|
Abstract:
|
Biomarkers are used routinely for diagnosing sepsis among patients with systemic inflammatory response syndrome (SIRS). However, it is still unclear how useful they really are for diagnostic purposes, to some extent due to a lack of statistical methodology. While classical approaches consider only one parameter at a time, latent class regression models allow for the inclusion of several biomarkers as indicators of a latent class variable, which may be interpreted as the true disease classification. Other variables, such as body temperature and respiratory frequency, may be included as covariates predicting the latent class membership. The ability of the proposed model to classify the patients correctly is compared with existing methods, such as the ROC curve, on a dataset of SIRS patients collected at Hvidovre University Hospital in Copenhagen, Denmark.
|
