Identifying Critical Design Parameters for Improved Body Temperature Measurements: A Clinical Study Comparing Transient and Predicted Temperature Measurements

Abstract

Readily available store brand, or “home,” thermometers are used countless times in the home and clinic as a first diagnostic measure of body temperature. Measurement inaccuracies may lead to unnecessary medical visits or medication (false positives), or, potentially worse, lack of intervention when a person is truly sick (false negatives). A critical first step in the design process is to determine the shortcomings of the existing designs. For this project, we evaluated the accuracy of three currently available store brand thermometers in a pediatric population. The accuracies of the thermometers were assessed by comparing their body temperature predictions to those measured by a specially designed and calibrated and fast-responding reference thermometer. The reference thermometer was placed at the measurement site simultaneously with the store brand thermometer and recorded the temperature at the measurement site continuously. More than 300 healthy or sick pediatric subjects were enrolled in this study. Temperatures were measured at both the oral and axillary (under the arm) sites. The store brand thermometer measurements characteristically deviated from the reference thermometer temperature after 120 s, and the deviations did not follow a consistent pattern. The Brand C thermometers had the greatest deviations of up to 3.7 °F (2.1 °C), while the Brand A thermometers had the lowest deviations; however, they still deviated by up to 1.9 °F (1.1 °C). The data showed that the tested store brand thermometers had lower accuracy than the ±0.2 °F (0.1 °C) indicated in their Instructions for Use. Our recorded reference (transient) data showed that there was a wide variation in the transient temperature profiles. The store brand thermometers tested stated in their documentation that they are able to predict a body temperature based on transient temperature values over the first 5–10 s of measurements, implying that they use an embedded algorithm to extrapolate to the steady-state temperature. Significant deviations from the maximum temperature after time t = 4.6t0.63 illustrated that the transient temperature profiles may not be represented by an exponential function with a single time constant, t0.63. The accuracy of those embedded algorithms was not confirmed by our study, since the predicted body temperatures do not capture the large variations observed over the initial 10 s of the measurements. A thermometer with an error of several degrees Fahrenheit may result in a false positive or negative diagnosis of fever in children. The transient temperature measurements from our clinical study represent unique and critical data for helping to design the next generation of readily available, highly accurate, home thermometers.