Statistical analyses were performed using the R software package (version 2.10). Descriptive statistics were used for demographic data. Continuous data are presented as means (SD) for normally distributed data or medians (IQR) for data not normally distributed. Categorical data presented as percentages (n). Unpaired t-test was performed for variables presented using a mean and Wilcoxon rank sum test used for variables presented using a median. Chi-squared tests or Fisher's exact test were used for variables presented using a percent.
The independent variable tested was before versus after MEWS implementation. We adjusted for severity of illness (APACHE II) between the two groups. The number of observations completed was modelled using Poisson regression. An offset was used of the potential number of observations in 24 h, for some patients who were observed for less than 24 h in the ICU, on their day of discharge, before being sent back to the ward. A p value of less than 0.05 was considered statistically significant.