How AI Helps Children’s Hospitals Bring Clarity and Calm

Mom, I’m in the ER with Olivia. She has a fever of 104. I’m so worried and don’t know what to do. I just wish you were here.

As an emergency department physician, cases involving children are some of our most challenging. Not because their conditions are inherently more difficult to treat. On the contrary; children are amazingly resilient. It’s because many of us who work in the ED—physicians, nurses, techs, etc.,—are parents ourselves, and there’s nothing that weighs on us more than seeing a worried mom or dad with a sick child.

While those working in children’s hospitals are among the most resourceful health care professionals I’ve ever worked with, many lack the most important resource of all: time. 

But I’ve also seen some pretty innovative ways to solve for this throughout my 11-year career. More recently, I have seen artificial intelligence (AI) used to excellent effect in reducing administrative burden, enhancing communication with parents, and supporting the kind of health literacy that leads to more productive conversations with clinicians.

Challenges facing children’s hospitals

Before I share my thoughts around using AI in pediatric EDs, let’s review what children’s hospitals are up against today.

There are approximately 131 million visits to the ED in the US every year. Thirty million of these visits are for children. Writ large, about 1 in 5 children will need to visit the ED every year. Within the pediatric group, children under 1 year old make up the most common visitors to the ED.  

That’s a lot of ED visits. And although many result in an inpatient stay, many hospitals are actually shutting down their pediatric units—a development that is all too familiar. From 2008 to 2018, there was a 20% decline in the number of hospitals with inpatient pediatric care units. While this trend may have helped children’s hospitals capture a bigger market share, the result is the same for parents: they must travel farther to obtain care for their children.

Another challenge has to do with surges. Last year’s “tripledemic” of Covid, seasonal influenza, and RSV resulted in huge ED overflows, with some children’s hospitals even setting up tents outside their facilities. Now with positive Covid tests and Covid-related emergency department visits ticking up, the CDC is concerned that the U.S. may face another triple whammy this year.

Meanwhile, our nation faces a nursing shortage. Nearly 80% of nurses already report that their units are inadequately staffed—a particularly troublesome statistic for children’s hospitals where nurses usually need additional training beyond what they would get in nursing school. The effects of this dearth of talent (poorer outcomes, more medical errors, additional expenses, etc.) are likely to increase exponentially with as many as one-third of nurses reporting they plan to leave the profession in the next few years.

A note from the silver lining department

When it comes to responding to challenges like these, I’ve found children’s hospitals to be very creative at problem solving. While other provider types are slow to adopt new technology for increased efficiency and better patient experiences, for example, children’s hospitals are more likely to embrace it. 

Perhaps this is because they have such a good read on what parents expect during an ED visit. The median age for new mothers is now 30 years old, meaning that many parents of infants and small children came of age when Amazon’s 1-click feature, Uber, and Netflix had already been well established. Today’s parents expect that technology to be part of the care experience, just as it is in the other “one-tap moments” of their daily lives.

Truly, children’s hospitals are well positioned to adopt new technologies, including those powered by today’s AI. In fact, I’d say that children’s hospitals stand at the forefront of innovation and thought leadership for other facilities that regularly care for children. So, let’s jump in!

The role of AI in children’s hospitals

Good communication is vital to improving patient experience. In fact, one-third of complaints in the ED are attributed to poor communication. There’s no better proving ground for this than the ED in children’s hospitals, especially if yours is overcrowded or understaffed. AI can relieve some of that burden while contributing to a positive patient experience.

Let’s start with wait times. Predicting wait times is difficult. Do you give the patient an average, the “longest wait on the board,” a gestalt, a guess? Many EDs shy away from providing wait times because, honestly, it is very difficult for people to do! The operational inputs in the web of emergency department flow are many and not always visible to ED staff in the waiting area.   

But AI can easily calculate personalized wait times for each and every patient in the waiting area.  Not only that, but it can do it without being governed by staff. It doesn’t need to be constantly monitored and updated, and it can seamlessly recognize and adjust to changes in the departmental flow: Closing a few beds because of a staff call-out? Hospital full and boarding is worsening? A bus pulled up with a poly-trauma situation? None of these scenarios is a problem for AI). 

That’s a good start. But as my colleague Stephanie Frisch points out, there’s no such thing as an “average” ED patient, especially in children’s hospitals. To get a personalized idea of individual wait times, factors unique to each patient (e.g., emergency severity index score, reason for visit, age, etc.) and the volumes coming in the ambulance bay should be taken into account. AI can do this with lightning speed and provide parents with real-time updates with up to 97% accuracy.

AI can also help address service gaps—an important objective for children’s hospitals who know that it won’t be long before a family returns with the same child, a brother, or a sister with a broken arm, a high fever, or an unexplained rash. Powered by AI, parent-facing mobile apps can automatically route service requests (e.g., “My child is thirsty,” or “Can we get a blanket?”) to the appropriate personnel. Oftentimes, that’s someone in environmental services, concierge care, or the HUC. This means nurses can respond to more clinical needs, thereby allowing them to operate at the top of their license.

Outside the hospital walls, patient-centered mobile apps can be an invaluable partner. Consider the story of Olivia’s visit to the ED that began this article. You can bet her grandma was on pins and needles. With the right mobile app, though, Olivia’s mom could very easily share updates with family members about her daughter’s progress in real time. This can include things like lab results, inpatient transfer updates, or discharge summaries. 

But what about all the confusing medical terminology contained in those lab reports and clinician notes that hospitals must make available to patients? As my Co-Founder Aaron Patzer pointed out, few patients understand medical terms like “cerebral infarction” or can interpret jargon like, “NPO at 00:00.” Fortunately, large language models (LLMs) like those used in generative AI applications like ChatGPT can translate this confusing language into “a stroke” or “don’t eat or drink after midnight.”

Given the right training, validation and testing, generative AI can also parse information (think 15-page discharge summaries) to pull forward the most relevant details for each patient. Nearly 80% of the content is boilerplate information (e.g., COVID policies) or things the patient already knows (e.g., their body mass index). In mere seconds, AI can sift through all that stuff and present the patient with, for example, the top three things they need to do after leaving the hospital: pick up medications, avoid certain foods, set up an appointment with a specialist, etc.

Head spinning yet? Then let’s end with a why-didn’t-anyone-think-of-that-before application of AI in health care. 

Consider how Netflix uses algorithms to recommend content. Well, very similar algorithms can automatically serve up patient education, including video content, based on information contained in the EHR and clinician notes. Especially in children’s hospitals where parents are often more concerned about their child’s health than they are of their own, many find themselves frantically Googling information that should be—that must be—presented to them with empathy and compassion. I can think of nothing more important than using plain language when discussing a child’s health with a worried parent. AI can help with that. Parents ask better questions and we have much more productive conversations.

Why patient experience matters

We all know that poor patient experiences can negatively impact a children’s hospital’s bottom line. More importantly, though, it , too. When parents leave your hospital feeling their child didn’t get the attention they deserved, they’re less likely to follow your post-discharge instructions, use medications properly, or set up a follow-up appointment.

This is what excites me most about applying AI to EDs, especially those in children’s hospitals. Its full potential for enhancing the doctor-patient relationship remains to be seen. But human empathy, compassion, and expertise will always be irreplaceable, and AI should complement these qualities to ensure the delivery of patient-centered care. 

Like many of my colleagues, I am optimistic about the role of AI in healthcare. With thoughtfulness and care, I believe we can mold AI into one of the most powerful tools we have to care for our patients. To do so, we must consider the different roles that AI can play in the healing process and approach them individually with diligence and bravery.

Learn how a top 10 children’s hospital in Pennsylvania uses AI to engage 69% of families that enter their ED.

Read the case study.