Using machine learning to predict deterioration of symptoms in COPD patients within a telemonitoring program

Abstract

Abstract COPD exacerbations have a profound clinical impact on patients. Accurately predicting these events could help healthcare professionals take proactive measures to mitigate their impact. For over a decade, telEPOC, a telehealthcare program, has collected data that can be utilized to train machine learning models to anticipate COPD exacerbations. The objective of this study is to develop a machine learning model that, based on a patient’s history, predicts the probability of an exacerbation event within the next 3 days. After cleaning and harmonizing the different subsets of data, we split the data along the temporal axis: one subset for model training, another for model selection, and another for model evaluation. We then trained a gradient tree boosting approach as well as neural network-based approaches. After conducting our analysis, we found that the CatBoost algorithm yielded the best results, with an area under the precision-recall curve of 0.53 and an area under the ROC curve of 0.91. Additionally, we assessed the significance of the input variables and discovered that breathing rate, heart rate, and SpO2 were the most informative. The resulting model can operate in a 50% recall and 50% precision regime, which we consider has the potential to be useful in daily practice.