The World Health Organization recommends human touch to detect hypothermia in resource-limited settings. This method involves simultaneous palpation of the baby at the abdomen and soles of their feet with the dorsum of the observer's hand (1). Warm abdomen and feet correspond to a warm baby (36.5–37.5°C), warm abdomen and cold feet to mild hypothermia (36–36.4°C) and cold abdomen and feet to moderate hypothermia (<36°C) (1). The benefits of human touch are that it is simple, quick, inexpensive and easy to implement (14). However, the accuracy of this method varies widely depending on who is conducting the measurements and whether they have been trained.

When conducted by untrained mothers, health workers and field workers, the accuracy of human touch to detect hypothermia is poor when compared to mercury-in-glass axillary thermometry. One study reported a sensitivity of 11–42% and specificity of 93–100% depending on the observer (15), whilst another found that only 24.6–34.4% of hypothermic babies were correctly identified (16). However, the accuracy of human touch improves when used to detect moderate hypothermia compared to mild hypothermia (15) and hence, it is recommended as a screening tool in a community setting where observations are made by untrained workers and mothers. Ellis et al. (15) emphasize the need for further palpation and monitoring if the baby is thought to be cold based on human touch. The accuracy of human touch increases when conducted by trained workers (14, 17) and pediatricians (18), with all pediatricians in one study found to be capable of correctly identifying all hypothermic babies.

Mercury-in-glass thermometers have traditionally been considered the gold standard for temperature measurement and their use is still prevalent throughout low- and middle-income countries (19). However, they are no longer used in most high-income countries due to concerns about the risk posed by the mercury in them (20). Furthermore, when compared to newer electronic and infrared thermometers, mercury-in-glass thermometers take significantly longer to reach a stable temperature reading (21). Their accuracy is unclear as their recorded temperatures are often considered the gold standard against which other thermometers are compared (22, 23). Within clinical practice, their efficacy is often suboptimal due to uncertainties regarding where they should be placed and insufficient placement times (24).

The ideal location to place a mercury-in-glass thermometer in neonates is debatable, with rectal thermometry remaining the gold standard in many low- and middle-income countries (22). Whilst many have traditionally believed rectal temperatures to accurately reflect core temperature, numerous studies have now shown that rectal temperatures lag during rapid changes in core temperature (25) and are affected by the presence of feces and bowel organisms (26). The main advantage of placing mercury-in-glass thermometers in the rectum is the shorter placement time required for temperatures to stabilize. Kunnel et al. (24) found that 90% of rectal measurements reached their optimal temperature after 5 min, compared with 11 min for axillary measurements. However, rectal thermometry has been reported to cause significant complications in neonates. In multiple hospitals that experienced neonatal infection outbreaks, rectal thermometers were identified as the route of transmission (27), with organisms subsequently isolated from disinfected thermometers (28) and the disinfectant solutions in which the thermometers were stored (29). Rectal thermometers are also known to cause rectal perforations (30) when the probe is inserted too far or with too much force, leading to peritonitis, pneumoperitoneum (31) and mortality (32).

It is difficult to compare the accuracy of mercury-in-glass thermometers placed in the rectal and axillary sites due to a combination of reasons. Rectal temperatures vary depending on how deep into the rectum the probe is inserted and only reach a constant temperature at a depth of 5 cm (33). However, within studies comparing rectal and axillary thermometry, the depth of rectal probe insertion is often not stated (34, 35) or probes are only inserted to a depth of 2 to 3 cm (36, 37). Within these studies, the use of axillary thermometers is also often suboptimal (34) as they are kept in place for less than the 11 min required to reach stabilization (24). Two studies that compared these measurements after stabilization showed that axillary and rectal temperatures did not differ to a clinically significant degree of 0.2°C or more (38), with differences ranging from 0.02 to 0.1°C (12, 39). However, neither study inserted the rectal thermometer to a depth of 5 cm and hence comparison of the two sites based on existing studies remains difficult.

Electronic thermometers are gradually replacing mercury-in-glass thermometers for routine monitoring especially in high-income countries. The advantages are that they pose minimal risk to the neonate, provide rapid temperature readings (40) and allow for continuous temperature monitoring when electronic probes are used (20). However, their accuracy is variable depending on which site measurements are taken from.

Although axillary and rectal sites both allow for continuous temperature monitoring in neonates, current evidence surrounding electronic thermometry supports axillary measurements whilst further research is needed regarding rectal measurements. When placed in the axilla, measurements by electronic thermometers have been shown to highly correlate to those made by traditional mercury-in-glass thermometers in preterm (36) and term (41, 42) neonates, with a mean difference of only 0.02°C (36). Indwelling rectal probes are commonly used for the continuous measurement of core temperature in neonates undergoing therapeutic hypothermia for the treatment of hypoxic-ischaemic encephalopathy (43). However, evidence regarding electronic rectal measurements in neonates is limited. Two studies which compared electronic and mercury-in-glass rectal thermometry found that many of the measurements differed by a clinically significant (38) 0.4 to 0.5°C (44, 45). However, in both studies rectal probes were only inserted 2 cm and mercury-in-glass thermometers were kept in place for 3 to 4 min rather than the 5-min optimal placement time needed (24). Hence, further studies with adequate insertion depths and placement times are required.

Electronic thermometers have also been designed for placement on the skin, with these probes commonly used for continuous temperature monitoring of neonates in incubators and under radiant warmers. Benefits of this site are that it poses minimal risk and allows for intermittent and continuous monitoring (20). The accuracy of skin measurements is highly variable however, as factors such as swaddling, clothing (46), the environmental temperature, how closely the thermometer is placed to the skin and the peripheral perfusion of the baby (20) affect the recorded temperature. Furthermore, skin temperatures have been shown to vary across the body, with higher temperatures over areas with large amounts of brown fat including the liver and intrascapular regions (47). Because different studies measure skin temperatures from different sites (48–50), the overall accuracy of electronic skin thermometry is difficult to determine. Early findings have shown that limiting placement of skin probes to areas of zero-heat-flux, such as between a neonate's skin and the mattress, allows recording of temperatures that reflect the neonate's core temperature (51) by creating an area of skin that is almost perfectly insulated (50).

Their ability to measure skin temperatures also means that electronic thermometers can continuously monitor two sites at once. Simultaneous monitoring of central and peripheral temperatures has been shown to provide valuable information on the health of newborns, particularly those that are premature or sick (20). Sites appropriate for the recording of central temperature include the abdomen or axilla, whilst the sole of the foot is typically used for peripheral measurements (52). Sustained changes in the central-peripheral difference in temperature are defined as a thermal gradient >2°C that is maintained for over 4 h (53) or which cannot be corrected with air temperature modifications (52), and have been shown to be an early indication of late-onset sepsis (53). Electronic probes can also measure nasopharyngeal (43), oesophageal (50, 54), bladder (55) and pulmonary artery (56) temperatures in neonates. However, although these sites are more representative of core temperature, their invasiveness often limits their use to neonates undergoing surgery (47) or, in the case of nasopharyngeal measurements, those who require nasogastric feeding tubes (57).

Infrared tympanic thermometers measure the infrared energy emitted from the tympanic membrane and surrounding tissue and convert it into a temperature reading through electronic thermal transducers (38). They are commonly used to measure the temperature of children and adults in high-income countries as they are rapid and painless (58) and have been shown to be accurate in these populations (59). Their use in neonates is currently limited and there is inconsistent information regarding their accuracy, which varies according to the population studied and the specific model used. Many different models of infrared tympanic thermometers are available for neonates, however only the FirstTemp Genius (model 3000A, Intelligent Medical Systems, Carlsbad, CA, USA) and Thermoscan (PRO-1 Instant Thermometer, Thermoscan, Inc., San Diego, CA, USA) models are discussed here as they have been the most widely researched.

The FirstTemp Genius infrared tympanic thermometer closely reflects axillary temperatures when used in sick neonates. Within a large-scale study conducted on sick newborns, temperatures measured by FirstTemp Genius and mercury-in-glass axillary thermometers differed by an average of 0.03°C (36), which did not reach clinical significance (38). Existing studies show that the accuracy of the FirstTemp Genius thermometer in healthy neonates, in either the rectal- or oral-equivalent modes, is low. In the rectal-equivalent mode, the FirstTemp Genius measured temperatures that were significantly higher than those measured by a mercury-in-glass rectal thermometer in both term (58) and preterm (12) neonates. No study compares the oral-equivalent mode to oral temperatures taken by another method of thermometry, most likely because oral temperatures require cooperation of the subject and are therefore difficult to obtain in neonates (60). However, when compared to mercury-in-glass rectal measurements, the FirstTemp Genius in the oral-equivalent mode was found to differ by 0.3°C or more for over 50% of measurements (58).

The Thermoscan infrared tympanic thermometer has been shown to accurately reflect core temperatures when used in a pediatric population aged 6 months to 15 years (61). The limited data regarding its accuracy in neonates have reported promising results. When compared to electronic axillary thermometry in term neonates, no clinically significant difference was found between temperatures measured by the two methods. One study found no statistically significant difference (62) whilst another reported a difference of 0.15°C (63), which still falls within the limits of clinical acceptability (38). In preterm neonates, the clinical utility of the Thermoscan thermometer is still unclear with Weiss (63) reporting that tympanic measurements were significantly higher than electronic axillary temperatures by 0.19 to 0.22°C. However, only 12 neonates were included within this study and hence larger-scale studies are needed to determine its accuracy in preterm babies.

Further research is needed to determine whether the accuracy of infrared tympanic thermometers varies across different clinical circumstances. Early studies have shown that tympanic temperatures in pediatric populations were not affected by postnatal age, the presence of vernix (13), otitis media (64), cerumen (65, 66) and fluid within the middle ear (67). However, the current literature does not address other clinical situations, including neonates undergoing certain treatments such as oxygen therapy or therapeutic hypothermia. Furthermore, within relevant studies, placement of neonates under radiant warmers or in incubators was shown to cause tympanic temperatures to be consistently higher than temperatures at other sites when compared to neonates nursed in open cots (46, 68). Another finding which must be considered when using infrared tympanic thermometers in neonates, was that the temperature measured differed depending on which ear was used. This finding was applicable to both the FirstTemp Genius (12) and Thermoscan (62, 63) models. All studies that reported this found that the ear that the neonate was lying on, also known as the protected ear, produced a higher temperature reading than the exposed ear (12, 63) by 0.2 to 0.3°C (62). It appears that using this ear may provide a better approximation of the rectal temperature (62) and allow closer estimation of the neonate's core body temperature.

Infrared thermometers are also used at sites such as the temporal artery, mid-forehead, axilla (69) and leg (70). These thermometers measure the heat that radiates from the subcutaneous blood supply (71), allowing temperature measurement without direct contact with the baby. The main advantage is the minimal disturbance it causes the neonate. Studies comparing different characteristics before and after temperature measurement found that infrared thermometry led to less disturbance than axillary thermometry in terms of behavioral states (71), pain profiles, heart rate variation and partial oxygen saturation (19). The accuracy of temporal artery and mid-forehead measurements remains unclear and seems to vary depending on the site of measurement and the population studied.

The temporal artery is one of the main sites used for infrared thermometry as its connection to the heart via the carotid artery means it has a constant blood flow (72). Most studies regarding temporal artery thermometry compare it to electronic axillary thermometers and consistently report temporal artery readings to be higher. Haddad et al. (73) compared the two methods in healthy term and late preterm neonates and found that although temporal artery temperatures were higher than electronic axillary measurements, the difference was not clinically significant (38). However, when used in sick neonates in neonatal intensive care units, studies have shown that the difference between temporal artery and electronic axillary measurements exceeds the threshold for clinical significance (74). Sim et al. (75) analyzed this difference according to the neonate's environment and found that the difference between the measurements varied depending on the environment, ranging from 0.10°C for neonates in cots to 0.97 and 1.15°C for those under radiant warmers and in incubators, respectively. Furthermore, when compared to digital rectal thermometry, temporal artery thermometers were found to have a sensitivity of 73.6% and specificity of 52.9% in detecting hypothermia (<36.5°C).

Infrared mid-forehead thermometry has been proposed as an alternative to infrared temporal artery thermometry as the temporal artery area is small in neonates and hence difficult to use (19). Current studies show that mid-forehead measurements are inaccurate due to the various factors that influence the reading, including birthweight, the baby's environment (i.e., incubator or cot) and type of ventilator support used (76). Despite large study sizes, Uslu et al. (36) and Can et al. (77) found that mid-forehead thermometry correlated poorly to mercury-in-glass axillary thermometry (77). Restricting use of infrared mid-forehead thermometry to a standardized environment may make it a viable method of thermometry. When mid-forehead measurements were compared to electronic axillary thermometry in neonates nursed in incubators, temperatures measured by the two methods did not differ to a clinically significant degree (38).

Liquid crystals selectively scatter light waves at specific wavelengths depending on the temperature (78). This is the basis of liquid crystal thermometry which changes color according to the baby's temperature. The most notable of these is ThermoSpot (Hallcrest, Glenview, IL, USA), a reusable plastic disc that sticks onto the baby's skin and changes from green to black when their temperature falls below 35.5°C (79). It can remain on for seven to 10 days and is marketed as a cost-effective hypothermia indicator that can be used by non-literate and non-numerate carers (79).

Within hospital studies, the reported accuracy of ThermoSpot differs due to variations in the definition of hypothermia used, which ThermoSpot defines as <35.5°C (79). Pejaver et al. (23) compared ThermoSpot to mercury-in-glass rectal thermometry in neonates using the ThermoSpot definition of hypothermia and found that ThermoSpot agreed with rectal temperatures 99.04% of the time. This was applicable across normothermic and hypothermic temperatures, with ThermoSpot correctly identifying every case of hypothermia. Kambarami et al. (80) reported that ThermoSpot had an overall accuracy of 57% and a sensitivity of 19% in detecting hypothermia, however they defined hypothermia as <36°C. Comparing studies regarding ThermoSpot is made even more difficult as the site of disc placement differs, with some placing it in the supraclavicular region while others place it over the liver area and axilla. Since it is known that skin temperatures vary across different parts of the body (47), this variation makes it difficult to compare relevant studies.

ThermoSpot is also a feasible and affordable temperature device for low-resource community settings. Green et al. (81) compared ThermoSpot to electronic axillary thermometry in babies that were born at home in Indian slum dwellings. Although they defined hypothermia as a temperature <35°C, ThermoSpot had a sensitivity of 88% and specificity of 97% in detecting hypothermia. The ThermoSpot device has not been reported to cause any skin damage or discomfort (81) and although it did occasionally fall off babies during the studies, it was easily reapplied with tape (23).

The BEMPU TempWatch (BEMPU Health, Bangalore, Karnataka, India) is a novel bracelet device that allows continuous temperature monitoring for 30 days (82). It consists of a thermistor metal cup within a plastic casing and a silicone band that is worn around the wrist of neonates who weigh between 800 to 3,300 g (83). The TempWatch indicates to carers when the neonate is hypothermic (<36.5°C) through an audio-visual alarm, at which point parents are encouraged to provide skin-to-skin contact to their babies (82).

Currently only one study has been conducted to determine the accuracy of the TempWatch in detecting hypothermia. Tanigasalam et al. (84) compared the TempWatch to mercury-in-glass axillary thermometry and reported it to have a sensitivity of 98.6% and specificity of 95% in detecting hypothermia. However, only neonates weighing <2,000 g were included in this study and the TempWatch was only worn for 24 h. Therefore, further studies are needed to determine whether this accuracy is applicable to neonates weighing 2,000 g or more and if the accuracy is maintained throughout the entire 30 days that the device is marketed for (82). Using the TempWatch for 4 weeks has been shown to promote parental compliance to skin-to-skin contact and neonatal weight gain during the first and fourth weeks of use (85). Across hospital and community settings, the device has also been shown to be accepted well by doctors and families (86), and has not been reported to cause any adverse effects to the neonate (85, 87). Since the TempWatch is a relatively new device, further research is needed to validate these early findings.

The advantages and disadvantages of the thermometry methods identified are listed in Table 2.

Other methods for temperature monitoring in neonates include the chemical dot and pacifier thermometers. Currently, limited studies have been conducted to determine the utility of these two methods.

The chemical dot thermometer is designed for use at the oral, axillary or rectal sites and is a flexible polystyrene plastic strip that consists of a matrix of 50 dots, each containing a specific chemical mixture that changes color from beige to blue (88) according to their melting point (47). Each dot represents an increment of 0.1°C and, after stabilization, is able to be read according to the last dot that changed to blue (88). The chemical dot thermometer can detect temperatures ranging from 35.5 to 40.4°C, with axillary temperatures generally available after a stabilization period of 3 min (88). A disadvantage of this method, however, is that instruction is required prior to use (89).

The pacifier thermometer measures supralingual temperatures (90) and consists of either temperature-sensitive crystals or a thermistor placed inside the nipple of a pacifier with a digital display at the front on which the temperature is displayed (47). Within studies that compared pacifier thermometers to other methods, no subgroup analysis of neonates was conducted (91, 92). However, results of these studies showed that many parents did not want their infants using pacifiers (90) and many infants were unable to suck on the pacifier long enough to allow a steady temperature reading (91). Although pacifier thermometers are easy to understand, these findings limit their utility in neonates.