In certain medical specialties, AI is already changing the way patients are cared for in the hospital. In radiology, a french startup is developing an algorithm which improves mammography screening to detect breast cancer. In cardiology, a french company has developed an extremely powerful algorithm, able to analyse electrocardiograms and improve diagnostics. In ophthalmology, the IDx-DR software developed in the United States identifies signs of diabetic retinopathy with an accuracy rate of 90 %. Needless to say that in France, with an average waiting period for an ophthalmologist appointment of 80 days, this kind of technology would definitely impact the landscape of ophthalmology. In clinical research, a french company builds mathematical models and algorithms that can interpret biomedical images, genomics and clinical data, to discover biomarkers and mechanisms associated with diseases and treatment outcomes. It will improve diagnosis and personalization of care.
The medical professions are not the only ones affected by AI. The care support functions are also concerned. A very likely hypothesis would be that AI solutions, already mature and present in other sectors (management of payroll, customer files, logistics, IT support), would be deployed to the support functions of the healthcare sector. For example, computer maintenance, inventory and flow management, delivery, etc. are gradually being carried out by robots or AI software in other sectors, such as mass distribution. At Nantes University hospital, intelligent software has the potential to save time for technical and administrative teams, in particular by optimizing the coding of hospital activity in order to anticipate patient flows. At Lyon, the University hospital has signed a partnership with Microsoft around AI. They aim to develop a vocal recognition system used during consultations, to assist medical secretaries with writing reports. The AI technology would also recognize keywords and be able to carry out a medical diagnosis.
How should we anticipate these disruptions?
Because AI will profoundly change medical practices and the back office function, an evolution and transformation of healthcare professions is now necessary. Indeed, it takes at least ten years to train a doctor, so we must ensure that the knowledge transmitted today will not be obsolete tomorrow. Thus, a methodology must be developed to assess the impacts on employment. The note published by Institut Montaigne and Ethik IA proposes a methodology that is divided into six steps, to determine the level of substitution by AI of each activity for a given job and then to define the rate of substitution for the job. For instance, for a medical secretary, it is possible to identify various tasks such as physical and telephone reception of patients and families, management and medico-administrative coordination, identification of patients' and families' needs and expectations, note taking, typing and document formatting, data entry related to medical activity, maintenance of patient records, etc. All these tasks must be reviewed to identify the level of substitution and then to determine whether the job would be completely or partially replaced by IA.