Table 2. Barriers and strategies in the planning phase of AI implementation.
| Barriers for planning |
|---|
|
Leadership • Lack of guidance and common vision among stakeholders Change management • Personal reasons by staff leading to resistance to change Buy-in • Concerns for self-esteem, social status, identity, gradual knowledge loss; Perceived irreplaceable experience, intuition, human judgement, potential unequal care; Not understanding value of AI, current care perceived as optimal. Engagement • Lack of communication, socialization and alignment among stakeholders; Stakeholders’ skepticism Workflow • Not anticipated conflicts between potential workflow changes and KPIs of units; Significant changes in workflow (disruption); Non-interoperable systems. Finance and human resources • Not affordable to dedicate clinicians’ time to implementation; Lack of published evidence on AI system hinders chances of funding; High costs for regulatory approval. Legal • Regulatory approval requirements; Procurement procedures; Uncertain legal framework for clinicians’ liability when using AI. Ethics • Concerns that AI would impersonalize care; Patients not willing to know predicted outcomes. Data • Insufficient quality of data; Variations in handling workflows that create data; Bias in data; Complicated access to data; EHR not supporting data processing and analytics. Training • Disparate extent of knowledge and skills (also IT skills) among potential users can affect the levels of use of AI and create misunderstanding of potential. |
| Strategies for planning |
|
Leadership • Involved leaders providing a vision and strong support to an implementation team. Buy-in • Achieving buy-in by top management can positively affect buy-in by employees; A request for developing an AI solution is initiated by local clinicians (a local problem). Change management • Communicating a sense of urgency, vision for change across organization; Raising initial awareness among staff about upcoming change and potential impact on processes; Engaging middle managers and identifying local ‘champions’. Engagement • Early involvement of stakeholders (also the ones interested in analytics); Inter-disciplinary group; Alignment with organizational priorities; Building partnerships with key staff inside a hospital. Workflow • Problem assessment with interdisciplinary team; Assess feasibility for workflow and infrastructure changes, and interoperability with existing systems at the hospital; Assess if current suppliers could be used or new collaborations need to be initiated. Finance and human resources • Budgeting for recruitment, involvement of staff, and for preparation of training materials. Ethics • Promoting AI as an assistant in ethical dilemmas; Transparency with patients that they are interacting with AI; Integrating a possibility for patients to choose how much data they want to share with AI. Data • Training staff how to ensure data quality during entries; Introducing data cleaning procedures; When training the AI model, knowledge about how long the data was stored, how changes were made. Training • Planning for training, budgeting and planning of the preparation of training materials and of training activities. Evaluation and monitoring • Planning a pilot intervention on AI system and new workflow; Having randomization and control groups; Assessing patient safety; Cost and benefit evaluation; Adherence metrics. |