Use of a modified early warning system to predict outcome in patients admitted to a high dependency unit

Introduction

The modified early warning system (MEWS) is a physiological scoring system that identifies patients at risk of deterioration who require increased levels of care [1]. The use of a patient's MEWS score to predict outcome in a high-dependency unit (HDU) has not been previously described.

Method

Approval for the study was granted by the local ethics committee. We reviewed MEWS scores from all patients (n = 2,974) admitted to a six-bed medical and surgical adult HDU in a general hospital from July 2002 to October 2005. The MEWS score was calculated from observations of heart rate, blood pressure, respiratory rate, urine output, conscious level and temperature recorded within 24 hours of admission to the HDU.

Results

Of the 2,974 patients reviewed, 2,447 patients had sufficient data. Analysis using logistic regression shows a strong relationship between the probability of death and the MEWS score: the odds of death increase by 1.48 (confidence interval 1.41–1.56; P < 0.001) for each unit increase in the MEWS score. However, there is no reason that these data should follow a logistic form and the estimates of uncertainty around the point estimates from logistic regression are poor. More accurate estimates of the death rate for each of the MEWS scores were achieved using a 'bootstrapping' technique (repeated sampling, with replacement, from the dataset) 1,000 times (see Figure 1). The median death rate (%) for each MEWS score was: 3 = 1.5%; 4 = 0.7%; 5 = 2.2%; 6 = 3.0%; 7 = 3.0%; 8 = 5.6%; 9 = 9.7%; 10 = 13.3%; 11 = 15.9%; 12 = 20.9%; 13 = 32.0%; 14 = 47.2%; 15 = 51.9%; 16 = 54.6%; 17 = 66.7%. MEWS scores of 2 or less (n = 19) had no deaths.

Figure 1

Distribution of bootstrapped estimates of death rates by MEWS score (median, interquartile and observed range).

Conclusion

The MEWS score can be used as a useful predictor of outcome in a HDU.