Diagnostic Accuracy of Three Methods of Body Temperature Measurement in Children: A Latent Class Approach

Introduction Fever is a common manifestation of acute illness among children, and it is essential to measure body temperature accurately in pediatric clinical practice. Various methods are in use, but no gold standard exists for body temperature measurement among this population. Latent class analysis (LCA) is increasingly used to assess diagnostic accuracy in the absence of a gold standard. LCA is a method that identifies unobserved groups in populations, allowing diagnostic evaluation even without a reference standard. This study aimed to assess the diagnostic accuracy of the axillary, forehead, and tympanic thermometers in children (one to five years) using LCA.  Methods A cross-sectional study was done to determine the diagnostic accuracy of axillary, forehead, and tympanic thermometers in diagnosing fever among children with LCA as a reference. The digital axillary thermometer and the infrared dual-mode (forehead and tympanic) thermometer were used for the measurements. The study was conducted among 728 in a tertiary care center in Kerala, India. Children were recruited from the pediatric medical wards after 24 hours of admission.  Results The study sample consisted of 728 children aged one to five years. LCA had identified two latent classes with a probability of 8.55% for fever and 91.45% for no fever. The sensitivity and specificity of the methods were: axillary method 96.97% and 97.67%, forehead thermometer 98.99% and 100%, and tympanic thermometer 99.7% and 67.44%. The probability of misclassification was minimal (0.10%), and a strong association between the test results and the latent class was observed. Conclusions The forehead thermometer demonstrated the highest diagnostic accuracy, followed closely by the axillary thermometer. The tympanic thermometer, though highly sensitive, exhibited a relatively high false-positive rate.