Our evaluation of 3 ITDS in emergency department settings found that the FLIR and OptoTherm reliably identified elevated body temperatures. The high AUCs for these 2 systems suggest that they can differentiate between febrile and afebrile persons with relatively high sensitivity and specificity at an optimal fever cutoff. The relatively high correlation with oral temperature measurement also supports the utility of these 2 ITDS, which predicted fever better than self-reports of patients and more accurately alone than in combination with self-reported fever.
Our study is one of few that simultaneously examined the effects of multiple external and internal factors (age, sex, time of day, room temperature, and antipyretic drug use) on ITDS and oral temperature measurement accuracy. We found that ITDS and oral temperature measurements were strongly influenced by site and time of day, which may be a real effect or a result of variations in oral measurement techniques. The effects of age and time of day on body temperature found in this study have been well established by previous research (26
). We observed strong associations between ITDS and room temperatures. Similar observations with room temperatures and extended exposure to hot or cold environments have been reported (22
). The unexpected association between higher temperature measurements (oral and OptoTherm) and recent antipyretic drug use may result from patients with higher fevers taking antipyretic drugs, inadequate antipyretic drug dosage, or both. The finding that men had relatively higher ITDS measurements than women has not been previously reported and may be because of differences in facial hair, use of cosmetics, or subcutaneous fat composition (31
). Similar associations across multiple ITDS underscore the strength of these findings. By controlling for these covariates, we were able to measure the relationship between ITDS and oral temperatures with greater precision.
Although the sensitivity, specificity, and AUC of the devices we tested were similar to those found in previous studies, we observed a higher correlation between ITDS temperature measurements and confirmatory temperature measurements (15
). Several factors may have contributed to these differences. The higher correlation between ITDS and body temperatures reported here may be related to the use of oral temperature measurement as reference. Although oral temperature measurements better reflect core temperatures than infrared tympanometric measurements, most previous investigations of ITDS have used the latter as reference (19
). The preference for oral temperatures as reference is supported by an evaluation of methods for measuring body temperature conducted by the American College of Critical Care Medicine and the Infectious Diseases Society of America; researchers found that rectal temperatures were the most accurate of the peripheral thermometry methods, followed by oral, tympanic, and axillary temperature measurements, respectively (32
Many types of ITDS are available, ranging from inexpensive hand-held point-and-shoot devices with laser sighting to hand-held cameras with light-emitting diode displays, wall-mounted cameras, and portable cameras on tripods such as the ones used in this study (19
). To maximize potential efficacy, we evaluated technically advanced ITDS that were recently developed for human temperature detection. Other studies used more basic systems and did not compare different devices. Although the costs of the OptoTherm and FLIR were comparable at $22,000 and $16,000 per system, respectively, the Wahl was relatively less expensive ($8,000). Testing 3 different models at various price ranges allowed us to demonstrate substantial differences among ITDS. These differences are likely to affect their sensitivity and utility for fever screening. The systems used in this study require the person to stand in front of the camera for ≈2–3 seconds to capture a temperature. Other differences, such as moving persons, could have further affected the sensitivity of ITDS for fever detection.
Although addition of a health declaration form would allow screening to also consider recent travel history, previous fever, and other symptoms or illness exposures, health declarations have variable compliance rates and depend on a person’s ability to understand questions and accurately assess symptoms as well as willingness to report (12
). In our study, in which patients had no disincentive to report, we found that one fourth of febrile patients did not report having fever, which suggests true unawareness of fever among some persons. Only one tenth of those who reported having a fever were actually found to be febrile. Our results, therefore, probably underestimated the benefit of ITDS over self-reports of fever. In other settings, ill persons may be less likely to report symptoms for fear of adverse consequences such as travel delays, involuntary isolation of ill persons, or quarantine of exposed contacts. In settings such as travel sites (e.g., airports) and the workplace, ITDS could provide an objective means for the mass detection of fever as part of a comprehensive public health screening strategy because ITDS had greater accuracy than self-reports.
Mass health screening during a pandemic will certainly be influenced by several other factors, including perceived and actual pandemic severity, as well as the potential consequences of illness detection, either negative or positive, which can affect the sensitivity of screening that uses self-report. If being detected as febrile is perceived as harmful, travelers may hide their symptoms (12
). Alternatively, during a pandemic with high mortality rates, incentives for reporting symptoms might be present, such as access to scarce antiviral medications and medical care. In both situations, a comprehensive screening approach may be necessary, which uses ITDS for fever screening and a health questionnaire to detect other symptoms or exposures that would increase specificity of the screening process. Finally, the usefulness of any infectious disease screening must take into account temperature fluctuations, use of antipyretic medications, transmission risks, prevalence of infections, and asymptomatic infections.
This study had several limitations. Measurement error resulting from variation in digital oral thermometer measurement and technique may have decreased the correlation between ITDS and oral temperature measurements (38
). For FLIR and Wahl, varying readouts by different operators may have led to increased variability. This method, although necessary for direct temperature comparisons, may have decreased the accuracy of FLIR and Wahl. Use of alarm features as recommended by the manufacturers could minimize these differences but might lead to more false-positive results. In addition, unlike the other 2 devices, Wahl required calibration to ambient temperature once per day, but room temperatures varied within the day. We evaluated only systems submitted by manufacturers to the bid process, thus limiting the generalizability of our results to other devices.
To assess the sensitivity and specificity of different ITDS for fever detection and to determine their optimal thresholds, we validated each measurement by oral thermometry, which required a clinical setting. Thus, generalizability to settings such as airports and border crossings may be limited. Substantial delays to travelers and ethical concerns such as follow-up treatment made it impractical to conduct this study in an airport setting. In addition, although a few studies have examined screenings in airports, they confirmed temperature only in febrile persons, thus sensitivity and specificity of ITDS could not be established from such studies.
The sensitivity and specificity of screening by using ITDS are determined by the selected fever temperature cutoff, which tends to be 2–3 degrees lower than the standard fever threshold because of differences between skin and core temperatures. Increasing or decreasing sensitivity causes a reciprocal change in specificity. For example, lowering OptoTherm’s threshold from the optimal 95.7°F to 94.5°F would achieve almost 100% sensitivity but would reduce specificity to 63.6% and increase the false-positive rate to 36.4%; to reach near 100% specificity with the OptoTherm by using a cutoff of 100°F for ITDS, sensitivity decreases to 6.4%.
Maximizing accuracy by choosing the optimal cutoff with the highest sensitivity and specificity may not be practical in a real-world setting, considering the relative costs of false-positive and false-negative results. In settings where secondary evaluation is available or during a pandemic with high illness severity, ITDS temperature can be set at a lower cutoff to ensure fewer false negatives, each of which represents a potential public health threat. However, setting the cutoff to achieve very high sensitivity can result in many false positives, which could have adverse consequences to the population being screened (e.g., unnecessary travel delays, missed work) and increase the workload of staff who are conducting the screening. In settings where confirmatory testing may not be feasible or high costs may be associated with a false-positive result, a higher ITDS temperature cutoff may be preferable.