Preventing heat injuries by predicting individualized human core temperature

Heat injury is a problem for the Armed Forces, especially during deployment to localities with very hot and humid climates. Early warning of a rising core body temperature (TC) can help prevent heat injuries. To this end, we developed an algorithm that, given a series of past TC measurements obtained using an ingestible temperature pill, accumulates evidence of a rising TC over time and provides ahead-of-time warning of an impending, dangerously elevated TC. Using data from a cohort of six Soldiers involved in field exercises whose TC exceeded 38.5C, we assessed the performance of the warning algorithm. The algorithm predicted rises in TC with a clinically useful lead time (> 18 min) and reasonable sensitivity and specificity (> 87%). However, because ingestible temperature pills are impractical for monitoring a large number of Warfighters during prolonged operations, we developed a mathematical model that uses non-invasive measurements of physiological variables, such as activity (AC), heart rate (HR), and skin temperature, as well as environmental information [ambient temperature (TA) and relative humidity (RH)], to provide individualized real-time TC estimates. Using the same cohort of Soldiers, we evaluated two variants of the individualized model, one that used all the measurements (original model) and another that used AC, HR, and monthaverage TA and RH values (reduced model). The original and reduced models yielded TC estimates with average errors of 0.31C and 0.35C, respectively, which are within the physiological intra-subject variability of TC. In addition, on average, the estimation time delays ranged from 3 min for the original model to 5 min for the reduced model. We conclude that the individualized TC estimation model can be used to replace ingestible temperature pills and enable the development of a field-deployable early warning system of an impending rise in core temperature.