Predicting the unpredictable in cardiothoracic surgery

Excerpt

The realisation of the value and merit of outcome research in the medical field developed very early in cardiac surgery. Risk prediction thus took pole position and has been a work in progression for almost half a century. As the computing powers evolved and magnified, the simple ‘risk scores’ of yesteryears have yielded to more complicated and detailed prediction models, which are usually based on multivariate logistic regression. Initially, these models used to be static, but as the knowledge dawned that their predictive value deteriorates over time, the concept of periodical revision of these systems evolved. With a further awakening that these models can be used for benchmarking of institutions and for aiding decision-making with regard to patient selection and the expected outcomes, besides being indispensable for quality monitoring, additional impetus has been imparted to the development of innovative hybrid models. …