The National Early Warning Score: Preceding Dynamics in the Score for Those Who Suffer an In-hospital Cardiac Arrest

Power:

A sample size of 300 patients in the control group and 150 patients in the group that suffered from an IHCA would generate a power estimate of 80 percent if the difference in the median is one point on the NEWS with a standard deviate of three points.

Analysis of the Research Data:

Categorical and nonparametric data will be presented in median (25-75 percentiles).

A hypothesis testing will be performed where the documented NEWS will be categorized into low-, medium- and high-risk and divided into different timespans. The timespans will be 0-6 h, 6-12 h, 12-18 and 18-24 h preceding the IHCA. In case of multiple NEWS measurements within each timespan, the worst NEWS measurement will be chosen. Each timespan during the 24 hours will not be treated as repeated measures and will be tested by the Chi square test. This test is chosen because the data is categorical. Another hypothesis test where all the NEWS measurements will be included and each documented NEWS on the same patient will be treated as repeated measures. A binary logistic regression analysis will be performed using Generalized Estimating Equations (GEE) and modelled to fit.

In order to select the covariates that is to be included in the regression analysis a non-parametric test including all the parameters in the NEWS will be executed. The level of significance for this test will be set to p=0.2 2. Collinearity within the NEWS parameters that is to be included in the regression analysis will be tested. The level of collinearity that is accepted will be <+0,6 and >-0,6. A binomial logistic regression analysis will be executed with the selected parameters in the NEWS The outcome of the regression analysis will be presented as odds ratio (OR).