While there are many similarities between the use of artificial intelligence in adult care and pediatrics, there are also some unique and important considerations for the latter, due to the complexity of health issues in childhood and adolescent care.
These include understanding the intricacies of patient-provider relationships – including the relationships with parents and caregivers – as well as privacy and confidentiality considerations as children transition to adult care.
Provider organizations need to recognize these distinctions and make sure AI is being used appropriately to enhance patient, family and provider experiences in pediatric settings.
AI applications powered by natural language processing and large language models are changing how providers engage with patients, says Dr. Natalie Pageler, chief health informatics officer at Stanford Medicine Children’s Health.
We spoke with her recently about the special considerations needed for safe and effective use of AI in pediatric settings.
‘Children change so much’
The complex relationships between doctors and their young patients as they grow up require that provider organizations “engage with the right person at the right time,” Pageler explained.
“Children change so much – both in age and when, where and how they take control of their healthcare as they transition into adulthood. Providers need to understand how clinical notes are written, what information is shared through patient portals and who can access it, as well as engaging in online conversations through these platforms.
“NLP is being used to better understand who is speaking through patient portals – is it a parent, caregiver, sibling or the patient themselves?” she continued. “Technology increasingly is being used to review clinical notes, helping to identify what information is being shared and what should remain private, depending on who is accessing it.”
LLMs are used to process large amounts of meta data to support patient experiences. They can scale the speed at which reports and other outcomes are reviewed, process that information, and better understand where attention needs to be focused.
This enables providers to improve patient experiences and quality of care, while also bringing attention to areas that can relieve administrative burden through automated support, where appropriate, she added.
“Moving forward, AI and other meaningful technological advances in healthcare will continue to be integrated into pediatrics,” Pageler said. “Where appropriate, AI also will impact the level of personalized care we can deliver.
“For example, tailoring precise care for pediatrics,” she added. “This historically has been a challenge as data sets are much smaller and, thankfully, the majority of children are generally healthy. We’re seeing early signs that AI can support more precise care and precision medicine, for example, in genomics.”
Pediatrics’ additional considerations
Pediatric use cases for AI typically are more difficult to implement and require an additional list of considerations before being deployed. One of the best examples is the use of ambient AI and scribe technologies to support note taking during patient-provider interactions.
“Here’s the challenge: Unlike the typical one-to-one conversations common in adult care, pediatric visits often involve many different people in the room of varying ages,” Pageler noted. “Appointments may include entire families, sick children who are screaming, multiple caregivers, and providers who serve multidisciplinary needs.
“As a result, proper testing is required to determine the effectiveness of ambient AI in more dynamic settings, ensuring it supports provider engagement with both the patient and the family,” she continued.
Ambient AI for scribing and notetaking frees providers to focus on the patient and their family. Particularly in pediatrics, a lot of rich data can be captured at the point of care – much of which is behavioral or non-verbal.
For example, important insights can be found in the interactions between parents and families, the subtle nuances of a child’s behavior, body language and emotions – and how it’s perceived by those in the room.
Are providers missing cues?
“Providers who are busy typing notes may miss cues critical to delivering accurate and specialized care,” Pageler explained. “Ambient AI scribes allow providers to spend more time with patients, strengthening relationships and relieving the burden of administrative tasks.
“Unlike in the adult setting where there are fewer privacy and confidentiality considerations, ambient AI and other automated note taking requires additional levels of oversight,” she continued. “Providers must be aware of sensitive information that is transcribed and determine if it is accurate – in addition to who can access that information.”
Pageler is seeing AI applications helping flag potential areas of concern and supporting providers in directing attention to the most important information.
When evaluating AI deployments and algorithms in pediatric populations, there are four key considerations – or barriers to overcome – currently top of the agenda, she said.
Asking the correct questions
“First, the industry must have engaged and educated parties at the table to ask the right questions during technology development and deployment,” she suggested. “We need pediatric providers, along with patients and families, to make sure we’re asking the right questions of IT leaders. This is critical because it helps close the current data gap in pediatrics and informs the direction of application development.
“Second, understanding the data sets we have and that applications are built on,” she continued. “Today, many of the industry’s data sets are focused on adults. This makes technological development in pediatrics more difficult. In addition to having smaller pediatric data sets, pediatric data is subject to stricter protections – and rightly so.”
It’s important to understand where existing pediatric data is insufficient, so providers need to consider what additional data is needed to effectively complete these projects, she added.
“Third, even with good data, it can be ineffective based on the complex nature of pediatric populations and the diversity of patients and families seeking care,” Pageler explained. “Models built on adult data may have useful applications in pediatrics, but they need to be tested extensively.
“Children’s physiology and care needs change dramatically with age, from newborns to teenagers,” she continued. “So, adult-trained algorithms must undergo rigorous testing across different age groups. Take smartwatches, for example. These are primarily built for adults, but our cardiologists are seeing successful applications in older children and teens who are old enough to reliably wear a smartwatch for extended periods.”
Right intervention required
Finally, no AI application is meaningful without the right intervention, she said.
“Pediatricians must be involved in the development of AI applications to help contextualize the information, understand the data and identify the appropriate intervention,” Pageler explained. “For example, AI algorithms can help predict when a patient is unlikely to show up for a clinic appointment.
“Instead of overbooking the clinic schedule to ensure efficiency, pediatric organizations may implement interventions like reminders, child care and transportation support to enable pediatric families to attend their critical clinical appointments,” she added.
The pediatric healthcare industry is buoyed by the potential of AI to strengthen connections between patients, families and providers to improve care outcomes. Pageler said Stanford has long argued that technology shouldn’t be deployed for technology’s sake.
“The ambient AI use cases we’ve explored present untapped opportunity and potential to reimagine the entire pediatric patient experience,” she noted. “The pandemic revealed the power of seeing patients and families in different environments like the home, where those interactions can present richer and more actionable data for providers.
“Ambient AI allows this trend to continue to evolve and strengthen during the in-person experience,” she continued. “This keeps providers focused on the family and the intricate dynamics at play. It also helps relieve the burden of documentation. Again, the complexity of pediatric care stands out. AI can help with the heavy lifting, freeing up providers to focus on delivering added value.”
Potential for positive outcomes
Looking beyond providers, this technology shows potential for nursing and other clinical applications within pediatric healthcare that contribute to positive outcomes, she added.
“This includes the use of AI for clinical decision support,” she noted. “Stanford has been at the forefront of clinical informatics and decision support tools for more than a decade – and the idea of AI helping to speed up and scale these applications to provide support in near real time is promising.
“We also are seeing early signs that AI can be an effective tool to help evaluate images, videos and other forms of visual data that add to the richness of every patient and provider engagement,” she continued.
“Similarly, in settings where providers can’t directly observe a child outside of the exam room or while attending to other patients, AI can help process and evaluate information to strengthen the ‘always on’ continuum of care – without adding burden to the provider.”
Stanford Medicine Children’s Health also has a team researching how AI can support early autism diagnosis and improve the development of therapeutics for children.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: [email protected]
Healthcare IT News is a HIMSS Media publication.
WATCH NOW: Generative AI advice from a top Chief AI Officer ↓↓↓
