Regularização de modelos para predição precoce: um estudo na predição de complicações na UTI

Abstract

Predictive models are proving to be important for decision-making in the ICU, but they often present the black box problem because they do not provide the information from the logic involved in patient-specific predictions. There are already techniques capable of analyzing the models and generating valuable explanations about their functioning. However, once the model is generated, it is important to ensure that it will continue with the same prediction logic originally intended. Knowing that complications can occur at any time during a patient’s ICU stay, we built our machine learning models using features obtained from the patient’s administrative data, laboratory results, and vital signs, available within the first hour after admission to the ICU.This enables our models to provide great anticipation because complications can occur at any moment during ICU stay. To build models that continue to work as originally designed we first propose to measure (i) how the provided explanations vary for different inputs (that is, robustness), and (ii) how the provided explanations change with models built from different patient sub-populations (that is, stability). Second, we employ these measures as regularization terms that are coupled with a feature selection procedure such that the final model provides predictions with more robust and stable explanations. Experiments were conducted on a data set containing 6000 ICU admissions of 5474 patients. Results obtained on an external validation cohort of 1069 patients with 1086 ICU admissions showed that selecting features based on robustness led to gains in terms of predictive power that varied from 6.8% to 9.4%, whereas selecting features based on stability led to gains that varied from 7.2% to 11.5%, depending on the target complication. Our results are of practical importance as our models predict complications with great anticipation, thus facilitating timely and protective interventions.