1. Interoperability Roadmap Comments on Common Clinical Data Set Content Standards Workgroup Andy Wiesenthal, chair Rich Elmore, co-chair March 23, 2015

2. Interoperability Roadmap Section J1 Consistent Data Formats and Semantics 2 Questions for Workgroup Discussion Are the actions the right actions to improve interoperability nationwide in the near term while working toward a learning health system in the long term? What, if any, gaps need to be addressed? Is the timing of specific actions appropriate? Are the right actors/stakeholders associated with critical actions? Category 2015-2017 Send, receive, find and use a common clinical data set 2018-2020 Expand interoperable health IT and users 2021-2024 Achieve nationwide LHS J1. Common, list of interoperability standards 1.ONC will annually publish a list of the best available standards and implementation specifications 2.Implementers and decision-makers should use ONC’s list of the best available standards to enable specific use cases 3.Implementers will update their systems to align with the list of the best available standards C-CDA 2.0 Associated vocabulary standards Associated code sets supporting common clinical data set 4. ONC will annually publish an updated list of the best available standards and implementation specifications. 5. ONC will annually publish an updated list of the best available standards and implementation specifications.

3. 1. Are the actions the right actions to improve interoperability nationwide in the near term while working toward a learning health system in the long term? 2. What, if any, gaps need to be addressed? 3. Is the timing of specific actions appropriate? 4. Are the right actors/stakeholders associated with critical actions? 3 2015-2017 Send, receive, find and use a common clinical data set Comments 3.Implementers will update their systems to align with the list of the best available standards C-CDA 2.0 Associated vocabulary standards Associated code sets supporting common clinical data set It is assumed that the list of best available represents a reference to the 2015 Interoperability Standards Advisory – Best Available Standards and Implementation Specifications Proposed time frame creates risk because of the time required to execute changes to conform / comply o Burden on vendors here, and the time lags for both vendors and then implementers are important Anything that can be done to expedite this should be specified in the roadmap As vendors move toward FHIR there could be a “double standard” or multiple standards that vendors will need to meet As standards transition a clear transition paths should be defined as some vendors / implementers will be at various levels of alignment with new and existing standards Accepted vocabulary standards are not necessarily updated at the frequency required for clinical practice thus implementers will need guidance on approaches to use required vocabulary standards while addressing the greater need to manage the issues presenting in clinical practice While C-CDA 2.0 is helpful, the need to share information for other than direct clinical care (e.g., with public health registries) remains tied to HL7 V2 messages o Requirements to align with multiple standards for the same information is duplicative. If EHR vendors are expected to transition and implement new standards, all users of the same data (including public health) should be expected to align with the same standards o Suggested vocabularies and code sets do not align well with widely used research and clinical standards, including those defined by FDA and other US and international agencies and SDOs such as CDISC. Specifically the need for MedDRA terminology to document and report clinical trial Adverse events is not addressed

4. 1. Are the actions the right actions to improve interoperability nationwide in the near term while working toward a learning health system in the long term? 2. What, if any, gaps need to be addressed? 3. Is the timing of specific actions appropriate? 4. Are the right actors/stakeholders associated with critical actions? 4 General Comments Page 77 (McDonald) The section on content/structure standards is not ideal. The content is often considered to be the codes, numbers, and text that lands in the fields. Better to call it message/structure standards. Then describe the fields or data elements to be packaged together for a particular purpose and add information about grouping and repeats, etc. Page 79 (Rallins) Background and Current State Additionally, other vocabulary and terminology standards are embedded within the implementation guides documenting the use of data formats such as HL7 v2 messages and C- CDA. In many cases, “value sets”, such as those published in the Value Set Authority Center (VSAC), are established to identify subsets of the standard vocabularies to be used for a specific purpose. For example, for the purposes of quality measurement, all the relevant codes from SNOMED-CT and/or LOINC that can be used to identify diabetic patients for quality measurements in EHRs are put in a specific list. (McDonald) Page 79, 2 nd to last paragraph: LOINC is not restricted to laboratory tests. It is specified for vital signs, and many other things via the cCDA so would change that sentence to: … (LOINC) for laboratory tests, vital signs and other observations.

5. 1. Are the actions the right actions to improve interoperability nationwide in the near term while working toward a learning health system in the long term? 2. What, if any, gaps need to be addressed? 3. Is the timing of specific actions appropriate? 4. Are the right actors/stakeholders associated with critical actions? 5 General Comments Page 78 (Elmore) Figure 8: The Categories of Standards describes two applicable standards categories: Vocabulary & Code Sets (Semantics) and Format, Content & Structure (Syntax). This appears to conflate concepts which should be logically separated. For example, content should be abstracted from how it is formatted and structured. Additionally, semantics/content/context (i.e., the “meaning”) is similarly distinct from vocabularies, dictionaries & code sets (i.e., the items used to create “meaning”). I’d suggest the following is a better representation of the applicable categories of standards (from highest level down): · Semantics, Context and Content · Vocabularies, Dictionaries and Code Sets · Format and Structure Taking this revised approach would emphasize the building blocks necessary for clinicians/users to understand the semantics, context and content. Also, the final roadmap should include more representative examples, e.g., - “RxNorm for Ibuprofen is 5640” is not semantics. - DNS + LDAP and X.509 relate more to security than to services.

6. 1. Are the actions the right actions to improve interoperability nationwide in the near term while working toward a learning health system in the long term? 2. What, if any, gaps need to be addressed? 3. Is the timing of specific actions appropriate? 4. Are the right actors/stakeholders associated with critical actions? 6 2015-2017 Send, receive, find and use a common clinical data set General Comments Page 80 (continued) (McDonald) Proposed Common Clinical Data Set Page 80 would add DICOM (and maybe X12) to the list of SDOs mentioned in that paragraph o The list of common clinical data sets on page 80 (copied below) is horribly incomplete. The only diagnostic studies it includes are Labs (Rallins) Regarding two data elements listed in the Common Clinical Data Set on page 80: Smoking status: as written this data does not include all uses of tobacco including smokeless tobacco. Consider revising to “Tobacco Use” to align with electronic clinical quality measures that also include smokeless tobacco. Care team members: agree this is an important data element to capture. However, are there existing vocabularies that can capture this information? Perhaps SNOMED CT? In the context of clinical quality measurement, the Quality Data Model standard may need revision to accommodate this element.

7. 1. Are the actions the right actions to improve interoperability nationwide in the near term while working toward a learning health system in the long term? 2. What, if any, gaps need to be addressed? 3. Is the timing of specific actions appropriate? 4. Are the right actors/stakeholders associated with critical actions? 7 2015-2017 Send, receive, find and use a common clinical data set General Comments Page 81 (Elmore) “…In order to support exchange across the continuum of care and in support of a learning health system, interoperable data formats must be adopted for care settings…” Backward compatibility is a must, as every vendor is on their own timetable. Asynchronous upgrades must also be allowable, as well, given the need for pragmatism. Footnote 59: I agree with expanding the set of vital signs. (Hull) In order to support exchange across the continuum of care and in support of a learning health system, interoperable data formats must be adopted for care settings such as behavioral health, long-term and post-acute care (LTPAC) and community service providers (e.g., schools, prisons and social services) and also by the research community. Need to add consumer centric care settings that may be tethered or untethered to brick and mortar settings. Potential Gaps: Page 81 “At a minimum, we must as a nation agree to a standardized common clinical data set that is consistently and reliably shared during transitions of care (and with individuals and their caregivers) to achieve our near-term goal of establishing a foundation of interoperability that can be expanded over time. Mobile health solutions are frequently addressing transitions of care, and need to be included Transitions of care standards need to address equivalence to a simpler lay language that individuals and caregiver’s can understand. (perhaps this would fit better in the 2018-2020 time period)

8. 1. Are the actions the right actions to improve interoperability nationwide in the near term while working toward a learning health system in the long term? 2. What, if any, gaps need to be addressed? 3. Is the timing of specific actions appropriate? 4. Are the right actors/stakeholders associated with critical actions? 8 2015-2017 Send, receive, find and use a common clinical data set General Comments Page 81 (continued) (McDonald)“Content standards” in 5 th line of 3 rd paragraph: There are two messages coming across. Because the word “content” is often taken to mean the codes, text, and numbers within a structure, I think the two meanings should be more clearly distinguished. In this sentence you are really talking about message/structure standards. Not the discrete value payloads they carry. Then you talk about two different problems. One is that a lot of the CDA content is sent as free text, and the other is the lack of requirements for standard codes in some parts of cCDA. (There is also the problem that even where they are required, producers do not always comply.) The problem is compounded by the fact that CCDA only accommodates the (HL7 v2-like) general observation structure for lab tests not for any other panel of test observation. Some diagnostic study services do send discrete results as HL7 v2 messages, but they don’t necessarily use standard codes --two different problems. Worse, the reports for some studies may start as machine- produced as discrete results (e.g. Spirometry, 12 Lead EKG) but get converted to a simple ASCII narrative when shipped out as CCDA or as HL7 v2. We should push mightily to have such reports stay in their discrete state when they get shipped to a practice. There is also the case exemplified by the provider who dictates his note. He will never succeed in converting that to discrete coded content. And we should not push him/her to do so.

9. 1. Are the actions the right actions to improve interoperability nationwide in the near term while working toward a learning health system in the long term? 2. What, if any, gaps need to be addressed? 3. Is the timing of specific actions appropriate? 4. Are the right actors/stakeholders associated with critical actions? 9 2015-2017 Send, receive, find and use a common clinical data set General Comments Page 82 (Elmore) “To do this, over the next six to ten years, the industry…” The report suggests that vendors will have to implement within the next three years with standards that are mature and robust. However, it is very clear that we need better alignment between Federal and State requirements. In the current state, the wide variability between various regulatory organizations (even those down to the city / county level) leads to significant additional work and complexity (with immunization registries a good example). (Eisenberg) The current landscape has enabled a growing industry and expense to purchase services that maps data captured during clinical care to standard terminologies to enable business operations and interoperability. While a common data set is helpful to enable software systems to use the same language and can increase usability and interoperable acceptance of structured data, a common data set is not a complete solution. Much of the rich clinical patient data exists in narrative form that provides context and meaning to the structured data. It is not feasible or desirable to require structured data entry for all elements. Also, a common data set is helpful as a “floor” for what products should address but documentation of patient care but not the “ceiling.” (Rallins) Comment: Agree with the shift from static code lists that define a concept within a single application or organization to systematic was of representing meaning. This shift would be helpful o quality measurement efforts. Recommend the use of structurally defined value sets (intensional) rather than enumeration or “cherry picking” to identify individual members of values sets. This recommendation is based on lessons learned as they relate to requirements for effective change management. When new versions of vocabularies, terminologies and code sets are released, the changes can present a considerable impact on maintenance processes. Intensional value sets with definitions developed from structured rules and semantics will facilitate change management and sharing information across applications and organizations.

10. 1. Are the actions the right actions to improve interoperability nationwide in the near term while working toward a learning health system in the long term? 2. What, if any, gaps need to be addressed? 3. Is the timing of specific actions appropriate? 4. Are the right actors/stakeholders associated with critical actions? 10 2015-2017 Send, receive, find and use a common clinical data set General Comments Page 81 (continued) (McDonald) Major kinds of clinical (and environmental) data can be sent in HL7’s observation structures. Indeed, HL7 has ways to distinguish population data, from patient data, from environmental data. And LOINC carries all kinds of environmental data. I hate to say it, but see little value in NIEMS for any healthcare data. It took a different path and is just different. (McDonald) I like the idea of getting away from document, but you can’t get away for groups of data elements that may repeat or not and have to go together. General Comment (Dinhofer) From my perspective as someone who is on the clinical side, I would like to focus on the perception of clinicians rather than the technical aspects. Clinicians are not programmers, nor do they understand the inner workings of software. They do not understand the complex issues of data mapping or lines of code that need to be written in order to make a system functional In that regard, even before the issues of semantics and syntax, Clinicians had problems with nomenclature from one region to another and one specialty to another. What will take time is developing a national nomenclature that crosses regions and specialties. The issue of nomenclature should involve all the major clinical areas and regions in the US. The timing as such will be a variable.

11. Slide 3 bullet comments: o Suggested vocabularies and code sets do not align well with widely used research and clinical standards, including those defined by FDA and other US and international agencies and SDOs such as CDISC. Specifically the need for MedDRA terminology to document and report clinical trial Adverse events is not addressed Humpreys: As discussed previously, all EHRs should NOT have to deal with MedDRA. The fact that some EHR products may be used to record clinical trial data should not lead to an additional and duplicative vocabulary requirement for all EHR products. This is a place where mapping is a logical approach. NLM and NCI are discussing how to do this for a subset. Rocca: FDA follows an open, consensus-based process to develop and maintain data standards in collaboration with accredited SDOs. Adverse event reporting is not captured as a use case in this roadmap and not all centers at FDA leverage MedDRA for clinical trial Adverse events. Office of the National Coordinator for Health Information Technology 11

12. Slide 4 comment: Office of the National Coordinator for Health Information Technology 12 General Comments Page 77 (McDonald) The section on content/structure standards is not ideal. The content is often considered to be the codes, numbers, and text that lands in the fields. Better to call it message/structure standards. Then describe the fields or data elements to be packaged together for a particular purpose and add information about grouping and repeats, etc. Page 79 (Rallins) Background and Current State Additionally, other vocabulary and terminology standards are embedded within the implementation guides documenting the use of data formats such as HL7 v2 messages and C-CDA. In many cases, “value sets”, such as those published in the Value Set Authority Center (VSAC), are established to identify subsets of the standard vocabularies to be used for a specific purpose. For example, for the purposes of quality measurement, all the relevant codes from SNOMED-CT and/or LOINC that can be used to identify diabetic patients for quality measurements in EHRs are put in a specific list. (McDonald) Page 79, 2 nd to last paragraph: LOINC is not restricted to laboratory tests. It is specified for vital signs, and many other things via the cCDA so would change that sentence to: … (LOINC) for laboratory tests, vital signs and other observations. Humphreys: On slide 4, I am not sure I am following the argument correctly, but if it is getting at the fact that structure (e.g., message format, common data element format, etc) is separate from semantics then I disagree.