Original Research The Admission Hamilton Early Warning Score (HEWS) Predicts the Risk of Critical Event during Hospitalization Benjamin Tam BHSc MD, Michael Xu BHSc, Michelle Kwong BHSc Cand., Christine Wardell BHSc Cand., Andrew Kwong BHSc Cand., Alison Fox-Robichaud MSc MD FRCPC

About the Authors: Lead author Benjamin Tam is an internist and a trainee in the Adult Critical Care Medicine residency program at McMaster University in Hamilton, Ontario. His interest is in patient safety, particularly the use of rapid response teams and early warning scores to reduce critical events and improve the patient experience. Co-authors Michael, Xu, Michelle Kwong, Christine Wardell, Andrew Kwong are BHSc Candidates at McMaster University. Principal investigator and corresponding author Alison Fox-Robichaud is with the Department of Medicine, Division of Critical Care, Faculty of Health Sciences, McMaster University. Correspondence may be directed to [email protected]. Benjamin Tam

Abstract Background: Early warning scores detect patients at risk of deterioration in hospital. Our objective was to first, demonstrate that the admission Hamilton Early Warning Score (HEWS) predicts critical events and second, estimate the workload required to identify critical events during hospitalization. Methods: We prospectively identified a consecutive cohort of medical/surgical patients for retrospective review. Critical events were defined as a composite of inpatient death, cardio-pulmonary arrest or ICU transfer. Likelihood of a critical event during hospitalization and the number needed to evaluate to detect a critical event was based on highest admission HEWS. Results: We found 506 critical events occurred in 7130 cases. HEWS identified graduated levels of risk at admission. We found 2.6 and 1.8 patients needed to be evaluated in the ‘high-risk’ and very ‘high-risk’ subgroups to detect a critical event. Conclusions: HEWS identified patients at risk for critical events during hospitalization at ward admission. Few patients with high HEWS required evaluation to detect a critical event.

Résumé Contexte : Les scores d’alerte précoce permettent de dépister les patients à risque de détérioration à l’hôpital. Notre objectif était de démontrer, d’une part, que le Hamilton Early Warning Score (HEWS) permet de prédire les épisodes critiques et, d’autre part, évalue la charge de travail nécessaire pour reconnaître les épisodes critiques au cours d’une hospitalisation. Méthode : Nous avons déterminé de manière prospective une cohorte de patients en vue d’un examen rétrospectif. Les épisodes critiques ont été définis comme étant soit le décès du patient hospitalisé, un arrêt cardio‑respiratoire ou un transfert au service de soins intensifs. La probabilité qu’un épisode critique survienne durant l’hospitalisation et le chiffre nécessaire pour prévoir un épisode critique ont été basés sur les scores HEWS les plus élevés à l’admission. Résultats : Nous avons recensé 506 épisodes critiques survenus pour 7130 cas suivis. Le HEWS a permis de déterminer les niveaux de risque à l’admission. Nous avons fait ressortir que les patients ayant un score de 2,6 et de 1,8 devaient être classés dans les sous-groupes respectifs à « risque élevé » et à « risque très élevé » en ce qui a trait à la probabilité d’un épisode critique. Conclusions : Le HEWS permet de cibler dès leur admission les patients à risque d’épisodes critiques durant l’hospitalisation. Peu de patients ayant un HEWS élevé ont fait l’objet d’une évaluation pour un épisode critique.

24 Volume 11, Issue 4, 2016 Canadian Journal of General Internal Medicine

CJGIM_Vol 11_4.indd 24 06/02/17 6:17 PM Tam et al.

In-hospital emergency systems are designed to detect and monitoring and alerted the charge nurse. For a score 4 or greater, respond to unexpected clinical deterioration. These systems a junior resident was called to assess, while at a score 5 or greater, are composed of an afferent arm and an efferent arm. The a senior resident and/or the rapid response team was called to role of the afferent arm is to detect patients at risk of decline assess. The most responsible physician was made aware of all while the efferent arm delivers critical care assessment scores greater or equal to 6. and treatment. Early warning scores may improve outcomes by standardizing and automating the detection arm of the system.1 Data Collection The Hamilton Early Warning Score (HEWS) was created Demographic data was recorded and pertinent history was reviewed based on all available scores in literature but also incorporated to calculate a Charlson Co-Morbidity Index (CCI). All charts delirium assessment, adjusted thresholds and were reviewed for a critical event identified as an inpatient death, accounted for variable oxygen delivery (see Table 1). Our objective inpatient arrest or ICU transfer. Inpatient deaths included all was to first, demonstrate that the admission HEWS predicts deaths occurring in hospital except patients specifically admitted critical events during hospitalization and second, estimate the for palliation. Inpatient arrests included ventricular tachycardia, workload required to identify critical events. ventricular fibrillation, -less electrical activity, or respiratory arrests. ICU transfers were specifically unexpected ICU transfers Methods and excluded patients transferred to the ICU for post-operative Study Design, Setting, and Patient Description monitoring. Critical events were mutually exclusive. As such, We prospectively identified a consecutive cohort for retrospective a patient may have experienced both an inpatient arrest and chart review over 6 consecutive months in 2014. The study included inpatient death. This counted toward 2 outcomes. The highest, patients admitted to the medical and surgical wards at the 2 academic complete HEWS within a 24-hour window was calculated for all hospitals affiliated with the Hamilton Health Sciences. HEWS was patients at time of admission to the ward. Data were documented built into the electronic medical record at both centres. Mandatory as missing if the pre-specified required to calculate electronic vital sign documentation was implemented to facilitate HEWS was incomplete. HEWS use. Patients were included into the study if they were admitted to the medical or surgical floor at the study hospitals. We Analysis defined 4 points of entry to the ward: the emergency department, The demographic data, CCI and HEWS at time of admission were post-ICU discharge, post-operating room and other. compared between patients who had a critical event and those who had an event-free hospitalization. Comparison was made Ethics across 4 groups of patients, low risk (HEWS 0–2), moderate risk We received ethics approval from the Hamilton Integrated (HEWS 3–5), high risk (HEWS 6–8) and very high risk (HEWS Research Ethics Board (13-724-C). > 9). Likelihood ratios were calculated for any critical event during hospitalization based on admission HEWS. Additionally, a positive HEWS Ramp-Up Protocol predictive value (PPV) and its inverse, which was defined as the HEWS was based on a ramp-up response system. For a score number of cases needed to be evaluated (NNE) to detect a critical of 3 or greater, the bedside nurse increased the frequency of event was calculated based on admission HEWS.2

Table 1. The Hamilton Early Warning Score 3 2 1 0 1 2 3 Heart Rate <40 41-50 51–100 101–110 111–130 >130 Systolic Blood Pressure <70 71–90 91–170 171–200 >200 <8 8–13 14–20 21–30 >30 Temperature <35° 35.1°–36.4° 36.5°–38° 38.1°–39° >39.1° Oxygen Saturation <85 85–91 >92 Oxygen Delivery Room air < 5 L/min or >5 L/min or < 50% by >50% by mask mask Neurologic Status CAM positive Alert Voice Pain Unresponsive Neurologic status based on CAM, Confusion Assessment Method Tool and AVPU assessment with outlines the patient’s responsiveness to stimuli.

Canadian Journal of General Internal Medicine Volume 11, Issue 4, 2016 25

CJGIM_Vol 11_4.indd 25 06/02/17 6:17 PM The Admission Hamilton Early Warning Score (HEWS) Predicts the Risk of Critical Event during Hospitalization

Results low-risk HEWS and a moderate risk HEWS was associated with Patient Characteristics a reduced risk of a critical event during hospitalization. We found 506 critical events occurred in 7130 cases over a six-month period (Table 2). We excluded 74 patients due Discussion to insufficient data from the HEWS score. The majority of We found that HEWS at the time of ward admission predicted the critical events were either death (49.3% of all critical events), risk of a critical event during hospitalization. Moreover, only 2.6 or unanticipated ICU transfer (44.0% of all critical events). The cases and 1.8 cases needed to be evaluated in the high-risk group inpatient arrest rate was 4.95 arrests/1000 hospital admissions. and very high-risk group to potentially prevent a critical event. We found that patients with critical events were more likely to Our work showed that the initial vital signs at the time be older, male, and have other co-morbidities. of ward admission predicted the risk of critical event during hospitalization. Similar work had been demonstrated with the Characteristics of the Admission HEWS National Early Warning Score and the Abbreviated VitalPac We used HEWS to stratify risk for critical event during hospitalization Early Warning Score (ViEWS).3,4 Identifying patients at high based on initial ward presentation (Table 3). We found that the risk of deterioration early in hospitalization provides a method high-risk group and very high-risk group comprised only 4% of for rapid response teams and ward healthcare teams to organize patients admitted to the ward (261/7130). High-risk patients were and allocate resources necessary to meet patient acuity. much more likely to have a critical event than low-risk patients. The In order to determine workload, we studied the number number needed to evaluate analysis revealed 1 critical event would of cases that needed to be evaluated to detect a critical event. be detected for every 2.6 cases evaluated in the high-risk group This is based on the seminal work published by Romero-Brufau and 1.8 cases evaluated in the very high-risk group. Conversely, a et al who outlined a method of using the PPV to evaluate early

Table 2. Comparing demographic characteristics between patients who experienced a critical event in hospital (n=506) and those who had an event free hospitalization (n=6624)

Critical Events Inpatient Inpatient ICU Event Free Death Arrest transfer Hospitalization p-value Number of events (%) 245 33 228 6624 N/A (3.6) (0.5) (3.2) (92.8) Age 79.1 ± 12.4 69.5 ± 16.9 68.3 ± 15.2 64.6 ± 18.5 <0.001 Male – no, (%) 141 24 126 3180 <0.05 (55.1) (68.6) (55.3) (47.6) Admitted at centre with RRT – no, (%) 89 12 84 2945 <0.05 (34.8) (34.3) (36.8) (44.1) Average CCI – no. 8.3 ± 3.1 6.1 + 3.3 6.8 ± 3.3 5.0 ± 3.2 <0.001 Heart disease – no, (%) 98 13 92 1350 <0.001 (38.3) (37.1) (40.4) (20.2) Chronic lung disease – no, (%) 76 9 56 1380 <0.05 (29.7) (25.7) (24.6) (20.6) Moderate/severe renal impairment – no, (%) 69 5 50 611 <0.001 (27.0) (14.3) (21.9) (9.1) Dementia – no, (%) 47 1 13 486 <0.001 (18.3) (2.9) (5.7) (7.3) Malignancy – no, (%) 100 7 82 1696 <0.001 (39.1) (20.0) (36.0) (25.4) AND prior to admission – no, (%) 112 2 12 643 <0.001 (43.8) (5.7) (5.3) (9.6) Highest HEWS at admission 3.8 ± 2.9 2.4 ± 1.9 3.1 ± 2.6 1.6 ± 2.5 <0.001 No. number, RRT, Rapid Response Teams, CCI, Charlson Co-morbidity Index, AND, Allow Natural Death, HEWS, Hamilton Early Warning Score x ± y denotes mean, standard deviation

26 Volume 11, Issue 4, 2016 Canadian Journal of General Internal Medicine

CJGIM_Vol 11_4.indd 26 06/02/17 6:17 PM Tam et al.

Table 3. Likelihood ratio of critical event and number needed to evaluate to detect critical event based on admission HEWS Admission Critical Event Likelihood Positive Number Needed HEWS Yes No Ratio Predictive Value to Evaluate Number Proportion Number Proportion (95% CI) (95% CI) (95% CI) Low Risk 221 222/506 = 0.437 5201 5201/6624 = 0.785 0.56 4.1 25 0–2 (0.50–0.61) (3.7–4.5) (22–27) Moderate Risk 177 177/506 = 0.350 1270 1270/6624 = 0.192 1.82 12.2 8.3 3–5 (1.60–2.08) (10.9–13.7) (7.2–9.2) High Risk 80 80/506 = 0.158 131 131/6624 = 0.020 7.99 37.9 2.6 6–8 (6.15–10.4) (31.2–44.3) (2.3–3.2) Very High Risk 28 28/506 = 0.055 22 22/6624 = 0.003 16.7 56.0 1.8 >9 (9.6–28.9) (42.3–68.8) (1.5–2.3) Total 506 6624 HEWS, Hamilton Early Warning Score

warning scores.2 In short, the number needed to evaluate, which Conclusions is the inverse of the PPV, outlines the number of evaluations HEWS identified patients at risk for critical events at the time required to detect critical events. In the high-risk and very-high of ward admission. Few patients with elevated HEWS required risk groups, we found that only 2.6 and 1.8 cases needed to be evaluation to detect a critical event. Number needed to evaluate evaluated to potentially prevent a critical event. This may seem is a method to anticipate resources required to treat a patient at like a reasonable tradeoff. However, critical events represent risk of critical illness. failure of the system to prevent deterioration. As such, more cases would be evaluated per critical event when the system Acknowledgements becomes more effective at preventing deterioration. The number Special thanks to Erin Kelleher and Terrance Rock for technical needed to evaluate provides a way to quantify and anticipate assistance in compiling and abstracting patient data. workload. With further development, a marker like number needed to evaluate will allow the health care system to rapidly Conflict of Interest and Funding adapt resources to pre-emptively manage high-risk patients. This project received funding from the Hamilton Health Sciences This study is important for several reasons. First, we Resident Research Grant in Patient Safety as well as the Hamilton demonstrate that HEWS is a simple and standardized method Health Sciences Department Quality and Patient Safety Award. that supplements a clinician’s identification of patients at risk of Funding parties were not involved in study design, interpretation critical illness. A simple and standard language supports situational or writing of the manuscript. awareness which is fundamental to the delivery of safe patient care in a complex environment.5 In addition, we demonstrate how References HEWS can be used to predict workload. This can be leveraged 1. McNeill G, Bryden D. Do either early warning systems or emergency to direct resources toward patients at high risk of deterioration. response teams improve hospital patient survival? A systematic review. Resuscitation 2013;84(12):1652–67. There are several limitations to this study. First, vital sign 2. Romero-Brufau S, Huddleston JM, Escobar GJ, Liebow M. Why the monitoring frequency was not mandated so higher admission C-statistic is not informative to evaluate early warning scores and what HEWS may have been undetected. Furthermore, though admission metrics to use. Crit Care 2015;19(1):285. 3. Alam N, Vegting IL, Houben E, et al. Exploring the performance of with a high HEWS predicted critical events, admission with a low the National Early Warning Score (NEWS) in a European emergency HEWS did not rule critical events out. There may have been vital department. Resuscitation 2015;90:111–15. sign abnormalities throughout a hospitalization that predicted 4. Kellett J, Wang F, Woodworth S, Huang W. Changes and their prognostic implications in the abbreviated VitalPAC™ Early Warning Score (ViEWS) deterioration that we did not include into our analysis. Finally, our after admission to hospital of 18,827 surgical patients. Resuscitation study presented the number of patients that required evaluation to 2013;84(4):471–76. detect, not prevent, a critical event. Outlining how to move from 5. Brady PW, Goldenhar LM. A qualitative study examining the influences on situation awareness and the identification, mitigation and escalation of detection to prevention was outside the scope of this study but recognised patient risk. BMJ Quality & Safety 2014;23(2):153–61. remains an important question to pursue to improve patient safety.

Canadian Journal of General Internal Medicine Volume 11, Issue 4, 2016 27

CJGIM_Vol 11_4.indd 27 06/02/17 6:17 PM