- Email: [email protected]
A single-centre observational cohort study of admission National Early Warning Score (NEWS)
Resuscitation 92 (2015) 89–93
Contents lists available at ScienceDirect
Resuscitation journal homepage: www.elsevier.com/locate/resuscitation
Rapid Response Systems
A single-centre observational cohort study of admission National Early Warning Score (NEWS)夽 Tom E.F. Abbott a,∗ , Nidhi Vaid b , Dorothy Ip c , Nicholas Cron d , Matt Wells e , Hew D.T. Torrance a , Julian Emmanuel a,f a
Queen Mary University of London, EC1M 6BQ, UK Consultant in Acute Medicine, Northwick Park Hospital, Harrow HA1 1UJ, UK c Consultant in Acute Medicine, Whittington Health, N19 5NF, UK d London School of Economics, London WC2A 2AE, UK e Specialist Trainee in Haematology, Cheltenham General Hospital, GL53 7AN, UK f Consultant in Acute and Metabolic Medicine, Barts Health NHS Trust, E1 1BB, UK b
a r t i c l e
i n f o
Article history: Received 28 November 2014 Received in revised form 10 March 2015 Accepted 15 April 2015 Keywords: Early warning score Monitoring Clinical outcomes Physiological parameters
a b s t r a c t Introduction: Early warning scores are commonly used in hospitals to identify patients at risk of deterioration. The National Early Warning Score (NEWS) has recently been introduced to UK practice. However, it is not yet widely implemented. We aimed to compare NEWS to the early warning score currently used in our hospital – the Patient at Risk Score (PARS). Methods: We conducted a prospective observational cohort study of all adult general medical patients admitted to a single hospital over a 20-day period. Physiological data and early warning scores recorded in bedside charts were collected on admission and a NEWS score was retrospectively calculated. The patient notes were reviewed at 48 h after admission. The primary outcome was a composite of critical care admission or death within 2 days of admission. The secondary outcome was hospital length of stay. Results: NEWS was more strongly associated with the primary outcome than PARS (odds ratio 1.54, p < 0.001 compared to 1.42, p = 0.056). A NEWS of 3 or more was associated with the primary outcome (odds ratio 7.03, p = 0.003). Neither score was correlated with hospital length of stay. Conclusion: NEWS on admission is superior to PARS for identifying patients at risk of death or critical care admission within the first 2 days of hospital stay. Current guidelines advocate a threshold of 5 for triggering a clinical review. However, since a score of 3 or more was associated with a poor outcome, this recommendation should be reviewed. Both scores were poor predictors of hospital length of stay. © 2015 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Hospitals commonly use early warning scores (EWS) or ‘track and trigger’ systems to identify patients at risk of deterioration and to prompt a clinical review.1 EWS are central to the concept of ‘critical care without walls’, delivered by nurse-led critical care outreach teams or physician-led medical response teams.2 There are many types of EWS and a variety are used in the United Kingdom.1 Some scoring systems are associated with patient outcome.3,4 However, many are poor predictors of survival and the impact on
夽 A Spanish translated version of the abstract of this article appears as Appendix in the final online version at http://dx.doi.org/10.1016/j.resuscitation.2015.04.020. ∗ Corresponding author. Critical Care Research Office, The Royal London Hospital, London E1 1BB, UK. E-mail address: [email protected] (T.E.F. Abbott). http://dx.doi.org/10.1016/j.resuscitation.2015.04.020 0300-9572/© 2015 Elsevier Ireland Ltd. All rights reserved.
hospital mortality is variable.5,6 In an attempt to standardise hospital EWS across the UK, the Royal College of Physicians introduced the ‘National Early Warning Score (NEWS)’ in 2012.7 NEWS assigns a score to each of the following physiological observations: level of consciousness, heart rate, systolic blood pressure, respiratory rate, temperature, oxygen saturation and the use of supplemental oxygen therapy (Table 1).8 An increasing score suggests worsening physiological derangement. Three trigger levels are recommended: 1–4, 5–6 (or a score of 3 in any one parameter) and 7 or more. It is intended that each level should attract an increasing degree of medical response, from ward nurse to critical care outreach team review.8 Several studies have already evaluated the NEWS.5,9–12 Current evidence from studies of acute medical patients suggests NEWS is more strongly associated with cardiac arrest, unexpected Intensive Care Unit (ICU) admission or death within 24 h of NEWS being calculated compared to other scoring systems.5,12 Similar findings
90
T.E.F. Abbott et al. / Resuscitation 92 (2015) 89–93
Table 1 Comparison of Patient at Risk Score (PARS) and National Early Warning Score (NEWS). 3 PARS Temperature (◦ C) Heart rate (beats/min) Systolic BP (mmHg) Respiratory rate (breaths/min) Oxygen saturation (%) CNS response (AVPU) Urine output (ml/kg/h) NEWS Temperature (◦ C) Heart rate (beats/min) Systolic BP (mmHg) Respiratory rate (breaths/min) Oxygen saturation (%) Supplemental oxygen CNS response (AVPU)
<40 <70
2
1
0
1
2
<35.0
35–35.9 40–49 80–89
36–37.4 50–99 100–179 10–19 >95 A 0.5–3
37.5–38.4 100–114
>38.5 115–129 >180 30–39
V >3
P
36.1–38.0 51–90 111–219 12–20 >96 No A
38.1–39.0 91–110
>39.0 111–130
<85
70–79 <10 85–89
Nil
<0.5
<35.0 <41 <91 <9 <92
91–100 92–93 Yes
90–94 Confused Dialysis 35.1–36.0 41–50 101–110 9–11 94–95
20–29
21–24
3
>130 >40 U
>130 >219 >25
V, P, U
Each parameter is graded 0–3 for both PARS and NEWS. Scores for each parameter are added together to give a total. For PARS, scores of 3 and 5 trigger reviews by the medical team or critical care outreach team respectively. For NEWS, scores of greater than 5 (or 3 in any one parameter) trigger an urgent medical review. A score of over 7 triggers a review by a critical care outreach team, medical response team or similar.8
are reported for septic patients in the emergency department and oncology ward inpatients.9,10 However, the current evidence has several limitations. Firstly, NEWS has only been evaluated using electronic patient observation databases. Since many hospitals continue to record observations on paper bedside charts, conclusions derived from electronic database studies may not be generalisable. Secondly, the majority of studies have evaluated the recommended thresholds for triggering a clinician response. However, alternative trigger thresholds may be superior.10 Finally, studies of other scoring systems have identified association between EWS on admission and hospital length of stay, but this has not yet been investigated for NEWS.6 This is important because the ability to predict hospital length of stay on admission could improve resource allocation and reduce bed pressures. We aimed to compare NEWS to the Patient at Risk Score (PARS), the existing early warning score in use at our hospital (Table 1). We chose two outcome measures: a composite of mortality and critical care unit (level two or three care) escalation within the first 48 h of the admission, and hospital length of stay. This composite outcome measure has previously been used in similar studies and will capture all patients suffering cardiac arrest at our institution.9,10,13 We hypothesise that admission NEWS is more strongly associated with the primary outcome compared to PARS and that admission NEWS is associated with hospital length of stay. 2. Method
(multi-organ support, renal replacement therapy, advanced respiratory support). Research ethics approval was provided by the National Research Ethics Service (12/LO/1985). 2.2. Data collection The patient notes and bedside observation charts were reviewed by researchers during the first 48 h of the hospital stay. The first physiological measurements and early warning score after admission to the Acute Assessment Unit (AAU) were recorded on paper data collection forms and subsequently entered into a computer database. These data were considered missing if they were not available during the first 24 h after admission. A random sample of ∼20% of the electronic database was independently validated against the paper data collection forms. The primary and secondary outcome measures were assessed by reviewing electronic patient records, discharge summaries and patient notes. 2.3. Calculation of early warning score Microsoft Excel (Microsoft Inc., Redmond, WA) was used to calculate PARS and NEWS for each set of physiological measurements available. The computer-calculated PARS was compared to the PARS recorded on the bedside observation chart to measure accuracy of EWS recording by healthcare workers.
2.1. Study design
2.4. Statistical analysis
This was a prospective observational cohort study of all adult patients admitted to the Acute Assessment Unit at a large London teaching hospital between 25th March and 13th April 2013. The Royal London Hospital is a busy teaching hospital serving a diverse population in London’s east end. The Acute Assessment Unit is the receiving area for all new adult medical admissions, excluding patients admitted directly to the critical care unit. Bedside observations including respiratory rate, oxygen saturation, temperature, blood pressure, heart rate and consciousness level are recorded on a paper chart by nursing staff at admission and throughout the hospital stay. The bedside nurse or healthcare assistant uses this information to calculate an early warning score. At our institution critical care comprises level two care (high frequency nursing care, single organ support, invasive monitoring, ‘step down’ from higher level care) and level three care
We used SPSS version 21 (IBM, Armonk, NY) to analyse the data. Statistical analysis was performed on computer-calculated scores, in order to provide a fair comparison between NEWS and PARS. The early warning scores were first considered as continuous variables. Multivariable logistic regression analysis assessed for association between the independent variables and the primary endpoint, corrected for age and gender. We chose minimal covariates to reflect the pragmatic nature of this study and to facilitate comparison with other research in this field.9,10 The sample was dichotomised according to early warning score and logistic regression analysis used to test the association of different thresholds with the primary endpoint. Linear regression was used to test association between early warning scores and hospital length of stay. Logistic regression analysis was used to test association between EWS and hospital stay of less than and greater than 7 and 14 days.
T.E.F. Abbott et al. / Resuscitation 92 (2015) 89–93
91
Table 3 Descriptive data stratified by correct or incorrect admission PARS.
Total cases Mean age in years (s.d.) Female gender Weekend cases Weekday cases Mortality and critical care admission Length of stay in days, median (IQR) PARS under-scored PARS over-scored
Correct PARS
Incorrect PARS
73 61 (22.4) 53 33 67 2.6 4 (7) – –
27 60 (22.3) 54 32 68 4.3 5 (7) 76 24
Results given as percentages of each group, unless stated otherwise.
Table 4 Association between early warning score and primary outcome measure (critical care admission or death within 48 h). Odds ratio Fig. 1. A flow diagram of study recruitment and analysis.
3. Results Four hundred and fifty-three patients were included in the study. Eight patients were excluded from the primary analysis because a PARS was not recorded. A further 14 patients were excluded from the secondary analysis because length of stay was not available (Fig. 1). Sixteen patients (3.5%) met the primary outcome. The median length of stay was 4 (IQR 7) days. The mean age was 60.9 (s.d. 22.4) years and 53.5% of the sample were female. Table 2 outlines the diagnosis groups admitted for the entire cohort and for patients with NEWS of three of more. Early warning scores recorded on the bedside observation chart were different from computer-calculated scores in 116 cases (27.4%). Table 3 shows descriptive data for groups with correctly or incorrectly calculated PARS. Table 4 presents the results of logistic regression analysis of early warning score at admission, as a continuous variable, against the composite endpoint of death or admission to critical care as the dependent variable. NEWS is more strongly associated with this outcome measure than PARS. The odds ratio for NEWS is 1.54 (95% CI 1.26–1.91, p < 0.001) compared to 1.42 (95% CI 1.00–2.05, p = 0.056) for PARS. Analysis of individual NEWS thresholds identified that a score of 3 or more is associated with the composite
Table 2 Diagnosis groups and demographics for all patients and for patients with NEWS ≥3. Entire cohort
NEWS ≥3
Age (years) Female
60.9 (±22.4) 242 (53.5)
61.7 (±23.5) 88 (52.6)
Post-take diagnosis category General Medical Respiratory Cardiology Health care of the elderly Gastroenterology Neurological Haematology Endocrinology Psychiatry Oncology Surgery Rheumatology Nephrology Infection and immunology Dermatology Gynaecology
114 (25.2) 71 (15.7) 54 (11.9) 54 (11.9) 35 (7.7) 30 (6.6) 30 (6.6) 16 (3.5) 13 (2.9) 12 (2.7) 10 (2.2) 5 (1.1) 4 (0.9) 3 (0.7) 1 (0.2) 1 (0.2)
32 (19.2) 45 (26.9) 18 (10.8) 27 (16.2) 11 (6.6) 6 (3.6) 12 (7.2) – 2 (1.2) 11 (6.6) 1 (0.6) 1 (0.6) – – 1 (0.6) –
Values are presented as mean (± standard deviation) or n (%).
p-value
Univariable logistic regression without adjustment for covariates 1.54 NEWS 1.43 PARS
<0.001 0.051
Multivariable logistic regression with adjustment for age and gender NEWS 1.54 Age 1.01 1.23 Gender
<0.001 0.664 0.699
PARS Age Gender
1.42 1.01 1.36
0.056 0.559 0.566
Table 5 Association between NEWS and primary outcome measure, showing NEWS trigger thresholds from 1 to 7. NEWS threshold
Odds ratio
Multivariable logistic regression adjusted for age and gender 3.23 >1 >2 7.03 >3 8.12 >4 6.36 >5 6.02 11.66 >6 >7 15.11
p-value 0.073 0.003 <0.001 0.001 0.002 <0.001 <0.001
outcome measure (Table 5). Since PARS is not significantly associated with the primary outcome, further analysis of individual PARS thresholds was not performed. We did not identify significant associations between admission early warning score and hospital length of stay using linear regression (r2 = 0.037 (p = 0.44) for NEWS and r2 = 0.002 (p = 0.97) for PARS) or using logistic regression (p-values 0.153–0.999). 4. Discussion These results support the hypothesis that NEWS is superior to the PARS currently used in our hospital for identifying patients at risk of critical care admission or death. Patients with a NEWS of three or more were more likely to meet the primary endpoint; every one-point increase in NEWS was associated with a 55% increased risk. This is consistent with other published work in undifferentiated medical patients, patients with sepsis in the emergency department and oncology inpatients.5,9–12 We did not identify association between NEWS on admission and hospital length of stay. There is some evidence to support such a relationship for other EWS. However, a recent systematic review reports inconsistent results from a small number of studies. 6,13–15 Further research using data from a larger number of patients is recommended.
92
T.E.F. Abbott et al. / Resuscitation 92 (2015) 89–93
Most of the studies of NEWS have adopted the recommended trigger thresholds. The largest of these used a database of electronically recorded physiological measurements from ∼36,000 separate patient episodes at a single district general hospital. The authors found that NEWS was more strongly associated with cardiac arrest, death or ICU admission, compared to 33 other scoring systems.5 Only one study has attempted to re-examine the NEWS trigger thresholds. Corfield et al. examined whether NEWS was associated with death within 30 days, or ICU admission within 2 days, in septic patients. The authors stratified their sample according to the recommended trigger thresholds and used logistic regression analysis to compare patients with scores of 5–6, 7–8 and 9–20 to those in the lowest risk group (0–4). In contrast with our results, which demonstrate that a NEWS of three or more is associated with critical care admission or death, they report that a NEWS of greater than six was associated with ICU admission.10 This may reflect differences in statistical methods, since instead of comparing the recommended trigger groups, we dichotomised the sample for each NEWS value and compared the outcome of patients above and below that score. There may also be differences in the definition of the primary endpoint. We defined critical care as either level 2 or level 3 care, i.e. admission to the Intensive Care Unit or High Dependency Unit. However, it is unclear if Corfield et al. used the same definition or only included level 2 care. The latter may, in part, explain the lower NEWS threshold that we identified. We electronically calculated early warning scores from physiological data recorded on bedside charts and found that the calculated score differed to the score (PARS) recorded on the patient observation chart in ∼27% of cases. Our data suggest that incorrect PARS were more likely to be under-scored than over-scored. There were no patterns in the distribution of incorrect scoring and hospital lengths of stay were not impacted. Mortality was higher in the incorrect PARS group (Table 3). However, we suspect this is an artefact of the small number of cases, since sensitivity analyses suggest the results were not confounded by incorrect PARS calculation. In this study multiple observers did not independently review source data, so we must acknowledge the potential for transcription error during data collection. However, our observation is unlikely to be an isolated occurrence since other studies report similar error rates.16,17 The difference in these error rates may be explained by variations in the scoring systems studied, which could be due to the complexity of the scoring system or the number of physiological parameters used.17 Or, it may be a local phenomenon. Since our study was conducted at a single-site, we were unable to perform further sub-group analyses based on location. We suggest two explanations for the observed error. Firstly, mistakes could arise when interpreting the thresholds for each physiological parameter or when adding up the total score. Secondly, if the physiological measurements are not recorded directly on to the observation chart by clinical staff, transcription error may occur. Human error should be taken into account when designing the complexity of these scoring systems. However, this is likely to become less of a problem in the future as electronic observation charts become more common.18 Our study has several limitations. Firstly, only 16 patients reached the composite primary outcome of critical care unit admission or death, representing a 3.5% event rate. Whilst our statistical methods were robust and the relationship between NEWS and the primary outcome measure was statistically significant, we are conscious that statistical significance does not equate to clinical significance. Secondly, our study sampled medical patients with undifferentiated presenting complaints. However, we included patients from only one hospital, which could limit the generalisability of our results. Our sample appears heterogeneous, but there may be underlying characteristics common to this cohort that we are not aware of, which may bias the results. We performed
a sensitivity analysis excluding patients that received palliative treatment but this did not change our results and we are not aware of any organisational factors during the study period that could have skewed our results, for example bed availability. Thirdly, we included all patients admitted to the acute assessment unit during a fixed period of time to reduce selection bias. However, we did not include patients from the emergency department who were not admitted to the medical service or patients that were admitted directly to the critical care unit. This in itself is a form of selection bias, which will have restricted our sample to patients that were unwell enough to be admitted to hospital but not so unwell that they required level two or level three care from the outset. From a pragmatic point of view this form of selection is quite sensible since we tested early warning scores in patients on medical wards – the very cohort that should be targeted by these tools. However, the unwitting exclusion of critical care admission directly from the emergency department may explain the lower than expected event rate. The Royal College of Physicians recommends that: a NEWS of 1–4 should prompt a review by a registered nurse, a NEWS of 5–6 should prompt a review by a ward doctor or acute nurse, and a NEWS of 7 or more should prompt a review by a critical care outreach team.8 Our analysis did not examine the response to an abnormal EWS, either in terms of type or speed of clinical review. However, our results suggest that patients with a NEWS greater than two were more likely to die or be admitted to level two or level three care within 2 days. The threshold for triggering a review by a doctor could be revised to a score of three; we recommend that this be considered when NEWS is next reviewed or refined. However, this would require careful balance against the likely increase in resource requirements for hospitals as more patients trigger a medical review. Our institution will be introducing NEWS in the near future. Further research examining the type and speed of clinical response to abnormal EWS is needed. 5. Conclusion Track and trigger scores are an integral part of ward-based hospital care in the UK. This study sought to evaluate the National Early Warning Score in a single hospital. Our results suggest that NEWS on admission is associated with death or critical care admission within the first 2 days of hospital stay. However, we did not identify association between NEWS on admission and hospital length of stay. There may be benefit in lowering the trigger point of the medium risk category from 5 to 3, as a score of 3 was associated with the primary endpoint. However, this requires further evaluation in larger, multi-centre datasets. Conflict of Interest statement There are no conflicts of interest. References 1. Patterson C, Maclean F, Bell C, Mukherjee E, Bryan L, Woodcock T, et al. Early warning systems in the UK: variation in content and implementation strategy has implications for a NHS early warning system. Clin Med 2011;11:424–7. 2. McGinley A, Pearse RM. A national early warning score for acutely ill patients. Br Med J 2012;345:e5310. 3. Bleyer AJ, Vidya S, Russell GB, Jones CM, Sujata L, Daeihagh P, et al. Longitudinal analysis of one million vital signs in patients in an academic medical center. Resuscitation 2011;82:1387–92. 4. Cei M, Bartolomei C, Mumoli N. In-hospital mortality and morbidity of elderly medical patients can be predicted at admission by the Modified Early Warning Score: a prospective study. Int J Clin Pract 2009;63:591–5. 5. Smith GB, Prytherch DR, Meredith P, Schmidt PE, Featherstone PI. The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and death. Resuscitation 2013;84:465–70.
T.E.F. Abbott et al. / Resuscitation 92 (2015) 89–93 6. Alam N, Hobbelink EL, van Tienhoven AJ, van de Ven PM, Jansma EP, Nanayakkara PW. The impact of the use of the Early Warning Score (EWS) on patient outcomes: a systematic review. Resuscitation 2014;85:587–94. 7. Jones M, NEWSDIG:. The National Early Warning Score Development and Implementation Group. Clin Med 2012;12:501–3. 8. Physicians RCo. National Early Warning Score (NEWS): standardising the assessment of acute- illness severity in the NHS. Report of a working party London 2012. 9. Cooksley T, Kitlowski E, Haji-Michael P. Effectiveness of Modified Early Warning Score in predicting outcomes in oncology patients. QJM: Monthly J Assoc Physicians 2012;105:1083–8. 10. Corfield AR, Lees F, Zealley I, Houston G, Dickie S, Ward K, et al. Utility of a single early warning score in patients with sepsis in the emergency department. Emergency Med J 2014;31:482–7. 11. Tirkkonen J, Olkkola KT, Huhtala H, Tenhunen J, Hoppu S. Medical emergency team activation: performance of conventional dichotomised criteria versus national early warning score. Acta Anaesthesiol Scand 2014;58:411–9. 12. Romero-Brufau S, Huddleston JM, Naessens JM, Johnson MG, Hickman J, Morlan BW, et al. Widely used track and trigger scores: are they ready for automation in practice? Resuscitation 2014;85:549–52.
93
13. Peris A, Zagli G, Maccarrone N, Batacchi S, Cammelli R, Cecchi A, et al. The use of Modified Early Warning Score may help anesthesists in postoperative level of care selection in emergency abdominal surgery. Minerva Anestesiol 2012;78:1034–8. 14. Paterson R, MacLeod DC, Thetford D, Beattie A, Graham C, Lam S, et al. Prediction of in-hospital mortality and length of stay using an early warning scoring system: clinical audit. Clin Med 2006;6:281–4. 15. Bokhari SW, Munir T, Memon S, Byrne JL, Russell NH, Beed M. Impact of critical care reconfiguration and track-and-trigger outreach team intervention on outcomes of haematology patients requiring intensive care admission. Ann Hematol 2010;89:505–12. 16. Kolic I, Crane S, McCartney S, Perkins Z, Taylor A. Factors affecting response to National Early Warning Score (NEWS). Resuscitation 2015. http://dx.doi.org/10.1016/j.resuscitation.2015.02.009 17. Subbe CP, Gao H, Harrison DA. Reproducibility of physiological track-and-trigger warning systems for identifying at-risk patients on the ward. Intensive Care Med 2007;33:619–24. 18. McNamee M. Electronic keypad replaces early warning score paper charts. Nursing Children Young People 2013;25:9.
Contents lists available at ScienceDirect
Resuscitation journal homepage: www.elsevier.com/locate/resuscitation
Rapid Response Systems
A single-centre observational cohort study of admission National Early Warning Score (NEWS)夽 Tom E.F. Abbott a,∗ , Nidhi Vaid b , Dorothy Ip c , Nicholas Cron d , Matt Wells e , Hew D.T. Torrance a , Julian Emmanuel a,f a
Queen Mary University of London, EC1M 6BQ, UK Consultant in Acute Medicine, Northwick Park Hospital, Harrow HA1 1UJ, UK c Consultant in Acute Medicine, Whittington Health, N19 5NF, UK d London School of Economics, London WC2A 2AE, UK e Specialist Trainee in Haematology, Cheltenham General Hospital, GL53 7AN, UK f Consultant in Acute and Metabolic Medicine, Barts Health NHS Trust, E1 1BB, UK b
a r t i c l e
i n f o
Article history: Received 28 November 2014 Received in revised form 10 March 2015 Accepted 15 April 2015 Keywords: Early warning score Monitoring Clinical outcomes Physiological parameters
a b s t r a c t Introduction: Early warning scores are commonly used in hospitals to identify patients at risk of deterioration. The National Early Warning Score (NEWS) has recently been introduced to UK practice. However, it is not yet widely implemented. We aimed to compare NEWS to the early warning score currently used in our hospital – the Patient at Risk Score (PARS). Methods: We conducted a prospective observational cohort study of all adult general medical patients admitted to a single hospital over a 20-day period. Physiological data and early warning scores recorded in bedside charts were collected on admission and a NEWS score was retrospectively calculated. The patient notes were reviewed at 48 h after admission. The primary outcome was a composite of critical care admission or death within 2 days of admission. The secondary outcome was hospital length of stay. Results: NEWS was more strongly associated with the primary outcome than PARS (odds ratio 1.54, p < 0.001 compared to 1.42, p = 0.056). A NEWS of 3 or more was associated with the primary outcome (odds ratio 7.03, p = 0.003). Neither score was correlated with hospital length of stay. Conclusion: NEWS on admission is superior to PARS for identifying patients at risk of death or critical care admission within the first 2 days of hospital stay. Current guidelines advocate a threshold of 5 for triggering a clinical review. However, since a score of 3 or more was associated with a poor outcome, this recommendation should be reviewed. Both scores were poor predictors of hospital length of stay. © 2015 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Hospitals commonly use early warning scores (EWS) or ‘track and trigger’ systems to identify patients at risk of deterioration and to prompt a clinical review.1 EWS are central to the concept of ‘critical care without walls’, delivered by nurse-led critical care outreach teams or physician-led medical response teams.2 There are many types of EWS and a variety are used in the United Kingdom.1 Some scoring systems are associated with patient outcome.3,4 However, many are poor predictors of survival and the impact on
夽 A Spanish translated version of the abstract of this article appears as Appendix in the final online version at http://dx.doi.org/10.1016/j.resuscitation.2015.04.020. ∗ Corresponding author. Critical Care Research Office, The Royal London Hospital, London E1 1BB, UK. E-mail address: [email protected] (T.E.F. Abbott). http://dx.doi.org/10.1016/j.resuscitation.2015.04.020 0300-9572/© 2015 Elsevier Ireland Ltd. All rights reserved.
hospital mortality is variable.5,6 In an attempt to standardise hospital EWS across the UK, the Royal College of Physicians introduced the ‘National Early Warning Score (NEWS)’ in 2012.7 NEWS assigns a score to each of the following physiological observations: level of consciousness, heart rate, systolic blood pressure, respiratory rate, temperature, oxygen saturation and the use of supplemental oxygen therapy (Table 1).8 An increasing score suggests worsening physiological derangement. Three trigger levels are recommended: 1–4, 5–6 (or a score of 3 in any one parameter) and 7 or more. It is intended that each level should attract an increasing degree of medical response, from ward nurse to critical care outreach team review.8 Several studies have already evaluated the NEWS.5,9–12 Current evidence from studies of acute medical patients suggests NEWS is more strongly associated with cardiac arrest, unexpected Intensive Care Unit (ICU) admission or death within 24 h of NEWS being calculated compared to other scoring systems.5,12 Similar findings
90
T.E.F. Abbott et al. / Resuscitation 92 (2015) 89–93
Table 1 Comparison of Patient at Risk Score (PARS) and National Early Warning Score (NEWS). 3 PARS Temperature (◦ C) Heart rate (beats/min) Systolic BP (mmHg) Respiratory rate (breaths/min) Oxygen saturation (%) CNS response (AVPU) Urine output (ml/kg/h) NEWS Temperature (◦ C) Heart rate (beats/min) Systolic BP (mmHg) Respiratory rate (breaths/min) Oxygen saturation (%) Supplemental oxygen CNS response (AVPU)
<40 <70
2
1
0
1
2
<35.0
35–35.9 40–49 80–89
36–37.4 50–99 100–179 10–19 >95 A 0.5–3
37.5–38.4 100–114
>38.5 115–129 >180 30–39
V >3
P
36.1–38.0 51–90 111–219 12–20 >96 No A
38.1–39.0 91–110
>39.0 111–130
<85
70–79 <10 85–89
Nil
<0.5
<35.0 <41 <91 <9 <92
91–100 92–93 Yes
90–94 Confused Dialysis 35.1–36.0 41–50 101–110 9–11 94–95
20–29
21–24
3
>130 >40 U
>130 >219 >25
V, P, U
Each parameter is graded 0–3 for both PARS and NEWS. Scores for each parameter are added together to give a total. For PARS, scores of 3 and 5 trigger reviews by the medical team or critical care outreach team respectively. For NEWS, scores of greater than 5 (or 3 in any one parameter) trigger an urgent medical review. A score of over 7 triggers a review by a critical care outreach team, medical response team or similar.8
are reported for septic patients in the emergency department and oncology ward inpatients.9,10 However, the current evidence has several limitations. Firstly, NEWS has only been evaluated using electronic patient observation databases. Since many hospitals continue to record observations on paper bedside charts, conclusions derived from electronic database studies may not be generalisable. Secondly, the majority of studies have evaluated the recommended thresholds for triggering a clinician response. However, alternative trigger thresholds may be superior.10 Finally, studies of other scoring systems have identified association between EWS on admission and hospital length of stay, but this has not yet been investigated for NEWS.6 This is important because the ability to predict hospital length of stay on admission could improve resource allocation and reduce bed pressures. We aimed to compare NEWS to the Patient at Risk Score (PARS), the existing early warning score in use at our hospital (Table 1). We chose two outcome measures: a composite of mortality and critical care unit (level two or three care) escalation within the first 48 h of the admission, and hospital length of stay. This composite outcome measure has previously been used in similar studies and will capture all patients suffering cardiac arrest at our institution.9,10,13 We hypothesise that admission NEWS is more strongly associated with the primary outcome compared to PARS and that admission NEWS is associated with hospital length of stay. 2. Method
(multi-organ support, renal replacement therapy, advanced respiratory support). Research ethics approval was provided by the National Research Ethics Service (12/LO/1985). 2.2. Data collection The patient notes and bedside observation charts were reviewed by researchers during the first 48 h of the hospital stay. The first physiological measurements and early warning score after admission to the Acute Assessment Unit (AAU) were recorded on paper data collection forms and subsequently entered into a computer database. These data were considered missing if they were not available during the first 24 h after admission. A random sample of ∼20% of the electronic database was independently validated against the paper data collection forms. The primary and secondary outcome measures were assessed by reviewing electronic patient records, discharge summaries and patient notes. 2.3. Calculation of early warning score Microsoft Excel (Microsoft Inc., Redmond, WA) was used to calculate PARS and NEWS for each set of physiological measurements available. The computer-calculated PARS was compared to the PARS recorded on the bedside observation chart to measure accuracy of EWS recording by healthcare workers.
2.1. Study design
2.4. Statistical analysis
This was a prospective observational cohort study of all adult patients admitted to the Acute Assessment Unit at a large London teaching hospital between 25th March and 13th April 2013. The Royal London Hospital is a busy teaching hospital serving a diverse population in London’s east end. The Acute Assessment Unit is the receiving area for all new adult medical admissions, excluding patients admitted directly to the critical care unit. Bedside observations including respiratory rate, oxygen saturation, temperature, blood pressure, heart rate and consciousness level are recorded on a paper chart by nursing staff at admission and throughout the hospital stay. The bedside nurse or healthcare assistant uses this information to calculate an early warning score. At our institution critical care comprises level two care (high frequency nursing care, single organ support, invasive monitoring, ‘step down’ from higher level care) and level three care
We used SPSS version 21 (IBM, Armonk, NY) to analyse the data. Statistical analysis was performed on computer-calculated scores, in order to provide a fair comparison between NEWS and PARS. The early warning scores were first considered as continuous variables. Multivariable logistic regression analysis assessed for association between the independent variables and the primary endpoint, corrected for age and gender. We chose minimal covariates to reflect the pragmatic nature of this study and to facilitate comparison with other research in this field.9,10 The sample was dichotomised according to early warning score and logistic regression analysis used to test the association of different thresholds with the primary endpoint. Linear regression was used to test association between early warning scores and hospital length of stay. Logistic regression analysis was used to test association between EWS and hospital stay of less than and greater than 7 and 14 days.
T.E.F. Abbott et al. / Resuscitation 92 (2015) 89–93
91
Table 3 Descriptive data stratified by correct or incorrect admission PARS.
Total cases Mean age in years (s.d.) Female gender Weekend cases Weekday cases Mortality and critical care admission Length of stay in days, median (IQR) PARS under-scored PARS over-scored
Correct PARS
Incorrect PARS
73 61 (22.4) 53 33 67 2.6 4 (7) – –
27 60 (22.3) 54 32 68 4.3 5 (7) 76 24
Results given as percentages of each group, unless stated otherwise.
Table 4 Association between early warning score and primary outcome measure (critical care admission or death within 48 h). Odds ratio Fig. 1. A flow diagram of study recruitment and analysis.
3. Results Four hundred and fifty-three patients were included in the study. Eight patients were excluded from the primary analysis because a PARS was not recorded. A further 14 patients were excluded from the secondary analysis because length of stay was not available (Fig. 1). Sixteen patients (3.5%) met the primary outcome. The median length of stay was 4 (IQR 7) days. The mean age was 60.9 (s.d. 22.4) years and 53.5% of the sample were female. Table 2 outlines the diagnosis groups admitted for the entire cohort and for patients with NEWS of three of more. Early warning scores recorded on the bedside observation chart were different from computer-calculated scores in 116 cases (27.4%). Table 3 shows descriptive data for groups with correctly or incorrectly calculated PARS. Table 4 presents the results of logistic regression analysis of early warning score at admission, as a continuous variable, against the composite endpoint of death or admission to critical care as the dependent variable. NEWS is more strongly associated with this outcome measure than PARS. The odds ratio for NEWS is 1.54 (95% CI 1.26–1.91, p < 0.001) compared to 1.42 (95% CI 1.00–2.05, p = 0.056) for PARS. Analysis of individual NEWS thresholds identified that a score of 3 or more is associated with the composite
Table 2 Diagnosis groups and demographics for all patients and for patients with NEWS ≥3. Entire cohort
NEWS ≥3
Age (years) Female
60.9 (±22.4) 242 (53.5)
61.7 (±23.5) 88 (52.6)
Post-take diagnosis category General Medical Respiratory Cardiology Health care of the elderly Gastroenterology Neurological Haematology Endocrinology Psychiatry Oncology Surgery Rheumatology Nephrology Infection and immunology Dermatology Gynaecology
114 (25.2) 71 (15.7) 54 (11.9) 54 (11.9) 35 (7.7) 30 (6.6) 30 (6.6) 16 (3.5) 13 (2.9) 12 (2.7) 10 (2.2) 5 (1.1) 4 (0.9) 3 (0.7) 1 (0.2) 1 (0.2)
32 (19.2) 45 (26.9) 18 (10.8) 27 (16.2) 11 (6.6) 6 (3.6) 12 (7.2) – 2 (1.2) 11 (6.6) 1 (0.6) 1 (0.6) – – 1 (0.6) –
Values are presented as mean (± standard deviation) or n (%).
p-value
Univariable logistic regression without adjustment for covariates 1.54 NEWS 1.43 PARS
<0.001 0.051
Multivariable logistic regression with adjustment for age and gender NEWS 1.54 Age 1.01 1.23 Gender
<0.001 0.664 0.699
PARS Age Gender
1.42 1.01 1.36
0.056 0.559 0.566
Table 5 Association between NEWS and primary outcome measure, showing NEWS trigger thresholds from 1 to 7. NEWS threshold
Odds ratio
Multivariable logistic regression adjusted for age and gender 3.23 >1 >2 7.03 >3 8.12 >4 6.36 >5 6.02 11.66 >6 >7 15.11
p-value 0.073 0.003 <0.001 0.001 0.002 <0.001 <0.001
outcome measure (Table 5). Since PARS is not significantly associated with the primary outcome, further analysis of individual PARS thresholds was not performed. We did not identify significant associations between admission early warning score and hospital length of stay using linear regression (r2 = 0.037 (p = 0.44) for NEWS and r2 = 0.002 (p = 0.97) for PARS) or using logistic regression (p-values 0.153–0.999). 4. Discussion These results support the hypothesis that NEWS is superior to the PARS currently used in our hospital for identifying patients at risk of critical care admission or death. Patients with a NEWS of three or more were more likely to meet the primary endpoint; every one-point increase in NEWS was associated with a 55% increased risk. This is consistent with other published work in undifferentiated medical patients, patients with sepsis in the emergency department and oncology inpatients.5,9–12 We did not identify association between NEWS on admission and hospital length of stay. There is some evidence to support such a relationship for other EWS. However, a recent systematic review reports inconsistent results from a small number of studies. 6,13–15 Further research using data from a larger number of patients is recommended.
92
T.E.F. Abbott et al. / Resuscitation 92 (2015) 89–93
Most of the studies of NEWS have adopted the recommended trigger thresholds. The largest of these used a database of electronically recorded physiological measurements from ∼36,000 separate patient episodes at a single district general hospital. The authors found that NEWS was more strongly associated with cardiac arrest, death or ICU admission, compared to 33 other scoring systems.5 Only one study has attempted to re-examine the NEWS trigger thresholds. Corfield et al. examined whether NEWS was associated with death within 30 days, or ICU admission within 2 days, in septic patients. The authors stratified their sample according to the recommended trigger thresholds and used logistic regression analysis to compare patients with scores of 5–6, 7–8 and 9–20 to those in the lowest risk group (0–4). In contrast with our results, which demonstrate that a NEWS of three or more is associated with critical care admission or death, they report that a NEWS of greater than six was associated with ICU admission.10 This may reflect differences in statistical methods, since instead of comparing the recommended trigger groups, we dichotomised the sample for each NEWS value and compared the outcome of patients above and below that score. There may also be differences in the definition of the primary endpoint. We defined critical care as either level 2 or level 3 care, i.e. admission to the Intensive Care Unit or High Dependency Unit. However, it is unclear if Corfield et al. used the same definition or only included level 2 care. The latter may, in part, explain the lower NEWS threshold that we identified. We electronically calculated early warning scores from physiological data recorded on bedside charts and found that the calculated score differed to the score (PARS) recorded on the patient observation chart in ∼27% of cases. Our data suggest that incorrect PARS were more likely to be under-scored than over-scored. There were no patterns in the distribution of incorrect scoring and hospital lengths of stay were not impacted. Mortality was higher in the incorrect PARS group (Table 3). However, we suspect this is an artefact of the small number of cases, since sensitivity analyses suggest the results were not confounded by incorrect PARS calculation. In this study multiple observers did not independently review source data, so we must acknowledge the potential for transcription error during data collection. However, our observation is unlikely to be an isolated occurrence since other studies report similar error rates.16,17 The difference in these error rates may be explained by variations in the scoring systems studied, which could be due to the complexity of the scoring system or the number of physiological parameters used.17 Or, it may be a local phenomenon. Since our study was conducted at a single-site, we were unable to perform further sub-group analyses based on location. We suggest two explanations for the observed error. Firstly, mistakes could arise when interpreting the thresholds for each physiological parameter or when adding up the total score. Secondly, if the physiological measurements are not recorded directly on to the observation chart by clinical staff, transcription error may occur. Human error should be taken into account when designing the complexity of these scoring systems. However, this is likely to become less of a problem in the future as electronic observation charts become more common.18 Our study has several limitations. Firstly, only 16 patients reached the composite primary outcome of critical care unit admission or death, representing a 3.5% event rate. Whilst our statistical methods were robust and the relationship between NEWS and the primary outcome measure was statistically significant, we are conscious that statistical significance does not equate to clinical significance. Secondly, our study sampled medical patients with undifferentiated presenting complaints. However, we included patients from only one hospital, which could limit the generalisability of our results. Our sample appears heterogeneous, but there may be underlying characteristics common to this cohort that we are not aware of, which may bias the results. We performed
a sensitivity analysis excluding patients that received palliative treatment but this did not change our results and we are not aware of any organisational factors during the study period that could have skewed our results, for example bed availability. Thirdly, we included all patients admitted to the acute assessment unit during a fixed period of time to reduce selection bias. However, we did not include patients from the emergency department who were not admitted to the medical service or patients that were admitted directly to the critical care unit. This in itself is a form of selection bias, which will have restricted our sample to patients that were unwell enough to be admitted to hospital but not so unwell that they required level two or level three care from the outset. From a pragmatic point of view this form of selection is quite sensible since we tested early warning scores in patients on medical wards – the very cohort that should be targeted by these tools. However, the unwitting exclusion of critical care admission directly from the emergency department may explain the lower than expected event rate. The Royal College of Physicians recommends that: a NEWS of 1–4 should prompt a review by a registered nurse, a NEWS of 5–6 should prompt a review by a ward doctor or acute nurse, and a NEWS of 7 or more should prompt a review by a critical care outreach team.8 Our analysis did not examine the response to an abnormal EWS, either in terms of type or speed of clinical review. However, our results suggest that patients with a NEWS greater than two were more likely to die or be admitted to level two or level three care within 2 days. The threshold for triggering a review by a doctor could be revised to a score of three; we recommend that this be considered when NEWS is next reviewed or refined. However, this would require careful balance against the likely increase in resource requirements for hospitals as more patients trigger a medical review. Our institution will be introducing NEWS in the near future. Further research examining the type and speed of clinical response to abnormal EWS is needed. 5. Conclusion Track and trigger scores are an integral part of ward-based hospital care in the UK. This study sought to evaluate the National Early Warning Score in a single hospital. Our results suggest that NEWS on admission is associated with death or critical care admission within the first 2 days of hospital stay. However, we did not identify association between NEWS on admission and hospital length of stay. There may be benefit in lowering the trigger point of the medium risk category from 5 to 3, as a score of 3 was associated with the primary endpoint. However, this requires further evaluation in larger, multi-centre datasets. Conflict of Interest statement There are no conflicts of interest. References 1. Patterson C, Maclean F, Bell C, Mukherjee E, Bryan L, Woodcock T, et al. Early warning systems in the UK: variation in content and implementation strategy has implications for a NHS early warning system. Clin Med 2011;11:424–7. 2. McGinley A, Pearse RM. A national early warning score for acutely ill patients. Br Med J 2012;345:e5310. 3. Bleyer AJ, Vidya S, Russell GB, Jones CM, Sujata L, Daeihagh P, et al. Longitudinal analysis of one million vital signs in patients in an academic medical center. Resuscitation 2011;82:1387–92. 4. Cei M, Bartolomei C, Mumoli N. In-hospital mortality and morbidity of elderly medical patients can be predicted at admission by the Modified Early Warning Score: a prospective study. Int J Clin Pract 2009;63:591–5. 5. Smith GB, Prytherch DR, Meredith P, Schmidt PE, Featherstone PI. The ability of the National Early Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive care unit admission, and death. Resuscitation 2013;84:465–70.
T.E.F. Abbott et al. / Resuscitation 92 (2015) 89–93 6. Alam N, Hobbelink EL, van Tienhoven AJ, van de Ven PM, Jansma EP, Nanayakkara PW. The impact of the use of the Early Warning Score (EWS) on patient outcomes: a systematic review. Resuscitation 2014;85:587–94. 7. Jones M, NEWSDIG:. The National Early Warning Score Development and Implementation Group. Clin Med 2012;12:501–3. 8. Physicians RCo. National Early Warning Score (NEWS): standardising the assessment of acute- illness severity in the NHS. Report of a working party London 2012. 9. Cooksley T, Kitlowski E, Haji-Michael P. Effectiveness of Modified Early Warning Score in predicting outcomes in oncology patients. QJM: Monthly J Assoc Physicians 2012;105:1083–8. 10. Corfield AR, Lees F, Zealley I, Houston G, Dickie S, Ward K, et al. Utility of a single early warning score in patients with sepsis in the emergency department. Emergency Med J 2014;31:482–7. 11. Tirkkonen J, Olkkola KT, Huhtala H, Tenhunen J, Hoppu S. Medical emergency team activation: performance of conventional dichotomised criteria versus national early warning score. Acta Anaesthesiol Scand 2014;58:411–9. 12. Romero-Brufau S, Huddleston JM, Naessens JM, Johnson MG, Hickman J, Morlan BW, et al. Widely used track and trigger scores: are they ready for automation in practice? Resuscitation 2014;85:549–52.
93
13. Peris A, Zagli G, Maccarrone N, Batacchi S, Cammelli R, Cecchi A, et al. The use of Modified Early Warning Score may help anesthesists in postoperative level of care selection in emergency abdominal surgery. Minerva Anestesiol 2012;78:1034–8. 14. Paterson R, MacLeod DC, Thetford D, Beattie A, Graham C, Lam S, et al. Prediction of in-hospital mortality and length of stay using an early warning scoring system: clinical audit. Clin Med 2006;6:281–4. 15. Bokhari SW, Munir T, Memon S, Byrne JL, Russell NH, Beed M. Impact of critical care reconfiguration and track-and-trigger outreach team intervention on outcomes of haematology patients requiring intensive care admission. Ann Hematol 2010;89:505–12. 16. Kolic I, Crane S, McCartney S, Perkins Z, Taylor A. Factors affecting response to National Early Warning Score (NEWS). Resuscitation 2015. http://dx.doi.org/10.1016/j.resuscitation.2015.02.009 17. Subbe CP, Gao H, Harrison DA. Reproducibility of physiological track-and-trigger warning systems for identifying at-risk patients on the ward. Intensive Care Med 2007;33:619–24. 18. McNamee M. Electronic keypad replaces early warning score paper charts. Nursing Children Young People 2013;25:9.