Predicting Non-Invasive Ventilation in ALS Patients using Time Windows

Amyotrophic Lateral Sclerosis (ALS) is the most common neurodegenerative disorder of young adults. ALS patients usually present a rapidly progressive motor weakness caused by motor neuron demise. Death occurs in a few years, mainly due to respiratory failure. Therefore, adequate prediction of respiratory insufficiency has great importance in ALS management. We propose a prognostic model based on time windows, where a set of initial evaluations is used to predict the probability that an ALS patient will require non-invasive ventilation (NIV) k days after the last observation. Data includes clinical features, functional impairment scores and different respiratory measures. Our experimental results suggest a good performance, achieving an area under the receiver operating characteristics curve (AUC) of 84.64%, 75.86% and 77.06% for, respectively, the windows of 90, 180 and 365 days, using a Bayesian classifier, which can provide a risk measure associated to respiratory failure.