Description
Project abstract (1000 Characters): The Compass application is part of a larger standards initiative called mCODE (minimum common oncology data elements). The American Society of Clinical Oncologists (ASCO) and The MITRE Corporation lead the consortium of healthcare organizations that collaborate on development and deployment. mCODE provides an open FHIR-based (Fast Healthcare Interoperability Resources) standard that enables the storage and access to cancer data.
Compass is used at the point-of-care to provide decision support and shared decision-making with patients. When launched within the electronic health record (EHR), Compass accesses the patient's demographics and cancer characteristics via mCODE elements. The application then matches these data elements to a database of deidentified cancer cases provided by CancerLinQTM, an oncology data and services company. The "patients like this" function enables the comparison in the matched cohort between treatment options based on outcomes, such as overall survival.
Project rationale, impact and innovation (3500 Characters): Every year there are over 1.6 million new cases of cancer, but only 3% of these join the clinical trials which generate structured data for research. The vast majority of cancer cases in the US produce low-quality real-world data, much of which is stored in free text form in pathology, imaging, and clinical notes. mCODE (minimal common oncology data elements) addresses this by providing FHIR-based (Fast Healthcare Interoperability Resources) resources for a core set of cancer data elements to store, evaluate, and share.
mCODE is a core set of structured data elements for oncology. The goal is to improve the overall quality and consistency of cancer data. This will accelerate advances in patient care, inform research, and improve surveillance by enabling analyses of data across the lifetime of a single cancer patient and across patient cohorts. The repeatable approach applied here is to establish mCODE as the standard, FHIR-based use-case-driven, data element set that will be used to populate all electronic health records (EHRs) for patients with cancer. mCODE is currently being piloted at cancer centers around the country to assess its value in two use cases – one focused on improving research and one on improving care.
mCODE has been developed by a group of oncologists under the auspices of the ASCO and is supported by the Oncology Center of Excellence of the US Food and Drug Administration (FDA), the Alliance for Clinical Trials in Oncology Foundation (Alliance Foundation), and The MITRE Corporation. mCODE is currently undergoing the HL7 ballot process for its FHIR implementation guide (release 1). The number of collaborators and potential pilot sites among health systems, EHR vendors, pharmaceutical companies, and payers continues to grow.
The Compass application is the use case for mCODE at the point of care in oncology. The impetus for Compass is the need for oncology decision support using real world data (as opposed to that from highly constrained clinical trials). The rapidly growing number of new genomic and therapeutic advances present challenges to oncologists active in clinical care. The first step toward progress is the establishment of a semantically interoperable cancer data model to standardize cancer characteristics (addressed above in mCODE description). Compass organizes the cancer findings in the user interface (UI) for rapid comprehension. This prevents the disruption of searching through various pathology reports and clinical notes to understand the nature of the case.
ASCO and MITRE initiated the Compass project to show the value of mCODE by using real world data to produce a matched cohort for the patient being seen. Compass gives the oncologist and the patient the means to compare treatment options by data-driven outcomes from other patient cases. Treatment options encompass the full range of types of therapy: surgery, radiation, chemotherapy, and targeted drugs.
The treatment option/outcome user interface contains visualizations that include patient friendly graphics to display difference between the therapies being compared. This shared decision aspect allows both oncologist and patient to view the application, making Compass a helpful center of discussion.
Project design and implementation (7000 characters): Product Design
Compass is built on top of FluxNotes, a prototype also developed by MITRE that is designed to demonstrate the concept of a clinician entering a clinical note that includes structured data embedded within it. FluxNotes provides a solid platform for a SMART-on-FHIR since it uses mCODE as its core data model and provides clinician-reviewed health record visualization capabilities.
Compass was developed with three core principles in mind. First, Compass should enable better care for patients that is more in line with their personal values. Second, Compass should increase patient engagement by reducing the barrier to accessing relevant information. Finally, Compass should improve over time as more patient data is captured in structured mCODE format and available for research and comparison. To support the first and second goals, Compass clearly highlights the differences between treatment options for a given patient cohort and makes that information visible in multiple different visual formats. While Compass does not yet support direct data capture and writeback, as the use of mCODE grows and other tools supporting mCODE data entry emerge, the value of Compass will increase.
One of the biggest challenges in developing Compass is that the user interface must provide relevant information without overwhelming the users. Providers will likely familiarize themselves with the UI within a couple uses, but every day new patients will be seeing it for the first time. To enable those conversations between patients and providers and allow them to make data-driven decisions on care, the UI must be both intuitive and comprehensive. Several rounds of user interface design iteration and refinement were performed with clinicians providing valuable feedback on how to make the application more accessible to patients.
Because the world does not yet run on mCODE, currently Compass must be configured on a site-by-site basis to map that sites standards to mCODE. This mapping may be as simple as applying profiles, or as complex as converting resources from one type to another. Compass is designed to be eminently configurable to minimize the level of development effort needed to install Compass at a given site and tailor it to that site’s specific needs.
In addition to the software itself, the second biggest challenge in developing Compass is the data. While small health systems would likely need to call out to an external data service to enable comparisons, large health systems may have enough data within their own data warehouses to provide sufficient data for meaningful comparisons. While the first use case of Compass calls out to a webservice at CancerLinQTM, Compass can be configured to use alternative data sources or even a predefined static data set.
Technical details: Compass and FluxNotes are built on a JavaScript platform, using modern technologies such as Node.js, ReactJS, Redux, d3, and Material UI.
Implementation
Patient and Oncologist user groups were established early on during Compass' development. Using agile methodology, incremental UI prototypes were developed to help shape the UI and fulfill requirements.
The FHIR-based mCODE profiles during this time were also refined to produce the current version: HL7 Ballot/mCODE FHIR Implementation Guide version 1. The Compass implementation is based on this release.
Intermountain Healthcare (IMH) is the first pilot site for Compass and uses Cerner as its EHR. Implementation consists of two workstreams for Cerner Integration. The first is Cerner mapping in which mCODE elements were mapped to Cerner data elements. The second workstream is FHIR conformance, which is a detailed analysis of each mCODE FHIR profile by MITRE and IMH interoperability/FHIR experts. One lesson learned is that a subject matter expert on Cerner EHR infrastructure and FHIR mapping should be involved early at the integration planning stage to promote an efficient workflow process and avoid delays.
The level of effort for training oncologists to use Compass is much lower than that of a full EHR implementation. The main workflow begins with launching the SMART-on-FHIR application from the Cerner FHIR RESTful API during a patient encounter. Compass is streamlined in that it requires few clicks to get to a shared decision-making section comparing treatment options by cancer outcomes.
Oncologists learn how to use the criteria options for producing a matched cohort for the "patient like this" function. The oncologists also learn how to change the treatment options being compared. For the IMH implementation, a MITRE subject matter expert was onsite for the clinical champion training. Following this, other IMH oncologists will be trained on the application by their clinical champion colleagues, supplemented by training content on the mCODE website (https://mcodeinitiative.org/), and support from The MITRE mCODE staff as needed.
This first implementation contributes valuable feedback for the Compass development roadmap. The application use in the production environment provides a rich source of enhancement requests to optimize its value in cancer care. The enhancement requests are characterized and prioritized for the product roadmap.
Project evaluation and sustainability (3500 characters): The Compass project evaluation consists of three main areas:
--Application integration into a Cerner environment.
--Application performance in several parameters (e.g. load time)
--Oncologist use and feedback
Application Integration: mCODE mapping to the Cerner environment involves both the availability of data elements in the source system, and FHIR conformance/function. Since this was the first Compass installation into a Cerner environment, there was the need to explore both aspects in detail. FHIR testing using Argonaut profiles was done first to confirm proper functioning of the Cerner FHIR API. Then mapping proceeded with adjustments made to handle aspects of FHIR-based terminologies (mainly LOINC and SNOMED). Overall evaluation showed that Cerner subject matter experts in FHIR-based integration will accelerate the process. The integration effort gave opportunity to refine certain mCODE elements to better match clinical or genomic concepts. These refinements to mCODE are tracked to be considered as modifications for a future version of mCODE.
Application Performance: the follow metrics are measured:
----Load time
---Time to add treatments to the comparison/outcomes section
---Time of data refresh of the "patients like us" function when a user changes the match criteria
---Time to add treatments to the comparison/outcomes section
---Total time of usage by a clinician
---Number of application launches per day by clinician
Oncologist Feedback
During the development process an oncologist user group reviewed the Compass user interface and iterative changes made to optimize usability and clinical impact. After implementation, oncologist feedback is collected by IMH and MITRE staff. Software defects are sometimes discovered during initial use and will be managed according to the existing process. Suggested user interface changes and workflow changes by oncologists are tracked. Some of these can be addressed during the implementation process. Others will be managed according to severity by MITRE's software development process.
Sustainability
The underlying de-identified database of cancer cases provided by CancerLinQTM will have periodic updates to maintain relevancy to current cancer treatments and outcomes. We anticipate that Compass use will grow as more cancers are added (currently covers breast cancer only). Two additional upcoming pilots will bring further maturity to the application. This will lead to an open source general release to the US healthcare ecosystem.
Twitter project summary (140 characters): Compass is the first SMART on FHIR application demonstrating the use of mCODE for oncology decsion support. See https://mcodeinitiative.org/
How is FHIR used in the App being demonstrated (500 characters)? : Compass is a user-facing application designed to enable better conversations between cancer patients and providers. Compass loads one patient’s full record via SMART-on-FHIR and displays targeted information relevant to that patient and their cancer. Selected criteria from that patient’s record can then be used to query a large de-identified data source to find similar patients, and display those matched patients’ aggregate outcomes, in terms of survival and side effects.
1. What FHIR release does your application use? (500 characters)?: Compass currently supports FHIR DSTU2, selected specifically to target available FHIR APIs at select pilot sites. However, the toolchain used to develop the internal data model classes supports FHIR DSTU2, STU3, and R4, so upgrading to support newer versions of FHIR does not require major effort. FHIR resources that drive the Compass UI include Patient, Condition, Observation, MedicationOrder (DSTU2), MedicationStatement, Procedure, Practitioner, and Encounter.
What is the data source for the FHIR resources and how are the FHIR resources accessed? (500 characters): FHIR resources are accessed via SMART-on-FHIR API. Compass supports arbitrary “mapping” and conversion of FHIR resources, driven by configuration, so while data conforming to the mCODE Implementation Guide is preferred, it is not required.
Any other information about the project we should know about (1500 characters)?: ©2019 The MITRE Corporation. ALL RIGHTS RESERVED
Approved for Public Release; Distribution Unlimited. Public Release Case Number 19-2577
Authors:
James O'Connor (Presenter)
The MITRE Corporation