Epic's pediatric sepsis model misses two-thirds of ED cases

Epic's sepsis models historically underperform

These findings follow scrutiny of Epic's sepsis model for adults, a saga that dates back to 2021, when researchers reported the model correctly flags sepsis risk only 63% of the time.

In August 2023, researchers writing in JAMA Network Open reported that the vendor's Sepsis Prediction Model was less timely and missed more cases than other clinical decision support tools. Another 2024 report from the University of Michigan found Epic's sepsis model for adults could only flag some high-risk patients after sepsis had been clinically recognized.

Most recently, the Epic Sepsis Model came in at a 0.65 area under the receiver operating characteristic curve, or AUROC. The AUROC is a one-point scale that measures a model's ability to distinguish positive from negative cases; a score of 0.65 is only slightly better than random guessing. This indicates Epic's sepsis model is still underperforming for adults.

That's notable, given Epic's significant market share.

According to 2025 KLAS data, Epic controls 43.7% of the acute care EHR market and has a 56.9% market share of hospital beds. This makes Epic's pediatric and adult sepsis models the default for millions of patients nationwide.

That market share, plus federal Honor Roll programs designed to increase adoption of clinical decision support tools and sepsis models, could create an adverse situation, the researchers said.

"Our findings suggest that reliance on a low-sensitivity PESM alerting system to satisfy screening mandates may provide a false sense of security, potentially delaying clinician evaluation when alerts fail to fire while contributing to alert fatigue through low-yield signaling," the research team concluded.

"If a hospital chooses to implement a screening tool or is required to do so, published tools with higher sensitivity or demonstrated utility in the setting of quality improvement programs would be better supported by existing evidence."

According to Epic, the Pediatric Sepsis Early Detection (PSED) model, the one investigated in the JAMA Pediatrics report, has been effective in other settings.

"The Pediatric Sepsis Early Detection (PSED) model alerts clinicians to the possibility of sepsis in pediatric patients, in whom it’s rare," an Epic spokesperson said in an emailed comment. "A 2025 study (published in JAMIA) found that after Children's Healthcare of Atlanta put the model in place, children with sepsis received IV fluids significantly faster as part of sepsis treatment."

Moreover, the EHR vendor flagged the varying sepsis measures available for organizations to use. PSED was developed at Nationwide Children's Hospital and made available to Epic users, but each organization validates and configures PSED for its own patients and workflows, the Epic spokesperson said.

"The JAMA Pediatrics study measured PSED against the Phoenix Sepsis Criteria (PSC). They each use a different set of criteria for identifying sepsis. Organizations can use PSED, PSC, or another set of criteria to identify sepsis in line with their specific clinical practices," the spokesperson added.

Finally, Epic pushed back on notes about federal programs designed to incentivize sepsis model use, clarifying that those Honor Roll programs do not incentivize the Epic model specifically, but rather any model of the organization's choice.

"The study also states that hospitals receive Honor Roll incentives for using PSED, which is inaccurate," the spokesperson explained. "Honor Roll requires the use of any sepsis detection tool -- Epic's model or a different tool -- because it's important for patient outcomes to use technology that catches sepsis early.

Sara Heath is an executive editor at Xtelligent Healthcare Media, where she covers patient engagement, healthcare policy and health IT.