Never has there been more interest and excitement around the potential of Artificial Intelligence (AI) in healthcare. The advent of Generative AI and LLMs, popularized by OpenAI with ChatGPT, has opened many people’s eyes to what might be possible. Yet, the translation of this promise to patient impact is only in its infancy, and while it is a path with huge potential, as healthcare professionals, we must tread carefully and responsibly.
In this article, we discuss the importance of the ecosystem in which healthcare AI must exist, to be successful. That ecosystem is made up of people, process, and technology.
However, before we can get to successful AI deployment – you must consider the problem you are trying to solve. Once that has been identified – we hope the rest of this article will help in putting in place the pieces to enable you to be successful in your AI deployment!
1. People: The Right Stakeholders
To build and deliver successful AI, multidisciplinary team engagement is a must.
Clinical Experts
Input from clinical experts is critical – both in the design and in the way that the tool will be used. The domain expertise of these HCPs ensures that AI solutions align with real-world needs. For example, when developing an AI-powered screening tool for social determinants of health, involving social workers that will use the output is key.
IT and Operational Teams
Collaboration between IT experts and operational staff is essential. These teams are needed to ensure that you can put in place reliable processes (see below) to make your AI impactful. The IT expertise needed will depend on the technology, and how it will be deployed.
Ethical Considerations
Make sure you engage with governance, privacy and ethics teams early on, as their knowledge and input in the quickly evolving field of ethical AI will be crucial to ensure compliance of the technology. Making sure AI is defensible (e.g. it meets the highest standards of integrity and reliability) should be a key part of your strategy.
2. Process: Integration and Workflow
Seamless Integration
AI systems must fit seamlessly into existing workflows. Consider interoperability with electronic health records (EHRs) and other systems. For instance, working with hospital EHR teams on techniques such as Smart-on-FHIR can ensure a more seamless process for integration.
Workflow Optimization
Identify bottlenecks and inefficiencies in current processes. Anything you can do to increase clinician-patient face-to-face time will be well received. Removing onerous or repetitive tasks will allow you to focus on patient care. For example, tasks that involve lots of medical record review can be mitigated with technologies like NLP, especially if these insights can be served to end users within their existing workflow.
Change Management
Prepare staff for AI adoption through training and clear communication. Change management strategies should address concerns, dispel myths and emphasize the benefits of AI. One of the biggest concerns you will hear is “will this take my job?” In just about every example of AI that I have seen successfully implemented – the AI is not replacing people (there are already not enough people!) – but it is instead working hand in hand with them to enable those people to put their unique skills and expertise where they have most value.
3. Technology: the AI!
With the above in place – you will be positioned for your AI to have a positive impact. If you have involved the stakeholders in the design of your solution, it will be well placed to succeed, and with their involvement in the rollout – the AI will be able to impact patient lives for the better.
Case Study: Endeavor Health's AI-NLP Solution
Endeavor Health (previously NorthShore – Edward-Elmhurst Health) is a large health system in Illinois, providing care across multiple hospitals and clinics. The organization identified that there was very low capture of SDOH within their EMR, despite there being significant evidence at a census level that social determinants of health were impacting patient outcomes. Having successfully piloted NLP algorithms and software from IQVIA for identifying these features from patient’s unstructured medical records, they set about deploying this technology within the ED.
Described as a “game changer,” the deployment allowed social workers to screen 56% more patients. The success relied upon deliberate interaction with a broad multidisciplinary team, putting in place a process that enabled the social care workers to do more with the time and resource they had.
Data scientists, social care workers, data warehouse technicians and EMR analysts worked alongside IQVIA NLP team members to put in place a process that leverages the AI at the point of care. This is a perfect example of people, process and technology coming together for patient benefit.
Conclusion
In the dynamic landscape of AI in healthcare, success hinges on collaboration, process optimization, and thoughtful technology adoption. By embracing these principles and learning from real-world case studies, healthcare organizations can harness the full potential of AI while keeping patient care at the center of what we do. You can learn more about the risks and benefits of AI in our on-demand webinar.