Jeremy Meller on the Future of AI in Pediatrics

Jeremy Meller has served as the Chief Information Officer for Children's Healthcare of Atlanta for over 10 years. He serves as the deciding doorman, carefully selecting which new AI technologies enter CHOA.

Tell me about your background and what gravitated you toward the pediatric field?

I come at healthcare IT from a mix of military, operational, and health system leadership experience. Early in my career, I served as a Captain in the U.S. Army Medical Service Corps, where I was responsible for logistics and IT across multiple military hospitals. That experience shaped how I think about reliability, mission‑critical systems, and the responsibility that comes with supporting caregivers in high‑stakes environments.

After transitioning to civilian healthcare, I held CIO leadership roles at organizations like Marshfield Clinic, Iowa Heart Center, and Spectrum Health before joining Children’s Healthcare of Atlanta in 2012. What ultimately pulled me into pediatrics, and kept me here, is the clarity of purpose. I tell people that there is a special urgency in pediatrics and a dedication that goes beyond typical medicine. Technology has become vital to how we deliver care and has the potential to greatly improve its effectiveness. That combination creates both a challenge and a calling—it aligns closely with why I went into healthcare leadership in the first place.

What is a recent success and a work‑in‑progress project in your role as CIO?

A recent success I’m particularly proud of was the opening of the Arthur M. Blank Hospital. From a technology perspective, this was one of the most complex pediatric hospital launches in the country—over 60 new systems, more than 5,500 integrations, and one of the world’s largest deployments of autonomous delivery robots. The foundational infrastructure performed exceptionally well and positioned us for long‑term AI and innovation work.

At the same time, that experience reinforced an important lesson: success isn’t just about implementing advanced technology—it’s about adoption. Some of our most advanced capabilities initially saw slower utilization, not because the technology failed, but because workflows, training, and change management needed more attention. A major work‑in‑progress since then has been tightening the loop between technology design and real‑world clinician behavior so that innovation translates into sustained value.

As CIO, what do you look for in high‑quality technology architecture and clinical systems?

First and foremost, I look for architecture that is resilient, interoperable, and designed for scale. Pediatric healthcare is complex by nature—multi‑specialty, data‑dense, and increasingly digital. Our systems must integrate cleanly, support real‑time decision‑making, and remain secure without becoming barriers to care.

Equally important is human‑centered design. Clinical applications have to fit into how clinicians actually think and work, not the other way around. If technology adds steps, increases cognitive load, or lives outside the workflow, it has a harder time delivering value. This principle increasingly guides how we evaluate everything from EHR enhancements to AI‑enabled decision support.

What motivated your desire to integrate AI into CHOA’s systems, and where is it currently implemented?

The world is changing faster than we think, and autonomous AI will fundamentally change the nature of how we work. The organizations that will lead in pediatric healthcare over the next decade are the ones building their AI foundation now — not reacting to it later. So we want to be well-positioned, make reasonable investments, and learn quickly.

We’ve worked for the last year to develop a comprehensive AI strategy to advance our organization and align with our goals. We've deployed generative chat, are working on deploying a custom assistant/agent framework, and have a large number of product evaluations going on. Additionally, our predictive analytics team is very busy working on models and tools to speed up and improve decision-making.

What improvements—and areas for improvement—have you seen since implementation?

In addition to some of the same early gains that many organizations have seen with ambient AI scribes and generative chat, I would highlight a couple of really impactful AI projects at Children’s:

ED Sepsis Detection — This is one of our most mature and impactful deployments. We combined two predictive models — an Epic model at ED triage and a custom homegrown model at PICU disposition — to identify septic children faster. The result: a measurable, special-cause reduction in time to first fluid bolus and time to first antibiotic, both directly linked to reduced organ dysfunction and mortality. We've published three manuscripts on this work. The model fires 130–200 times per month in our three EDs.

Asthma Exacerbation Risk — Using Epic's predictive model in our pulmonology and allergy clinics, we identify children at risk of a severe asthma exacerbation in the next 90 days. Implementation was associated with a 25% reduction in asthma-related ED visits — a published outcome. This is a great example of a predictive model that shifts care from reactive to proactive.

Non-Accidental Trauma Recognition — A natural language model that identifies high-risk fractures in radiology reports for young children, then surfaces that risk to ED clinicians to trigger a skeletal survey and social work consult. This was associated with a 40% reduction in missed evaluations for non-accidental trauma.