• 
• 

Three Part Question

In [adult ED patients] do [Early Warning Scores] predict [mortality]?

Clinical Scenario

Whilst working in the Emergency Department you assess a 44 year old male patient with a large intracerebral haemorrhage. He only scores 2 on his early warning score due to his decreased level of consciousness, his other physiological variables being normal. Based on clinical indications, he is intubated and ventilated and taken to intensive care but dies two days later. Whilst reflecting on this case you wonder whether early warning scores are sensitive and specific enough to predict mortality in ED patients.

Search Strategy

A combined database search of Cinahl (1981 to present), Embase (1980 to present) and Medline (1950 to present from PubMed) was carried out using NHS Evidence Health Information Resources. The following search terms were used: {[scoring systems].ti,ab OR [track and trigger].ti,ab OR [mortality prediction].ti,ab OR [early warning scores].ti,ab OR [MEWS].ti,ab OR [rapid emergency medicine score or [REMS].ti,ab} AND {[emergency medicine].ti,ab OR [emergency department].ti,ab}. No limits were placed on the search.

Search Outcome

Altogether a total of 144 papers were identified, of which 10 were relevant to the study question. These are shown in the table

Relevant Paper(s)

Author, date and country	Patient group	Study type (level of evidence)	Outcomes	Key results	Study Weaknesses
Olsson et al 2004 Sweden	11,751 consecutive non-surgical patients presenting to the ED	Prospective cohort	In-hospital mortality	AUROC REMS: 0.852 (SEM ± 0.014)	Single-centre study in a non-surgical ED. Data was 7-8 years old when published. No timing relating to mortality is given
Goodacre et al 2006 UK	5583 medical patients either admitted to hospital or died in ED	Secondary analysis	In-hospital mortality	AUROC REMS: 0.74 (95% CI 0.70-0.78)	Large amount of missing data (REMS only calculated in 39.7% of patients). Not all emergency medical admissions were included. Ambulance data rather than first ED recording
Howell et al 2007 USA	2132 ED patients with an diagnosis of infection or possible infection	Prospective cohort	28-day in-hospital mortality	AUROC REMS: 0.802 (95% CI 0.752-0.852)	Only patients with infection were included. Missing or inaccurate admission data. Modified version of REMS
Cattermole et al 2009 Hong Kong	330 consecutive adult resuscitation room patients	Prospective observational	30-day mortality	AUROC MEWS: 0.754 (95% CI 0.703-0.799) AUROC REMS: 0.696 (95% CI 0.722-0.816)	Secondary outcome (30-day mortality)used. No use of split sample technique for a derivation study. Only resuscitation room patients used. Data collection from patient documentation. No mention of availability of data given. Small numbers. Patients identified over a single month
Burch et al 2008 South Africa	790 medical patients presenting to ED	Prospective observational	In-hospital mortality	MEWS 3-4: Risk ratio 2.8 (95% CI 1.7-4.8) MEWS ≥5: Risk ratio 4.6 (95% CI 2.7-7.8)	Only medical patients included. Large amount of missing data (data was only collected for 70.2% of patients). Timing of the MEWS score is not clear. In-hospital mortality is not clearly defined
Vorwerk et al 2009 UK	307 ED patients with sepsis. Divided into low and high risk based on MEWS score (≥5)	Retrospective cohort	28 day mortality	AUROC MEWS: 0.72 (95% CI 0.67-0.77)	Data from patient notes. Patients may have been missed due to incorrect ED coding. Timing of MEWS unclear. Restricted cohort.
Groarke et al 2008 Ireland	225 consecutive medical admissions, divided into four score categories based on their EWS (0-1, 2-3, 4-5, >5)	Prospective cohort	In-hospital mortality	Estimated OR for each rise in score category 2.19 (95% CI 1.41-3.39)	Small numbers. Only includes medical admissions
Vorwerk et al 2007 UK	1024 ED patients who required hospital admission	Prospective cohort	28 day mortality	AUROC MEWS: 0.8	No analysis of age or co-morbidity. No mention of how many patients were removed from the results. Only two weeks of data is used. Confidence intervals are not given for the AUROC.
Armagan et al 2008 Turkey	309 patients presenting to ED, split into high risk (mEWS >4) and low risk (mEWS≤4	Prospective cohort	Death in hospital	OR 14.81 (95% CI 5.52-39.73)	Small numbers. Multiple exclusions

Comment(s)

Both MEWS and REMS can predict mortality in adult ED patients. Area under ROC curve was, on the whole, better for REMS than MEWS (0.74-0.911 compared with 0.67-0.8) If it is generally accepted that an AUROC of ≥0.8 has good predictive value, then REMS reaches this target in 3 papers and MEWS in only one. However in the only paper to directly compare the two scoring systems, AUROC for 30-day mortality was almost the same: REMS 0.771 and MEWS 0.75413. REMS is a score derived from ED patients; it was compiled using a split-sample method which is the method generally accepted by the medical community as standard. It was then validated in a further study of almost twelve thousand patients by the same authors, much greater numbers than the work involving MEWS Ideally REMS needs to be validated in a multi-centre UK ED population. The future probably lies in a modified version of REMS that is appropriate to the local population and a method of calculating it more easily.

Clinical Bottom Line

It must be remembered that critically ill patients may have a low early warning score and that EWS should be used alongside, rather than as a replacement for senior clinical expertise

References

1. Olsson T, Lind L Comparison of the Rapid Emergency Medicine Score and APACHE II in nonsurgical emergency department patients Acad Emerg Med 2003;10:1040-1048
2. Olsson T, Terent A, Lind L Rapid emergency medicine score: a new prognostic tool for in-hospital mortality in nonsurgical emergency department patients. Journal of Internal Medicine 2004;255:579-587
3. Goodacre S, Turner J, Nicholl J Prediction of mortality among emergency medical admissions Emerg Med J 2006;23:372-375
4. Howell MD, Donnino MW, Talmor D, Clardy P, Ngo L, Shapiro NI Performance of severity of illness scoring systems in emergency department patients with infection Acad Emerg Med 2007;14(8);709-714
5. Cattermole GN, Mak SKP, Liow CHE, Ho MF, Hung KYG, Keung KM, Li HM, Graham CA, Rainer TH Derivation of a prognostic score for identifying critically ill patients in an emergency department resuscitation room Resuscitation 2009;80:1000-1005
6. Burch VC, Tarr G, Morroni C Modified early warning score predicts the need for hospital admission and inhospital mortality. Emerg Med J 2008;25:674-678
7. Vorwerk C, Loryman B, Coats TJ, Stephenson JA, Gray LD, Reddy G, Florence L, Butler N Prediction of mortality in adult emergency department patients with sepsis Emerg Med J 2009;26:254-258
8. Groarke JD, Gallagher J, Stack J, Aftab A, Dwyer C, McGovern R, Courtney G Use of an admission early warning score to predict patient morbidity and mortality and treatment success Emerg Med J 2008;25:803-806
9. Vorwerk C, Gray L, Florence L, Fudge T, Goss C, Coats T Modified early warning (MEW) score predicts in-hospital mortality in unselected emergency department patients Emerg Med J 2007;24(Suppl 1):A4
10. Armagan E, Yilmaz Y, Fatih Olmez O, Simsek G, Bulen Gul C Predictive value of the modified Early Warning Score in a Turkish emergency department European Journal of Emergency Medicine 2008;15:338-340