Continuously, novel digital health care applications are being developed, such as clinical decision support or recommender systems. Especially with this increasing availability of data from the electronic healthcare records (EHR) and the possibility to train complex multi-model AI, there is a considerable potential for AI-models to enhance clinical care. With the large amount of data currently becoming accessible for research, tools are continuously being developed, but implementation of tools remains scarce. This may be due to (technical) feasibility, generalizability, or clinical applicability issues, incomplete compliance to quality standards of tools or sometimes even due to irrelevance of the novel tool to clinical practice, as the tool is not likely to improve clinical care or streamline clinical processes. Thus, whereas the opportunities for AI in healthcare seem promising, prior to implementation, developed models require careful evaluation on their quality (performance, safety), applicability and usability to ensure safe and responsible application and implementation of AI in healthcare.

Therefore, ideally when developing this model, ultimately the end-goal (i.e., clinical implementation) should be taken into account. Therefore, from the very start of the development (i.e., idea generation), regulatory, ethical, clinical and technical aspects should be taken into account. To streamline the development of novel tools and make sure that tools being developed are required and applicable in clinical practice, we summarize important clinical and healthcare requirements for implementation to aid the successful application and added value for clinical care. We therefore focus on various aspects of clinical and health-care requirements, including ethical, privacy, and clinical aspects and map this to the current laws and guidelines for the development and implantation of AI-tools. We propose a first draft of the general acceptance criteria for developed models, and map this to the different phases for model development and implementation.