with scribes at the 3 to 6 month mark (Table 2).

Limitations

There are several significant limitations of our study. First, there was a short time period between scribe onboarding and the study period.

Door to provider (minutes) 16 21 b0.001

Treatment room (minutes) 123 130 0.020

Provider to disposition 95 98 0.20

Scribe training/onboarding concluded in May with the study beginning July 1. The scribes were new to their role and most providers had no previous experience working with scribes. Furthermore, no prescrip- tions were made as to how the scribes should function with each pro- vider. The study does not look at the impact of scribes with a specific provider, thus does not control for across provider differences. July is a difficult month to evaluate throughput in an academic ED given the ad- dition of new junior learners and senior resident transitions. Additional- ly, although no significant changes were made to practice, our department was in the midst of a major physical space renovation throughout the entire study period.

Discussion

At our high acuity academic ED, the Early operational impact of scribes 2 to 5 months after completion of scribe training allocated across all age groups, treatment areas and providers on ED LOS was negative. This was true across patient age group and almost all provider types. Our findings suggest that scribes are not an immediate solution to patient-specific throughput inefficiencies for a high acuity, high volume academic ED. The impact over time remains to be seen.

This study’s strength was in Methodological rigor with a prospec- tive allocation scheme, making our study results notably different from those previously published. Allen et al. showed improved throughput metrics (door to disposition, door to provider, door to exit) after scribe program implementation across a tertiary academic

Table 2

Adult ED

Provider group Non-scribed Scribed P value

Median

N = 5924 N = 2091

Length of stay (minutes)		255	265	0.028
All providers Door to provider (minutes)		21	23	0.29
	Treatment room (minutes)	210	208	0.14
	Provider to disposition (minutes) Length of stay (minutes)	149 N = 599 297	153 N = 314 322	0.15 0.057
Attendings	Door to provider (minutes)	92	117	0.051
	Treatment room (minutes)	199	204	0.17
	Provider to disposition (minutes)	151	149	0.67
	Length of stay (minutes)	N = 771 249	N = 612 263	0.55
PGY 2 residents	Door to provider (minutes)	16	17	0.15
	Treatment room (minutes)	220	215	0.56
	Provider to disposition (minutes) Length of stay (minutes)	156 N = 1062 262	153 N = 860 244	0.77 0.021
PGY3 residents	Door to provider (minutes)	16	16	0.17
	Treatment room (minutes)	223	208	0.44
	Provider to disposition (minutes) Length of stay (minutes)	152 N = 215 288	155 N = 183 282	0.92 0.39
Nurse practioners, Physician Assistants	Door to provider (minutes)	89	90	0.68
	Treatment room (minutes)	173	171	0.31
	Provider to disposition	125	129	0.93

H.A. Heaton et al. / American Journal of Emergency Medicine 34 (2016) 1982-1985 1985

ED [7]. Other studies report efficiency in terms of patients per hour with estimates ranging from 0.1 to 0.8 increase in patients per hour [8,10].

There are several potential reasons for the differences noted be- tween our study and published literature, with some being weaknesses of this study. Providers included in this study had limited exposure to scribes prior to the study, and some of our scribes had only been in their role for weeks when this study started. Further, our department has relatively short length of stays and high admission rates when com- pared to other academic EDs. Our department’s acuity mix does not lend itself to the creation of a “fast track” area where lower acuity pa- tients could be seen. Additionally, we chose to evaluate our program shortly after scribe onboarding.

Our study did not evaluate other reported benefits of scribes discussed in the literature including decreased provider documen- tation time after a shift, or improved revenue capture [8,10]. We choose to look solely at the impact of scribes on patient movement within our adult and pediatric departments.

Patient-specific throughput metrics in most cases were not sig- nificantly improved with scribes in our study. Future studies will focus on understanding what provider types are best paired with scribes in an academic ED, and the impact on patient-specific throughput metrics as the program matures. Additionally, the training and learning curve associated with scribe onboarding is yet to be defined and likely influences the marginal effects seen in this study.

References

1. Tegen A, O’Connell J. Rounding with scribes: employing scribes in a pediatric inpa- tient setting. J AHIMA 2012;83:34-8 [quiz 9].
2. Conn J, Meyer H. More docs get EHR help. Medical scribes move beyond the emer- gency room. Mod Healthc 2013;43:40-1 [3].
3. Koshy S, Feustel PJ, Hong M, Kogan BA. Scribes in an ambulatory urology practice: patient and Physician satisfaction. J Urol 2010;184:258-62.
4. Bank AJ, Obetz C, Konrardy A, et al. Impact of scribes on patient interaction, produc- tivity, and revenue in a cardiology clinic: a prospective study. Clinicoecon Outcomes Res 2013;5:399-406.
5. Baugh R, Jones JE, Trott K, Takyi VE, Abbas JT. Medical scribes. J Med Pract Manage 2012;28:195-7.
6. Clarification: safe use of scribes in clinical settingsJt Comm Perspect 2011;31:4-5.
7. Allen BBB, Weeks E, Payton T. An assessment of emergency department throughput and provider satisfaction after the implementation of a scribe program. Adv Emerg Med 2014;2014:7.
8. Arya R, Salovich DM, Ohman-Strickland P, Merlin MA. Impact of scribes on perfor- mance indicators in the emergency department. Acad Emerg Med 2010;17:490-4.
9. Bastani A, Shaqiri B, Palomba K, Bananno D, Anderson W. An ED scribe program is able to improve throughput time and patient satisfaction. Am J Emerg Med 2014;32: 399-402.
10. Hess JJ, Wallenstein J, Ackerman JD, et al. Scribe impacts on Provider experience, op- erations, and teaching in an academic emergency medicine practice. West J Emerg Med 2015;16:602-10.
11. Walker K, Ben-Meir M, O’Mullane P, Phillips D, Staples M. Scribes in an Australian private emergency department: a description of physician productivity. Emerg Med Australas 2014;26:543-8.
12. Grimshaw H. Physician scribes improve productivity. Oak street medical allows doc- tors to spend more face time with patients, improve Job satisfaction. MGMA Connex 2012;12:27-8.
13. Hafner K. A busy Doctor’s right hand. N Y Times 2014;163:D1-5.
14. Heaton HA, Samuel R, Farrell KJ, Colletti JE. Emergency department scribes; a two- step training program. Ann Emerg Med 2015;66(4):S159.