Generative AI has increasingly value for pediatric healthcare — and academic children’s hospitals have a key role to play. Learn what they need to do to prepare and how they can have an impact.
By Dr. Amol Rajmane, HPE WW AI Solutions Lead, Healthcare and Life Sciences
The generative AI revolution is here. Gartner placed it at the tallest peak of its 2023 AI hype cycle.1 McKinsey has called 2023 the "breakout year” for generative AI, following a cross-industry survey that finds 22% of businesspeople are already using it for their own work and 79% have had at least some exposure to it.2
In healthcare, generative AI is gaining significant traction. Compared with more traditional AI methods, it has shown potential to:
- Increase speed to market, enabling faster creation of competitive new products and services.
- Simplify data preparation, requiring less labeled data and handling multimodal data more seamlessly.
- Provide smarter interfaces such as those that offer helpful prompts to healthcare professionals.
- Increase predictive accuracy leading to better health outcomes.
And this potential is inspiring rapid market growth — faster in healthcare than in any other industry, with compound annual growth of 85% through 2027 forecast by BCG.3
The revolution hasn’t reached pediatrics — yet. Most of the traction is focused on adult healthcare today. But as challenges to pediatric AI development are overcome, academic children’s hospitals can take an important role in developing and evaluating specialized AI solutions and helping to bring them to market.
Overcoming challenges to solution development
There are significant challenges in using current AI solutions for the pediatric population:
- Difficult to generalize AI solutions developed for adults using adult data across the age spectrum. Children cannot be treated as small adults; their bodies have significant physiological and anatomical differences. Pediatrics therefore presents a distinct set of clinical problems to diagnose and treat. Unlike adult healthcare, very few AI solutions have been cleared by the FDA for pediatric use. For example, only 3% of the 193 FDA-cleared, AI-based radiology tools have been declared suitable for children.4
- Children account for a smaller proportion of healthcare resources. As a result, EHR-derived datasets from children are frequently smaller. Also, pediatric research and associated data collection present unique ethical and practical challenges. Strict consenting requirements, parental or guardian roles, and added data protection requirements introduce many challenges when it comes to conducting pediatric research and collecting pediatric data. As a result, most of the pediatric datasets are typically smaller than adult datasets. Training AI models with these smaller datasets is difficult and may create models that show bias and are not robust.
- There is still scope for improving the quality of Clinical Practice Guidelines (CPGs) in pediatrics, and that in turn is leading to practice pattern variations in care and sub-optimal management of many common pediatric conditions.5 This impacts the ability to collect uniform data for use in AI development.
Due to the renewed focus that generative AI has placed on AI in recent months, efforts to overcome these challenges are likely to grow. The transformative potential of pediatric AI is likely to invite regulatory changes, requiring that related solutions are developed using pediatric data and are rigorously evaluated before they are used to support clinical care.
We have witnessed such a shift in the past for pediatric drug development. Before the year 2000, children were being treated with medications approved for adults with limited or anecdotal pediatric experience through “off-label” prescription. Around the year 2000, drug development programs started to include studies based on pediatric patients. The change was driven by two main pieces of US legislation: the Best Pharmaceuticals for Children Act (BPCA) of 2002 and the Pediatric Research Equity Act (PREA) of 2003, which aimed to improve and incentivize pediatric drug development.6
In the light of current efforts to regulate AI application in healthcare and ensure equitable benefits of AI, it is fair to expect lawmakers to take the same kinds of action if they want children to receive the benefits of AI-based healthcare. With such a legislative action, it may soon become mandatory for AI solutions used for the pediatric population to be developed using pediatric data.
This will help the wider adoption of AI solutions in the pediatric domain and create opportunities to use generative AI to detect, monitor and treat diseases in children.
Generative AI is here to stay and create long-term value
Current generative AI adoption in healthcare mainly aims to streamline administrative and operational tasks, helping to alleviate physician burnout and improve operational efficiency. Use cases include automating medical transcription services, providing decision support for patient triage, and drafting insurance pre-authorization and appeal letters. These will be joined by many other use cases in the coming years.
New large language models (LLMs), which are the foundation of generative AI, are being created to address highly specialized use cases. These LLMs are trained on specialized medical data, evaluated through studies, and retrained to continuously improve outcomes.
The vital role of academic children's hospitals
Academic children’s hospitals and research centers can therefore play a central role in pediatric AI solution development and evaluation, as well as benefiting from broader AI solutions.
Academic children’s hospitals typically:
- Maintain large, pediatric patient datasets for different age groups and disease conditions that can act as valuable training datasets for development of pediatric AI solutions.
- Have strong clinical and ethical expertise to provide oversight through all stages of solution development.
- Are able to perform evaluation studies needed before bringing pediatric AI tools to market, generating evidence necessary for regulatory approval and/or clinical dissemination.
- Have capability to develop effective practices for workflow integration of newly developed pediatric AI solutions, to ensure that these tools deliver scalable and sustained value for patient care in the real-world setting.
To play this pivotal role, academic children’s hospitals and research centers need a reliable AI infrastructure platform and the expertise to advance from experimenting with AI all the way to developing and bringing robust AI solutions to market. They need a machine learning (ML) platform that can help them disengage with current fragmented AI solutions and instead offer them reliability, flexibility, scalability, security, data protection and reproducibility of a robust ML platform.
HPE, through our enterprise-grade, end-to-end machine learning platform, is helping healthcare professionals solve the frustrating problems that block the majority of institutions from successfully developing generative AI models and adopting AI and ML. These problems include choosing the right tools from a myriad of options, supporting a diverse set of use cases and users, scaling to handle increasingly massive datasets, reproducing workflows, and meeting compliance needs.
Along with this HPE machine learning platform, we also provide the partnership and services to guide and support you through each step of your AI journey. If your organization has a vision to be a leader and steer the AI revolution in pediatric health, we can partner with you to explore the possibilities.
Discuss the future of pediatrics with HPE
Sometime in the near future, generative AI’s potential to improve pediatric healthcare will become too great for lawmakers to ignore. When pediatric studies start to be encouraged or even required in AI solution development, pediatric research centers will play a vital role in bringing specialized tools to market. HPE is ready to offer expert partnership and enterprise-grade AI platform solutions today.
Click below to learn more and talk to an HPE representative.
- Learn more about HPE Machine Learning Development Environment
- Talk live with an HPE representative
Amol is Worldwide Healthcare and Life sciences AI Solutions Lead at HPE. In this role, he supports AI-based solution development for HLS vertical and supports formulation of AI team’s GTM strategy for HLS vertical. Prior to joining HPE, he worked as a Sr. Program Lead for AI, Research and Evaluation within the IBM Watson Health team and held many positions in the HLS sector in clinical research. Amol completed medical school at MGM Medical College and Research Center and an FT-MBA in Life Sciences Innovation and Management from Rice University. He is passionate about using the power of technology to revolutionzie healthcare and life sciences with an aim of making modern healthcare accessible and affordable in every part of the world.
Compute Experts
Hewlett Packard Enterprise
twitter.com/HPE_Cray
linkedin.com/showcase/hpe-servers-and-systems/
hpe.com/supercomputing
1. Evolution of pediatric drug development (2019, May 1) https://www.contemporarypediatrics.com/view/evolution-pediatric-drug-development
2. American College of Radiology Data Science Institute (2022, August 10). AI Was Not Designed for Pediatric Patients: What Does That Mean for Them?https://www.acrdsi.org/DSIBlog/2022/08/10/AI-Was-Not-Designed-for-Pediatric-Patients#:~:text=Of%20the%20193%20FDA%20cleared,been%20cleared%20for%20pediatric%20use.
3. Liu, Y., Zhang, Y., Wang, S. et al. Quality of pediatric clinical practice guidelines. BMC Pediatr 21, 223 (2021). https://doi.org/10.1186/s12887-021-02693-1
4. Gartner (2023, August 17). What’s New in Artificial Intelligence from the 2023 Gartner Hype Cycle. https://www.gartner.com/en/articles/what-s-new-in-artificial-intelligence-from-the-2023-gartner-hype-cycle.
5. McKinsey (2023). The state of AI in 2023: Generative AI’s breakout year. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-in-2023-generative-AIs-breakout-year.
6. BCG. Medtech’s Generative AI Opportunity. https://www.bcg.com/publications/2023/generative-ai-in-medtech.
ComputeExperts
Our team of Hewlett Packard Enterprise server experts helps you to dive deep into relevant infrastructure topics.
