Identifying Episodes of Hypovigilance in Intensive Care Units Using Routine Physiological Parameters and Artificial Intelligence: a Derivation Study (Preprint)

Abstract

BACKGROUND Delirium is a prevalent condition in intensive care units (ICUs), often leading to adverse outcomes. Hypoactive delirium is particularly difficult to detect. Despite advancements, timely identification of hypoactive delirium remains challenging due to its dynamic nature, lack of human resources, lack of reliable monitoring tools, and subtle clinical signs that include hypovigilance. Machine learning detection models could support the identification of hypoactive delirium episodes by better detecting episodes of hypovigilance. OBJECTIVE In this study, we aim to develop a machine learning algorithm capable of detecting hypovigilance events using routinely collected physiological data in the ICU. METHODS This derivation study used prospective observational data collected from eligible ICU patients in Lévis, Québec. We included patients admitted to the ICU between October 2021 and June 2022 who were at least 18 years old and had an anticipated ICU stay of at least 48 hours. ICU nurses identified hypovigilant states every hour using the Richmond Agitation and Sedation Scale (RASS) or the Ramsay Sedation Scale (RSS). Routine vital signs (heart rate, respiratory rate, blood pressure, and oxygen saturation), as well as other physiological and clinical variables (premature ventricular contractions, intubation, use of sedative medication, and fever), were automatically collected using a GE CARESCAPE Gateway or manually collected through chart review. Time series were generated around hypovigilance episodes for analysis. Random Forest, XGBoost, and LightGBM classifiers were then used to detect hypovigilant episodes on the basis of analyzing time series. Hyperparameter optimization was performed using random search in a 10-fold group cross-validation setup. We report the results of this study using the TRIPOD+AI guidelines and potential biases were assessed using PROBAST. RESULTS Out of 146 potentially eligible participants, data from 30 patients (mean age: 69 years old; 63% male) were collected for analysis. Of the group, 30% were admitted to the ICU for surgical reasons. Following data preprocessing, the study included 1,493 hypovigilance episodes and 764 non-hypovigilant episodes. Among the three sets of models evaluated, LightGBM demonstrated the best performance. It achieved an average accuracy of 68% to detect hypovigilant episodes, with a precision of 76%, a recall of 74%, an AUC of 60%, and an F1 score of 69%. Notably, the model was particularly influenced by intubation, respiratory rate, and systolic blood pressure acquired non-invasively. CONCLUSIONS All of the classifiers showed promising precision and recall, with slightly different yet comparable performance in classifying hypovigilant episodes. Machine learning algorithms designed to detect hypovigilance have the potential to support early detection of hypoactive delirium in ICU patients.