Background and aims In an area of the UK covered by a busy district general hospital 6.4% of households have no adult who has English as their main language [1]. This can pose a problem for the consent of procedures, such as caesareans and epidurals. In view of the medico-legal requirement to communicate effectively with our patients [2], and the time-consuming nature of using telephone translation services, we evaluated the effectiveness of the online translation service Google Translate (GT).

Methods Eight phrases commonly used during epidural consenting and placement were typed into GT. the audio translation of six languages was then listened to by a non-medical native speaker. Their understanding of the audio was then recorded in English. This was compared to the original English statement and assessed for meaning. the accuracy of the translation was ranked: correct; minor errors with no effect on understanding; and major errors with significant effect on understanding or factually incorrect.

Results Both Asian languages had major error rate of 50%, Romanian 25%, Spanish and French 12.5% and Polish 0%.

Conclusions GT is a statistical machine translation service; therefore, some of the major errors in translation were completely nonsensical. In Tamil ‘paralysis’ was translated to ‘possessed’ and in Spanish ‘ventouse’ to ‘ventricular’, rendering it inappropriate for medical consent. There could be scope for its use in procedural instructions if no specialist terms are used. However, care is needed to maintain a patient’s confidence particularly during this vulnerable period. In future a precompiled set of accurate audio translations could be evaluated.