Introduction:
The global market for Artificial Intelligence (AI) in healthcare is estimated to grow to USD 355.78 billion by 2032, expanding at a double digit CAGR of 37.66 per cent from 2023 (1). Traditionally a laggard in technology adoption, the healthcare industry is now responding to the rising demand from consumers for personalised healthcare, adopting advanced technologies and innovating clinical research. Health systems also see the potential for AI to efficiently manage ever-growing clinical workforce shortages, reduce burnout, enhance patient experiences, improve care outcomes, and reduce the cost of care. With the easy availability of Generative AI (GenAI) tools and resources like ChatGPT, the field of AI is now accessible to many. At the same time, clinicians and hospital management are navigating challenges around meaningful and responsible AI integration.
Common challenges of integrating AI in healthcare:
Data quality and governance: In recent years – particularly during the pandemic – the urgency to standardise healthcare data collection and exchange methods has intensified. New and complex challenges have emerged with widespread adoption of electronic health records (EHR) and the availability of vast amounts of structured and unstructured patient data from various sources such as medical devices, laboratories, diagnostics, imaging, genome sequencing, and social determinants. Inconsistency and inaccuracy in managing this data significantly impacts the performance and reliability of AI models, leading to potentially flawed insights.
Cost of implementation: Integrating AI successfully requires robust information systems, additional computing and infrastructure, efficient processes, advanced data management tools, and licenses, which come at a cost. This, along with the cost of technical and domain experts to manage AI models for optimal performance can add to the financial burden but without adequate investment in these areas, the return on investment can be significantly diminished. For smaller health systems, or those lacking the necessary foundational information systems, the cost can be prohibitive.
Compliance and ethics: Healthcare regulatory bodies, at the regional, national, and international level are evolving their thinking around AI, resulting in frequent new rules and amendments to establish data privacy and safeguard patient medical information. This makes healthcare compliance and regulations even more complex and multifaceted. Healthcare providers must stay up to date to avoid the severe financial and legal penalties that come from non-compliance.
Insights to practice: Implementing AI in core healthcare workflows, such as care orchestration, care delivery, and clinical decision-making, is significantly more complex in comparison to areas such as enhancing experiences or operations. The intricacies involved in core areas require a deeper integration of AI technologies, robust data management, and stringent regulatory compliance. Translating theoretical research findings into real-world clinical practice is also a significant challenge that must be addressed for healthcare professionals to seamlessly apply insights into their practice.
Paving the way for AI integration
Build a solid foundation for AI implementation:
Health system leaders are likely to see more success by investing at the outset in a holistic enterprise data strategy, with the right technical talent and a foundational integrated healthcare data platform to compute data from various sources. A robust data governance system can facilitate interoperability for seamless data exchanges across systems. Adhering to common data models like OMOP (Observational Medical Outcomes Partnership) ensures that information from varying sources, such as encounters, patients, providers, diagnoses, drugs, measurements, and procedures, is consistently captured. A service-oriented design approach for enabling data consumption for training and deploying AI models allows for control over data access, data lineage, and quality.
Take a measured and meaningful approach to implementation:
The ability to identify a well-defined problem space that aligns with business priorities, risk, and compliance is crucial to successfully implementing AI. Organisations can assess their AI preparedness by testing the maturity of their existing IT systems through low-risk AI use cases such as medical journal recommendations and summarisation. Leaders must also weigh varying approaches to AI integration. For example, open-source models are relatively easier to deploy but may have higher associated costs and risks to data privacy, ethics and compliance. Alternatively, custom-built AI solutions are harder to implement, but can significantly reduce costs and control risks.
Build a robust compliance program:
Healthcare leaders must be dynamic in their approach to building compliance frameworks that adapt to the changing regulatory complexities of AI in healthcare. This includes providing constant regulatory updates to teams to ensure that AI systems meet all compliance requirements, and engaging with regulatory bodies on policy developments. Leaders must ensure robust data privacy measures around patient data de-identification in adherence with regulatory guidelines, such as India’s Digital Information Security in Healthcare Act (DISHA), or United States’ HIPAA, which protect patients’ privacy. At the solution-level, technical teams must ensure that they mitigate bias while preparing AI training model datasets by introducing enough noise and diverse datasets that prevent AI system recommendations from violating ethics and equity guidelines.
Simplify AI insights to ease the way for users:
Generative AI has expanded our possibilities to learn from a vast amount of data. However, it can also make recommendations inconsistent and difficult to understand. It is therefore critical to partner early on with patients and healthcare professionals when implementing AI in the core workflows of healthcare decisions and practices. By building transparency and easy-to-understand recommendations, practitioners feel empowered to use AI insights to make informed decisions.
To fully harness the potential of AI in healthcare, providers must establish robust integrated systems, processes, and tools, with a strong emphasis on data management and deep domain expertise. It is crucial to understand the complexities of AI use case implementation and the associated risks related to compliance, ethics, and equity. Providers must weigh in the balance the return on investment and total cost of ownership for sustainable AI integration.
References:
(1) 1 AI in Healthcare Market Size | 37.66% – CAGR Growth by 2032 (towardshealthcare.com)