Despite rapid developments in the field of artificial intelligence (AI), successful implementation in healthcare remains a challenge. In a recently published preprint, researchers from Halmstad University (Sweden), together with Julie Swillens (IQ Health), analysed 65 empirical studies on the implementation of AI applications in everyday healthcare practice.
The key conclusion: successful AI implementation is not just a technical challenge, but above all an organisational issue. The researchers demonstrate that organisational support, multidisciplinary collaboration, stakeholder engagement and effective integration with existing work processes are crucial for successful implementation. At the same time, poor performance of AI systems, an increased workload for healthcare professionals, limited IT infrastructure and difficult integration into healthcare processes constitute significant barriers.
Furthermore, it appears that most AI applications are used for automation, image analysis and natural language processing, primarily to support doctors, nurses and radiologists. It is striking that patients are considered direct users in only a small proportion of the implementations.
The study also highlights a gap between theory and practice: only one in five studies made use of existing frameworks. Furthermore, patient perspectives, administrative constraints (such as data protection and reimbursement structures) and early implementation phases (such as needs analysis and stakeholder engagement) still receive relatively little attention.
According to the authors, AI implementation should therefore be viewed as a socio-technical process requiring technical, organisational and systems knowledge, rather than merely a technical exercise.