Diagnostic accuracy of oral thermometry for fever detection in adult patients: literature review and meta‐analysis

Aims and objectives To synthesise the evidence on the accuracy and precision of oral thermometry in adult patients, as well as on its sensitivity and specificity for fever detection. Background Oral thermometry has long been used in various clinical settings thanks to its rapid, safe and convenient measurements, which are easy to obtain and minimally prone to operator errors. Design Literature review and meta‐analysis that adhered to the PRISMA statement. Methods By using key terms, literature searches were conducted in CINAHL, PubMed, Web of Science, Scopus and Cochrane Library. Method‐comparison studies, which were published from January 1990 to December 2020 in English‐language, peer‐reviewed journals, compared oral temperature measurements with invasive thermometry ones, and were conducted on patients ≥18 years, were included. Methodological quality of selected studies was evaluated with QUADAS‐2. Results Sixteen articles were selected for inclusion. Risk of bias was assessed as low in most of them. Quantitative synthesis indicated that pooled mean oral temperature was lower than core temperature by .07℃, with 95% limits of agreement ranging between −.22℃ and .08℃. Pooled sensitivity and specificity for fever detection (defined as core temperature ≥38℃ in most studies) were .53 (95% confidence interval, .39–.66) and .98 (95% confidence interval, .97–.99), respectively. Sensitivity analysis indicated larger temperature underestimation in case rectal temperature was used as reference standard. Conclusion Despite its satisfactory accuracy, precision and specificity, oral thermometry has low sensitivity for fever detection, which entails a high number of false‐negative readings and uncertainty for excluding fever in patients found to be non‐febrile. Relevance to clinical practice Oral thermometry cannot be recommended for replacing invasive thermometry methods in hospitalised adult patients, considering the high incidence of fever in them and possible negative effects of missing fever for patient diagnosis and outcomes.