Background: Following the severe acute respiratory syndrome (SARS) outbreak, remote-sensing infrared thermography (IRT) has been advocated as a possible means of screening for fever in travelers at airports and border crossings, but its applicability has not been established. We therefore set out to evaluate (1) the feasibility of IRT imaging to identify subjects with fever, and (2) the optimal instrumental configuration and validity for such testing.
Methods: Over a 20-day inclusive period, 176 subjects (49 hospital inpatients without SARS or suspected SARS, 99 health clinic attendees and 28 healthy volunteers) were recruited. Remotely sensed IRT readings were obtained from various parts of the front and side of the face (at distances of 1.5 and 0.5 m), and compared to concurrently determined body temperature measurements using conventional means (aural tympanic IRT and oral mercury thermometry). The resulting data were submitted to linear regression/correlation and sensitivity analyses. All recruits gave prior informed consent and our Faculty Institutional Review Board approved the protocol.
Results: Optimal correlations were found between conventionally measured body temperatures and IRT readings from (1) the front of the face at 1.5m with the mouth open (r=0.80), (2) the ear at 0.5 m (r=0.79), and (3) the side of the face at 1.5m (r=0.76). Average IRT readings from the forehead and elsewhere were 1 degrees C to 2 degrees C lower and correlated less well. Ear IRT readings at 0.5 m yielded the narrowest confidence intervals and could be used to predict conventional body temperature readings of < or = 38 degrees C with a sensitivity and specificity of 83% and 88% respectively.
Conclusions: IRT readings from the side of the face, especially from the ear at 0.5 m, yielded the most reliable, precise and consistent estimates of conventionally determined body temperatures. Our results have important implications for walk-through IRT scanning/screening systems at airports and border crossings, particularly as the point prevalence of fever in such subjects would be very low.