Explainable time-series deep learning models for the prediction of mortality, prolonged length of stay and 30-day readmission in intensive care patients

Abstract

BackgroundIn-hospital mortality, prolonged length of stay (LOS), and 30-day readmission are common outcomes in the intensive care unit (ICU). Traditional scoring systems and machine learning models for predicting these outcomes usually ignore the characteristics of ICU data, which are time-series forms. We aimed to use time-series deep learning models with the selective combination of three widely used scoring systems to predict these outcomes.Materials and methodsA retrospective cohort study was conducted on 40,083 patients in ICU from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database. Three deep learning models, namely, recurrent neural network (RNN), gated recurrent unit (GRU), and long short-term memory (LSTM) with attention mechanisms, were trained for the prediction of in-hospital mortality, prolonged LOS, and 30-day readmission with variables collected during the initial 24 h after ICU admission or the last 24 h before discharge. The inclusion of variables was based on three widely used scoring systems, namely, APACHE II, SOFA, and SAPS II, and the predictors consisted of time-series vital signs, laboratory tests, medication, and procedures. The patients were randomly divided into a training set (80%) and a test set (20%), which were used for model development and model evaluation, respectively. The area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and Brier scores were used to evaluate model performance. Variable significance was identified through attention mechanisms.ResultsA total of 33 variables for 40,083 patients were enrolled for mortality and prolonged LOS prediction and 36,180 for readmission prediction. The rates of occurrence of the three outcomes were 9.74%, 27.54%, and 11.79%, respectively. In each of the three outcomes, the performance of RNN, GRU, and LSTM did not differ greatly. Mortality prediction models, prolonged LOS prediction models, and readmission prediction models achieved AUCs of 0.870 ± 0.001, 0.765 ± 0.003, and 0.635 ± 0.018, respectively. The top significant variables co-selected by the three deep learning models were Glasgow Coma Scale (GCS), age, blood urea nitrogen, and norepinephrine for mortality; GCS, invasive ventilation, and blood urea nitrogen for prolonged LOS; and blood urea nitrogen, GCS, and ethnicity for readmission.ConclusionThe prognostic prediction models established in our study achieved good performance in predicting common outcomes of patients in ICU, especially in mortality prediction. In addition, GCS and blood urea nitrogen were identified as the most important factors strongly associated with adverse ICU events.

Clinical nomogram prediction model to assess the risk of prolonged ICU length of stay in patients with diabetic ketoacidosis: a retrospective analysis based on the MIMIC-IV database

“…Our study also found that the use of vasoactive agents was an independent risk factor for prolonged ICU LOS in patients with DKA, consistent with findings in other diseases [ 26 ]. The use of vasoactive agents suggests a state of hypoperfusion and hemodynamic instability [ 27 ], requiring prolonged ICU monitoring compared to patients not using such agents, ultimately resulting in prolonged ICU LOS.…”

Predicting length of stay ranges by using novel deep neural networks

“…In addition to being a major indicator of the consumption of hospital resources, LOS can be considered a metric that can be used to identify the severity of illness [ 21 ] and provide an enhanced understanding of the flow of patients through hospital care units and environments; understanding the flow of patients is important in the evaluating the operational functions of various care systems [ 38 ].…”

“…There have been several studies on predicting LOS [ 4 ]. The methods used in these studies included human prediction by experienced physicians [ 5 ], prediction by using regression models [ [6] , [7] , [8] ], prediction by using machine learning models [ [9] , [10] , [11] , [12] , [13] , [14] ], and prediction by using deep learning models [ 1 , [15] , [16] , [17] , [18] , [19] , [20] , [21] ].…”

“…This type of task is also known as a classification task. In studies of classification tasks, the 5 days [ 24 ], 7 days [ [25] , [26] , [27] ], and the 3rd quartile of LOS [ 21 , 28 ] are the most regularly used dichotomous time intervals. The LOS ranges were the same for all patients in the same classification, as they were preset.…”

A comparison of random survival forest and Cox regression for prediction of mortality in patients with hemorrhagic stroke

“…Traditional scoring systems commonly used in clinical practice include acute physiology and chronic health evaluation(APACHE II) [ 6 ], sequential organ failure assessment(SOFA) [ 7 ], Oxford acute severity of illness score(OASIS) [ 8 ], and simplified acute physiology score(SAPSII) [ 9 ], which include various variables with their respective point assignment scheme [ 10 ]. However, these traditional scores are applicable to a wide population, whose effectiveness in predicting specific diseases’ prognosis is not always satisfactory [ 11 , 12 ], the application of these scores in HS is limited. Many scholars have made efforts to construct predictive tools for HS.…”