Every outbreak response of the past two decades has reached for the same first-line tool: point a thermal scanner or a non-contact infrared thermometer at arriving travelers and pull aside anyone running a fever. It is fast, non-invasive, and photographs well for a press briefing. It is also, according to the accumulated evidence from SARS, Ebola, H1N1, and COVID-19, one of the least reliable ways to find an infected person in a crowd.
The gap between the visibility of fever screening and its actual detection rate is now well documented enough to draw firm conclusions, and it has direct implications for anyone designing outbreak triage protocols rather than airport optics.
Fever Is a Late, Optional, and Suppressible Signal
The core problem is biological, not technical. Fever screening can only work if the infection reliably produces a fever, if that fever is present at the moment of screening, and if nothing is masking it. None of those conditions hold consistently.
A meaningful share of infections are asymptomatic or pre-symptomatic at the point of travel — for COVID-19, on the order of 40 percent of infected people may never develop a fever at all, and transmission from those who are pre-symptomatic undercuts the entire premise of catching people at the door. Even among those who do develop fever, timing varies: it can appear days into an infection, well after a traveler has already passed through screening and dispersed. And self-medication with common antipyretics can flatten a genuine fever into a normal-looking reading, whether or not the traveler intends to conceal anything.
What the Numbers Show Across Four Outbreaks
Retrospective analyses of screening programs from four separate outbreaks converge on the same conclusion: sensitivity is low, and it is low by a wide margin, not a narrow one.
| Outbreak | Screening context | Reported detection performance |
|---|---|---|
| SARS (2003) | Airport and school thermal screening | Some analyses reported sensitivity as low as 0 percent; compulsory school checks in Singapore detected no cases |
| Ebola | Point-of-entry screening | Screening alone estimated to detect only about 4.0 percent of infected travelers |
| H1N1 influenza (2009) | Airport arrival screening | Sensitivity of roughly 22.2 percent for detecting cases on arrival |
| COVID-19 | Modeled airport entry/exit screening at 86 percent device sensitivity | Still projected to miss an estimated 46 percent of infected travelers |
A separate clinical evaluation of non-contact infrared thermometers, using the common 37.3°C cutoff, found a sensitivity of only 9.43 percent against PCR-confirmed COVID-19 cases — meaning the device correctly flagged fewer than one in ten actual infections at that threshold. The European Centre for Disease Prevention and Control reached a similarly stark figure from a different angle, finding that thermal scans combined with symptom questionnaires still failed to identify up to 75 percent of infected individuals passing through.
The Devices Add Their Own Error on Top
Biology explains most of the gap, but the hardware compounds it. Regulatory guidance in the United States has been unusually blunt on this point: thermal scanners are not effective at determining whether any individual definitively has an infection, and they have not been shown to be accurate when used to take the temperature of multiple people in rapid succession, which is exactly how they are deployed in a border or terminal setting. Ambient temperature, humidity, recent physical exertion, and the angle and distance of the scan all introduce measurement drift. Some devices have even been documented to compress a genuinely elevated reading toward normal, which is the worst possible failure mode for a tool whose entire purpose is catching outliers. None of this is a case against the equipment as such — it is a case against asking a single low-cost sensor to carry the full weight of a screening decision it was never validated to make on its own.
Respiratory Rate: The Vital Sign Triage Programs Underuse
If temperature is an unreliable solo indicator, the practical response is not to abandon vital-signs screening but to widen it. Respiratory rate is, by most clinical accounts, one of the most sensitive and most neglected vital signs available to a triage team. It responds to a wide range of pathological states — respiratory compromise, sepsis, shock, early pneumonia — and it frequently shifts before heart rate or blood pressure do, giving screeners an earlier window than fever alone provides. In deteriorating patients, an abnormal respiratory rate has been linked to downstream events including cardiac arrest and ICU admission, which is a stronger evidentiary basis than thermal screening can currently claim.
Respiratory rate is not without its own limitations: manual counting is prone to observer error and short counting windows, and it demands a trained eye in a way that pointing an infrared gun does not. But combining it with temperature, rather than relying on temperature alone, is consistently described in the literature as producing a more robust screening signal than either measure in isolation.
Designing Triage Around What the Evidence Actually Supports
None of this argues for discarding temperature checks; a fever is still a useful data point when it is present, and the equipment is inexpensive enough to deploy at scale. What the evidence argues against is treating a thermal scan as a screening program in itself, or presenting a clean thermal-camera pass as meaningful reassurance to travelers or staff. Programs built around outbreak triage get more defensible results by pairing temperature with respiratory-rate observation, treating a normal thermal reading as inconclusive rather than clearing, and reserving genuine confidence for whatever diagnostic or laboratory confirmation follows behind the initial screen. The scanners have a role. It is a much smaller one than the queue of travelers filing past them tends to suggest, and treating it as anything more invites the same false confidence that let SARS, H1N1, and COVID-19 travelers pass through screening lines undetected.