1

Fixing the Front End: Using ESI Triage v.4 to Optimize Flow

David R. Eitel, MD, MBA for the ESI Triage Research Team:

Dave Eitel, Nicki Gilboy, Alex Rosenau, Paula Tanabe, Debbie Travers, and Richard Wuerz ( Deceased)

Corresponding Author: David R. Eitel, MD, MBA Department of Emergency Medicine Wellspan Health – York Hospital 1001 South George Street York, PA 17405 Work phone: (717) 851-2450 Business Cell Phone (717) 495-5314 FAX (717) 851-3420 Email: daveitel@suscom.net In Memory Of: Richard Wuerz, MD (1960 – 2000) Associate Clinical Director Department of Emergency Medicine Brigham and Women’s Hospital Harvard Medical School

Richard C. Wuerz, MD1960-2000

2

Introduction: In this paper and accompanying power point presentation, we introduce you to “the Emergency Department (ED) Problem” and ask three questions about ED triage. We then give a brief background about ESI Triage, its history and development, and make clear the versions of ESI (4). We show what it is, how it should be implemented, and what is new in version 4. We then address some of the service management opportunities available once it is implemented, methods and tools with which this audience should be familiar. We close with how you can obtain at no cost the ESI version 4 “Everything You Need To Know” implementation manual and training video to help you to improve ED and hospital patient flow at your health system. AHRQ: www.ahrq.gov/reserach/esi The ED Problem: This is how Rich liked to explain Emergency Medicine: 1) patient arrives, 2) stuff happens, 3) patient leaves. Our research team decided to spend a fair amount of time on the front end, so we could help out with steps 2 and 3. There is some good news for us in Emergency Medicine: business is booming! In 2003 the number of yearly visits to ED’s in the United States (U.S.) climbed to nearly 114 million. That is good news; isn’t it? That is a lot of business, and it continues to grow! However, there is also some bad news for Emergency Medicine. In 1993, the U.S. Government Accounting Office (GAO) published the following, which is now felt to be truth by every hospital administrator, hospital board member and every third party payer in the U.S. Of all the patients who come into the ED in the U.S., 17 % are emergent; 40% are urgent; and fully 43% of patients, who come into your ED and my ED, are non-urgent - and should not be there; their visits should not be covered. Is that a problem? And where did this data come from anyway? We ask you to remember the following graphic:

The Bad NewsThe Bad News……U.S. GAO, 1993U.S. GAO, 1993

17%

40%

43%

urgent

emergent

non-urgent

The following was published back in the mid-sixties, in the Journal of the American Medical Association (JAMA) no less: “The Emergency Department Problem”, Silver and Manegold, JAMA Oct 24, 1966. In their paper, Silver, et. al, stated that ED visits rose 175% from 1955-1965. They wrote that 42% of ED patients were “non-urgent” problems. They suggested that the following factors

3

contributed to “the ED problem”: 1) mobility – no primary doctor; 2) difficulty finding a physician at night; 3) indigent populations and; 4) 24/7 diagnostic facilities at hospital. This kind of discussion continued through the 1990’s. Mr. Clinton called us “…the most costly care of all…” Dr. Bob Williams painted us in a somewhat better light when he published his data about the marginal costs of minor emergencies – approximately twenty-five dollars. His argument, however, assumed excess capacity. We all know that the use of ED as source of primary care is ongoing. Forty-three million people (at least) in the U.S. are without health insurance; and we know that an insurance card does not equal access to health care. And then there is the ACEP prudent layperson language definition of “emergency”. What Is an Emergency? What is an emergency anyway? Is an event an “emergency” if it is a life threat? What about a life or limb threat? Is an event an emergency if it eventually results in hospital admission or an operation? Is it an emergency if it requires care within two hours or if it requires care within twenty-four hours? (however you determine that). Is an event an emergency if it causes severe pain? What about: “My lawyer sent me in to get checked”? Other ED Problems There are other problems for us in Emergency Medicine, one of the biggest the perception that we “cost way too much”. We also have a “Quality/ Satisfaction” problem. We are all aware of the variation in timeliness to care perceived by ED patients and we know that the single biggest thing that our patients complain about is wait time. There is a new term suddenly part of the lexicon of Emergency Medicine: “ED overcrowding”, although perhaps better called “access block” by our Australian colleagues. For those of us who work in and around the ED, we know we have a frank patient safety problem and nursing exodus from the ED. Is that a problem? I would now like to ask the following questions: 1) What is ED Triage? 2) Why do we do it? 3) What does ED triage have to do with any of this, anyway? ESI Triage The driver of my (Eitel’s) interest in ED triage came about in 1994 while doing an operations management assignment with my MBA team. In my MBA we were assigned to teams for the three year MBA - I was on an incredible team. For our operations management class our team was assigned a reengineering project. Reengineering 101:

1. Pick a business that is in trouble. We decided to pick the York Hospital ED.

2. Identify its key business processes. I did not know what that meant going into it, but I do now. As this audience knows, the way one identifies a business’s “key business processes” is to get into the shoes of its customers and map out all the steps those customers are forced to go through - to get through it.

4

3. If something is broken – FIX IT! Do not TQM it or tweak it – fix it. Therefore, as my MBA team watched patients go through the ED at York Hospital in the autumn of 1994, we all saw ED triage in action for the first time. We noticed that everyone did triage but everyone did it differently. Even the same nurse later on the same shift did triage differently than she/he had done it earlier in the shift. Our team paper for that operations management assignment was “Reengineering the ED – Fixing Triage”. The premise for the paper was that ED triage had to become about more than just sorting; ED triage had also to be about streaming. In fact, ESI later was initially developed so that we could move toward predictive management and modeling in the ED care delivery setting. ESI triage was developed (Wuerz and Eitel) so we could flow (map) and then model the complex ED service delivery setting. A Brief and Important Aside: Service vs. Products We in health care delivery are in a service business and we must begin to manage it as such. There exists a science of service management, within the discipline of operations management/industrial engineering. We should begin to train all of our hospital /health care managers in the core concepts, content and tools of service management. We would like to introduce you to a great, easy-to-read textbook with most of the quantitative stuff at the end of the chapters: Service Management, Third Edition, Fitzsimmons & Fitzsimmons, ISBN 0-07-231267-X. Note the following chapter titles: Chapter 10 Forecasting demand for services Chapter 11 Managing waiting lines Chapter 12 Queuing Models (Server) & Capacity Planning Chapter 13 Management Capacity and Demand Do you think these kinds of solutions and associated tools could be of help to us in hospitals and health care delivery settings? This is a good time to remind the reader/audience that there is out a new report from the National Academy of Sciences and Institute of Medicine (NAS/IOM): “Building a Better Delivery System: A New Engineering/Healthcare Partnership”. Effective January 2005 there also is out the new JCAHO Hospital Patient Flow standard. ESI Triage The following are the ED triage levels with definitions in place at the Brigham and Women’s hospital, Rich’s ED, before April of 1999: Emergent: 1%- requires immediate evaluation and treatment Urgent: 65% - can tolerate a period of time in the waiting room Non-urgent: 35%- minor illness/injury that can be treated within six hours Check out again the definition of “urgent”: “can tolerate a period of time in the waiting room”. (Whose opinion?)

5

At about the same time The Emergency Nurses Association suggested the following definitions for a three-level ED triage approach: Emergent/Level 1 – Life or limb-threatening illness/injury Urgent/Level 2 - requires prompt care, but will not cause loss of life or limb if left untreated for several hours Non-urgent/Level 3 - time is not a critical factor; minor illness or injury In the spring of 1999, Dave Eitel at York Hospital asked their ED Information Technology guy, Lance, if he could give Dave the breakdown of patients who had come into their tertiary care community teaching hospital and trauma center for the first several months of 1997. Lance gave him the first three columns of data as in the following table:

Triage Data Report YH ED 1997Triage Data Report YH ED 1997

22 % 22 % admitsadmits

18,029 18,029 visitsvisits

JanJan--Apr Apr 9797

11 %11 %73 %73 %13,15013,150Level 3Level 3

51 %51 %25 %25 %4,5774,577Level 2Level 2

69 %69 %2 %2 %302302Level 1Level 1

ADMIT %ADMIT %%%VOLUMEVOLUMETRIAGETRIAGE

This report told Dave -and many others at the hospital – that of all the patients that came into their ED in the first four months of 1997: 2% were level one; 25% were level 2; and fully 73 % of all patients were level 3’s: the low “acuity” patients. Seventy-three percent of all patients coming into the York Hospital ED were declared level 3’s, or very “low acuity” patients. Dave personally had NEVER seen a shift like that in the York Hospital ED, and he had been there since 1982. WHERE did this data come from? (Remember that pie chart?) Dave then asked Lance for another piece of data: what was the percent admit rate for each of the 3 triage levels. Lance gave Dave the fourth column: 69% of the 1’s were admitted; 51% of the 2’3; and 11% (more than 1 in 10) of the Level 3’s were actually admitted to hospital. There was something terribly wrong with this data, even though it was being used to make many decisions or confirm “assumptions” about the ED’s case mix. When this was shared with Rich, and he looked at his own data at that time at Hershey, Rich did what he always did. He designed a study. The following study was published a year later: “Inconsistency of Triage,” Wuerz et al: Annals of Emergency Medicine, October 1998. In this study, Rich asked eighty-seven volunteer nurses at two academic EDs (the Hershey Medical Center and

6

the York Hospital) to triage five standardized patient scenarios using each of their three level scales. Hershey was using an emergent, urgent, non-urgent approach; York a level 1-2-3 approach, like you just saw. The five scenarios were given in a scripted was by medical student Joe Alacron to each of the volunteer ED triage nurses. Rich then measured reliability. Results:

a) inter-rater (between raters) reliability: there was only a 35% agreement beyond chance between the nurses.

b) test-retest (intra-rater) reliability. Rich asked the nurses to repeat the classification of the five standardized cases two to four weeks later: Only 25% of the time did the triage nurses agree with themselves from their first triage assignments.

The problem was not with the nurses. They were great nurses, especially then during a time of stability in ED staffing, with very experienced nurses at the front desk. The problem was with the tool (instrument) they had been “given”. The real conclusion is that the instrument was too “blunt”. Actually, they really had no validated instrument (tool) with which to do ED triage. What else was out there? We then spent a lot of time finding out what was out there regarding ED triage methods. We cannot detail that in this paper/presentation but there existed (then and now) three other established ED triage methods: 1) The Australian National Triage Scale (NTS) – established in 1994 by Fitzgerald; 2) The Canadian Triage and Acuity Scale (CTAS) established by Beverige, et. al, and promulgated as the tool to be used in Canadian ED’s starting in 1996 and; 3) Manchester Triage – established by a consensus working group in Manchester, England and promulgated to be used as the triage method for the United Kingdom’s National Health Service starting in1997. Manchester Triage is a different approach from the NTS and CTAS. Both the NTS and CTAS triage methods are organized around the following statement: “This patient can wait no longer than (pick a time) time, to see a physician” as in the table below:

This patient This patient can wait no longercan wait no longer thanthan……to see to see a a physicianphysician

Australian & Canadian Triage

120 min120 min120 min120 min55

60 min60 min60 min60 min44

30 min30 min30 min30 min33

1515 minmin10 min10 min22

0 min0 min0 min0 min11

CTASCTASNTS NTS Triage levelTriage level

7

We ask again: What is ED triage and why do we do it? Our research team thinks that a principal goal of ED triage should be to determine who should be seen first – right? Yes, most would agree. The problem is, if that is the only question asked by a triageur, the functional operational question becomes: “how long do you think everyone should/could wait?” There just is no science, medicine or nursing evidence around this “how long can everyone wait?” question. There is nothing to drive a reliable, operationally relevant, reproducible classification of ED presenting patients around this “wait” question. Our research team also thinks that a second major goal for ED triage should be not to just “sort” but to “stream”: to get the right patient to the right resources in the right place and at the right time, from the get-go. Rich made a very important observation while playing what he came to call “the triage game” at the Brigham with his nurses. If he gave case scenarios to his triage nurses and asked, “What will this patient need to get to a disposition?” hands would rise in unison to address what a particular case (scenario) would need to get through their ED visit. To make explicit another important observation: there are “big” emergencies and there “little” emergencies. Experienced ED nurses are excellent at this - especially those potentially “big emergencies”. In the same vein of thinking, “If your little girl, niece or granddaughter falls and cuts her forehead, her face and dress are all bloody, and she needs stitches to close the laceration on her forehead – is that an emergency?” Of course it is; it might be complex, but it does not require the resources of an ED resuscitation room. Next, a short train of thought about ESI triage: ED Triage is not just about time – it’s about resources! We also want to begin to manage the ED by thinking flow 1st, not capacity 1st (beds), as in The Goal by Goldratt. However, to manage by flow, we have to first decide how to stream incoming patients. So, in ESI triage two questions are asked: 1) “Who should be seen first,” and 2) “What does the patient in front of you need, in terms of resources, to reach a disposition?” We never ask the “how long must/can everyone wait?” question in ESI triage. ED triage with ESI becomes about sorting and streaming: getting the right patient to the right resources and at the right time – from the get-go. Those in need of few resources but the doc-nurse team can bypass the main ED. The parallel processing of patients can occur if patient categorization is done reliably at the front end. Remember that “Bad News” pie chart? Just because data is lousy does not mean that it won’t be used to make decisions or support assumptions. The data from 3 level triage approaches produce lousy information about ED case mix. We absolutely must do better than this.

8

ESI Versions: We next clarify the various “versions” of ESI found in the published literature. The ESI v.1 Triage Algorithm Over time, Rich and Dave embedded their logic into five levels with explicit definitions, and, like others, embedded this logic into complex tables. In August 1998 there came a breakthrough! The logic of ESI was embedded into a flow chart-based algorithm and tool after reading the work of Edmund Tufte, which includes “The Visual Display of Scientific Information”. The first version of ESI was an “adults only” tool, for patients > age fourteen. The tool was first applied in a retrospective way to the triage charts of patients who had presented to the Brigham. The ESI Triage Research Team also began to come together.

none one many

vital signs

1

2

5 4

3

yes

yes

no

no

yes

patient dying?

shouldn’t wait?

no

how many resources?

ESI v.1 (conceptual) In ESI we used vital sign criteria to up triage patients who had gone through four previous categorization steps. We did not want to miss any potentially ill patients; we wanted to increase the sensitivity of the tool. However, there was (and remains) no clear consensus in the literature on what constitutes an “abnormal vital sign” and its cut-offs. Because of work done at the Henry Ford Hospital, we adopted the SIRS Systemic Inflammatory Response (Not SARS) criteria for version one. Comments: Reliability versus Validity Reliability: for us in this operations research, reliability means the reproducibility & repeatability of a measurement tool (instrument). We looked to the industrial engineering literature for guidance rather than just traditional biomedical texts and literature regarding reliability testing. “Inter-rater agreement” and “test-retest agreement” are the operational definitions we use.

9

Validity: The “so what?” question Within the biomedical statistical literature there are several definitions of validity. For us in this operations research validity means predicative validity. What are the major outcomes associated with each ESI level? We remind the reader that reliability in classification assignments begets predictability. We look at several operational outcomes associated with each triage level. ESI v.1 (Adult) Retrospective The retrospective work we completed from October – December 1998 with ESI v.1 produced the following paper: “Reliability and Validity of a New Five-Level Triage Instrument”, Wuerz, Milne, Eitel, Travers and Gilboy, Academic Emergency Medicine (AEM), 2000;7(3):236-42. In this initial work with the new ESI algorithm, the operational outcomes - such as ED length of stay and inpatient admission rates - made sense by triage class. We were excited! ESI version 1 (Adult) Implementation ED leaderships at University of North Carolina (UNC) at Chapel Hill and The Brigham decided to replace their existing three level triage approaches with the new ESI v.1 five level triage algorithm. On April 1, 1999 at UNC - Chapel Hill and on April 15, 1999 at the Brigham the new ESI method was implemented at these two adults-only facilities. The research team had developed a standardized training program for implementing ESI v.1. Nurses at the two sites were trained in the 1.5 hour standardized education package. The package included a didactic presentation, a group discussion of triage case scenarios, and a twenty case test set that included photos. (We no longer use photos in implementation training). Everyone in the department was informed about the implementation a priori. By the way, this is the way that you too should implement ESI triage: train all your triage nurses in a standardized training program, and inform everyone. This adult, two-hospital implementation produced feedback from nursing staff at both locations. This resulted in some new triage language in the ESI v.2, All Age version. ESI version 2 (All Age) 1999 ESI v.2 contained the same five levels and same explicit definitions as ESI v.1. Pediatric triage criteria were added to handle the potentially bacteremic child using existing literature-based criteria. Vital signs criteria were upgraded based upon nursing and other feedback. We had completed a research planning retreat in April of 1998, and by the spring of 1999 the research team was fully in place. We had written a grant to the Agency for Healthcare Research and Quality (AHRQ) with Rich as the Principal Investigator. The grant was awarded in August 1999 so a multi-site implementation of ESI version 2 was able to begin. The team had produced new implementation training materials, case scenarios for training

10

and testing, standardized approaches to reliability and validity testing, and pocket cards. The training materials were field-tested at the York Hospital; two cases were changed. The materials were then distributed to the other six sites where training and implementations occurred, and data collection began. This multi-site implementation resulted in the following paper: “Eitel, D, Travers, D, Rosenau, A, Gilboy, N, and Wuerz, R. “The Emergency Severity Index Triage Algorithm Version 2 is Reliable and Valid”, Academic Emergency Medicine. 2003; 10(10)1070-1080. Rich had died suddenly and unexpectedly on October 6, 2000, as we were assembling the data and comments for this paper. ESI Triage Version 2 versus 3 Based upon feedback from nurses at all seven sites in the study the language regarding the usage of vital signs to up-triage a patient was changed from “yes” to “consider”. All nurses involved felt strongly that a one-time vital sign recording, separate from the pediatric triage criteria, should not supersede the judgment of the triage nurse.

none one many

vital signs

1

2

5 4

3

yes

**consider

no

no

yes

patient dying?

shouldn’t wait?

no

how many resources?

ESI Triage Version 3 Distribution via the ENA Handbook Using that which we learned from the multi-site implementation and validation study, the research team produced: “The Emergency Severity Index Implementation Handbook: A Five-Level Triage System” authored by Nicki Gilboy and Paula Tanabe, with contributions from Debbie Travers, Alex Rosenau and David Eitel. It was published and distributed by The Emergency Nurses Association [ENA] Deplains, IL: 2003. This handbook contained ESI version 3 (“consider”). This is no longer available from the ENA. ESI version 4 is out and, by mutual agreement, with a new publisher, the Agency for Healthcare Research and Quality ( www.ahrq.gov/research/esi). ESI Triage v.4

11

What is New in Version 4? Based upon work principled by Paula Tanabe the level 1 criteria have been expanded: Tanabe, etal., AEM, June 2005 – “Refining Emergency Severity Index Triage Criteria”. The definition of a Level 1 patient is “requires immediate life-saving intervention”. The pediatric fever criteria have been updated to match the new American College of Emergency Physicians (ACEP) Pediatric Fever Criteria: “The American College of Emergency Physician’s Clinical Policy for Children Younger than three years Presenting to the Emergency Department with Fever” - 2003.

requires immediate

life-saving intervention?

high risk situation?

or

confused/lethargic/disoriented?

or

severe pain/distress?

1

2

yes

no

yes

ESI Triage Algorithm v.4

A

B

ã ESI Triage Research Team 2005

What can you do with ESI v.4 triage? First a quote from one of the author’s favorite scientists: “The job of management is prediction” (Dr. Deming). We sincerely believe that this should be the case in health services delivery and its management, including the ED. We also hope that we have convinced the reader/audience that reliability begets predictability, operationally. The following opportunities are available to you immediately upon the successful and reliable installation of ESI v.4 Triage; 1) The real time management of patient flow: Level 1’s and 2’s in general go to your critical care area. Most level 4‘s and 5’s go to another area of your ED (often called urgent care or fast track), at the same time. The parallel processing of patients can begin. Note: we on the ESI Triage Research Team have zero interest in triaging patients away from the ED as some have advocated. Zero interest. 2) Communicating the ED workload to others

12

The definitions used to differentiate patients with ESI triage are explicit and thus easily understood - by clinicians and non-clinicians alike - such as hospital administrators. You are on your way to a meeting where you will discuss ED staffing and the negative effects overcrowding is having on patient safety and staff retention. Last evening you had six level 2 patients who had to remain for five hours in your waiting room. You remind your administrator that the definition of a Level 2 patient is: “A high risk situation; acutely confused/lethargic/disoriented; or in severe pain or distress”. This was of great concern to your competent and motivated staff last night, all of whom felt terrible that they could not provide better patient care. You can immediately begin to have much more meaningful discussions with your administrators about your ED resourcing needs. 3) Physical plant and staffing decisions If nearly 40% of your ED’s presentational case mix are 4’s and 5’s, as for the York ED – do you really need a bigger ED to handle your volume, or do you need a simple re-design of your existing space? 4) Physical plant, staffing and staffing mix decisions If nearly 40% of your ED’s presentation case mix are 4’s and 5’s, how many types of docs vs. NP’s/PA’s are you likely to need for that kind of case mix? What if you also knew that 65-70% of 4’s and 5’s as triaged by ESI are “boo-boo’s” (trauma related?). And: do insurance companies, in general, pay well for boo-boo management? Yes they do, at least for ED docs. NP/PA reimbursement is highly state and region specific, so you should investigate the financial implication for using ED docs vs. mid-levels in your Fast track as determined by ESI triage. 5) Multiple Hospital Capacity Planning If you have several ED’s in your system (country, or consulting mix) how might you think about staffing at each site if you had ESI-driven, reliable ED case mix data available to you across those ED’s? Or, if you are a health planner how could ESI’s reliable ED case mix data help you?

13

Case Mix by Site

0%

10%

20%

30%

40%

50%

60%

1 2 3 4 5

ESI© Triage Level

% P

ati

en

ts

BW

FH

17th

MH

YH

UNC

LVCC

6) Downstream Hospital Readiness Since you began using ESI triage a few months ago, you’ve been able to collect the following data:

Presentational Case Mix DataPresentational Case Mix Data(“can manage the waiting room…”)

8,063TOTAL

1.414%.003%812 (10%)Level 5

2.047%2%2,197 (27%)Level 4

3.473%24%3,173 (39%)Level 3

4.090%54%1,756 (22%)Level 2

2.480%73%125 (2%)Level 1

ED LOS

(hours)

Resource

Intensity

Admit

Rate

Case Mix

(% total)

Triage

Level

These numbers (from an ESI v.3 triage system) indicate that just over ½ of the level 2 patients and nearly three quarters of the level ones in your department are admitted to the hospital. Because the admission rate for each group is remarkably consistent over time, knowing the ESI acuity classification of the patients in your ED enables you to estimate the number of patients who are likely to require a bed upstairs. While you can’t know for certain the disposition of any particular patient, comparing the ESIs of any group of patients will allow you to

14

make fairly accurate predictions about the number of inpatient beds that will be needed. By tracking which inpatient units your level 1, 2 and 3 patients are admitted to over a period of time, you can compile destination distribution data. This allows the hospital’s management team to proactively prepare for the anticipated patient service loads on those units as well as to anticipate the inevitable “downstream ancillary demand” that occurs as additional diagnostic studies are ordered. 7) Services Operations Management Concepts, Contents, Tools With ESI case mix data, some service (operations) management tools that are available to you are: Demand analysis and statistical forecasting Capacity to serve planning: optimize staff scheduling (rostering) to predicted demand ED workflow diagramming (ED service unit mapping) and conceptual [static] modeling The Lean (Process Excellence) business improvement method Enhanced discrete event simulation modeling How can you get ESI version 4 triage? The ESI v.4 Implementation Handbook and Training DVD

You can download a fully-licensed PDF version of the implementation handbook from this site: www.ahrq.gov/research/esi. You can call 800-358-9295 and request one free copy (was 3) of the “Everything You Need To Know” spiral bound handbook and the “Everything You Need to Know” Training DVD to be mailed to you, wherever you are located. Thank you for the opportunity to present.

15

Respectfully, David Eitel MD MBA For the ESI Triage Research Team February 12, 2006. In Memory Of: Richard Wuerz, MD (1960 – 2000) Associate Clinical Director Department of Emergency Medicine Brigham and Women’s Hospital Harvard Medical School