A real time heat strain risk classifier using non-invasive measures of heart rate and skin temperature

For the military and civilian first responders heat injury is a very real concern. Several real-time physiological monitoring systems exist that can utilize heat strain indexes [e.g. physiological strain index (PSI)] and provide alerts to medical personnel. However, these systems depend on core temperature measurement using ingested pill thermometers which can suffer inaccuracies from ingested water. In order to find a better solution and to overcome this problem we suggest the use of a layered heat strain management system which identifies individuals “at risk” from heat injury using noninvasive measures. The intent is to identify individuals that need closer monitoring or heat injury mitigation strategies. This paper proposes a logistic regression classification model built from a data set containing 81 bouts of exercise from 49 subjects with and without personal protective equipment. Labels of “at risk” and “not at risk” were determined a priori based upon a PSI threshold of 7.5. The model has a classification error rate of 10% with only one false negative. An earlier classification model and a least squares regression model had classification errors of 21% and 14%, respectively. In addition classifying “at risk” subjects the model provides a decision boundary that can be set based upon mission needs and risk. We conclude that the logistic regression model would provide a valuable tool in a layered heat strain management system.