An automated machine learning-based model predicts postoperative mortality using readily-extractable preoperative electronic health record data

Abstract

Background: Rapid, preoperative identification of patients with the highest risk for medical complications is necessary to ensure that limited infrastructure and human resources are directed towards those most likely to benefit. Existing risk scores either lack specificity at the patient level or utilise the American Society of Anesthesiologists (ASA) physical status classification, which requires a clinician to review the chart. Methods: We report on the use of machine learning algorithms, specifically random forests, to create a fully automated score that predicts postoperative in-hospital mortality based solely on structured data available at the time of surgery. Electronic health record data from 53 097 surgical patients (2.01% mortality rate) who underwent general anaesthesia between April 1, 2013 and December 10, 2018 in a large US academic medical centre were used to extract 58 preoperative features. Results: Using a random forest classifier we found that automatically obtained preoperative features (area under the curve [AUC] of 0.932, 95% confidence interval [CI] 0.910e0.951) outperforms Preoperative Score to Predict Postoperative Mortality (POSPOM) scores (AUC of 0.660, 95% CI 0.598e0.722), Charlson comorbidity scores (AUC of 0.742, 95% CI

Trends in Clinical Information Systems Research in 2019

“…Due to COVID-19 restrictions, the selection meeting of the IMIA Yearbook Editorial Committee was held as a videoconference on Apr 17, 2020. In this meeting, three papers [6][7][8] were finally selected as best papers for the CIS section (Table 2). A content summary of these three CIS best papers can be found in the appendix of this synopsis.…”

“…All three of the best papers in the CIS section can be assigned to these two clusters. The contribution by Brian L. Hill and colleagues [8], who successfully created a fully automated machine-learning-based model for postoperative mortality prediction is in the yellow cluster. We selected this paper from the British Journal of Anaesthesia as the approach is innovative, uses only preoperative available medical record data, and can better predict in-hospital mortality than other state-of-the-art methods.…”

Development and validation of an interpretable neural network for prediction of postoperative in-hospital mortality

“…We and others have recently shown that deep neural networks (DNNs) and random forest algorithms, using only readily available information extracted from the electronic health record before or at the end of surgery, can successfully predict postoperative inhospital mortality with area under the curve (AUC) ranging from 0.87 to 0.93 [1][2][3] . While DNNs and other machine learning models often have higher accuracy than simpler models like logistic regression (LR), they are often considered to be "black box" models and this lack of interpretability and transparency is considered a challenge for clinical adoption 4 .…”

Multidimensional Prognostic Index (MPI) score has the major impact on outcome prediction in elderly surgical patients with colorectal cancer: The FRAGIS study

“…Risk scores commonly used for surgical patients (i.e., ASA, APACHE, P‐POSSUM) have substantial limitations when applied to elderly patients: (I) they were built and validated in past decades mainly on young adults; (II) they consider one single organ function at a time, and (III) they do not estimate the patient's physiologic reserve, functional, cognitive, and social factors, which are all recognized as relevant outcome determinants in geriatric medicine 7 . Thus, despite the widespread adoption of these scores, their ability to predict complications appears low among elderly CRC patients scheduled for surgery 8 …”

“…7 Thus, despite the widespread adoption of these scores, their ability to predict complications appears low among elderly CRC patients scheduled for surgery. 8 The comprehensive geriatric assessment (CGA) is the "evidencebased" reference tool to assess the overall health of elderly patients. 9,10 CGA is usually delivered by a multidisciplinary team led by a geriatrician alongside specialist nurses, physiotherapists, occupational therapists, and social workers.…”