NLM R&D Tools for Standardization of Electronic Medical Records and Clinical Genetic Test Reports for Care and Research

National Library Of Medicine

Investigators

Linked publications & trials

Abstract

This project reflects a longstanding NLM interest in clinical terminology and message standards. It uses and tunes message and vocabulary standards that NLM has supported to facilitate interoperability, defined as communicating data from a source to a destination where it can be used for decision making and analysis. (HIMSS Dictionary of Healthcare Information Technology Terms, Acronyms and Organizations, 2nd Edition, 2010, Appendix B, p190.) NLM sits at the nexus of standards needed to make clinical data flow from sources for clinical care and research. The major code systems required by Meaningful Use regulations, LOINC, UCUM, SNOMED and RxNorm have been developed and/or supported by NLM. NLM continues to contribute to their development and adoption. Most of my career has been committed to developing, studying and improving electronic medical records and the standards needed to deliver clinical data to them. I was the instigator of the first two coding systems in the preceding list and one of the founders of HL7 and authored HL7s orders and observations chapter during HL7s early years. My engagement in HL7 has continued and I recently authored standards for clinical genetics, cytogenetics and LIVID which defines mappings between the laboratory instrument vender codes LOINC which is the standard code for laboratory tests within medical record standards. I have also been deeply involved with FHIR, the new HL7 message standard, and am an author on two recently approved FHIR standards, one for reporting clinical genetic studies and the other for defining Structured data capture forms (SDC). My involvement and knowledge of FHIR, gave me the opportunity to inform NIHs leadership about, it and its, surging support from industry and federal agencies (Apple, Google, Microsoft, Amazon, the health insurance industry, CMS, ONC and more.) This led to an NIH guidance document, which I had the privilege of announcing at the White House this summer. Within Lister Hill, I lead a group that has developed apps and tools that implement and demonstrate many FHIRs capabilities. These tools which can be used in EHRs, PHRs, and research systems can be demonstrated from a web page (https://lhcforms.nlm.nih.gov/fhir). They are open source, Section 508-compliant (i.e., accessible to screen readers), and freely-available from GitHub where we deposited them. Groups or institutions can customize many of the features for their specific needs. We describe them below. A) LHC-Forms: Form Rendering Widget (http://lhncbc.github.io/lforms/) LHC-Forms creates input forms for Web-based medical applications, EHRs, PHRs, and mobile health apps. It renders a formal description of a form into a fully functioning data input form and does this on the fly. Users can generate input forms from 2,000+ pre-existing LOINC panels, or they can user our form building tool to develop their own unique forms. LHC-Forms is rich in functionality. User can specify the data type, cardinality, default value, units of measure (if numeric), answer lists, relationships (in a nested hierarchy), and scoring of survey instruments. It includes validation checks to ensure quality data collection, skip logic and help messages. LHC-Forms uses the NLM-developed autocomplete package (http://lhncbc.github.io/autocomplete-lhc/). LHC-Forms has the ability to accept, store, and display forms as FHIR Questionnaire resources and to store the collected data in a FHIR server. We lead a new HL7 FHIR work group (Structured Data Capture (SDC)) to standardize questionnaires and associated tools. This specification was balloted under an ANSI approved process in September 2018, and we anticipate having a trial use version published soon, in early fall. We have a new SMART on FHIR app that show cases our e SDC Questionnaire profile, including pre population of input fields with data from a FHIR medical record at: https://lhcforms.nlm.nih.gov/sdc. With other collaborators, we also developed a JavaScript version of the FHIRPath library, https://github.com/lhncbc/fhirpath.js. which is required for implementing calculation and other special functions in the FHIR Questionnaire. Users can build and/or customize forms using our form builder tool, which can deliver and retrieve finished newly built forms form to/from a user selected FHIR server. Try it: https://lhcformbuilder.nlm.nih.gov. We created a web widget that generates a time oriented clinical flowsheet from observations in FHIR server. Each row represents a different variable and each column a different time. It can display records with any, duration or number of observations. We have tested it with record of 30 years duration with 28,000 observations. It also has a number of special features for compressing rows and columns to facilitate faster review of such large records. It has special applicability to research data collected in longitudinal studies. One can demo it on a prototype web application at https://.nlm. aih.gov. We have produced two other tools 1) A Clinical Table Search Service, which serves up contents of more than 25 important clinical coding systems. The supported coding systems include ClinVar, dbSNP and Gene from NCBI, and from other sources, e.g. COSMIC, ICD9, ICD10, RxTerms. 2) UCUM is a standard and computable unit of measure. It is the required unit standard for HL7, IEEE (instrument measurements), DICOM (radiology measurement), and ISO-11240 (units for strength of medicine). It is required by meaningful use for HL7 for Public Health laboratory reporting, and most measurements in HL7s CDA. We have developed a UCUM units validator and converter, which converts values specified in an any unit of measure into values in any other commensurate unit. The conversion function can also convert values reported in mass units to molar units and vice versa. Try it: https://ucum.nlm.nih.gov. B) Research database of de-identified FHIR resources A set of deceased, de-identified patient data for 10,000 patients was obtained from Regenstrief Institute. These were then loaded into a HAPI FHIR server to test its performance. Significant performance issues were encountered both with loading the data and searching, and various attempts were made to solve the issue including switching databases. We are working with the HAPI server developers to fix these performance bottleneck This de-identified data is now being used to test and demonstrate the LHC-Flowsheet and a new tool for pulling observations from set of medical records https://lhcforms.nlm.nih.gov/fhir/obs-viewer.html.