Cross-sectional analysis of site- and ambient temperature–related variability in noncontact infrared thermometry among healthy Black African adults

Abstract

Objectives

Noncontact infrared thermometers are widely used for rapid screening, particularly during infectious disease emergencies. Although noncontact use is time-efficient and helps maintain a safe distance, their readings may be influenced by the anatomical measurement site and ambient temperature, and evidence from Black African populations remains limited. This study examined the variability and agreement of noncontact infrared thermometer–derived surface temperature readings obtained from different anatomical sites under two controlled ambient temperature conditions in healthy Black African adults.

Methods

In this cross-sectional, repeated measures study, noncontact infrared thermometer–derived temperatures were obtained from 216 healthy adults using sites including the forehead, neck, temple, and wrist (in that order) in an environmental chamber maintained at 18°C and subsequently at 28°C. Differences across anatomical sites and ambient conditions were assessed using the Wilcoxon signed-rank and the Friedman tests with Bonferroni correction. Bland–Altman analysis and Cohen's kappa were used to assess agreement between readings and temperature classification obtained under the two ambient conditions.

Results

Increasing the ambient temperature from 18°C to 28°C significantly increased forehead, neck, and temple readings (all p < 0.001), whereas wrist readings did not change significantly. Temperature readings differed significantly among anatomical sites at both ambient conditions (p < 0.001). Most neck (94.9%) and forehead (74.5%) readings were classified within the predefined normal range, whereas wrist and temple readings were frequently classified below this range. Agreement in temperature classification between ambient conditions was fair for the neck (κ = 0.42, p < 0.001) and lower for the other sites.

Conclusion

Noncontact infrared thermometer–derived surface temperature readings vary substantially according to anatomical site and ambient temperature in healthy Black African adults. The findings confirm the relevance of previously identified site- and environment-related measurement variability in this underrepresented population and support the use of standardized, site-specific noncontact infrared thermometer measurement protocols.

Introduction

Background

Body temperature is a fundamental vital sign routinely assessed, with normal values typically ranging between 36.5°C and 37.5°C. ¹ Multiple physiological and environmental factors influence body temperature measurements, including age, sex, physical activity, circadian rhythms, health status, measurement location, and, critically, ambient temperature conditions. ² Among these factors, the measurement site and environmental conditions represent controllable variables that can significantly impact the accuracy and reliability of temperature assessments.

The core body temperature, defined as the temperature of the internal organs, such as the heart, liver, and brain, provides a more accurate reflection of overall physiological status compared with peripheral measurements.^3,4 Although blood temperature measurement remains the gold standard for core temperature assessment, alternative sites, including the tympanum, rectum, urinary bladder, and nasopharynx, may serve as clinically acceptable proxies.^4–6 However, these methods may be inadequate for balancing the requirements of accuracy and rapid hygienic temperature screening in clinical and public health environments.

Given the need for temperature assessment in clinical and public health decision making, coupled with the need for infection control and rapid screening coverage capabilities, noncontact infrared thermometers (NCITs) have emerged as important tools. The use of NCITs gained popularity during the coronavirus disease 2019 (COVID-19) pandemic.^7–9 As close contact needed to be avoided to prevent cross-infection, NCITs provided a contactless screening tool to prevent spread of infections. ¹

Currently, NCITs are widely adopted for rapid temperature screening in both clinical and public health settings, ranging from hospitals and outpatient clinics to airports and other mass-entry points. In practice, several anatomical sites are commonly used for measurement, with the same reference ranges used to define febrile and hypothermic states regardless of the anatomical site or ambient conditions. ¹⁰ We hypothesized that NCIT-derived temperature readings would differ significantly based on ambient temperature conditions and anatomical site of measurement. Several studies have investigated the performance of NCIT at specific sites. According to the study by Lai et al., neck measurements yielded the most reliable NCIT-derived temperature readings, although performance decreased at temperatures <18°C. ⁹ However, in their study, ambient temperature was an observed variable rather than a controlled one. Chen et al. reported that wrist temperature measurements were as reliable as, or more reliable than, forehead temperature measurements during the COVID-19 pandemic. ¹¹ However, their research was limited by non-manipulated temperature conditions and included febrile individuals, which might have introduced confounding factors. Furthermore, Black African adults have been inadequately represented in studies evaluating the effects of anatomical site and controlled ambient temperature conditions on NCIT-derived measurements. To address these limitations, this study was designed to evaluate how NCIT-derived surface temperature readings differ across four anatomical measurement sites (forehead, wrist, temporal, and neck) in healthy Black African adults under controlled ambient temperatures of 18°C and 28°C.

Methods

Study design, settings, and sampling

This cross-sectional descriptive study, conducted from November 2022 to February 2023, included 216 participants. Participants were recruited consecutively from among undergraduate students of Muhimbili University of Health and Allied Sciences (MUHAS) in Dar es Salaam, Tanzania, and predominantly comprised a Black African university population. Race was not used as an inclusion or exclusion criterion. All enrolled participants were identified by the recruitment team as Black African, although racial identity and skin tone were not formally self-reported or objectively measured. Participants were included if they were apparently healthy with self-reported good health. Individuals who were ill or febrile in the past 72 h, had recently experienced >5% weight change, or were consuming medications that affect thermoregulation were excluded. All processes were performed in accordance with the Declaration of Helsinki. All participants provided written informed consent prior to enrolment.

Body temperature measurements

The body temperature was measured at different body surface sites (forehead, wrist, temporal, and neck) in a fixed sequence (forehead, neck, temporal, and wrist) for all participants, using the NCIT (TP-500, Contec Medical Systems Co., Ltd; China) in an environmental chamber. All measurements were performed by a single trained operator to ensure consistency. The device was held at the manufacturer-recommended distance of 3 cm from the skin surface, perpendicular to the measurement site. A single reading was obtained at each site per ambient condition. Participants were instructed to sit quietly for at least 10 min before measurements began in a setting with a temperature of 18°C and to remove any clothing, jewelry, or cosmetics from the measurement sites. Following completion of measurements at 18°C, the ambient temperature was raised to 28°C, and participants were allowed a 5-min acclimatization period before repeat measurements.

Data analyses

All collected data were entered and cleaned, study participant identifiers were removed, and analyses were performed using the STATA 18 software. NCIT-derived temperatures at each of the four anatomical sites (forehead, wrist, temporal, and neck) were compared in three ways: (a) between the two ambient conditions (18°C vs. 28°C) for each site separately; (b) across all four sites at each ambient condition; and (c) by assessing agreement in temperature classification between the two ambient conditions for each site. First, readings obtained at 18°C and 28°C ambient temperatures were compared for each anatomical site using the Wilcoxon signed-rank test. Second, differences across anatomical sites within each ambient temperature condition were evaluated using the Friedman test, followed by post hoc Wilcoxon signed-rank pairwise comparisons with manual Bonferroni correction for multiple tests. Third, temperature readings were categorized as below normal (<36.5°C), normal (36.5 to 37.5°C), or elevated (>37.5°C), and agreement in classification between the two ambient conditions was assessed for each anatomical site using Cohen's kappa. ⁵ Bland–Altman analysis was used to assess agreement in continuous temperature readings between the two ambient conditions for each anatomical site. A p-value <0.05 after correction for multiple tests was considered statistically significant. The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. ¹²

Results

Participants’ demographic characteristics

In total, 216 participants were enrolled into the study. Most participants were male (n = 120; 55%). The median age of the participants was 22 (20 to 25) years, with 76.4% (n = 165) aged ≤25 years.

Variation in body temperature measured at different body surface sites at different ambient temperatures

At 18°C, the highest median body temperature was recorded at the neck (36.6°C (36.4°C to 36.9°C)), and the lowest was recorded from the temporal area (36.1°C (35.9°C to 36.5°C)). When the ambient temperature was increased to 28°C, the temperatures at all sites increased (Figure 1). The temperature measured at the neck was the highest median temperature (36.8°C (36.4°C to 37.7°C)), whereas that recorded at the wrist was the lowest (36.35°C (34.0°C to 36.7°C)). The median difference in body temperature between ambient conditions was 0.1°C (−0.2°C to −0.6°C) at the forehead, 0.1°C (−0.2°C to −2.6°C) at the neck, 0.4°C (−0.1°C to −0.9°C) at the temporal region, and 0°C (−2.3°C to −1.1°C) at the wrist.

Figure 1.

Body temperatures measured at different body surface sites at 18°C and 28°C ambient temperatures (N = 216, Wilcoxon signed-rank test).

The paired median comparisons revealed that temperatures recorded at an ambient temperature of 28°C were significantly higher at the forehead (36.6°C (36.4°C to 36.9°C) vs. 36.5°C (36.1°C to 36.8°C), z = 9.4, p < 0.001), neck (36.8°C (36.4°C to 37.7°C) vs. 36.6°C (36.1°C to 36.8°C), z = 12.2, p < 0.001), and temporal region (36.5°C (36.1°C to 36.9°C) vs. 36.1°C (35.9°C to 36.6°C), z = 12.6, p < 0.001) but not at the wrist (36.35°C (34.0°C to 36.7°C) vs. 36.3°C (35.5°C to 36.5°C), z = 1.6, p = 0.117). The Friedman multiple comparisons test revealed significant differences in body temperature readings across all four sites at an ambient temperature of 18°C (X² (3, n = 216) = 539, p < 0.001) and 28°C (X² (3, n = 216) = 528, p < 0.001).

Categorization of body temperature measured at different body surface sites at different ambient temperatures

At 18°C ambient, temperatures measured at the forehead (77.8%, n = 168) and neck (96.3%, n = 208) were predominantly classified as normal, whereas those measured at the temporal region (97.2%, n = 210) and wrist (89.8%, n = 194) were primarily classified as below normal. When the ambient temperature was raised to 28°C, the measurements at the forehead and neck sites maintained their normal classification patterns, with 74.5% (n = 161) of the measurements at the forehead and 94.9% (n = 205) of those at the neck classified as normal. In contrast, most participants (67.6%, n = 146) with below-normal temporal temperatures at 18°C demonstrated normal temperatures at 28°C. Elevated temperatures were rare across all sites, with only neck measurements showing any above-normal classifications (1.3%, n = 3). Most participants with below-normal wrist temperatures at 18°C (75%, n = 162) maintained this classification at 28°C.

Agreement between body temperature measurements at different ambient temperatures

Based on Bland–Altman analysis (Figure 2), forehead measurements had a mean bias of 0.1°C, with 95% limits of agreement (LoA), ranging from −0.16°C to 0.36°C, without proportional bias (p = 0.224). Neck measurements had a mean bias of 0.19°C, with 95% LoA, ranging from 0.3°C to 0.69°C, and no evidence of proportional bias (p = 0.829). Temporal measurements had a mean bias of 0.41°C, with 95% LoA, ranging from −0.09°C to 0.91°C and proportional bias (p < 0.001). Wrist measurements had a mean bias of 0.02°C, with 95% LoA, ranging from −0.50°C to −0.54°C and proportional bias (p < 0.001). When measurements were categorized against the normal range, Cohen's Kappa agreement between 18°C and 28°C ambient temperature on classifying body temperature was achieved only for neck measurements (κ = 0.42, p < 0.001), representing fair agreement (Table 1).

Figure 2.

Bland–Altman plots showing agreement in body temperature measurements between 18°C and 28°C ambient temperatures (N = 216).

Table 1.

Cross tabulation of temperature classification at 18°C ambient temperature versus 28°C (N = 216, Cohen's kappa agreement)

18°C 28°C	Below normal	Normal	Above normal	Kappa	p-value
Forehead
Below normal	3 (1.4)	7 (3.2)	0 (0.0)	0.03	0.545
Normal	45 (20.8)	161 (74.5)	0 (0.0)
Elevated	0 (0.0)	0 (0.0)	0 (0.0)
Neck
Below normal	3 (1.4)	0 (0.0)	0 (0.0)	0.42	<0.001
Normal	5 (2.3)	205 (94.9)	0 (0.0)
Elevated	0 (0.0)	3 (1.4)	0 (0.0)
Temporal
Below normal	64 (29.6)	0 (0.0)	0 (0.0)	0.02	0.107
Normal	146 (67.6)	6 (2.8)	0 (0.0)
Elevated	0 (0.0)	0 (0.0)	0 (0.0)
Wrist
Below normal	162 (75.0)	16 (7.4)	0 (0.0)	0.08	0.208
Normal	32 (14.8)	6 (2.8)	0 (0.0)
Elevated	0 (0.0)	0 (0.0)	0 (0.0)

Discussion

The findings of this study demonstrate significant variations in body temperature readings obtained using a NCIT across different body surface sites and under varying ambient conditions. The neck consistently recorded the highest temperatures, whereas the wrist and temporal regions produced lower and more variable readings. Ambient warming from 18°C to 28°C resulted in a significant rise in forehead, neck, and temporal temperatures, whereas wrist measurements remained largely unaffected. Importantly, only the neck demonstrated fair statistical agreement in temperature classification between the two ambient conditions. Our findings are consistent with previous reports suggesting that noncontact infrared thermometry is highly sensitive to measurement site and environmental factors.^9,13–16 Importantly, the present findings demonstrate that these site- and ambient temperature–related variations are also evident in a healthy Black African university cohort, a population that has been inadequately represented in previous NCIT studies.

Studies in both clinical and public health settings have shown that exposed and peripheral regions, such as the wrist, are more susceptible to environmental cooling, yielding lower values than central sites such as the forehead or neck.^9,17 The fair agreement observed at the neck may be explained by its anatomical proximity to the body's core, rich vascular supply, and the presence of large superficial vessels such as the carotid arteries, which facilitate thermal exchange and provide a more physiologically stable temperature signal. These characteristics may have contributed to its relative insensitivity to ambient fluctuations under the tested conditions, suggesting that the neck is a more stable site for NCIT measurement in variable environments.^9,18–20 However, these findings do not establish its superiority for fever detection without comparison to a core temperature reference. Moreover, although fair “statistical” agreement at the neck indicates a degree of classification consistency across ambient conditions, it may also imply a potential proportion of consistent misclassifications. This underscores that the utility of NCIT-based screening, even at the most consistent site, is contingent on standardized ambient protocols. Conversely, the poor agreement at the forehead, wrist, and temporal areas highlights their unreliability in standard screening protocols.^9,14,19,20

The physiological basis for these findings may lie in skin perfusion and thermoregulation. The neck is richly vascularized and in proximity to major arteries, which may buffer temperature fluctuations and better approximate core temperature.^4,5,21–23 In contrast, the forehead and distal regions, such as the wrist, exhibit greater vasomotor responsiveness to ambient changes, explaining their persistently lower and unstable readings. ⁵ Several empirical studies support this. Berksoy et al. found that in children, neck/carotid skin temperature measurements more closely estimated rectal temperature than forehead measurements. ²⁴ Suhail et al. observed that carotid-site infrared readings outperform forehead measurements for approximating core body temperature, and a pediatric triage study has shown significantly greater sensitivity of neck measurements over forehead measurements for fever detection (2–6 years age group).^25,26 The fact that temporal readings shifted from predominantly “below normal” at 18°C to “normal” at 28°C suggests that this site is highly sensitive to environmental warming and may not reliably indicate febrile states at cooler ambient conditions.

In our study, the Bland–Altman analysis (Figure 2) further revealed that sites without proportional bias (forehead and neck) exhibited consistent mean differences across ambient temperatures, reflecting stable measurement behavior. In contrast, the proportional bias observed at the temporal and wrist sites indicates temperature-dependent measurement errors, which introduce a risk of misclassification during mass screening under varying real-world temperatures. However, Chen et al. compared wrist and forehead NCIT readings against tympanic temperature and concluded that wrist measurements showed superior performance for temperature screening. ¹¹ However, their study was conducted in a single ambient environment without systematic manipulation of temperature conditions, and participants included individuals with active illness, which may have introduced thermoregulatory confounders absent in a healthy cohort.

Although significant differences were noted across various sites and conditions, the clinical relevance of these findings must be assessed. Median temperature variations ranged from 0.02°C at the wrist to 0.41°C at the temporal site. Differences may not be individually clinically significant but could influence the classification in mass population screenings, especially for readings close to the normal–elevated temperature threshold.

Strengths of the present study include the use of controlled ambient conditions and recruitment of healthy young adults. Previous studies evaluating NCIT performance have predominantly compared infrared readings against a reference standard (e.g. tympanic, rectal, and blood temperature) in clinical or febrile populations, commonly with a focus on a single measurement site.^9,14,17,20 Fewer studies have examined multiple peripheral sites systematically, and an even smaller number have done so under controlled ambient conditions in healthy adults. Methodologically, the present study differs in that it did not compare NCIT readings against a core temperature reference; instead, it used a controlled environmental chamber to isolate the effects of ambient temperature on site-specific readings. The study added important external relevance by simulating both cooler and warmer environments typically encountered in real-world screening contexts, such as hospitals, airports, and other public health settings. The study also extends existing evidence to a Black African population underrepresented in the NCIT literature.

These strengths should be interpreted in consideration of certain study limitations. First, tympanic temperature was not measured due to logistical constraints. Although tympanic thermometry is frequently considered the most reliable infrared-based proxy for core temperature, its requirement for contact, proper probe placement, and disposable covers limits practicality for large-scale or rapid screening applications. ⁵ Second, core body temperature was not directly measured, which prevented a definitive comparison with the gold standard. Third, only two ambient conditions (18°C and 28°C) in young adults were investigated; therefore, the findings may not be generalizable to more extreme environments or other populations, such as pediatric populations and older people. Fourth, the absence of a simultaneous contact-type reference measurement prevented the evaluation of absolute accuracy against a core temperature standard, introducing uncontrolled variability due to individual differences in autonomic thermoregulatory responses. Fifth, the 5-min acclimatization period might have been suboptimal, which may have influenced temperature readings at 28°C in a fixed sequence. Sixth, although all enrolled participants were identified by the recruitment team as Black African, racial identity and skin tone were not formally self-reported or objectively measured.^15,27 Therefore, the study cannot conclude whether the observed findings were related to skin pigmentation or other race-associated characteristics. Finally, perspiration was also not objectively assessed and may have influenced surface temperature readings.

Conclusion

Body surface temperatures measured using NCITs vary according to anatomical site and ambient temperature in healthy Black African adults. The neck yielded the most consistent measurements at 18°C and 28°C, whereas the wrist and temporal region exhibited greater variability, particularly under cooler conditions. These findings extend previous evidence to an underrepresented Black African population and support the need for standardized, site-specific NCIT measurement protocols. Where NCIT screening is used within these temperature ranges, the neck may provide more consistent readings; however, its superiority for fever detection requires confirmation against a core temperature reference. NCIT readings should therefore be interpreted cautiously, particularly when obtained from peripheral sites or under varying environmental conditions.