Generative AI to bring ‘transformative change,’ says Froedtert/Inception Health CTO

Froedtert & Medical College of Wisconsin Health Network is an academic health system based in eastern Wisconsin. It embarked on a journey to foster disruptive innovations by establishing Inception Health as an independent vehicle to drive innovation and digital transformation, focusing on digital health technology.

Artificial intelligence is at the heart of some of its most disruptive innovations.

So too is Dr. Melek Somai, vice president and chief technology and product officer at Inception Health and Froedtert & Medical College of Wisconsin Health Network, and assistant professor of medicine at the Medical College of Wisconsin. 

Somai knows AI, and strongly believes it is an integral part of the healthcare ecosystem and poised to bring about powerful, transformative change to the industry.

Healthcare IT News spoke with Somai to discuss AI in healthcare overall and AI at Inception Health and Froedtert & Medical College of Wisconsin Health Network. 

In this – part one of our two-part interview with the innovator – we talk about generative AI in healthcare and how hospital chief information officers and other health IT leaders should be preparing for a fast-evolving future for the technology.

Q. What role do you think generative AI should play in healthcare today? And what are the challenges of working with generative AI?

A. Generative AI in healthcare is poised to bring the really transformative change that was envisioned with the advent of information technology. It signifies a move beyond just the capability of traditional AI and machine learning models, which have already made significant contributions to medicine and healthcare.

If we think about what I call “traditional” AI and machine learning models, they have been instrumental in enhancing the safety and efficacy of randomized clinical trials, identifying disease risk factors, and supporting clinical decision making through decision support systems and others like image processing.

The introduction of generative AI represents a paradigm shift. It’s introducing a new operating model for healthcare technology. And like its predecessor, generative AI processes some unique characteristics that far surpasses its predecessor. Things like adaptability, multimodality, context awareness, generalizability, and, to some extent, originality.

Those characteristics enable it to produce what I call non-static output and results. By integrating genAI into the fabric of healthcare, we stand really on the brink of a new frontier in medicine and life sciences. Much like the revolution brought about by the scientific method, genAI has the potential to extend our human discovery in medicine – to open up really new avenues for research, diagnosis, treatment and patient care that were previously unimaginable.

However, like most technology, this is more of an evolution than a revolution. It’s an evolution that underscores the importance of genAI not as a tool, but actually as an integral component of our healthcare ecosystem.

As we move forward, the focus would be most probably on harnessing this technology responsibly and ethically, ensuring that it serves to improve patient outcomes. It has to enhance the efficiency of our healthcare delivery and must contribute to advanced medical science when we focus specifically on healthcare providers.

The integration of genAI into healthcare systems represents a transformative opportunity to enhance our patient care, improve our operational efficiency and foster innovation in diagnostics. If we take a historical perspective, the last decade has focused on digitizing healthcare, with the healthcare industry swiftly implementing EHRs.

However, we all know that while the implementation has been phenomenal, with EHRs becoming quasi-universal across the healthcare industry, we fail to fully grasp the value for patients and we equally fail to support our providers. While EHRs were intended to streamline documentation, improve patient coordination and enhance patient safety, the adoption has often led to unintended consequences.

Providers have faced increased administrative burdens, spending more time on data entry and documentation tasks than on direct patient care, frankly. Furthermore, the use of EHRs during clinical encounters can disrupt the natural flow of communication between providers and patients, with clinicians’ attention divided between screen interaction and face-to-face interaction.

So, this has created a kind of a digital divide that eroded the patient/provider relationship. This decade, AI offers us an opportunity to resolve these shortcomings. But also, it can certainly bring more challenges and potentially more unintended consequences. In other words, we can think of AI as version 2.0 of our initial foray in implementing technology such as EHRs and AI, and machine learning models that focused on clinical decision support.

So those were the first generation of clinical computing I would call transitioning from that era of an EHR to genAI. We have an opportunity to improve care delivery if it augments the provider capability, if it enhances the patient engagement, and if it restores the focus on personalized, patient-centric care.

That promise is not a technology problem. If we would like to take the lessons we learned over the last decades in terms of EHRs, we must build a path where AI is a catalyst for a better healthcare ecosystem for providers and patients alike.

And we must unlock the benefit of digitizing medicine and the care practice. So, it’s essential for us to acknowledge that the successful integration of genAI in healthcare requires us to do more than just implement this technology. While AI holds tremendous potential to improve patient outcomes and streamline healthcare processes, it is really not the panacea that can magically solve all the challenges we face today as an industry.

Q. What do you feel CIOs and other health IT leaders at hospitals and health systems should be doing today regarding AI, as the technology is just exploding all over the industry?

A. This is something that must be the priority of health systems. Our role as healthcare providers is really to understand, evaluate and assess this wave of innovation and seek to benefit the patients and providers. So, what we must consider as we are implementing genAI, we must act and we must understand first that AI technology must be tailored to the specific needs and workflows of healthcare providers and patients.

AI algorithms are not going to improve care by themselves. They must be trained on high-quality, diverse data sets that accurately reflect the complexity of real-world healthcare scenarios. So, without robust data governance practices and data quality insurance that must be put in place, AI models will produce biased and unreliable results, undermining their utility and trustworthiness.

The importance of healthcare data cannot be overstated and the responsibility of us as health systems and healthcare providers is to make sure this is resolved specifically as we enter the next phase of generative AI.

Healthcare data serves as the bedrock upon which AI-driven advancements are constructed. It’s furnishing the essential material needed for training algorithms, validating the models and extracting actionable insights.

Presently, there is a notable enthusiasm surrounding foundation model builds, which have been constructed using extensive sets of general-purpose data, allowing further repurposing with no to minimal retraining in healthcare. These foundation models, like GPT, which is a generative, pretrained transformer model, have demonstrated remarkable versatility in various domains beyond their original training scope, showcasing a potential of generative AI to revolutionize healthcare application without too much training.

However, it’s important to note that while foundation models offer tremendous potential for healthcare applications, we know on the path forward we’re going to need to be able to fine tune and validate those models, which is extremely crucial. So, health systems play a pivotal role in contributing to the era of AI.

It’s not only about building the AI foundation, but also building the data foundation. As healthcare providers, we must be well-positioned, and we are well-positioned, to create the necessary infrastructure, expertise and resources to the unique requirement of patient care and medical research in the area of generative AI.

The role of healthcare providers in AI extends far beyond clinical practice to encapsulate leadership, privacy, security and ethical governance. So, by establishing the organizational foundation to support patient privacy, data security and contributing domain expertise to the AI development efforts, we as healthcare providers can contribute to genAI in healthcare in a manner that prioritizes patient welfare, fosters trust and promotes equitable access to high-quality care.

Another point is that successful adoption of AI in healthcare depends on effective organizations. Healthcare providers must be adequately trained and ready to use AI tools effectively and understand their limitations and potential risk.

Moreover, patients should play a critical role in shaping the role of AI in their care and be empowered to participate in this process and the decision-making process as we evolve. So, another item health systems are starting to evolve into and have done a lot of work today across the industry is the ethical and regulatory considerations that must be prioritized throughout the AI implementation process and journey.

This is a long journey. It’s going to take several decades before we actually find the right organization and right structure around how we ensure the right use of AI in healthcare organizations.

We must ensure AI technology complies with privacy regulation, security standards and ethical guidelines to safeguard patient confidentiality and autonomy, and additionally transparency and accountability for patients. And building those mechanisms should be established to monitor the performance and impact of AI systems.

Beyond evaluation and implementation, we must be able to build the process to mitigate potential biases and address concerns related to algorithmic decision making. To do that, we need to evolve. If you consider the Health Insurance Portability and Accountability Act, this was a critical regulation that provided data privacy and security provisions for safeguarding medical information for patients. But it was enacted in 1996.

It was before the advent of the EHR and the Internet as we know it. Because HIPAA is antiquated, there is a general consensus today that we need to operate a much more advanced model that is aligned with the current scope and capability of information technology.

So that means that implementing AI is going to require us to build a new governance structure and additional capability. One approach I favor is rather than relying solely on internal governance structures, a collaborative approach involving NDC stakeholders and regulators should be advocated to promote quality care and patient safety.

In the era of AI, it has to be a multifaceted governance framework that will provide much-needed guidance and support for health systems tasked with overseeing AI technologies in healthcare settings. We must be able to ensure responsible and effective utilization of these powerful tools as we evolve, but that’s our role. It’s a great opportunity, but it’s going to require massive improvement and a transformation of how we deliver care and how we approach implementing technology in healthcare today.

So, to summarize all of this, we need to be an active participant today in shaping AI innovation and shaping generative AI implementation. Today in healthcare, we must have health systems harness the full potential, but also make sure that it enhances clinical decision making. It streamlines our workflows and ultimately transforms the delivery of care services to improve our patient outcomes.

So that’s our mission, that should be our guiding principle.

IT leaders in healthcare had to respond to market dynamics enforced by incentive programs like meaningful use and the HITECH Act that actually drove health IT implementation at an incredible wave and speed. And because of that rapid wave and the strict requirements that were part of meaningful use driving it, adoption has been critical, but led to some unintended consequences.

So, health IT suffered from that lack of a strategic vision of the role of technology in healthcare. This has led most health IT to be considered an organizational unit separate from clinical care and more of a back office supporting the care organization and operations.

However, as we move today, health IT leaders have a responsibility internally to support their employees to adapt to the new technology shift required in order to bring the value of AI. It means also adapting the IT organization to respond more swiftly and more aptly to the changing requirements and the needs of the healthcare organization.

It also means more strategic investment in cloud computing, data mesh architecture, zero trust security, digital infrastructure and modernization of the technology stack.

More importantly, we need to support the workforce to embrace this wave and support increasing the skills and the capability of our workforce in terms of data and AI literacy. Another challenge for health IT leaders is to enable organizational alignment. We need to be able to be part of the digital transformation, to proactively engage and not be a bottleneck.

The wave of AI is so fast. Every month, every week, there is a new AI model, there are new capabilities that can actually transform. But at the same time, we need to build the right guardrails around it. So, there is this challenge, between the urgent needs of today, but also the important needs of tomorrow.

And we need to be evolving in that nature. The role of health IT is only going to expand, and it’s going to merge, honestly, as we build generative AI across the entire spectrum of healthcare. There is going to be a new way of how we organize, how we train and how we deliver technology.

To watch a video of this interview that contains bonus content not in this story, click here.

PART 2 OF THIS INTERVIEW WILL APPEAR TOMORROW

Editor’s Note: This is the sixth in a series of features on top voices in health IT discussing the use of artificial intelligence in healthcare. To read the first feature, on Dr. John Halamka at the Mayo Clinic, click here. To read the second interview, with Dr. Aalpen Patel at Geisinger, click here. To read the third, with Helen Waters of Meditech, click here. To read the fourth, with Sumit Rana of Epic, click here. And to read the fifth, with Dr. Rebecca G. Mishuris of Mass General Brigham, click here.

Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.