They plan to further refine and test the algorithm and have their dietitian see 10 more patients next week. This process has increased the reliability of the algorithm and now allows consultations to be actually ordered into the electronic medical record (EMR).
“We’re seeing six to 10 people a week with undiagnosed malnutrition. If you think about it from the perspective of the child’s family, that’s a huge difference. That has huge implications from a practical AI perspective. Other ideas have come up, but that was one of our great use cases,” Higginson explained.
An AI integration journey to create new clinical algorithms
Five years ago, Phoenix Children’s Hospital embarked on a journey to harness the power of AI to solve clinical challenges. Traditional approaches that rely on biostatisticians to develop algorithms have proven to be time-consuming and often inefficient. He said teams could spend months working on an algorithm and ultimately find it doesn’t work. So Higginson’s team took a different route, using automated machine learning. This approach involves feeding datasets to AI systems that autonomously create algorithms, allowing hospitals to start using them within hours rather than weeks.
One of the key lessons from using AI in healthcare is that it’s rarely possible to get it right on the first try. Therefore, an iterative approach is essential to fine-tune the algorithm over time.
Currently, there are many vendors selling commercially available AI algorithms, many of which are generalized to hospitals’ needs, Higginson said, leading to the development of highly customized proprietary algorithms. That’s another reason why I decided to do it.
“One of the things I’ve learned over the years about AI is that it doesn’t translate very well. So I’m always very skeptical of vendors who say, ‘We have an AI model that works.’ It’s a target. “Because geography has a big impact as well. There are obviously some clinical conditions that it affects, but we’ve seen recent examples where a model is trained in one state and then lifted to another location and doesn’t work. I think I saw some,” he said. .
For example, he said, they have created AI models for operations that require completely unique, highly local and customized elements, such as managing donors and their employees. “Understanding how far is too far for employees to commute to work all depends on road density and where employees are traveling from. I think the concepts and ideas are transferable. But , I’m a little skeptical about just accepting that black box as is, “just hoisting it somewhere else,” Higginson explained.
Customize your solution for pediatrics
Pediatric healthcare presents unique challenges that often require customized solutions. Phoenix Children’s Hospital has developed a unique patient portal because we recognize that pediatric patients and their families have a distinct relationship with their healthcare providers. This patient portal addresses the complexities of patient relationships within families and guardianship scenarios. This includes who has access in a divorce or nursing home situation and the age at which patient information must be shared with the patient.
Additionally, the hospital has adapted to the post-pandemic situation by implementing telemedicine services, which have been particularly well received by pediatric patients and their caregivers. Higginson said the introduction of hybrid telemedicine, where patients and their caregivers participate in virtual visits, has changed the healthcare experience for families.
The future of AI in healthcare
Higginson encourages more general applications of AI in healthcare, highlighting its adaptability to a wide range of scenarios. He used the example of AI helping determine no-show rates to improve emergency room staffing. As another example, AI can be used to screen patient emails sent to physicians through patient portals and determine the most appropriate recipients within the medical team. This streamlines communication and improves efficiency, allowing physicians to practice to the fullest extent of their license without having to spend significant time sorting through basic email requests. Higginson said he has been told time and time again by doctors that more than 80 percent of these messages to him are about schedules, medications and billing that have nothing to do with doctors.
“So how cool would it be if you could run a message you received into a GPT prompt and ask, ‘Which help desk should I contact?'” he said.
Phoenix Children’s Hospital’s innovative approach to AI demonstrates the technology’s tremendous potential in healthcare. By taking a strategic and iterative approach, we have successfully developed clinical algorithms that not only improve patient care, but enhance the overall healthcare experience for pediatric patients and their families.
Learn more about HIMSS.