Risk Prediction for Myocardial Infarction via Generalized Functional Regression Models


Statistical learning
Health Analytics
Risk Prediction for Myocardial Infarction via Generalized Functional Regression Models
Saturday 22nd December 2012
Ieva, F.; Paganoni, A.M.
Download link:
In this paper, we propose a generalized functional linear regression model for a binary outcome indicating the presence/absence of a cardiac disease with a multivariate functional data among the relevant predictors. In particular the motivating problem is an analysis of Electrocardiographic (ECG)traces of patients whose prehospital ECG has been sent to 118 Dispatch Center of Milan (the Italian free-toll number for emergencies) by life support personnel of the basic rescue units. The statistical analysis starts with a preprocessing step of ECGs, treated as multivariate functional data. They are reconstructed from noisy observations, then the biological variability is removed by a nonlinear registration procedure based on landmarks. Thus, a Multivariate Functional Principal Component Analysis (MFPCA) is carried out on the variance-covariace matrix of the reconstructed and registered ECGs as well as of their first derivatives, in order to perform a data-driven dimensional reduction. The scores of the principal components that result to be significant are then used within a generalized functional regression model, together with other standard covariates of interest. Hence, a new semi-automatic diagnostic procedure is proposed to model the probability of disease (in the case of interest, the probability of being affected by Left Bundle Brunch Block) and to classify patients. Finally, the performance of this classification method is evaluated through cross validation and compared with other methods proposed in literature.
This report, or a modified version of it, has been also submitted to, or published on
Statistical Methods in Medical Research (2013) vol. 1, p. 1-17 doi: 10.1177/0962280213495988