AI in Pediatrics Lags Behind Adult Medicine – Experts Weigh In

The gap between innovation in adult and pediatric medicine is widening when it comes to artificial intelligence, experts say. While AI is rapidly transforming healthcare, its adoption in pediatric settings is lagging, raising concerns about potential disparities in care. This disparity isn’t due to a lack of interest, but rather a complex interplay of factors ranging from the unique challenges of applying adult-focused AI models to children, to financial constraints within pediatric healthcare systems.

The “Kids Aren’t Just Little Adults” Problem

The fundamental principle guiding pediatric care – that children are not simply smaller versions of adults – extends to the realm of artificial intelligence. As Clara Lin, MD, vice president and chief medical information officer at Seattle Children’s Hospital, explained in a recent Healio Community webinar, tools designed and trained on adult populations often don’t translate well to pediatric patients. “As pediatricians, we grow up in medical school and training reciting ‘kids are not just little adults,’ and you can’t retrofit adult solutions into pediatric populations,” Lin said. “We learned that in the medical context, but We see so much more true in IT and particularly in AI.”

This presents a significant hurdle. Ryan McAdams, MD, a neonatal intensivist at University of Wisconsin-Madison and cofounder of NeoMIND-AI, highlighted the safety and privacy concerns that arise when applying AI models developed for adults to a vulnerable pediatric population. He noted that many AI tools are initially invented for and trained on adult datasets, leaving unanswered questions about their suitability and potential risks for children. Research underscores the need for careful consideration of these risks before widespread implementation.

Financial Barriers and Infrastructure Gaps

Beyond the technical challenges, financial realities are significantly hindering AI adoption in pediatrics. James S. Barry, MD, MBA, medical director of the level III neonatal ICU at Children’s Hospital Colorado and cofounder of NeoMIND-AI, pointed to the often-precarious financial state of children’s hospitals. Many operate on thin or even negative margins, making it difficult to invest in expensive AI infrastructure and development. Barry estimates that in the next few years, only around 20% of children’s hospitals will have the necessary resources to effectively utilize AI in pediatric care.

This financial disparity could exacerbate existing inequalities in healthcare access. If AI-powered tools primarily become available at well-funded institutions, it could widen the gap in quality of care between those hospitals and those with fewer resources. Lin echoed this concern, warning that without careful attention, the disparity between adult and pediatric innovation will only grow as AI continues to evolve.

Taking a Proactive Approach to Pediatric AI

Despite these challenges, there is a growing movement to proactively address the gap in pediatric AI innovation. Lin and her team at Seattle Children’s Hospital are taking a lead role, developing and testing AI tools specifically designed for pediatric care. More than 150 clinicians at the hospital are currently participating in these trials, demonstrating a commitment to tailoring AI solutions to the unique needs of children.

McAdams highlighted the potential to leverage existing hospital infrastructure to integrate AI. He pointed out that many neonatal intensive care units (NICUs) already have cameras and other monitoring equipment in place. “The good news is cameras are already in the NICUs,” he said. “You can add a microphone, you can have a luminometer… It’s all technically there. You could then coordinate all of that with [AI] devices as well.” Utilizing unstructured data, such as clinical notes, also presents opportunities for AI applications.

Advocacy and the Future of Pediatric AI

Experts emphasize the importance of proactive engagement from clinicians and advocates. Barry encourages healthcare professionals to acquire involved in the development and implementation of AI tools, drawing a parallel to the early days of electronic health records (EHRs). “I equate it to the EHRs,” he said. “I think all of us can agree that EHRs don’t work as well for us and our patients as we would like them to, and part of that reason is we were never involved in the development of our EHRs.”

Advocacy at the state and national level is also crucial, particularly when it comes to securing increased funding for children’s hospitals and reforming reimbursement models for pediatric care. Barry believes that a fundamental change in how pediatric healthcare is financed is necessary to unlock the full potential of AI in this field. He stressed that addressing these systemic issues requires a collective effort from healthcare professionals, policymakers, and advocates.

Looking ahead, McAdams anticipates that financial pressures will eventually drive hospitals to integrate AI into their existing infrastructure within the next 3 to 5 years, as the benefits of improved care and reduced length of stay become increasingly apparent. AI models predicting sepsis in pediatric emergency departments are one example of promising developments.

James S. Barry, MD, MBA, can be reached at [email protected]. Clara Lin, MD, can be reached at [email protected]. Ryan McAdams, MD, can be reached at [email protected].