The aim of this study was to compare the body temperature measurements at tympanic, forehead and temporal sites using infrared thermometers. A total of 1576 consecutive visitors to Singapore General Hospital at two entry locations were included in this study. Pearson correlation and Bland–Altman mean difference between sites (95% confidence interval for limits of agreement) were calculated for the relationship between the three different sites of temperatures recorded (i.e. temporal, forehead and tympanic). Of all the visitors, 27 (1.7%) had fever. Moderate positive correlation was found between temporal and forehead temperature readings ( r=0.602, mean difference (temporal – forehead), (95% limits of agreement) = 0.1 (−0.8, 0.7)), and there was very weak positive correlation between tympanic and temporal temperature readings ( r=0.177, mean difference (temporal – tympanic), (95% limits of agreement) = −0.3 (−1.7, 1.1)). Sensitivity for temporal temperature readings (⩾37.5°C) to detect febrile visitors was 3.7%, specificity was 99.6%, positive predictive value was 14.3% and negative predictive value was 98.3%. Our results demonstrate that tympanic temperature readings should be used for fever screening instead of temporal or forehead readings.