Use of telemedicine with artificial intelligence for patient monitoring after transcatheter aortic valve implantation with a virtual voice assistant

Abstract Introduction Population ageing and the increase in the demand for care due to the progression of acute pathologies, which previously had reduced survival rates, to chronic pathologies, lead to the necessity for the classic consultation model to adapt and coexist with telematic consultation. Its basic objectives should include an adequate prioritization of patients and the identification of the best way to guide their care. Objective The application of a virtual voice assistant, based on artificial intelligence, could be an effective and safe alternative to monitor patients after TAVI implantation, allowing a more efficient use of the resources available in Cardiology services. Material and methods Observational, prospective, single-center and consecutive study of the implementation of a technology based on artificial intelligence and natural language processing to monitor all patients undergoing transfemoral TAVI implantation, according to usual clinical practice, in the hemodynamics unit of our hospital in the year 2023. The virtual voice assistant, "Lola", carries out a telephonic follow-up during week 1, week 2, month 1, month 3 and 12. In these calls, it asks a series of questions mainly related to vascular access and with the patient's cardiovascular situation. After finishing the call, all the information collected is uploaded to a web platform where we monitor the collected data and thus take the appropriate measures. Results The option of this type of follow-up was offered to all patients discharged after TAVI implantation and who did not present a language barrier during the year 2023. A total of 274 patients were included. Only 2 patients refused this follow-up system. A total of 1039 calls were made, involving 385 hours of autonomous conversation, with an average duration of 4 minutes and 3 seconds per call. A total of 93.92% of the calls were completed, with the degree of adherence always remaining above 85% throughout this period. In 89% of the calls, the patient himself answered the calls, and only 11% were answered by family members or caregivers. In 44% of the calls, no alerts were detected and, therefore, did not need to be revisited. Of the remaining calls, a total of 926 alerts were obtained, resulting in at least one intervention in 57% of the calls (593 in total). The number of alerts decreases as the follow-up progresses, reflecting a need for closer follow-up in the early stages after implant, specifically during the first month. The general satisfaction of patients has been good or very good in 88.89% of cases (Customer Satisfaction score 4.68/5) and 86.3% of patients would recommend "Lola" according to the "Net Promoter Score". Conclusions Clinical monitoring with artificial intelligence can be applied to the vast majority of patients after TAVI with a high degree of satisfaction and allowing optimized use of available resources.