Effect of introducing the Modified Early Warning score on clinical outcomes, cardio-pulmonary arrests and intensive care utilisation in acute medical admissions*
Some of the data has been presented at the Intensive Care Society state of the Art Meeting in London, December 2001 as an oral presentation.
Summary
The effects of introducing Modified Early Warning scores to identify medical patients at risk of catastrophic deterioration have not been examined. We prospectively studied 1695 acute medical admissions. All patients were scored in the admissions unit. Patients with a Modified Early Warning score > 4 were referred for urgent medical and critical care outreach team review. Data was compared with an observational study performed in the same unit during the proceeding year. There was no change in mortality of patients with low, intermediate or high Modified Early Warning scores. Rates of cardio-pulmonary arrest, intensive care unit or high dependency unit admission were similar. Data analysis confirmed respiratory rate as the best discriminator in identifying high-risk patient groups. The therapeutic interventions performed in response to abnormal scores were not assessed. We are convinced that the Modified Early Warning score is a suitable scoring tool to identify patients at risk. However, outcomes in medical emergency admissions are influenced by a multitude of factors and so it may be difficult to demonstrate the score's benefit without further standardizing the response to abnormal values.
Suboptimal care of critically ill patients on general wards may result in potentially preventable cardio-pulmonary arrests [1]. Suboptimal care prior to intensive care unit (ICU) admission may be associated with increased mortality measured as either ICU or hospital mortality [2].
Identifying patients at risk is possible, using simple algorithms based on bedside observations that include respiratory and mental function [3,4]. The Modified Early Warning score is an algorithm that has been validated for clinical use in medical emergency admissions [5]. Use of the Modified Early Warning score on surgical wards has been shown to decrease Acute Physiology and Chronic Health Evaluation (APACHE) II scores on admission to intensive care [6]. The Department of Health [7], Intensive Care Society [8] and Royal College of Physicians (London) [9] have recommended the Modified Early Warning score as an aid in identifying patients at risk on general wards. However, no prospective study has examined the impact of introducing the Modified Early Warning score as standard clinical management of medical emergency admissions.
The primary aim of our study was to measure the effect of introducing the Modified Early Warning score on the rates of ICU and high dependency unit (HDU) admission, cardio-pulmonary arrest and mortality. A secondary aim was to collect physiological data from patients prior to critical care admission, cardio-pulmonary arrest or death, in order to improve the discrimination of the score.
Methods
The Research Ethics Committee of the North-East Wales NHS Trust reviewed and approved the study. The study was performed in the 56-bedded medical admissions unit of a district general hospital serving a population of 300 000 in North Wales between 1 February and 31 April 2001 (3 months).
Patients
Patients were referred to the medical admissions unit by general practitioners or the Accident and Emergency Department. Patients diagnosed with myocardial infarction were admitted directly to the coronary care unit. The study examined all patients aged above 15 years but excluded patients admitted for palliative care only and patients admitted directly to other wards. Patients who were admitted more than once during the study period were entered only once, on their first admission.
Scoring tool
All medical admissions unit nursing staff were trained by the investigators and the critical care outreach team to collect bedside observations and to calculate the Modified Early Warning score (Table 1). Blood pressure and pulse rate were measured non-invasively (DINAMAPTM, Critikon Inc, Tampa, FL). The pulse rate was recorded manually in patients with arrhythmias. The respiratory rate was counted over 1 min. The AVPU score (AVPU denoting A = patient Alert, V = responsive to Voice, p = responsive to Pain and U = Unresponsive) was recorded using nail-bed pressure as a standardised stimulus for pain where appropriate. Temperatures were measured sublingually using Temp-PlusII® (IVAC-Corporation, San Diego, CA). All medical staff caring for emergency medical admissions were briefed concerning the Modified Early Warning score, its interpretation and their role in the management of a patient identified as being at risk of deterioration. The nursing staff were instructed to alert appropriate medical staff and the critical care outreach team if the Modified Early Warning score was 5 or more. Doctors were instructed to examine and assess patients as soon as possible but not later than ‘within 60 minutes’ with regard to further therapy and possible transfer to critical care.
| Points | |||||||
|---|---|---|---|---|---|---|---|
| 3 | 2 | 1 | 0 | 1 | 2 | 3 | |
| Systolic blood pressure; mmHg | < 70 | 71–80 | 81–100 | 101–199 | ≥ 200 | ||
| Heart rate; beats.min−1 | < 40 | 41–50 | 51–100 | 101–110 | 111–129 | ≥ 130 | |
| Respiratory rate; breath.min−1 | < 9 | 9–14 | 15–20 | 21–29 | ≥ 30 | ||
| Temperature; °C | < 35 | 35–38.4 | ≥ 38.5 | ||||
| Neurological score | Alert | Reacting to Voice | Reacting to Pain | Unresponsive | |||
Data
Basic demographic data and bedside observations on admission were collected for all patients admitted to the medical admissions unit. Two trained data collectors gathered data from observation charts once each day. Daily comparison of the collected data with hospital admission lists ensured that all patients who met the inclusion criteria were studied.
Data was collected on admission to medical HDU, ICU and following cardio-pulmonary arrest or death. Nursing staff on HDU and ICU who were trained for this study prospectively recorded basic demographic data and bedside observations from all medical admissions transferred to their unit from other wards. The investigators collected additional data from hospital case notes including diagnosis, pre-existing illness, basic laboratory data, social circumstances prior to admission and bedside observations 24 h, 6 h and immediately prior to transfer to critical care. Intensive care unit and hospital discharge data were collected retrospectively from the hospital's Patient Administration System. Time from hospital to ICU admission, length of stay on ICU and ICU mortality were calculated. Data on cardio-pulmonary arrests were collected in a standardised manner by the cardio-pulmonary arrest service. The investigators collected all additional data for these patients from hospital case notes as if the patient had been admitted to critical care. Data from patients who died on the medical admissions unit were collected in the same way. Data on 30-day hospital mortality were calculated from the Patient Administration System.
Data sets for which no outcome (i.e. death or hospital discharge) could be identified were excluded from analysis.
Control group
Data from a prospective observational study published previously [5] was used as a control group. This control group was admitted to the same admissions unit during February 2000.
Patients were classified on the basis of the Modified Early Warning score as being at low risk (Modified Early Warning score 0–2), intermediate risk (Modified Early Warning score 3–4) or high risk (Modified Early Warning score > 4) of catastrophic deterioration. Rates of admission to critical care, cardio-pulmonary arrests and death were calculated for each risk band.
Statistical analysis
Data was analysed using the Statistical Package for Social Sciences (Version 10, SPSS Inc., Chicago, IL). For normally distributed data, results are given as means and standard deviations (SD). For non-parametric data, medians and interquartile ranges (IQR) are given. Unpaired t-tests were used to compare mean variables in control and intervention groups and the Mann–Whitney U-test to compare medians in non-parametric variables. The Chi-squared test and Fisher's exact test were used to compare categorical variables. A p-value of less than 0.05 was considered significant.
Results
Patient outcomes
Admission, clinical and outcome data were available for 1695 patients. Demographic data of patients are summarised in Table 2. Patients' ages, sex and Modified Early Warning scores were not significantly different between the study and control groups. In the study group, 166 (9.7%) patients died, 9 (0.5%) patients were admitted to ICU, 79 (4.6%) were admitted to the medical HDU and 40 (2.3%) patients had cardio-pulmonary arrests.
| Parameter | Study group | Control group |
|---|---|---|
| n | 1695 | 659 |
| Age; years (SD) | 64 (19) | 63 (SD 20) |
| Sex; male/female; (%) | 45/55 | 45/55 |
| Systolic arterial blood pressure; mmHg (SD) | 139 (27) | 139 (27) |
| Pulse rate; beats.min−1 (SD) | 86 (21) | 87 (20) |
| Respiratory rate; breath.min−1 (SD) | 20 (6) | 20 (5) |
| Temperature; °C (SD) | 36.5 (1.3) | 36.7 (0.9) |
| Modified Early Warning score; median (IQR) | 2 (1–3) | 2 (1–2) |
The rates of ICU and HDU admissions and in-hospital mortality were similar in the study and control groups (Table 3) and for the three levels of risk as classified by the Modified Early Warning score. Overall, mortality was unchanged irrespective of risk band. There was an increased incidence of cardio-pulmonary arrests in the study group in patients with a Modified Early Warning score of 3 or 4 (i.e. intermediate risk).
| Modified Early Warning score risk band | Study group No. of patients with event/ no. of patients in risk band (%) | Control group No. of patients with event/ no. of patients in risk band (%) | p-values for Chi-square$ or Fisher's exact test* |
|---|---|---|---|
| Death | |||
| 0–2 | 70/1214 (6%) | 28/491 (6%) | 0.97$ |
| 3–4 | 59/348 (17%) | 15/117 (13%) | 0.29$ |
| 5–15 | 37/133 (28%) | 10/51 (20%) | 0.25$ |
| ICU or HDU admission | |||
| 0–2 | 45/1214 (4%) | 16/491 (3%) | 0.66$ |
| 3–4 | 22/348 (6%) | 7/117 (6%) | 0.90$ |
| 5–15 | 18/133 (14%) | 4/51 (8%) | 0.45* |
| Cardio-pulmonary arrests | |||
| 0–2 | 20/1214 (2%) | 3/491 (1%) | 0.11* |
| 3–4 | 16/348 (5%) | 0/117 (0%) | < 0.016* |
| 5–15 | 4/133 (3%) | 1/51 (2%) | 1.0* |
Intensive care utilisation
Only nine patients in the study group and six patients in the control group were admitted to ICU. Patients were admitted to ICU after 24 (IQR 12–84), hours in hospital in the study group and after 24 (IQR 18–186) hours in the control group (p = 0.9). APACHE II scores on admission were 15 (SD 8) in the study group and 23 (SD 7) in the control group (p < 0.06). Length of stay in ICU was 2 (IQR 1–30) days in the study group and 4 (IQR 1–8) days in the control group (p = 0.3). ICU mortality was 33% in the study group and 67% in the control group (p = 0.21).
Physiological changes prior to death, cardio-pulmonary arrest or critical care admission
Data from patients dying, suffering cardio-pulmonary arrest or being admitted to critical care (‘sick patients’) were compared with admission data from all other patients in the study (‘stable patients’). This ‘sick’ patient subgroup had statistically significant lower blood pressure, higher heart rates, higher respiratory rates and lower temperatures on admission (Table 4). Most physiological parameters in this patient group became more abnormal during the course of their illness. For the last set of observations prior to cardio-pulmonary arrest, death or critical care admission, the systolic blood pressure was 12 (9%) mmHg lower, the pulse rate 17 (20%) beats.min−1 faster, the respiratory rate 8 (41%) breaths.min−1 higher and the temperature 0.2 (0.5%) °C lower than the mean temperature in the ‘stable’ patients. For the ‘sick’ patients the mean drop in blood pressure was 9 (SD 26) mmHg from admission to the last measurement prior to death, cardio-pulmonary arrest or critical care admission.
| Parameter | On admission | 24 h prior to event | 6 h prior to event | Last set of observations prior to event |
|---|---|---|---|---|
| Systolic blood pressure; mmHg (SD) | 135 (30) | 120 (16) | 120 (31) | 128 (35) |
| n = 259 | n = 24 | n = 46 | n = 176 | |
| >Δ; mmHg (%), p-values | 5 (−4%) | 20 (−14%) | 20 (−15%) | 12 (−9%) |
| p < 0.005 | p < 0.001 | p < 0.001 | p < 0.001 | |
| Mean systolic blood pressure of ‘stable’ patients: 140 mmHg | 94 (26) | 106 (20) | 102 (20) | 101 (27) |
| n = 262 | n = 24 | n = 42 | n = 178 | |
| Pulse rate; beats.min−1 (SD) | 94 (26) | 106 (20) | 102 (20) | 101 (27) |
| n = 262 | n = 24 | n = 42 | n = 178 | |
| Δ; beats.min−1 (%), p-values | 10 (+ 12%) | 22 (+ 26%) | 18 (+ 21%) | 17 (+ 20%) |
| p < 0.001 | p < 0.001 | p < 0.001 | p < 0.001 | |
| Mean pulse rate of ‘stable’ patients: 84 beats.min−1 | 24 (8) | 27 (6) | 27 (7) | 27 (9) |
| n = 252 | n = 21 | n = 39 | n = 172 | |
| Respiratory rate; breaths.min−1 (SD) | 24 (8) | 27 (6) | 27 (7) | 27 (9) |
| n = 252 | n = 21 | n = 39 | n = 172 | |
| Δ; breaths.min−1 (%), p-values | 5 (+ 27%) | 8 (+ 44%) | 8 (+ 42%) | 8 (+ 41%) |
| p < 0.001 | p < 0.001 | p < 0.001 | p < 0.001 | |
| Mean respiratory rate of ‘stable’ patients: 19 breaths.min−1 | 36.3 (1) | 36.6 (0.8) | 36.2 (1.2) | 36.4 (1.3) |
| n = 258 | n = 25 | n = 39 | n = 170 | |
| Temperature; °C (SD) | 36.3 (1) | 36.6 (0.8) | 36.2 (1.2) | 36.4 (1.3) |
| n = 258 | n = 25 | n = 39 | n = 170 | |
| Δ; °C (%), p-values | 0.3 (−0.7%) | 0 (0%) | 0.4 (−1%) | 0.2 (−0.5%) |
| p < 0.001 | p = 0.961 | p < 0.056 | p < 0.052 | |
| Mean temperature of ‘stable’ patients: 36.6 °C. | ||||
Discussion
Introduction of a scoring system using bedside observations and a simple protocol to trigger medical and critical care review did not change outcomes in acute medical admissions. More importantly, outcomes in the patient group with the highest risk (Modified Early Warning score > 4) were not improved.
The proportion of patients admitted to critical care did not increase with the introduction of Modified Early Warning score but there was a trend towards earlier ICU admission. The increase in cardio-pulmonary arrests in the study group might be explained by the low rate of arrests in the smaller control group and a higher proportion of sick patients in the study group, as suggested by the difference in interquartile range of the Modified Early Warning score in control and study groups. Patients were not randomised for Modified Early Warning score use and the control group was a historic control from a shorter period of time than the study period. It was felt that randomisation of patients within the admissions unit would have been technically difficult.
Our medical admission unit is comparable to units in other district general hospitals in England and Wales. The Modified Early Warning score had previously been validated in our unit. However, our study did not standardise the response to high scores by medical and nursing staff. Delayed responses, faulty assessment of disease severity and inadequate treatment could have contributed to the negative outcome of this study. Other authors have found that suboptimal treatment prior to intensive care can be caused by failure of organisation, lack of knowledge, failure to appreciate urgency, lack of supervision and failure to seek advice [2]. Whilst the Modified Early Warning score identifies sick patients and encourages staff to appreciate the severity of their condition, we were not able to control any of the other factors that might have adversely affected patient's outcome.
Changes in physiological parameters prior to cardio-pulmonary arrest have been previously described [1]. Our study shows that changes in blood pressure and pulse rate are comparatively small and could potentially be clinically missed. Relative changes in respiratory rate are of a much greater magnitude and are therefore more likely to be better at discriminating between stable patients and patients at risk. Consistent recording of respiratory rate was achieved in this study for most patients in the medical admissions unit. This represents a crucial first step in recognizing patients at risk.
We are convinced that the Modified Early Warning score is a suitable scoring tool to identify patients at risk. However, outcomes in medical emergency admissions are influenced by a multitude of factors. To impact on outcomes the Modified Early Warning score has to be placed into an educational context of improved training in emergency medicine of junior and senior medical and nursing staff. Training can be delivered at the bedside or in courses such as the ALERTTM course (The Open Learning Centre, University of Portsmouth, Buckingham Building, Portsmouth, PO1 3HE, UK). Systematic feedback of adverse outcomes and near misses might further enhance care and show the true potential of the Modified Early Warning score in the management of critically ill patients on general medical wards.
Acknowledgements
We would like to acknowledge the invaluable help of our data collectors and data entry clerks Dawn Griffiths, Linda Edge and Pat Autiero, and the contribution of the medical and nursing staff participating in the study. We would like to thank specifically the ward sisters Helen Williams and Fiona Wilkinson for their trust, advice, expertise and leadership. The study was supported by a grant from the North-east Wales NHS Trust Research and Development Fund.
