ResultsMyWay: combining Fast Healthcare Interoperability Resources (FHIR), Clinical Quality Language (CQL), and informational resources to create a newborn screening application

Abstract

Newborn screening (NBS) can be life-changing for the families of infants who test positive for a rare condition. While resources exist to support these families, there can be delays in sharing these resources due to communication lag between the laboratory, result interpreting clinician, family of the newborn, and additional care providers. This delay can also be exacerbated when additional health history is required from the mother and infant. ResultsMyWay is a proof-of-concept application that uses Clinical Quality Language (CQL) to automate the search for this additional health history. It also translates the NBS results into Fast Healthcare Interoperability Resources (FHIR), increasing both the ease of exchange and the future utility of these data points. After the families are given the NBS results, ResultsMyWay then acts as a hub for several types of informational resources about the recently diagnosed condition.

Introduction

Newborn screening (NBS) is a public health program that screens infants shortly after birth for diseases and disorders that can severely harm the infant if early detection and care management are not provided. Of the approximately four million infants born per year in the United States, about 12,900 of them are diagnosed with a disorder via the NBS process¹.

In the event of a positive rare condition result, the guardians of the infant are contacted to inform them of the result and to provide the assurance that qualified professionals will be in contact to help establish an appropriate care plan for the child². Difficulties in the delivery of information to the care providers and the guardians can lead to problems such as loss to follow-up³, distress⁴, confusion and strong emotional reactions². Parental knowledge of the condition varies⁴ and the knowledge gap is another pain point that must be addressed. The internet provides an opportunity for parental education about NBS screening with health department web pages⁵, and there are resources such as Baby's First Test⁶ that aim to collate information. But a Google search of 'nbs cystic fibrosis test result' brings back 142,000 pages which is obviously overwhelming.

Although NBS is a routine clinical service, there are still improvements that can be made in the typical NBS workflow. For example, there is the need, in many cases, to manually collect additional health information about the mother and the infant for cases where the interpretation of the results could change depending on the presence or absence of these additional factors. This collection is a burden to the interpreting clinician who now must wait for what can be a slow health information exchange process, to the guardian who now knows that something isn't right but must continue to wait until verified results are returned, and for the care provider of the mother who must search through that mother's health records as well as the infant's health history (which may include several previous clinical care providers who need to be contacted). When the results are finalized, they are typically compiled into a static portable document format (PDF) report. Representing clinical data within a PDF locks the data out of the health system and renders it only human-readable. Storing the data as computable data points however, would increase the likelihood of the data being used to enhance the care of the patient in future encounters⁷.

Background

FHIR

Fast Healthcare Interoperability Resources (FHIR) is a rapidly-growing Health Level Seven International (HL7) standard that represents specific data artifacts found in typical clinical scenarios (Patient, Condition, Encounter, Observation, etc.) as standardized "resources". These resources are linked together via identifiers to provide a standardized and robust representation of clinical events⁸. The resources themselves can be thought of as both the content model (how the thing is represented) and the artifact (the thing that is exchanged). They are exchanged via representational state transfer and can be tailored to specific needs using a strict system of published profiles and extensions. There are currently 145 FHIR resources, each at one of seven different levels of maturity ranging from 0 (draft), 1-5 (multiple aspects of "completion" considered), to N (normative)⁹. Resources and profiles are developed, balloted, and published by ~40 HL7 working groups¹⁰. These working groups are chaired by and made up of top health professionals from a variety of medical backgrounds worldwide.

CQL

Clinical Quality Language (CQL) is an HL7 specification for defining standardized and shareable clinical logic. CQL authoring typically involves the creation of several distinct expressions. These expressions can query clinical artifacts and employ a wide variety of operators and filters before returning a result. A collection of CQL expressions is called a Library. CQL was designed for use within both the clinical decision support and clinical quality measurement domains¹¹. It has been predominantly developed and used by stakeholders in the HL7 Da Vinci Project, whose goal is to help payers and providers to positively impact clinical, quality, cost, and care management outcomes¹². They have piloted solutions that combine CQL with the HL7 CDS Hooks¹³ specification and HL7 FHIR Questionnaire resource in reducing burden within the specific contexts of Coverage Requirements Discovery¹⁴, Documentation Templates and Rules¹⁵, and Prior-Authorization Support¹⁶. These implementations use CQL to search the patient record and determine if that individual has the required pre-existing conditions to qualify for coverage of some type. NBS has a similar need in that results can occasionally be interpreted differently if certain pre-existing conditions apply to the infant or mother. ResultsMyWay pioneers the use of CQL in aiding clinical laboratory test interpretation.

Methods

While ResultsMyWay is a proof-of-concept application and not yet implemented live in the Utah NBS process, it was designed with interoperability and future implementation in mind. Therefore, several real-world scenarios were considered in the architecture design.

Account Creation

The first step was to ensure that a user can be uniquely matched to an infant in all foreseeable cases¹⁷ (e.g., adoptive parents or other forms of non-maternal guardianship). It is anticipated that a ResultsMyWay activation code will be given to the guardian at the time of specimen collection. This will allow that guardian to match to the infant (who will have that code added as an identifier to their Patient resource) and create an account to steward the NBS process for that infant. Additional data collected during account activation include name, relationship to the newborn, email address, phone number, an option for designating a call or text when results are ready to be viewed in the application, and the option to add a secondary user on behalf of the infant. These data are used to create a FHIR RelatedPerson resource which ResultsMyWay uses to manage users.

Automated Retrieval of Factors that Impact NBS Interpretation

The Utah NBS data model includes three LOINC panels that can be drawn upon in the interpretation of certain NBS results: Feeding type (67704-7), Infant factors that affect newborn screening interpretation (57713-0), and Maternal factors that affect newborn screening interpretation (67707-0). The health factors covered by these panels range include procedures (blood transfusion, thoracic surgery involving thymectomy, etc.), conditions (liver disease, biliary atresia, septicemia, etc.), and medications (parenteral steroids, systemic antibiotics, dopamine, etc.). As part of this application, the answer lists for each of these panels were converted to CQL expressions. These expressions rely on custom FHIR ValueSets that contain synonym codes from several terminology systems (e.g., LOINC, SnomedCT, ICD-10, RxNorm) for each Condition, Observation, Procedure, or MedicationRequest required to retrieve these additional factors. The first CQL expression in Figure 1 shows a simple retrieval of all MedicationRequest resources with an RxNorm code contained in the custom ValueSet called "Dopamine Set". This expression will return the ID of any existing Dopamine medication orders (stored as FHIR MedicationRequest resources) within the infant's health record. Single direct reference codes can also be declared and used as shown by the second expression in Figure 1 which retrieves all Condition resources for the infant and filters for any whose "severity" is acute. These expressions can also return structured objects as shown in the third expression in Figure 1 where the "performed" date of the Procedure resource (representing a blood transfusion) is returned along with the ID of the resource. CQL Libraries can be stored, exchanged, and used via the FHIR Library resource¹⁸. A FHIR Library resource contains a base-64 encoded string representing the CQL Library, references to terminology requirements, and various other fields. The CQL Library can then be executed within the data server by sending a request including the $evaluate operation and several context parameters. The server then returns the results of the CQL Library evaluation as shown in Figure 2. In following the example of the implementation guides published by the Da Vinci project, ResultsMyWay uses FHIR Questionnaire resources as companions to the Library resources. These questionnaires display the CQL retrieval results and capture additional needed data that were not found in the patient's medical record. These Questionnaire resources are constrained by profiles that have been created by the HL7 Structured Data Capture workgroup¹⁹ to facilitate their use alongside CQL Libraries with each question having an optional extension field containing the name of the matching CQL expression. The application then extracts the CQL response to auto-fill that corresponding question for the user.

Figure 1.

Selections from a CQL Library used by ResultsMyWay.

Figure 2.

A visual representation of how a CQL Library can be stored and executed on a FHIR server.

NBS Results Message Mapping

In the Utah NBS system, results are formatted in the fields of an HL7 Version 2.4 ORU-R01 message before being rendered as PDF files. One aim of this work was to provide the Utah NBS Program with mappings from HL7 v2.4 to FHIR v4.0.0. This effort to map from HL7 Version 2.4 to FHIR has a dedicated HL7 working group whose findings informed the mappings used for ResultsMyWay²⁰. A selection from the final mappings used for ResultsMyWay is shown in Table 1.

Table 1. A selection from the final mappings used for ResultsMyWay.

V2 Field	V2 Field Description	Utah NBS Result Segment	FHIR Resource.Field
ORC-3	Filler Order Number	F1590009202099	DiagnosticReport.identifier
ORC-5	Order Status	CM	ServiceRequest.status
ORC-12	Ordering Provider	LW3166^DUFFY^ TIMOTHY^^^^^^HC^^^^^	Practitioner.name
OBR-4	Universal Service Identifier	54090-6^TSH Panel^LN	DiagnosticReport.code
OBR-14	Specimen Received Date/Time	202006040028	Specimen.receivedTime
OBR-15	Specimen Source	DBS&Dried blood spot card [79566-6]^^^&	Specimen.type
OBR-22	Results Rpt/Status Chng - Date/Time	202006041722	Specimen.processing[x].time DateTime
OBR-25	Result Status	F	DiagnosticReport.conclusion
OBR-27	Quantity/Timing	^^^^^U	DiagnosticReport.conclusion
OBR-32	Principal Result Interpreter	39&Wallis& Heidi	DiagnosticReport.resultsInterpreter
NTE-2	Source of Comment	L	DiagnosticReport.extension
NTE-3	Comment	This is a TSH Panel Rpt comment.	DiagnosticReport.extension
NTE-4	Comment Type	RE	DiagnosticReport.extension
OBX-2	Value Type	NM	(determines valueCoding OR valueQuantity)
OBX-3	Observation Identifier	29575-8^TSH^LN	Observation.code
OBX-5	Observation Value	62	Observation.valueCoding OR valueQuantity.value
OBX-6	Units	uIU/mL	Observation.valueQuantity.units
OBX-7	Reference Range	0-40	Observation.referenceRange
OBX-8	Interpretation Codes	H	Observation.interpretation
OBX-11	Observation Result Status	F	Observation.note
OBX-14	Date/Time of the Observation	20200601	Observation.effective

Information Resource Content Management

One of the primary goals of ResultsMyWay is to act as a hub for educational, community outreach, and clinical trial resources that relate to an infant's condition. These resources are constantly changing and it is expected that a content manager within each implementing system would be assigned to monitor current content associated with each condition as well as approving new user-suggested resources via a submission system in the application. To ease this management, a simple schema was designed to represent each condition and its associated resources. A sample from this schema is shown in Figure 3.

Figure 3.

The resource content for the Phenylketonuria (PKU) condition.

Results

A demo of ResultsMyWay has been created to demonstrate the utility of the implementation (http://hematite.genetics.utah.edu/ResultsMyWay/).

Results Page

ResultsMyWay is meant to be used both before and after an infant's results are returned. Users are able to create accounts as soon as the original specimen is taken and the official process begins. Results can take days to finalize and, in the interim, ResultsMyWay provides useful general information about NBS for users and acts as a valuable data collection tool for providers. Consider the following use case: An expecting 37-year-old mother at 33 weeks gestation with known hypothyroidism is being observed in the hospital with sudden onset nausea, vomiting, and severe headache. She is shortly diagnosed with HELLP syndrome and started on fluid resuscitation, 1 unit of packed red blood cell transfusion, and intravenous steroids. She is stabilized and prepared for immediate delivery. She gives birth to a male infant who is found to have low birth weight, significant hypoxia, and lethargic at delivery. He responds to supplemental oxygen and is transferred to the neonatal intensive care unit (NICU) where a blood test is collected. Lab results showed low thyroxine in the newborn as well as anemia. The newborn is blood typed and started on a blood transfusion while in the NICU. A sample is also collected for NBS. While this can be considered an atypical birth experience, it highlights the value of the automated retrieval of factors that could influence the interpretation of the NBS results for this infant and is shown in Figure 4.

Figure 4.

The initial view of the results page--showing the CQL results for the use case given above.

These questionnaire forms can be added to for clinical data that perhaps weren't yet recorded or that were missed by the CQL library. It is anticipated that a FHIR QuestionnaireResponse resource, which contains all of these responses and is generated when the user completes the form and clicks "Submit," would be sent to the appropriate result-interpreting clinician or lab technician to assist them in the interpretation of the infant's NBS results.

Once the results are finalized and ready to be communicated back to the patient, ResultsMyWay contains three options for viewing and exploring those results. The first is the "Summary" view which gives the results for each panel while visualizing relevant data ranges and providing links to explanatory resources for panels, terms, codes, and comment sources and types. This represents a significant amount of metadata that can now provide added understanding and perspective for concerned guardians. This first view is shown in Figure 5. The second view is a "PDF" view which displays the classic PDF report which is stored as a base64 string in each FHIR DiagnosticReport resource. This allows for a lossless transition from the current process to this updated process. This is intended to improve the change management for care providers who may have a wide spectrum of attitudes toward new clinical innovations. The third view allows the relevant FHIR resources to be reviewed and has been initially included in the application for the benefit of clinicians who may want to review the actual raw data artifacts as they appear in the health system.

Figure 5.

The summary view of the ResultsMyWay results page--showing the results of one NBS panel.

Information Resource Pages

After the results have been communicated and the users are given access to four pages which contain several different kinds of resources specific to that infant and condition.

The first is a knowledge library. This is a collection of websites, factsheets, feeds, videos, etc. that have been separated into four categories: Basic Knowledge (resources designed for all levels of background education and experience), Nutrition (resources specifically involving potential food and lifestyle adjustments), Research (a custom PubMed RSS feed), and Media (spotlights or other news media attention for the condition or others dealing with that condition). Users may suggest additional resources to be added. This knowledge library is shown in Figure 6.

Figure 6.

The knowledge library page of ResultsMyWay--showing media results for the PKU condition.

The second page uses FHIR to assemble profiles for each of the significant members of that infant's care team. With a picture (when available) and phone number, this page provides a convenient reference to the user who may be managing communication between several clinicians.

The third page provides a list of various social media or other community groups that have been organized to support those living with that certain condition and is shown in Figure 7. The fourth page shows a custom ClinicalTrials.gov RSS feed that users can scroll through if they are interested in treatment options of that kind. This would likely increase participation in these trials and be a valuable resource to the research community.

Figure 7.

The community page of ResultsMyWay-showing links to the social media pages of NPKU.

Discussion

In this demonstration, we have used CQL to automate the retrieval of health factors which could influence the interpretation of NBS results. Automation would reduce the time and effort involved in the necessary health information exchange and could lead to earlier diagnoses of potential diseases and disorders. This could be critical for cases where early care management significantly decreases the likelihood of adverse effects. While CQL is an innovative and burden-reducing technology, manual input of clinical data points by users can be a dangerous option²¹. The health literacy of the user and the reliability of data entered would be unknown and could have harmful effects. Another option would be to restrict changing or adding answers to the ResultsMyWay pre-results form to clinicians only. This would increase the burden on the clinician who retrieves the NBS blood sample because they would need to take time to review the medical history of the patient in addition to educating the guardian about the app itself.

While FHIR is an interoperable and effective way to represent NBS results, this application also requires a significant amount of FHIR data server transactions. This could become an expensive burden to the implementing health system in situations where these transactions are meant to be employed with more thrift. However, these considerations were out of scope for this proof-of-concept application. Another critical consideration is the security of protected health information. While the SMART specification²² can ensure the secure launching of the app and the security of the data for authorized users, the user onboarding process will likely need additional measures to verify identity and relation to the newborn.

Conclusion

NBS is a routine clinical process that can result in complete lifestyle readjustment for the families of infants who test positive for a rare condition. ResultsMyWay has been designed to empower the guardians of these infants by providing informational resources and to assist the clinicians involved by streamlining the interpretation and exchange of NBS results.