Explainable Machine Learning on AmsterdamUMCdb for ICU Discharge Decision Support: Uniting Intensivists and Data Scientists

Abstract

Objectives: Unexpected ICU readmission is associated with longer length of stay and increased mortality. To prevent ICU readmission and death after ICU discharge, our team of intensivists and data scientists aimed to use AmsterdamUMCdb to develop an explainable machine learning–based real-time bedside decision support tool. Derivation Cohort: Data from patients admitted to a mixed surgical-medical academic medical center ICU from 2004 to 2016. Validation Cohort: Data from 2016 to 2019 from the same center. Prediction Model: Patient characteristics, clinical observations, physiologic measurements, laboratory studies, and treatment data were considered as model features. Different supervised learning algorithms were trained to predict ICU readmission and/or death, both within 7 days from ICU discharge, using 10-fold cross-validation. Feature importance was determined using SHapley Additive exPlanations, and readmission probability-time curves were constructed to identify subgroups. Explainability was established by presenting individualized risk trends and feature importance. Results: Our final derivation dataset included 14,105 admissions. The combined readmission/mortality rate within 7 days of ICU discharge was 5.3%. Using Gradient Boosting, the model achieved an area under the receiver operating characteristic curve of 0.78 (95% CI, 0.75–0.81) and an area under the precision-recall curve of 0.19 on the validation cohort ( n = 3,929). The most predictive features included common physiologic parameters but also less apparent variables like nutritional support. At a 6% risk threshold, the model showed a sensitivity (recall) of 0.72, specificity of 0.70, and a positive predictive value (precision) of 0.15. Impact analysis using probability-time curves and the 6% risk threshold identified specific patient groups at risk and the potential of a change in discharge management to reduce relative risk by 14%. Conclusions: We developed an explainable machine learning model that may aid in identifying patients at high risk for readmission and mortality after ICU discharge using the first freely available European critical care database, AmsterdamUMCdb. Impact analysis showed that a relative risk reduction of 14% could be achievable, which might have significant impact on patients and society. ICU data sharing facilitates collaboration between intensivists and data scientists to accelerate model development.

Intensive Care Unit Physicians’ Perspectives on Artificial Intelligence–Based Clinical Decision Support Tools: Preimplementation Survey Study

“…Lastly, training will be needed to educate physicians on the interpretation of the mortality or readmission risk predictions, as we observed a range of answers regarding the threshold at which patients would or would not be discharged to lower care wards (Figure 3). Due to a significant imbalance in the number of patients that were or were not readmitted or died after discharge, risk predictions are skewed along the 0% to 100% scale, being concentrated around an event rate of 5.3% [25]; the respondents were not informed of this. Therefore, attention should be paid to the interpretation of these calibrated risk predictions during training, as perceptions clearly differed on what constituted high and low risks for this outcome.…”

“…This survey study is part of preimplementation research for Pacmed Critical [24]. Pacmed Critical is a machine learning-based AI-CDS tool that predicts a patient's combined readmission and mortality risk within 7 days of ICU discharge to support physicians in their decisions to discharge patients to lower care wards [25,26]. The Pacmed Critical software is intended for use as a complementary tool by qualified ICU medical professionals and will be accessed on hospital premises; it will not be used on mobile devices.…”

Modern Learning from Big Data in Critical Care: Primum Non Nocere

“…Besides, the individual effects of ML techniques are relatively hard to interpret. Although more explainable algorithms exist, effects are less intuitive to interpret than regression coefficients [103]. There are some extensions of ML techniques that allow the researcher to assume relationships between variables, for example, through graph-based neural networks [104] or Bayesian networks [102,105].…”

Optimal discharge of patients from intensive care via a data-driven policy learning framework

“…A recent review of the application of machine learning in predicting hospital readmission identified 43 relevant studies employing a variety of modeling methods (Huang et al, 2021). Additional studies have focused on predicting multiple outcomes simultaneously including both post-discharge mortality and readmission (Badawi and Breslow, 2012;Campbell et al, 2008;Ouanes et al, 2012;Thoral et al, 2021), as well as length of stay and readmission (Hilton et al, 2020).…”