AI in Healthcare: Diagnostic Algorithms in Medical Workflows

Topic
The recent developments in healthcare data management and technology are pivotal, there is an alignment on the transformative potential of medical algorithms to increase diagnostic accuracy, reduce errors, increase operational efficiency, and improve patient outcomes. In reality, the adoption of diagnostic algorithms and AI is lacking. This research examines how diagnostic algorithms and AI solutions can be effectively integrated into multi-rational clinical workflows.

Relevance

Existing literature often focuses on the technical capabilities and theoretical clinical value, rather than on the holistic implementation into clinical practice, from implementation to routine use. This research provides actionable insights for both healthcare providers but also the diagnostic industry and policy makers with a “Diagnostic Algorithms and AI Adoption Framework”. Healthcare providers can leverage these outcomes to better understand the behaviors of their employees and challenges of new technology adaptation, exemplary diagnostic algorithms and AI. It guides hospital senior management for holistic change management approaches to ensure sustainable adoption success.

Results

On an individual layer, healthcare staff’s resistance, lack of trust, and limited knowledge and education on algorithms and AI are key challenges for adoption. While a new generation of medical staff tends to be more open to workflow and decision support through algorithms and AI, this is not yet established throughout.

On the institutional layer, the research concludes that successful integration predominantly depends on the organizational capability to drive change, independent from individual acceptance or the systemic environment. The complexity of clinical workflows demands seamless integration into both infrastructure systems and proven medical routines, which is challenged by the hospital's lack of in-depth and holistic understanding of its workflows. This challenge is complemented by a not-AI-ready infrastructure. A shortage of resources, both financially and of skilled IT staff, additionally limits the institutions’ ability to adopt innovations. Organizationally, the lack of cross-functional roles or innovation support structures driving holistic innovation and data strategies further slows adoption. Change management was concluded as the most important institutional success factor. Institutions that view AI implementation as an organizational transformation rather than a technology rollout at the department level tend to succeed more often.

At the systemic layer, regulatory frameworks and environments are a foundational enabler and a consistent obstacle. Existing frameworks are outdated, overly restrictive on data use or technologies like cloud, and highly inconsistent across and even within countries. The lack of universal standards, such as monitoring procedures, liability rules, clear reimbursement plans, and shared clinical evaluation criteria, creates uncertainty for hospitals and healthcare professionals.

In addition, this research reaffirms the impact on multiple rationalities that are impacted as part of this implementation process. Clinical, economic, and regulatory needs are partially conflicting. Holistic organizational change management can help to navigate this change.

Implications for practitioners

To address the adoption gap holistically, the thesis proposes a Diagnostic Algorithms & AI Adoption Framework. The framework provides hospital leaders with an actionable guide to manage the complexities of the implementation of algorithms and AI solutions. Building strong internal capabilities and a solid infrastructure is key for sustainable adoption. This multi-layered artefact emphasizes:

1.     Institutional Readiness: Organizational clarity on workflows, infrastructure, resource allocation, organizational set-up, strategy and objectives.

2.     Change Management: Structured and continuous education, cross-functional alignment, hands on experience and stakeholder engagement.

3.     Dynamic Adoption: Post-implementation monitoring, trust-building mechanisms, and iterative validation of clinical and operational value.

Methods

Based on the theory of multi-rationality and three layers of adoption, this research deploys a qualitative approach to generate insights through eight expert interviews with senior clinicians, industry leaders, healthcare consultants, and senior academics. All interviews were conducted remotely using a semi-structured guide and open discussion. This approach provided in-depth, practice-oriented insights into challenges and opportunities of the adoption of algorithms and AI in routine clinical workflows.