Each year over 75,000 children develop sepsis in the United States resulting in substantial morbidity, up to 20% mortality, and billions in US health care expenditures. There have been substantial advances that demonstrate improved patient outcomes with adherence to early aggressive emergency care. However, methods to accurately, reliably, and rapidly identify children who require these resource-intensive therapies are lacking. In addition, understanding the impact of these therapies on near-term outcomes, before significant morbidity occurs, is also lacking. Current algorithms do not reliably discriminate between patients who develop sepsis and those who are clinically similar upon initial presentation but do not progress to sepsis. As a result, children requiring life-saving treatments do not receive them, or do not receive them in a timely fashion, and others may be over-treated, wasting healthcare resources and potentially diverting emergency care from those in need. With the advent of electronic health records (EHR), there are now information-enabled solutions that offer unique opportunities to identify non-biased, heterogeneous samples of children and allow us to accurately and reliably measure risk factors and near-term outcomes for sepsis. This work addresses the critical need to improve pediatric sepsis outcomes by developing methods to accurately identify at-risk children presenting for emergency care. Utilizing the infrastructure of the Pediatric Emergency Care Applied Research Network (PECARN), this proposal will innovatively capture EHR data to create a multi-center registry with the ultimate goal to improve the detection and treatment of pediatric sepsis in the ED setting. To accomplish this, we propose the following specific aims: We will develop an expanded multicenter sepsis registry for pediatric patients from merged electronic health record clinical data from different hospitals with different EHR data sources. We will automate the determination of organ dysfunction in children with sepsis directly from structured and narrative data within the multicenter EHR registry. From the registry and outcome data, we will derive and validate a prediction model of pediatric sepsis using emergency department EHR data from the first 4 hours of care that predicts subsequent organ dysfunction within 48 hours. Each of these aims works to the goal of improving the emergent care for pediatric sepsis with innovative deliverables from this project including the existence of a broad and rich EHR registry, an automated process of outcome determination, and a prediction model of risk of sepsis. We will also have a strong foundation for future projects to implement and evaluate decision support tools, improve diagnostic techniques, engage in comparative effectiveness studies, measure quality of care, establish linked bio-repositories, and guide clinical trial design.The proposed project, thus, has enormous potential to improve our ability to improve the quality of care provided to our most acutely ill children.

Public Health Relevance

Sepsis is a leading cause of pediatric morbidity and mortality with life-saving treatment dependent on early and accurate identification. We will establish a multi-center data registry from electronic health records (EHR), identify a multi-center cohort of pediatric patients at risk for sepsis, automate sepsis-related pediatric organ dysfunction directly from the registry EHR data, and develop an emergency department based prediction model of sepsis related organ dysfunction. Each of these aims has the ultimate goal of improving the emergent care for pediatric sepsis with innovative deliverables from this project including the existence of a broad and rich EHR registry, the automated process of important proximal outcome determination, and an emergency department prediction model of risk of sepsis.

Agency
National Institute of Health (NIH)
Institute
Eunice Kennedy Shriver National Institute of Child Health & Human Development (NICHD)
Type
Research Project (R01)
Project #
5R01HD087363-03
Application #
9626417
Study Section
Biomedical Computing and Health Informatics Study Section (BCHI)
Program Officer
Maholmes, Valerie
Project Start
2017-02-01
Project End
2022-01-31
Budget Start
2019-02-01
Budget End
2020-01-31
Support Year
3
Fiscal Year
2019
Total Cost
Indirect Cost
Name
Northwestern University at Chicago
Department
Internal Medicine/Medicine
Type
Schools of Medicine
DUNS #
005436803
City
Chicago
State
IL
Country
United States
Zip Code
60611