FAQs

Digital tool supports Artificial Intelligence that targets improvements in nutrition-related outcomes

 

Background

Inadequate nutrition early in life is a significant contributor to preterm infants’ poor growth and neurodevelopmental impairment. Hospitals collect vast amounts of patient data stored in the electronic health record (EHR). Artificial Intelligence (AI) can help transform healthcare by improving diagnosis, treatment, and patient care delivery by drawing on the power of large amounts of data. To realize the potential of AI, clean, accurate, and standardized data is needed to fuel the development of algorithms. Developing tools that harness data from the EHR, identify and correct information gaps, and provide high-quality data to support the development of AI applications to improve preterm infant nutrition-related outcomes is urgently needed. 

Purpose

The purpose of this work was to provide the clinical team with a digital tool to access feeding and nutrition data directly from the EHR, analyze the data to identify missing, incorrect, and inconsistent data, and structure the data for feeding models targeting nutrition-related outcomes and quality improvement initiatives. 

Methods

A .NET application collects nutrition and feeding data directly from the EHR using secure Fast Healthcare Interoperability Resources Application Programming Interfaces (FHIR APIs), a standard for exchanging private health information. JavaScript frameworks are used to interpret and present data to the user in an intuitive, user-friendly visual display. 

Results

The EHR-integrated solution provided longitudinal granular-level data on feeding and nutrition for 1,135 patients including 1.18 million feeding events with high resolution details and revealed 2,550 unique feeding order combinations. The data fields collected included enteral feeding type and volume, fortification, parenteral and oral nutrition volumes, and anthropometric measurements. Visualization of this data can highlight charting inconsistencies, missing data points, and inaccurate information in the EHR, which help educate clinicians on best practices and provide high-quality data to support the development of AI applications to improve care.

Conclusion

Implementing software that harnesses high-quality EHR data can fuel the development of AI applications designed to identify complex clinical patterns and assist clinicians in decision making, diagnostics, and individualizing feeding and nutrition plans for preterm infants.

 

Laura Carroll1, Dave Genetti2, Aamir Nayeem2, Tammi Jantzen2, Claudette Ripley3, Ashley Ross1, Misty Virmani1

1Department of Pediatrics, University of Arkansas for Medical Sciences, 2Astarte Medical, 3Arkansas Children’s Hospital