This is a comparative diagnostic test study. A total of 400 sets of blinded simultaneous temperature readings were measured from 21 haemato-oncology in-patients with neutropenia following chemotherapy. Three-hundred sets were then randomly sampled. Agreements were analysed using random two-way intraclass correlation (ICC). Sensitivity and specificity were analysed using contingency 2 × 2 table.
Both right and left TM thermometry have good correlation with rectal thermometry; 0.810 (95 % CI, 0.748–0.855) and 0.770 (95 % CI, 0.713–0.815) respectively. Axilla thermometry has weak agreement (ICC 0.486 (95 % CI, 0.118–0.689)) with rectal thermometry. The sensitivity (sn) and specificity (sp) in detecting rectal fever (≥38 °C) were: right TM (sn) 0.712 (95 % CI, 0.586–0.814), (sp) 0.957 (95 % CI, 0.920–0.978); oral (sn) 0.561 (95 % CI, 0.433–0.681), (sp) 0.983 (95 % CI, 0.954–0.995); and axilla (sn) 0.348 (95 % CI, 0.238–0.477), (sp) 0.996 (95 % CI, 0.973–0.999).
Single tympanic membrane thermometry is in good agreement with rectal thermometry. It is more sensitive than oral or axillary thermometry in detecting rectal fever.