Project abstract (1000 Characters): Managing newborn bilirubin levels is important because untreated severe hyperbilirubinemia can lead to brain damage. Timely administration of phototherapy can prevent dangerous consequences of hyperbilirubinemia. However, collecting the relevant data and evaluating that data against age- and risk-based phototherapy treatment thresholds can be tedious and time consuming. By developing an app for neonatal bilirubin management, we hoped to both ensure newborn safety trough reliable and quick responses to elevated bilirubin levels and to save physicians time. The Neonatal Bilirubin Management SMART on FHIR application, which we call the BiliApp, retrieves, summarizes and visualizes relevant data automatically. The BiliApp promotes patient safety and relieves providers’ cognitive load. After two years in clinical use, the app is used extensively by providers, saves time, improves patient care, and is highly rated by clinicians.

Project rationale, impact and innovation (3500 Characters): From 1986 to 2001 there were 5-11 cases of disabling brain damage caused by severe hyperbilirubinemia per year reported in the USA (Bhutani et al, 2009). To minimize the risk of disabling brain damage, the American Academy of Pediatrics (AAP) developed guidelines on neonatal bilirubin management (AAP, 2004). The AAP guidelines recommend bilirubin screening for all newborns. Without assistance by health IT tools, screening results and relevant risk factors must be retrieved manually by physicians in nurseries and outpatient clinics, adding to their already overwhelming work load. A technical solution is needed which could both ensure the safety of the newborns and reduce physicians’ cognitive load. These goals could be achieved by automatically retrieving, summarizing and evaluating the data related to hyperbilirubinemia risk. Recent development and adoption of the SMART on FHIR framework allowed us to develop an app to meet these needs.

This app is innovative in several ways. First, the app leverages custom FHIR APIs developed into the EHR to retrieve relevant data not available in the base set of FHIR data supported by the EHR, such as laboratory data pulled from the mother’s record that influences the management of the baby. Second, the app adjusts to differences in EHR data availability; for example, the app adapts to situations in which the baby’s birth time is not available due to the patient being born outside of the health system. Third, the app incorporates predictive analytics in predicting the risk of rebound hyperbilirubinemia following phototherapy.

In the USA, there are over 4 million new birth yearly and over 90,000 pediatricians care for those infants. Based on our evaluations (manuscript under review), if deployed nationwide, our BiliApp has potential to improve the safety of newborns and save over 300,000 hours of provider time every year. Currently, the app has already changed the care for newborns at University of Utah Health and was used over 20,000 times in 2018 by newborn care providers.

Project design and implementation (7000 characters): In 2016, the BiliApp was requested by pediatric physicians at the University of Utah Health, an academic health system. The bilirubin management app available in the SMART on FHIR gallery at that time did not fully meet our physicians’ needs. Thus, an enhanced version of this app was developed by the University of Utah ReImagine EHR team with iterative feedback from physicians. The main challenge we had to overcome was that some data elements were not supported by the FHIR APIs of our EHR vendor, Epic®. We had to balance clinicians’ request for more features with portability restrictions produced by adding custom APIs. We decided to focus initial implementation on clinician needs, with custom FHIR services implemented as needed. Following significant enhancements by University of Utah Health, the BiliApp was released system-wide for clinical use on April 12, 2017.

The application is written in Java version 8 programming language. The clinical logic was defined using Drools business rule management system (https://www.drools.org/). The rules were written against a QICore FHIR logical model (https://github.com/HL7/fhir-qi-core). We initially used the Highcharts JavaScript package (https://www.highcharts.com/) for visualization, then replaced it with the Plotly package (https://plot.ly/javascript/) to facilitate our desired goal of enabling free access and use of the tool by any healthcare system. The rules are executed inside of OpenCDS. OpenCDS exposes a CDS Hooks service interface which is then consumed by the SMART on FHIR application.

We sought to improve patient safety by providing a graphical visualization which minimizes the potential of overlooking high bilirubin results. This approach was validated by physicians in interviews. We addressed the problem of tedious and error-prone data gathering from EHRs and transfer to an external website by automatically retrieving the relevant data from the EHR and visualizing them in an easy to digest form.

We are actively working to disseminate the BiliApp as a free tool through EHR “app stores.” We are adapting the app for dissemination to simplify its adoption by other institutions. The minimum requirement for the app implementation is that the EHR supports SMART on FHIR. We are considering enabling differential features based on EHR FHIR capabilities to promote app usage by other institutions.

Project evaluation and sustainability (3500 characters): We evaluated the outcomes of our project after the app was in clinical use for two years. To fully evaluate the impact, we combined qualitative and quantitative methods. We estimated app usage rates for the newborn cohort (>=35 weeks gestational age, admitted to the nursery unit) from EHR log data. We evaluated changes in guideline compliance by comparing the percentage of newborns for whom phototherapy was ordered within 4 hours of the first inpatient serum bilirubin result above the AAP treatment threshold before and after the introduction of the app. We estimated time saving from a randomized experiment recording time required for completing bilirubin management tasks using the BiliApp versus the previous standard of care. We assessed user satisfaction through interviews and surveys, including the system usability scale (SUS).

We cannot share detailed results of the evaluation here because our manuscript is currently under review. However, we will report the overall results here. The usage of the app surpassed 80% 8 months after the app was introduced and remained stable around 90%. Guideline compliance improved significantly after the introduction of the app. Estimated time saving for clinical providers translated to 365 hours per year. User satisfaction was exceptionally high.

Twitter project summary (140 characters): New Neonatal Bilirubin EHR App provides an exceptional user experience and saves clinicians time: https://reimagineehr.utah.edu/products/

How is FHIR used in the App being demonstrated (500 characters)? : EHR FHIR services are used to retrieve patient data. We also created additional FHIR services to access data elements not currently available through standard EHR FHIR services. The app passes these FHIR data to a CDS Hooks service, which executes its clinical logic against a QICore FHIR logical model. Ultimately, the data and patient-specific recommendations are provided through a SMART on FHIR user-facing application that is accessed through the EHR.

1. What FHIR release does your application use? (500 characters)?: The app uses FHIR version STU3 (3.0.1), US Core FHIR profiles version 1.0.1 and QI-Core FHIR profiles version 2.1.0.

What is the data source for the FHIR resources and how are the FHIR resources accessed? (500 characters): We use EHR FHIR server APIs as well as custom EHR FHIR APIs through EHR Web Services.

Any other information about the project we should know about (1500 characters)?: A screenshot from the app is available as the first app in https://reimagineehr.utah.edu/products/. Clicking on an image of the app there, or directly via the following URL, brings up a system screenshot (https://daqy2hvnfszx3.cloudfront.net/wp-content/uploads/sites/21/2018/02/15013549/Screen-Shot-2018-02-15-at-1.34.50-AM.png).

A related presentation, AMIA Control # 3203036, ‘Balancing Functionality versus Portability for SMART on FHIR Applications: Case Study for a Neonatal Bilirubin Management Application’, will be presented as part of session S90: Oral Presentations on November 19, 2019 from 3:30 PM to 5:00 PM.

A related manuscript, ‘Impact of an Electronic Health Record Add-on App for Neonatal Bilirubin Management on Provider Efficiency and Patient Care’, is currently under peer review for publication.

Authors:

Polina Kukhareva (Presenter)
University of Utah

Julie Shakib, University of Utah
Carole Stipelman, University of Utah
Salvador Rodriguez-Loya, University of Utah
Phillip Warner, University of Utah
David Shields, University of Utah
Kensaku Kawamoto, University of Utah