Your browser is no longer supported

For the best possible experience using our website we recommend you upgrade to a newer version or another browser.

Your browser appears to have cookies disabled. For the best experience of this website, please enable cookies in your browser

We'll assume we have your consent to use cookies, for example so you won't need to log in each time you visit our site.
Learn more

An investigation into the accuracy of different types of thermometers

  • Comment

Dr Dawn Dowding, BSc (Hons), PhD, RN.

Programme Leader, Nursing Research Initiative for Scotland, University of Stirling

A patient’s temperature is a crucial piece of clinical data. In the context of other data it can guide diagnostic and therapeutic measures by determining the presence of illness and the extent to which the patient is responding to treatment (Giuliano et al, 1999; Henker and Coyne, 1995). Temperature must be measured accurately to identify fluctuations fast and intervene early (Carroll, 2000; Fulbrook, 1993).

Traditionally in acute care, temperatures have been measured using mercury-in-glass thermometers. However, concerns are growing about the health and safety risks, such as glass breakage and the potential for mercury poisoning (Blumenthal, 1992). Mercury-in-glass thermometers have been implicated in episodes of cross-infection and outbreaks of diarrhoea caused by salmonella and Clostridium difficile (Brookes and Veal, 1992; Cutter, 1994). A growing number of alternatives are available, such as disposable, electronic and tympanic membrane thermometers. But these substitutes need to be accurate. Tympanic thermometers calculate temperature using an infra-red reading from the patient’s ear. Electronic thermometers take readings from the axilla or orally and use an algorithm to calculate the temperature.

Previous research

Research in this area is scarce. One study that compared a disposable thermometer (3M TempaDot) to core pulmonary artery readings found that oral mean readings were significantly higher than core (0.438C) (Henker and Coyne, 1995). However, at the axilla the average difference was only 0.018C, not significantly different to the core temperature. One study found a mean difference of 0.468C between core temperature and measurements taken using a disposable thermometer in the axilla (Fulbrook, 1997). In studies which examined the accuracy of electronic thermometers, various models were used, including:

- IVAC TempPlus II (Erickson and Yount, 1991; Manian and Griesenauer, 1998; Henker and Coyne, 1995; Weiss et al, 1998)

- Diatek 600 (Flo and Brown, 1995; Summers, 1991)

- Welch-Allyn (Giuliano et al, 1999).

Studies that have compared electronic thermometers to core temperature via a pulmonary artery catheter have reported variable results:

- At the oral site, results have been reported of an average of 0.188C above (Henker and Coyne, 1995), 0.158C below (Giuliano et al, 1999) and 0.198C below (Schmitz et al, 1995) core temperature

- At the axilla, studies have shown equally varied results, with electronic thermometers measuring an average of 0.458C above (Henker and Coyne, 1995), 0.338C above (Giuffre et al, 1990) to 1.128C below core temperature (Schmitz et al, 1995).

Studies have compared tympanic thermometer readings to core measurements taken by pulmonary artery catheter probes. Some suggest no significant differences between this type of thermometer and core probes, with results giving mean differences in readings of 0.098C (Ferrara-Love, 1991) or 0.088C (Edge and Morgan, 1993). Others provide differences of 0.158C (Schmitz, 1995) and 0.11-0.528C below core readings (Giuliano et al, 1999) and 0.478C to 0.58C above core readings (Fulbrook, 1997), suggesting significant differences.

Studies examining tympanic thermometers found variations (Fulbrook, 1997; Giuliano et al, 1999) and even differences in readings between different types of tympanic thermometer (Weiss et al, 1998; Henker and Coyne, 1995). Inconsistency is reported in the accuracy of all alternatives to mercury-in-glass. Some inconsistencies are large enough to cause significant clinical error if they were to be repeated in the clinical environment.

The study

These inconsistencies prompted this study within a local acute trust. We aimed to investigate the accuracy of tympanic thermometers, digital thermometers and disposable thermometers, when compared to mercury-in-glass temperature readings taken at the axilla. The researchers recognised that the ‘gold standard’ temperature measurement is that taken by a pulmonary artery catheter. But in this study, which aimed to obtain a large sample size and use several repeated measures, it was deemed impractical to use this device. Previous research suggests mercury-in-glass readings are a reasonably accurate ‘proxy’ measure for core temperature (Giuffre et al, 1990), so this was our gold standard.


The study used a repeated-measures experimental design to compare three devices in various body sites (oral, axilla and tympanic membrane) with a mercury measurement taken at the axilla and timed for 10 minutes, for each participant:

- Digital thermometer (DigiTemp) readings were taken orally and at the axilla (disposable probe covers were used to prevent cross-infection)

- Disposable thermometer (3M TempaDot) readings were taken orally and at the axilla

- Tympanic thermometer (Genius) readings were taken from the left and right ears (disposable probe covers were used to prevent cross-infection).

Genius supplied the tympanic thermometers and associated equipment and training for data collectors; and 3M TempaDot provided disposable thermometers and training for data collectors. DigiTemp electronic thermometers (British Rototherm) were already used within the trust. The manufacturer did not provide training in their use for the study.

All devices were used according to manufacturer instructions. The order in which measurements were taken varied, to avoid order effect in the results. All data collectors were registered nurses or nurse teachers. The tympanic and mercury thermometers were calibrated before the study by the hospital medical physics department. It was not possible to calibrate the digital or disposable ones.

A convenience sample of 145 were recruited as subjects. These comprised university students and staff members from the local acute hospital. All were healthy adults and fully consented to take part. Recruitment was in two phases over four months. It was not felt necessary to target patients.

The departmental ethics committee at the university gave permission. All data collected were anonymous. Consideration was given to the fact that participants may present with pyrexia. If so they were offered advice, for example if they had a cold, to drink plenty of fluid and take paracetamol.


Out of the 145 participants, 143 had complete data suitable for analysis. Incomplete data was due to failure of more than one of the thermometers. Of the 143, 123 were female and 20 male, with an average age of 33 years (range 17 to 54 years). Each dataset comprised seven readings, using the four types of thermometer for one subject. The data were analysed using SPSS computer software by a qualified statistician based at the University of Stirling.

The mercury readings were taken as the benchmark. The difference between this reading and the others was calculated for each participant. Negative differences indicate a reading below the mercury benchmark; a positive one indicates a reading above the mercury one. Small, random differences are to be expected given the inherent variability in instruments and data collectors. However, the mean difference between any one thermometer type and mercury can be assessed in terms of whether it could have arisen by chance or indicates a statistically significant difference between the readings. A t-test compared the mean differences between thermometer types and mercury readings (Table 1).

The two digital thermometer readings under-recorded compared with mercury to the extent of -0.278C and -0.348C respectively. In both cases, these are statistically significant differences. Further analysis shows that, in the case of the first digital thermometer (DigiTemp 1 oral), for 68% of subjects the reading was below that for mercury. In the case of the second digital thermometer (DigiTemp 2 axilla), the comparable figure was 79%.

One of the two disposable thermometer readings (oral) under-recorded by -0.418C and again this is statistically significant, with 71% of subjects’ readings below those of mercury. The second disposable (axilla), tended to over-record compared with mercury, with a mean difference of +0.138C, again statistically significant. In this instance 58% of subjects’ readings were above those of mercury. Although for one of the tympanic types (right ear), there is a slight under-recording, this was not statistically significant. We tried to assess whether the differences in readings could be attributed to how the data collectors used them rather than differences in the thermometers.

For three readings, a one-way analysis of variance found an influence on accuracy dependent on which data collector was using the thermometer. These were the second digital (axilla), the first disposable (oral) and the second disposable (axilla). This suggests these thermometers might be inherently more difficult to use accurately and reliably.

The variation in measurements according to data collector for the digital thermometers could, in part, be explained by the difficulties using these thermometers. The research began with 12 DigiTemp thermometers. By the end of the four months data collectors had discarded five because of problems with registering temperatures. It was common for these thermometers to be left in place for up to 30 minutes, without registering anything. The researchers perceived these to be the least reliable devices.


The results indicate that overall the DigiTemp thermometer on average under-recorded temperature significantly (p<0.05), both at oral and axilla sites, compared to a mercury-in-glass reading at the axilla. These mean differences could be considered clinically, as well as statistically, significant. For instance, a DigiTemp thermometer consistently under-recorded temperatures by 0.34°C, giving a measurement of 37°C, when the subject’s ‘real’ temperature was 37.34°C. On the basis of the reading, the patient’s temperature would appear normal, when in fact he or she had a pyrexia that would require nursing intervention (Heath, 1995). The disposable thermometers used to take oral temperature readings on average significantly under-recorded temperature compared to the mercury-in-glass reading.

The tympanic thermometers readings were not significantly different from mercury-in-glass. Research has indicated that the model in this study tends to provide readings not significantly different to core temperature (Weiss et al, 1998; Henker and Coyne, 1995).

The tympanic thermometers did not seem to depend as much on variability in data collectors, suggesting they can be used more reliably than other types of thermometer.

Limitations to the study

The subjects were healthy adults, and not representative of most hospital inpatients, who are unwell and older on average. Their chances of presenting with abnormal temperature readings were lower than in a hospital. The thermometers’ ability to highlight abnormal temperatures, particularly extreme pyrexia or hypothermia, has not been adequately tested.

The readings were compared to that taken at the axilla by a mercury-in-glass thermometer, although evidence is limited that this could be a proxy measure for the ‘gold standard’. Ideally, the study should be repeated using pulmonary artery catheter temperature measures.


Given the limitations outlined above, the tympanic thermometer seems to be a reasonable choice to replace mercury-in-glass thermometers. It provided readings not significantly different to those obtained by mercury-in-glass thermometers, and did not seem to depend on data collector variation. However, there was large variation in the readings obtained so, while overall mean readings were no different to the mercury, some individual results were different.






  • Comment

Have your say

You must sign in to make a comment

Please remember that the submission of any material is governed by our Terms and Conditions and by submitting material you confirm your agreement to these Terms and Conditions. Links may be included in your comments but HTML is not permitted.

Related Jobs