Sepsis prediction, early detection, and identification using clinical text for machine learning: a systematic review

Abstract

Objective To determine the effects of using unstructured clinical text in machine learning (ML) for prediction, early detection, and identification of sepsis. Materials and methods PubMed, Scopus, ACM DL, dblp, and IEEE Xplore databases were searched. Articles utilizing clinical text for ML or natural language processing (NLP) to detect, identify, recognize, diagnose, or predict the onset, development, progress, or prognosis of systemic inflammatory response syndrome, sepsis, severe sepsis, or septic shock were included. Sepsis definition, dataset, types of data, ML models, NLP techniques, and evaluation metrics were extracted. Results The clinical text used in models include narrative notes written by nurses, physicians, and specialists in varying situations. This is often combined with common structured data such as demographics, vital signs, laboratory data, and medications. Area under the receiver operating characteristic curve (AUC) comparison of ML methods showed that utilizing both text and structured data predicts sepsis earlier and more accurately than structured data alone. No meta-analysis was performed because of incomparable measurements among the 9 included studies. Discussion Studies focused on sepsis identification or early detection before onset; no studies used patient histories beyond the current episode of care to predict sepsis. Sepsis definition affects reporting methods, outcomes, and results. Many methods rely on continuous vital sign measurements in intensive care, making them not easily transferable to general ward units. Conclusions Approaches were heterogeneous, but studies showed that utilizing both unstructured text and structured data in ML can improve identification and early detection of sepsis.

Use of unstructured text in prognostic clinical prediction models: a systematic review

“…This demonstrates that the text data available in the EHR is a source of valuable information able to improve model performance in addition to structured data. Yan et al 14 found comparable results in a review of 9 studies predicting sepsis. Although on average text-based models did not outperform structured-data models, we did see an interesting difference between clinical settings.…”

“…Recently, Yang et al 13 performed a large review of 579 prognostic prediction models, expanding on the review by Goldstein et al, 2 but neither focused on the use of text data. Another review, by Yan et al 14 studied the use of unstructured text in, specifically, early sepsis prediction. To our knowledge, no broad systematic review has been conducted on the development of text-based prognostic prediction models.…”

Bending the patient safety curve: how much can AI help?

“…Indeed, a recent review found over 1800 published studies of AI programs developed to predict sepsis in patients hospitalized or in the emergency room. However, none of these models have been widely adopted 11 . The resulting vacuum has been filled by a large commercial EHR vendor that developed its own proprietary model which it deployed to hundreds of US hospitals without any published critical evaluation 10 .…”

“…The prediction of sepsis for inpatients, a common condition with a high mortality rate, is an area of intense AI focus in health care [10][11][12] . Many studies have shown early detection and treatment of patients with sepsis can markedly reduce mortality.…”

Research frontiers and trends in the application of artificial intelligence to sepsis: A bibliometric analysis

“…There was a evidence that early identification and diagnosis were critical, and appropriate interventions can significantly improve the prognosis of patients with sepsis ( 4 , 5 ). However, due to the complex pathophysiological conditions of sepsis, individual differences, and delayed laboratory results, it often leads to the lack of early detection and treatment of sepsis ( 6 , 7 ).…”