Article Text

Triage: making the simple complex?
  1. Ellen J Weber
  1. Department of Emergency Medicine, University of California, San Francisco, San Francisco, CA 94143, USA
  1. Correspondence to Dr Ellen J Weber, Department of Emergency Medicine, University of California San Francisco, San Francisco, CA 94131, USA; ellen.weber{at}ucsf.edu

Statistics from Altmetric.com

One of the hallmarks of an emergency department (ED), compared with an urgent care, or a physician’s office, is that patients are seen by the clinicians in order of acuity, not arrival. When there are plenty of caretakers and few patients, determining the order is less important; everyone gets seen quickly. But when there are more patients than there are staff, some form of prioritisation is needed to prevent a critically ill patient deteriorating in the waiting room. Thus, triage is a critical need when there are not enough clinicians to go around. And as our departments become more and more crowded, our waits longer, there appears to be a greater and greater need for triage.

At the same time, triage itself, or at least the way we are told to practise it, also soaks up resources. In most EDs where emergency medicine is practised, one form or another of a multilevel triage scale is used to sort patients into three, four or five categories. Usually, only experienced nurses are allowed to perform triage, but only after special training. However, despite their extensive experience, these nurses are then asked to follow an algorithm rather than utilising their clinical judgement. Moreover, we hand to triage tasks that could be done by others: in some cases a quick registration, taking observations which could be done by a technician, or handling complaints from waiting patients, which would be better solved by a security guard, or, better yet, a kindly volunteer offering a cup of tea. Moreover, the nurses are asked to sort to a degree that at best marginally contributes to an orderly flow of patients, requiring at times Talmudic judgements about whether a patient is a level 4 or a 5, a green or a blue.

In the EMJ, Iversen et al 1 now report results from an investigation from Denmark that suggests that we can do better, and if we cannot do better, we can pretty much do the same thing with less time and cost. These researchers compared the accuracy of triage decisions by nurses who adhered to the Danish Emergency Process Triage (DEPT) scale with a non-systematic ‘eyeball triage’ performed by phlebotomists and medical students working as phlebotomists from the Department of Clinical Biochemistry. The phlebotomists were instructed to look at the patient and if possible ask for a chief complaint, then assign the patient to a DEPT colour-coded acuity based on who needed the most immediate attention. The authors then compared the accuracy of the triage scores of the nurses and the phlebotomists in predicting mortality. In short, ‘eyeball triage’ was more accurate.

There are certainly several caveats to this study—it was performed at a single centre, with a triage system that is not widely used internationally. The authors used the proxy of mortality, because there is no gold standard for ‘acuity’—how quickly a patient needs treatment. However, this is a problem with all triage studies. There was no assessment of reliability (how often the nurses agreed with each other or the phlebotomists agreed with each other). Yet the study should make us rethink our current process and the evidence behind it.

Unlike many of our current decision rules, most triage algorithms were built by consensus, and their validity tested after. As described by Challen,2 they are difficult to evaluate because there is no gold standard for acuity and many different proxies have been used. While in general you find a correlation between acuity ratings and admission/mortality rate, most validation studies find ‘low acuity’ patients are admitted. Storm-Versloot and colleagues found that Emergency Severity Index (ESI) had an undertriage rate of 13.5%, and Manchester Triage Scale (MTS) an undertriage rate of 11.2%, with sensitivity for an acuity of 2 (using a reference standard) of 36% and 34%, respectively.3 This 2011 study compared ESI and MTS with an informally structured system of triage performed by nurses, and found the unstructured system had the lowest undertriage rate (8.3%), and highest sensitivity for level 2 (47%) without sacrificing specificity. A multicentre study conducted in the United States, Brazil and the United Arab Emirates found an overall accuracy of 59.2% and modest inter-rater reliability.4 Not surprisingly, triage systems work better on populations they were developed for.5 A recent review in our pages demonstrated that MTS had sensitivities in the range of 60%–70% for children compared with a reference standard of expert consensus.6 A study by Twomey et al found that the South African Triage Scale (SATS) was 91% sensitive, but 54.5% specific, risking overtriage.7 A further study on SATS showed only moderate sensitivity and specificity for this system in Haiti and Afghanistan.8 It is also not clear our current systems are fit for current purpose. A recent study in our pages demonstrated that the MTS had poor sensitivity and specificity for identifying sepsis.9 At the same time that all-out efforts are being made to minimise door to needle time, most triage systems would classify patients with ischaemic stroke as level 2.

While these scores may be able to distinguish the very sick from those with minor illness, they have more trouble distinguishing patients with worrisome complaints who may deteriorate or need admission from those who need a bit of fluid and can be discharged. Like a saggy bed, too many patients fall to the centre. In a five-level system, the majority of patients will be assigned a middle score, so that the patients with acute cholecystitis, small bowel obstruction, pneumonia and syncope are assigned the same acuity as a patient with gastroenteritis needing a litre or two of fluid and an antiemetic.

Sorting, and often resorting, patients causes delays. When I was researching the 4-hour target in 2009, several UK consultants told me that triage ‘just creates another queue.’ Studying this phenomenon in my own department in the USA, we found that nearly half of all high-acuity patients had not completed triage in the time frame that was recommended for them to see a physician.10 Using the terminology of ‘Lean Manufacturing’ , a production philosophy pioneered by Toyota and based on maximising value for customers, triage has been called‘muda’ (Japanese for waste), causing rework, overproduction and mistakes. (In Swahili, by the way, ’muda' means ‘time’, another relevant translation.) Patients become frustrated with the constant repetition of their histories, and the multiple providers they need to see before they reach the physician. Patients may assume (rightfully so) that something they have told a nurse in triage is now in the chart, and the doctor will see it. But that is not always the case.

In short, we have adopted complex systems that take up the time of highly qualified nurses, potentially delay care, to create what is probably, at best, a ‘meh’ result.

If CT scans performed this way, surely we would have continued to explore alternatives. We need to ask ourselves in these days of rising medical costs and rising patient numbers if we can afford to continue doing it the way we have always done it if we can do it just as well or better a simpler way.

As our departments become more crowded, there are several reasons to rethink triage. Yes, it is more vital than ever, but can we really continue to devote the time and resources to this function when simpler solutions may be just as good—or even better? We may not be quite ready to hand this vital task over to individuals with little emergency experience, but the study by Iversen and colleagues, along with the undistinguished performance of most widely used triage systems, suggests that we best keep a very open mind.

References

Supplementary materials

Footnotes

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent Not required.

  • Provenance and peer review Not commissioned; internally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles