FHIR enabling the EeMR has the potential to provide a pipe to all patient data that resides in Cerner in real-time. The reality of the current implementation is more modest, with only 29% or resources and about 8% of resources and operations supports. Hopefully, in time, this will make it possible to ask how a patient is trending over time. Do we see reciprocal trends in other data for this patient? It also allows one to compute what the expected result of therapy might be based on data that are in the warehouse, etc.
Cerner's solution is called the "Ignite APIs" for the Millennium system. This will be implemented by EHC in Q2 of 2018.
This article will explore the practicalities of using the Ignite APIs to FHIR enable a real-world application. The MTP app is the example that will be used because there is a real need to FHIR enable it and because it is a relatively simple app with only a few inputs.
Importnant Note: Because the FHIR specification is implemented differently and with different coverage by different vendors, the type of triage analysis that appears below is essential for any potential project. If an EMR supports FHIR or even supports a specific FHIR resource that is needed, that does not mean that the field values required for the potential application can actually be retrieved or updated. All such values must be specifically inventoried and verified to be supported and available.
Background on the MTP app
Dr. Chris Dente from the Department of Surgery designed and validated an algorithm that reliably identifies those patients who will need activation of a massive transfusion protocol. There are only a handful of inputs needed to do the calculation. The algorithm has already been implemented as a web-based application, a mobile application and as part of the EHR workflow. The result of the calculation made by the algorithm is called the "MTP Probability". The result is stored to a database to be used to, among other things, validate the predictive power of the algorithm against real outcomes. An MTP app can be either clinical or research in nature. For the purpose of this article, any application that implements the MTP algorithm or collects MTP data will be called an MTP app.
The link to the MTP webapp is https://qa.app.emory.edu/mtp/
The MTP algorithm requires only 4 inputs. Currently only 2 of the 4 inputs needed are in the Cerner FHIR sandbox. Cerner said they would enter into disucssions with Emory about how they might be able to support the other 2 inputs. The following table describes the MTP inputs.
|Input Name||Description||FHIR Resource||Notes|
|1||Systolic BP||Systolic blood pressure reading||Observation||Supported by searching for http://loinc.org|8480-6. This is a Laboratory type Observation.|
|2||Heart rate||Heart rate||Observation||Supported by searching for http://loinc.org|40443-4|
. This is a Laboratory type Observation.
|3||Base deficit||In physiology, base excess and base deficit refer to an excess or deficit, respectively, in the amount of base present in the blood. The value is usually reported as a concentration in units of mEq/L, with positive numbers indicating an excess of base and negative a deficit.||Observation||Needs to be "built out" in the FHIR Sandbox. There are 19 codes for base excess at https://loinc.org/?s=base+excess|
|4||MOI||Mode of Injury. This observation categorizes the wound as either "gunshot", "stabbing" or "blunt force".||Observation?||There is no code for this. It would have to be coded using an existing code or an extension. This would be a Patient Care type Observation.|
The output is the MTP Probability. Arguably, the logical place to store this would be in an Observation resource. However, Ignite doesn't support creating Observations, only retrieval. The DocumentReference resource could be used in lieu. This is a catch-all resource that can store unstructured text.
Smart-on-FHIR MTP implementation diagram
Research app using MTP data from FHIR source
Limits of Ignite APIs
The Cerner implementation is a step forward, but has limited coverage of approximately 8% of the FHIR API when all of the resources and their supported operations are considered. Cerner’s product forecast for expansion of their FHIR support presently only lists a few more resources; it is conservative at best.Emory could close this gap to spark innovation by implementing its own Emory Patient Data API Gateway that exposed both Cerner-supported FHIR resources and operations (presently 22 resources) and all remaining 71 FHIR resources for read-only operations.
Appendix A - EeMR MTP Implementation
The MTP is initiated by creating an MTP Order. The mode of injury is manually entered into the form, where as the most recent observations of BP, heart rate and base deficit are retrieved from the database. Workflow processing is used to process the order and return the MTP probability calculation in the order results.
Appendix B - Sample Smart-on-FHIR Webapp
Contact Tom Cervenka for credentials to the Cerner Developer Portal
Then the developer portal assigns test credentials to you to use to authenticate to the provider portal...
...when you click the launch button.
Login to the sandbox system with the test credentials.
Finally, the sample example launches with the Patient context/resource and Observations.
This is a link to a (draft) screen demo that shows the MTP research app that is currently in use in a research study with plans to expand the study to other institutions. It then demos a Smart on FHIR sample app based on the MTP and created as a reference app for SoF development.