Classification Offered by the Artificial Neural Networks (ANNs) for the Patient Information: An Overview

Purpose: Study of the classification offered by the artificial neural networks (ANNs) for the “Patient Information” variable in the “Not” category in all data groups: Training, Testing and Holdout.

Background: Informed consent is a principle of medical ethics, medical law and media studies because a patient must have sufficient information and understanding before making decisions on their medical care. Pertinent information may include risks and benefits of treatments, alternative treatments, the patient’s role in treatment, and their right to refuse treatment.

Methods: This study collects data from hospitals in the Burgos University Hospital, Spain, for two years, configuring a data file with 647 cases and 9 variables, 7 of them referred to the attitude to Informed consent, Sex and Age. We perform a descriptive analysis in order to have information about the variables that make up the classification/prediction model (Artificial Neural Network), how the data are distributed by category (“Yes” and “Not”) of the “Patient Information” variable.

Results: The structure of the most efficient artificial neural network found in the classification of the categories of the “Patient Information” variable (“Yes” and “Not” categories) is the binomial Hidden layer-Output layer: Hyperbolic tangent-Softmax Dependent variable: (“Patient Information”; Partition: Training 60%, Testing 20% and Holdout 20%).

Conclusions: The classification/prediction of the “Patient Information” variable by means of the artificial neural network, perceptron, offers us the low classification/prediction of the “Not” category, which is object of this study. An empirical investigation demonstrates that adding a new covariant variable, like "Consultation time," to the network enhances the classification of the "Not" category.