For fever, digital oral or rectal readings are the most reliable; ear and forehead models are faster but miss more fevers. 'Normal' isn't exactly 37°C.
Home thermometers come in two broad kinds. Digital contact thermometers (oral, underarm, rectal) read the temperature where the tip sits. Infrared thermometers read heat radiated from the eardrum or forehead: fast and easy on a sleeping child, but more sensitive to technique, sweat, room temperature and distance. Basal thermometers read to two decimal places for cycle tracking.
A fever is generally 38.0°C (100.4°F) or higher. 'Normal' varies: in 35,488 clinic patients, average oral temperature was 36.6°C, and individuals had their own baselines.
Taking a temperature is simple, but the site matters. In a meta-analysis of 75 studies, mostly in hospital patients, ear, forehead, underarm and oral thermometers often differed from core temperature by more than 0.5°C and caught only about 64% of fevers, though they rarely reported fevers that weren't there. In children, ear readings differed from rectal ones by up to about 1°C either way. Non-contact forehead devices picked up about 81% of fevers in a separate meta-analysis. For babies, a rectal digital reading remains the reference.
In 35,488 outpatients without infections (243,506 measurements), mean oral temperature was 36.6°C (95% of readings 35.7–37.3°C). Individuals had their own baselines, which varied with age, sex and health.
Across 75 studies (8,682 patients), ear, forehead (temporal artery), underarm and oral thermometers had limits of agreement with core temperature outside ±0.5°C; they detected fever with 64% sensitivity and 96% specificity.
In 31 comparisons covering 4,441 children, ear readings averaged 0.29°C below rectal ones, but individual differences ranged from about −0.7 to +1.3°C, too wide for situations that need precise measurement.
Pooling 19 studies, non-contact infrared thermometers detected fever with about 81% sensitivity and 92% specificity compared with conventional thermometers.