Tracking pediatric sepsis with clinical data exposes gaps in hospital billing records

When a child arrives at a hospital with a racing heart, plummeting blood pressure, and an infection spiraling out of control, the clinical team does not pause to consider how the case will be coded for billing. They treat. But weeks later, when that hospitalization becomes a line in a national database, the billing code assigned to it determines whether the child counts as a sepsis case at all. For years, that disconnect between bedside medicine and administrative record-keeping has clouded the picture of how many American children develop sepsis and how many die from it.

A study published March 22 in JAMA attempts to close that gap. Researchers led by Chanu Rhee at the Harvard Pilgrim Health Care Institute and Scott Weiss at Nemours Children's Hospital built a surveillance tool that bypasses billing codes entirely, drawing instead on the raw clinical data already sitting in hospital electronic health records.

The Pediatric Sepsis Event definition

The new tool, called the Pediatric Sepsis Event (PSE), adapts the CDC's existing Adult Sepsis Event framework for children. It incorporates the Phoenix pediatric sepsis clinical criteria - a consensus definition published in recent years - and translates them into data elements that can be extracted automatically from electronic health records: lab values, antibiotic prescriptions, and markers of organ dysfunction.

The logic is mechanical by design. A computer queries the record for specific thresholds - particular lab results within particular time windows, paired with antibiotic orders of a certain duration. If the pattern matches, the hospitalization is flagged as a sepsis event. No human coder needs to make a judgment call, and the same algorithm runs identically whether it is applied at a large academic medical center in Boston or a community hospital in rural Texas.

When validated against physician chart review, the PSE definition proved more accurate than billing codes at identifying true sepsis cases. That validation step matters because an algorithm is only useful if it catches real cases and does not generate excessive false positives.

The scale of pediatric sepsis in America

The researchers applied the PSE to 3.9 million pediatric hospitalizations from 2016 through 2023, drawn from hundreds of hospitals and health systems across the country. The results paint a consistent picture:

• Sepsis occurred in 1.3% of pediatric hospitalizations - about 1 in every 75 hospitalized children.
• More than 1 in 10 children with sepsis died during their hospital stay.
• Nearly 1 in 5 of all pediatric hospital deaths involved sepsis.
• National rates of pediatric sepsis and associated mortality remained relatively stable from 2016 to 2022.

Extrapolated nationally, these numbers suggest more than 18,000 children develop sepsis in US hospitals each year, with more than 1,800 dying before discharge.

Most cases were already present when the child was admitted, but a smaller proportion developed during hospitalization. That hospital-onset group had higher mortality rates, underscoring the importance of infection prevention programs within hospitals alongside efforts to recognize sepsis early at the front door.

Why billing codes fall short

The stability of the PSE-derived rates is itself noteworthy. Previous studies using billing codes had suggested large swings in pediatric sepsis incidence over similar time periods. But those swings may have reflected changes in coding behavior - hospitals becoming more or less aggressive about applying sepsis codes - rather than actual changes in how many children got sick.

Billing codes are created for reimbursement, not epidemiology. A hospital may have financial incentives to code certain conditions more frequently. Training on coding practices changes. New code sets get introduced. All of this introduces noise that makes it genuinely difficult to answer a basic question: is pediatric sepsis becoming more or less common?

A clinical-data-based approach strips away that noise. The vital signs, lab results, and medication orders in the record do not change depending on who fills out the billing form.

What this framework cannot do

The PSE is a measurement tool, not a treatment. It tells hospitals how many sepsis cases they have; it does not tell them how to prevent or treat those cases. The gap between counting a problem and solving it remains wide.

There are also practical limitations. The framework requires hospitals to have electronic health records that capture the necessary data elements in a structured format. Smaller facilities, particularly those in lower-resource settings, may not meet that bar. The study's validation was performed at participating academic centers, and its accuracy at community hospitals or critical access hospitals has not been independently confirmed.

The study captures only hospitalized children. A child who presents to an emergency department with early sepsis, receives antibiotics, and goes home without being admitted would not appear in these numbers. The true burden of pediatric sepsis across all care settings is likely higher than what hospitalization data alone can reveal.

Finally, the study period ended before the full impact of post-pandemic changes in childhood infection patterns could be assessed. Whether pediatric sepsis rates have shifted meaningfully in 2023 and beyond remains an open question.

From numbers to action

The CDC's investment in this work signals an intention to build pediatric sepsis surveillance into the broader infrastructure of public health monitoring. If the PSE definition is adopted widely, it could allow hospitals to benchmark their performance, track the impact of quality improvement initiatives, and identify disparities in sepsis outcomes across patient populations and geographic regions.

But surveillance without action is just bookkeeping. The 1,800 children who die from sepsis in US hospitals each year need more than accurate counting. They need earlier recognition, faster treatment, better infection prevention, and health systems willing to treat those numbers as a call to change practice rather than a statistic to file away.