Interesting article of noteŠ
http://stm.sciencemag.org/content/6/236/236ec87
Sci. Transl. Med. DOI: 10.1126/scitranslmed.3009310
- Machine Learning to the Rescue
Superman seems to know immediately when people need saving. For medical providers charged with rescuing patients after trauma, knowing when and how to intercede takes a bit more time. These clinicians rely on standard vital signs such as heart rate, blood pressure, and respiratory rate to monitor injured patients for abrupt changes that indicate the need for immediate lifesaving interventions. However, subtle changes in the complex behavior of heart rate driven by nervous-system activation may precede these abrupt vital-sign changes. Now, Liu et al. use heart-rate variability and complexity measurements to improve prediction of the need for lifesaving medical interventions such as blood transfusions, cardiopulmonary resuscitation, airway interventions, or tourniquets in trauma patients.
Heart-rate variability and complexity represent variation in the time between heartbeats and the change in the pattern of heartbeats over time, respectively. Healthy hearts tend to be both highly variable and complex over time. The authors borrowed from computer science, using machine learning via artificial neural networks, to find the right combination of vital signs, heart-rate variability, and complexity that most accurately identifies a “crashing” trauma patient in need of urgent medical intervention.
Data were captured from 104 trauma patients in real time and transferred wirelessly to the local trauma center by using a portable vital-sign monitor attached to patients during their helicopter transport. Twenty-three percent of these patients required a total of 75 lifesaving interventions. Patients who required at least one lifesaving intervention had lower heart-rate complexity than that of those who did not require intervention, implying a worse physiological insult.
The authors then used machine learning to create predictive models that could detect the need for lifesaving interventions without raising undue false alarms. They found that machine learning enabled creation of a near-perfect predictive model (receiver operating characteristics area under the curve = 0.99) using only standard heart rate, heart-rate complexity, and the Glasgow coma score—an estimate of the conscious state of a person based on eye, motor, and verbal responses. This study was limited by the small sample size and the acknowledged difficulty in extracting complex heart-rate signals from background noise. However, it appears that leveraging the power of both complex signal analysis and machine learning may be a promising way to identify those who require ongoing rescue after trauma. Such predictive power may soon save time, money, and most importantly, lives.
N. T. Liu et al., Utility of vital signs, heart-rate variability and complexity, and machine learning for identifying the need for life-saving interventions in trauma patients. Shock 10.1097/SHK.0000000000000186 (2014). [Abstract]
==
Abstract
To date, no studies have attempted to utilize data from a combination of vital signs, heart-rate variability and complexity (HRV, HRC), as well as machine learning (ML), for identifying the need for life-saving interventions (LSIs) in trauma patients. The objectives of this study were to examine the utility of the above for identifying LSI needs and compare different LSI-associated models, with the hypothesis that a ML model would be superior in performance over multivariate logistic regression models. 104 patients transported from the injury scene via helicopter were selected for the study. A wireless vital signs monitor was attached to the patient’s arm and used to capture physiologic data, including HRV and HRC. The power of vital sign measurements, HRV, HRC, and Glasgow coma score (GCS) to identify patients requiring LSIs was estimated using multivariate logistic regression and ML. Receiver-operating characteristic (ROC) curves were also obtained. 24 patients underwent 75 LSIs. After logistic regression, ROC curves demonstrated better identification for LSIs using heart rate and HRC (area under the curve [AUC] of 0.81) than using heart rate alone (AUC of 0.73). Likewise, ROC curves demonstrated better identification for LSIs using GCS and HRC (AUC of 0.94) than using GCS and HR (AUC of 0.92). Importantly, ROC curves demonstrated that a ML model using HR, GCS, and HRC (AUC of 0.99) had superior performance over multivariate logistic regression models for identifying the need for LSIs in trauma patients. Development of computer decision support systems should utilize vital signs, HRC, and ML in order to achieve more accurate diagnostic capabilities, such as identification of needs for LSIs in trauma patients.
(C) 2014 by the Shock Society