Data-driven decision support for individualised cardiovascular resuscitation in sepsis: a scoping review and primer for clinicians

Abstract

BackgroundWe conducted a scoping review of machine learning systems that inform individualised cardiovascular resuscitation of adults in hospital with sepsis. Our study reviews the resuscitation tasks that the systems aim to assist with, system robustness and potential to improve patient care, and progress towards deployment in clinical practice. We assume no expertise in machine learning from the reader and introduce technical concepts where relevant.MethodsThis study followed thePreferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviewsguidance. MEDLINE, EMBASE, Scopus, ClinicalTrials.gov, arXiv, bioRxiv and medRxiv were systematically searched up to September 2021. We present a narrative synthesis of the included studies, which also aims to equip clinicians with an understanding of the foundational machine learning concepts necessary to interpret them.Results73 studies were included with 80% published after 2018. Supervised learning systems were often used to predict septic shock onset. Reinforcement learning systems were increasingly popular in the last five years, and were used to guide specific dosing of fluids and vasopressors. A minority of studies proposed systems containing biological models augmented with machine learning. Sepsis and septic shock were heterogeneously defined and 63% of studies derived their systems using a single dataset. Most studies performed only retrospective internal validation, with no further steps taken towards translating their proposed systems into clinical practice.ConclusionsMachine learning systems can theoretically match, or even exceed, human performance when predicting patient outcomes and choosing the most suitable cardiovascular treatment strategy in sepsis. However, with some notable exceptions, the vast majority of systems to date exist only as proof of concept, with significant barriers to translation.

Can Machine Learning Personalize Cardiovascular Therapy in Sepsis?

“…Despite this promise, systems from our review ( 17 ) and the latest literature show a low overall level of technological readiness, with a paucity of clinical deployment and proven patient benefit. We identified only a single published randomized controlled trial of a decision support system for cardiovascular treatment in sepsis ( 29 ).…”

“…Compounding these problems, current systems for personalized sepsis resuscitation rely heavily on a small number of publicly available datasets from the United States and Western Europe, and most commonly on the Medical Information Mart for Intensive Care III (MIMIC-III) dataset, collected from 2001 to 2012 at a single U.S. hospital ( 17 , 35 ). Systems developed using these data may generalize poorly to practice and patient populations in low- and middle-income countries, where maternal and neonatal sepsis are more common ( 36 ) and choice of vasoactive drugs may be limited ( 37 ).…”

“…Such a system would not be susceptible to fatigue or cognitive overload, and could reduce harm due to errors in frenetic resuscitation scenarios. The systems included in our scoping review (17) were trained and validated using a median of 14,908 patient episodes, far exceeding the lifetime experience of human clinicians. Such systems could be deployed at scale to monitor patients with sepsis continuously where human resources are scarce, updating their recommendations in real time as the patient's hemodynamics evolve, and alerting clinicians when alarm thresholds are reached (28).…”

“…Below, we discuss the potential for machine learning systems to personalize cardiovascular resuscitation in sepsis. For an accessible introduction to underlying machine learning concepts, we direct readers to our recent scoping review ( 17 ).…”