Background: During the SARS epidemic, mass fever screening at border control points and public hospitals was done by measuring forehead temperature by non-contact infrared thermometry. However, its accuracy is not well documented.
Methods: We evaluated the agreement of non-contact infrared forehead temperature (NIFT) measurement by comparing NIFT readings with tympanic temperatures taken in children (1 mth to 18 yrs) admitted to the general paediatric wards of Kwong Wah Hospital, Hong Kong.
Results: A total of 567 patients were recruited and 1000 pairs of readings were obtained. The incidence of fever, defined as tympanic temperature (in rectal model) >38 degrees C (100.4 degrees F), was 12.3%. The mean difference between NIFT and tympanic temperature was 2.34 degrees C (4.21 degrees F) and the 95% limit of agreement between NIFT and tympanic temperature was 0.26-4.42 degrees C (0.47-7.96 degrees F). NIFT was significantly lower than tympanic temperature readings. The optimal cut-off point of NIFT derived from the receiver-operator characteristics curve for fever definition was 35.1 degrees C (95.2 degrees F). The sensitivity, specificity, positive predictive value and negative predictive value of this cut-off point for fever screening were 89.4%, 75.4%, 33.7% and 98.1%, respectively.
Conclusions: NIFT measurement has a reasonable accuracy in detecting tympanic fever in children. However, one should be aware of the high false-positive rate of fever screening using NIFT.