Maaike Swets

157 S/F94 as a proxy for COVID-19 severity 7 Our estimator for the expected change in yi associated with the treatment effect is obtained by subtracting Equation 2 from Equation 5, dividing by and averaging over the sample, This is a consistent estimator for , which follows from the fact that are consistent estimators for β0 , β1 , β. In the results presented in this paper, a treatment effect that is a relative risk reduction of e.g. 15% means that is defined by That is, the patient’s predicted risk of mortality is multiplied by 0.85. Protocolised S/F94 In the ISARIC dataset, S/F94 measurements are made opportunistically. We expect protocolised S/F94 measurements to be more precise and to be a better predictor of mortality that opportunistic measurements. This in turn implies that smaller sample sizes will be required when using protocolised S/F94 measurements as an intermediate endpoint in clinical trials. In order to calculate the magnitude of this effect, we assume a measurement error model of the following form where is a protocolised measurement of S/F94, are scalar coefficients, and is a residual. We will use the notation

RkJQdWJsaXNoZXIy MTk4NDMw