Semi-Parametric Models - An Application in Medicine

Modeling of medical data often requires the inclusion of non-linear forms of the predictors and, the Generalized Additive Models (GAMs) can provide an excellent fit in the presence of non-linear relationships and significant noise in the predictor variables. The accurate assessment of QT interval is of paramount importance since its prolongation (LQTS) is a life threatening condition. The QT interval is affected by heart rate and gender and may be adjusted to improve the detection of patients at increased risk. Bazett's formula is the most commonly used QT correction formula, and takes into account only the heart rate assessed by the RR interval, and cut-of values of corrected QT are defined according to gender. In this work we analyzed relevance of QRS, together with gender and RR to explain QT length using GAMs. Results showed that QRS and gender are significant to non-pathological QT modelling.