So, what's best? Accuracy and acceptance of thermometers in triage and inpatients in a low-resource tropical setting - The MaTe study.

Objectives: We searched for the most-suitable thermometry method in the low-resource, tropical setting of Sierra Leone, both in terms of accuracy and also patient and user acceptance. Methods: We conducted a prospective comparative study of different methods of body temperature measurement. Each participant had their temperature taken by four different methods: non-contact infrared temperature (NCIT), axillary, tympanic membrane and rectal measurements. Rectal temperature was considered clinical gold standard. Primary outcome was predicted sensitivity and specificity of thermometry methods in detecting fever (rectal temperature ≥38.0 °C). Questionnaires were used to explore patient and healthcare worker attitudes towards different methods of temperature-taking. Results: 824 rectal body temperature readings were taken from 562 participants. The mean rectal temperature was 37.4 °C (IQR 37 °C to 37.7 °C), with a minimum reading of 35.2 °C and maximum of 41.0 °C. Tympanic membrane thermometry showed the highest sensitivity of fever detection using the Genius3 TM thermometer (sensitivity 70.8 %, 95 % CI 60.2%-79.9 %; specificity 97.2 %, 95 % CI 95.5-98.4 %); and Braun TM (sensitivity 51.5 %, 95 % CI 42.6%-62.0 %; specificity 98.8 %, 95 % CI 97.7-99.5). NCIT thermometry sensitivity was low (36.8 %-41.4 % for the two devices used). Axillary thermometry sensitivity was 40.6 %. Participants ranked NCIT as the most and rectal as the least preferred method. Questionnaires from 32 participating nurses showed agreeability to using NCIT, TM and axillary methods routinely, but less so for rectal thermometry. Conclusions: When combining the accuracy of different thermometry methods in detecting fever with user and patient acceptability, tympanic membrane thermometry appears most suitable but also has limitations.