The national teleconsultation platform Qare offers a promising tool to collect data in primary care by sharing the same EHR among teleconsultants. Conclusions for every teleconsultation are thus collected by a text using an ICD10 code or not. Nevertheless, the use of the ICD10 classification by primary care doctors is scarce. We developed an algorithm based on bags of words that associates an ICD10 code for each textual conclusion and evaluated it. It shows good performances (88 to 89% concordance) to assign ICD10 category. It could be used to enhance the volume of coded teleconsultations ready to be reused.
Keywords: ICD coding; NLP; Teleconsultation; bag of words; evaluation.