Article Text


Use of an admission early warning score to predict patient morbidity and mortality and treatment success
  1. J D Groarke1,
  2. J Gallagher1,
  3. J Stack2,
  4. A Aftab1,
  5. C Dwyer3,
  6. R McGovern1,
  7. G Courtney1
  1. 1
    Department of Medicine, St Luke’s Hospital, Kilkenny, Ireland
  2. 2
    Department of Computing, Mathematics and Physics, Waterford Institute of Technology, Waterford, Ireland
  3. 3
    Medical Assessment Unit, St Luke’s Hospital, Kilkenny, Ireland
  1. Dr J D Groarke, Department of Medicine, St Luke’s Hospital, Kilkenny, Ireland; johngroarke1{at}


Background: Early warning scores (EWS) are used to identify physiological deterioration in patients. Studies to date have primarily focused on the correlation between trends in serially recorded EWS of inpatients and clinical outcomes. This study examined the predictive value of an EWS calculated immediately on presentation to hospital for acute medical patients.

Method: A prospective study of 225 consecutive medical admissions. Pulse, systolic blood pressure, respiratory rate, oxygen saturation and neurological status were used to calculate an EWS. Patients were divided into four score categories based on their EWS. The primary endpoints examined were intensive care unit (ICU)/coronary care unit (CCU) admission, death, cardiac arrest and length of hospital stay.

Results: For each rise in score category there was an increased risk of admission to ICU (odds ratio (OR) 3.35, CI 1.52 to 7.40, p = 0.003), admission to CCU (OR 1.82, CI 1.07 to 3.09, p = 0.027), death (OR 2.19, CI 1.41 to 3.39, p = 0.000) and reaching the combined endpoint of CCU/ICU admission or death (OR 2.19, CI 1.41 to 3.39, p = 0.000). The higher the score the longer the length of hospital admission (p = 0.04). A decrease in EWS between first presentation to hospital and transfer to the ward was associated with a decreased risk of reaching the combined endpoint of CCU or ICU admission or death (OR 2.56, CI 1.11 to 5.89, p = 0.028).

Discussion: Higher admission EWS correlate with increased risk of CCU/ICU admission, death and longer hospital stays independent of patient age. An improvement in serial EWS within 4 h of presentation to hospital predicts improved clinical outcomes. The EWS is a potential triage tool in the emergency department for acute medical patients.

Statistics from

Physiological decline in patients heralding clinical deterioration is common. For example, anticipating events occur in approximately 85% of patients in the period preceding an inhospital cardiopulmonary arrest.1 2 The severity of patients’ illness and clinical deterioration may go unnoticed in hospital or fail to secure timely and appropriate intervention.3 In one study 54% of patients received suboptimal care before intensive care unit (ICU) admission from the ward and 69% of patients were transferred to ICU late in the course of their illness.1 Early warning scores (EWS) have been developed in order to help identify patients at risk of deterioration.

An EWS is a simple bedside tool based on routinely recorded physiological parameters: respiratory rate, temperature, blood pressure, heart rate and a simple assessment of neurological status. Some variations of the score include urinary output and oxygen saturations. Each parameter receives an individual score identified from the reference table that is proportional to the deviation of that parameter from the normal range. The sum of the individual scores gives a total EWS. An algorithm is then used by staff to guide the need for more frequent observations and senior medical review. The benefit of these scores among inpatient populations is established.48 An increasing modified EWS is associated with worse outcome across a range of medical and surgical specialties.9

Therefore, the predictive value of EWS among inpatient populations has been demonstrated. However, this prospective study was carried out to determine if a single EWS calculated immediately on presentation to hospital can predict patient outcomes and identify those at risk of deterioration from the outset. Change in serially recorded scores following treatment as a predictor of outcome was also examined. It is proposed that an improvement in serial scores suggests effective treatment, whereas persistently high or worsening scores necessitate a review of patient management, from drug therapy to setting of care.


A prospective study of consecutive admissions via the Medical Assessment Unit (MAU) of a 315-bed Irish general hospital over a 30-day period was undertaken. Institutional ethical approval was obtained. This hospital offers acute medical, surgical, paediatric, psychiatric, intensive care, coronary care, obstetric and gynaecology services on site. All adult patients presenting between 08:00 and 19:00 hours each day with an acute complaint likely to be medical in nature as determined by triage nursing staff are reviewed in the MAU. These patients are either referred by their general practitioners or are self-referred, including patients arriving by ambulance. Patients with non-medical complaints (eg, fractured limb) are assessed in other departments. MAU patients will have a set of vital signs recorded within 10 minutes of presentation as per department protocol. Following appropriate assessment, the patient is discharged home or admitted to the hospital and transferred to a ward. The average length of patient stay in this MAU is 5 h.

All patients admitted to the hospital via the MAU over a 30-day period were entered into the study. Pulse, blood pressure, respiratory rate, temperature, oxygen saturation and conscious level (recorded using the AVPU score) were recorded for each patient by nursing staff on two occasions—on initial admission to the MAU (ie, within 10 minutes of presentation) and immediately before transfer from the MAU to the ward (ie, on average 5 h after initial presentation). Only those patients admitted to a hospital ward from the MAU were included in the study. As this institution did not have an EWS in use at the time of this study, staff were unaware of the parameters of the EWS when recording observations. The collected data were used to calculate the EWS at presentation and before transfer to a ward for each patient retrospectively. The EWS used in this study is outlined in table 1.

Table 1 EWS used in this study

This EWS was devised by the authors and is similar to other EWS reported in the literature.4 8 Urine output was not included in this study’s score. All patients were followed for the duration of their hospital admission (range 1–42 days). The primary endpoints of the study were length of hospital stay, survival to discharge, ICU or coronary care unit (CCU) admission at any point during hospital admission and cardiac arrest call during admission. ICU and CCU admissions were at the discretion of the attending doctors who were unaware of the EWS of the patient. Patients were divided into four categories based on their total score: 0–1, 2–3, 4–5 and over 5.

SPSS 13.0 for Windows was used for statistical analysis. Odds ratios were estimated using logistic regression, with EWS category as the independent variable and each of the following as the binary dependent variable: admission to CCU; admission to ICU; survival to discharge and combined CCU, ICU or death outcomes. These analyses were repeated controlling for the age of the patient.

The mean length of hospital stay was calculated for each of the four EWS categories and tested for statistical significance using analysis of variance. Those who died were excluded from analysis of length of stay. The effect of a change in EWS (between the initial score and the score on transfer to the ward) on endpoints was examined for patients with an initial EWS of 2 or more. Odds ratios were estimated by logistic regression.


Data were collected on 225 from a total of 225 consecutive medical admissions; 116 were male and 109 were female. The mean age of the patients was 64.7 years (SD 19.1). Collected datasets of 60 of the 225 (26.7%) patients had missing variables. However, 95% of these incomplete patient datasets were lacking a maximum of only two from a total of 12 variables (six from admission EWS and six from EWS calculated before transfer to the ward). No patient had more than three variables missing. Outcome data were available on all patients. Seven patients were admitted to ICU, 17 to CCU, eight patients died with a “Do not attempt resuscitation” order in place. No patient was the subject of a cardiopulmonary resuscitation attempt; 28 patients reached one or more of the endpoints of CCU, ICU or death. The frequency of each initial admission score is shown in table 2.

Table 2 Percentage of patients reaching endpoints per score category

For each rise in EWS category there was an increased risk of ICU admission, CCU admission, death and reaching the combined endpoint of CCU, ICU or death (table 3). The variation in risk (for each endpoint) for the different score categories is outlined in table 2. For some endpoints (eg, ICU admission), the risk increases uniformly by approximately the same factor for each step up in score category, as suggested by the logistic regression results (table 3). For other endpoints, however, the risk does not appear to increase uniformly with EWS. The risk of CCU admission is actually lower among patients in the 4–5 group than in the 2–3 group. This may be explained partly by the fact that many patients admitted to CCU with acute coronary syndromes demonstrate few physiological abnormalities and therefore have expectedly low scores. Nevertheless, for most outcomes, the risk increases with rising score and the risk is substantially greater in patients with high scores. For patients who survived to discharge, mean length of hospital stay was significantly longer in the higher score categories (fig 1). Analyses were repeated controlling for the age of the patient. Age did not influence the validity of the score.

Figure 1 Mean length of hospital stay for each score category.
Table 3 Odds ratio for each rise in score category (0–1, 2–3, 4–5, >5)

Patients with an initial score of greater than 3 whose score decreased before transfer to the ward were less likely to reach an endpoint than those whose score did not change or worsened. If the score improved, 23.1% reached an endpoint. If the score stayed the same, 40% reached an endpoint and if the score deteriorated, 33.3% reached an endpoint. As only 34 subjects in total, however, had scores in excess of 3, these differences in risk, although substantial, are not statistically significant. If this relationship between trends in serial scores and outcome is examined in patients with an initial score of greater than 2 (n  =  67) rather than greater than 3 (n  =  34), statistical significance is achieved. For this group, if the score improved 14.3% reached an endpoint. If the score remained unchanged 33.3% reached an endpoint and if the score worsened 50% reached an endpoint. Applying logistic regression to these data (with the categories of the independent variable coded as 1, improved, 2, no change and 3, worsened) the estimated odds ratio is 2.56 (95% CI 1.11 to 5.89, p = 0.028). In each case, patients with improving scores have the lowest risk of reaching an endpoint (table 4).

Table 4 Relationship between change in patients’ serial EWS following initiation of treatment and likelihood of reaching a combined endpoint


This study demonstrates that for unselected medical admissions, an increased EWS on admission predicts increased mortality, increased likelihood of admission to ICU or CCU, death and a longer length of hospital stay. The EWS could thus be used as a triage tool in the emergency department for acute medical patients and identify “at-risk” patients from the outset. The EWS can help nurses and doctors to identify vulnerable patients and aid decisions such as the type of medical bed required (ie, general ward versus high dependency bed) and the appropriate interval for nursing observations and physician review. Furthermore, in addition to its potential role as a triage tool, the EWS may be useful in the prehospital setting aiding paramedics in identifying those particularly ill patients and alerting emergency departments of their imminent arrival. The current systems in use by paramedics for pre-warning emergency departments of the arrival of critically ill patients have been highlighted as inadequate.10

A decreasing EWS would be expected to accompany an improvement in a patient’s condition. The admission EWS relative to the trend in successive scores may thus help evaluate the efficacy of the patient’s treatment and further highlight those at risk of deterioration. This study suggests that the score can be used in this context as there is a correlation between a falling score and a decreased risk of transfer to CCU, ICU and death. The NHS Plan (2000) set out firm targets to ensure all patients presenting to emergency departments should be admitted, discharged or transferred within 4 h of arrival.11 This study demonstrates that the trend between EWS on arrival and before transfer to a ward is an early predictor of outcome and treatment success. This could facilitate clinical decisions and highlight “at-risk” patient within the 4-h limit.

Study limitations

The small number of patients overall and in particular the small number of patients in higher score categories are limitations of this study. Furthermore, only acute medical patients are considered and patients from other specialties are not included. Despite these limitations, several statistically significant findings are demonstrated, calling for a similar larger study including patients from a wider range of specialties.

Not all patient datasets were complete. Unfortunately, in clinical practice not all variables will be recorded for every patient at all times and so to optimise the relevance of this study’s findings to day-to-day practice, these incomplete datasets were included for analysis. Missing variables were consistently scored as zero for analysis, which may lead to lower scores for individual patients than would have been the case if all variables had been available. However, we believe this demonstrates the robustness of the system in clinical practice.


As the evidence accumulates, it is acknowledged that early warning scoring systems can serve as useful clinical adjuncts. A key strength of the EWS is their simplicity and ability for them to be quickly calculated in busy clinical areas with reference to routinely collected data. This study identifies additional roles of this simple tool. A single admission EWS can identify patients at risk of deterioration upon presentation to the hospital and serves as an early prognostic indicator. It is a potential triage tool for patients presenting for medical assessment and its use should be considered in the prehospital setting to alert the receiving hospital of critically ill patients. The use of an EWS as a triage tool in emergency departments and in the prehospital setting warrants larger studies and validation of these uses across the range of specialties.

A decreasing EWS is an early independent predictor of outcome and treatment success. These findings are consistent for all ages. Therefore, EWS can assist in decisions on patient management such as the level of nursing care necessary, frequency of physician or nurse review and prompt discussions on resuscitation right from the outset of the patient journey. Worsening or persistently elevated serial scores should prompt an immediate revision of patient management. Further larger studies in this area are warranted.


The authors are indebted to the staff of the Medical Assessment Unit of the study centre for their assistance in the collection of data.


View Abstract


  • Competing interests: None.

  • Ethics approval: Institutional ethics approval was obtained.

  • Patient consent: Obtained.

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.