Toward Optimal Heparin Dosing by Comparing Multiple Machine Learning Methods: Retrospective Study

Abstract

Background Heparin is one of the most commonly used medications in intensive care units. In clinical practice, the use of a weight-based heparin dosing nomogram is standard practice for the treatment of thrombosis. Recently, machine learning techniques have dramatically improved the ability of computers to provide clinical decision support and have allowed for the possibility of computer generated, algorithm-based heparin dosing recommendations. Objective The objective of this study was to predict the effects of heparin treatment using machine learning methods to optimize heparin dosing in intensive care units based on the predictions. Patient state predictions were based upon activated partial thromboplastin time in 3 different ranges: subtherapeutic, normal therapeutic, and supratherapeutic, respectively. Methods Retrospective data from 2 intensive care unit research databases (Multiparameter Intelligent Monitoring in Intensive Care III, MIMIC-III; e–Intensive Care Unit Collaborative Research Database, eICU) were used for the analysis. Candidate machine learning models (random forest, support vector machine, adaptive boosting, extreme gradient boosting, and shallow neural network) were compared in 3 patient groups to evaluate the classification performance for predicting the subtherapeutic, normal therapeutic, and supratherapeutic patient states. The model results were evaluated using precision, recall, F1 score, and accuracy. Results Data from the MIMIC-III database (n=2789 patients) and from the eICU database (n=575 patients) were used. In 3-class classification, the shallow neural network algorithm performed the best (F1 scores of 87.26%, 85.98%, and 87.55% for data set 1, 2, and 3, respectively). The shallow neural network algorithm achieved the highest F1 scores within the patient therapeutic state groups: subtherapeutic (data set 1: 79.35%; data set 2: 83.67%; data set 3: 83.33%), normal therapeutic (data set 1: 93.15%; data set 2: 87.76%; data set 3: 84.62%), and supratherapeutic (data set 1: 88.00%; data set 2: 86.54%; data set 3: 95.45%) therapeutic ranges, respectively. Conclusions The most appropriate model for predicting the effects of heparin treatment was found by comparing multiple machine learning models and can be used to further guide optimal heparin dosing. Using multicenter intensive care unit data, our study demonstrates the feasibility of predicting the outcomes of heparin treatment using data-driven methods, and thus, how machine learning–based models can be used to optimize and personalize heparin dosing to improve patient safety. Manual analysis and validation suggested that the model outperformed standard practice heparin treatment dosing.

A Clinical Prediction Model to Predict Heparin Treatment Outcomes and Provide Dosage Recommendations: Development and Validation Study

“…As validated in a previous study [ 11 ], a shallow neural network model works best from among several machine learning models for the heparin outcome prediction task. In this study, we used a fully connected shallow neural network model [ 16 , 17 ] to predict the therapeutic effect (subtherapeutic, normal therapeutic, or supratherapeutic) of patients after 8 hours of heparin treatment.…”

“…The results show that the machine learning–based model can effectively predict the aPTT response after the initial dosing and in a step-by-step pattern, which can contribute to decreasing the duration of the therapeutic regime and avoiding treatment-related risks. In our previous work, we demonstrated that the shallow neural network algorithm performed best compared to algorithms such as extreme gradient boosting, adaptive boosting, and support vector machine [ 11 ]. Based on our local clinical database and modified treatment pattern, we validated our previously developed model and demonstrated the applicability of this machine learning–based algorithm for UFH treatment.…”

“…Therefore, retrospective analysis starts by extracting sequential dose response data as well as the concurrent laboratory and other clinical data, which may greatly contribute to the improvement on the feedback delay of UFH dosage optimization [ 10 ]. In our recent study [ 11 ], based on the public Multiparameter Intelligent Monitoring in Intensive Care III (MIMIC III) and electronic ICU databases [ 12 , 13 ], we compared several common models for predicting the effects of heparin treatment and showed that machine learning–based models, which outperformed the standard practices, can be used to optimize and personalize heparin dosing to improve patient safety.…”

Systematic review of machine learning models for personalised dosing of heparin

“…Study characteristics are shown in Table 1. Five studies 27,28,30,32,33 described ML models aimed at achieving a therapeutic aPTT in the intensive care unit (ICU) setting. Three studies 29,31,34 conference proceedings [28][29][30][31][32] and three in peer reviewed journals.…”

Early warning of citric acid overdose and timely adjustment of regional citrate anticoagulation based on machine learning methods

“…Therefore, to avoid the occurrence of adverse reactions, it is necessary to monitor the concentration of ionized calcium in the body, and timely adjust the calcium supplementation rate [17]. In recent years, with the rapid development of machine learning technologies, data-driven methods has been introduced in this fields to provide algorithm generated outcome predictions and dosing related clinical decision supports for clinicians [18,19]. These previous studies demonstrated that by closely monitoring the bedside patient data, it is possible to predict treatment outcomes, and to provide computer generated recommendations.…”