Anatomic Pathology Structured Reports

From IHE Wiki
Jump to: navigation, search

White Paper: Anatomic Pathology Structured Reports (APSR)

  • Proposal Editor: Christel Daniel, Thomas Schrader
  • Editor: Editor: David Booker, Victor Brodsky, Antoine Buemi, Christel Daniel, Bettina Fabiani, Dominique Henin, Mary Kennedy, Haitham Kussaibi, François Macary.
  • Domain: Anatomic Pathology

Introduction

The anatomic pathology report (APR) documents the pathologic findings in tissues removed from patients. Anatomic Pathology reports are medical consultation reports that contain data that are often the most critical for patient care and prognosis. Collectively, this data is also valuable for research, epidemiology, and education purposes. However, these reports currently lack adequate structure to support these needs accurately and efficiently. The need for structuring these reports is becoming ever greater with the current growth in the volume and complexity of their data content as well as for the need for clinical decision support. Information technology & knowledge engineering offer several solutions for optimizing APR collection and exchange in structuring, coding and standardizing their contents. Although AP reports typically constitute a relatively small portion of the total volume of data in patient medical records (particularly in comparison to clinical lab data, which may constitute 70% or more of the data in a typical patient chart), they have a major impact on patient care. AP reports are usually composed of surgical pathology, cytopathology and autopsy reports; however, they increasingly contain data from newer technologies such as flow cytometry, cytogenetics, and molecular studies. The data in these reports typically is found in the form of free text, generally precluding the ability to automate processes for sharing of the data for clinical and non-clinical uses. Pathology informatics has tended to focus on imaging, telepathology, voice recognition, data mining/warehousing, and natural language processing. A new area of current great interest is bioinformatics, particularly with regards to creating data standards for tissue microarrays. The data entry methods generally preclude the structuring of these reports and other benefits, such as decision support, rapid templating, and error avoidance.

Broad use of medical templates

In the past 10-15 years with the recognition of the need to standardize cancer reporting, pathologists have begun to use protocols* and outline* (synoptic*) formats for reporting. The other impetus for their use has been the increasing number of data elements that are of importance in pathology reporting in cancer case beyond the “diagnosis”; for example, the tumor size, cell type, grade, margin status, stage, prognostic markers, etc. Over time many pathologists have adopted this “synoptic” reporting of cancer cases, either by creating protocols internal to their labs or using vendor “synoptic” reporting tools, and these have gained considerable acceptance from clinicians receiving the reports.

Templates in anatomic pathology

  • Support pathologists to avoid significant errors in reports...There are several studies that have shown significant error rates in anatomic pathology reports. (See References) The major types of errors are :
    • Transcription -- common, well recognized error source, includes faulty proofing.
    • Voice Recognition -- this technology has an accuracy threshold. Homonyms, homophones, dialect, rapidly changing terminology, and many other challenges will continue to plague VR. Some errors will likely never be overcome (homonyms and homophones).
    • Omission -- Synoptic reporting is increasing but hampered by the need to develop complex checklists derived from an inconvenient variety of sources and have them easily available at sign out. Usually only an abbreviated checklist is available, lacking the decision support information in the more complete forms.
    • Interpretation -- At first blush this does not seem amenable to improvement through use of SDE/SD tools; however, consider that such tools will decrease missing historical information (through better use of the CPR) and will also provide decision support tools that can improve interpretation at the point of care.
    • Transposition -- Perhaps most difficult to correct. Cannot prevent in the absence of coupling sign out to point of care with SDE. For example, the uropathologist that sees 20 prostate biopsies in a day. Transposition of a diagnosis from one patient to another results in a double error that is almost impossible for the pathologist to detect at proofing.
  • ... are available in literature
    • Several studies regarding the quality of APR have attempted to outline recommendations delineating required, preferred, and optional elements in a report. These recommendations build a framework for an anatomic pathology report that contains all of the components required for optimal for patient care [ADASP06].
    • Recently published reporting guidelines for surgical pathology afford a framework for the creation of APRs containing all of the components that are required or optimal for patient care [Goldsmith 08]. **According to CEN TC 251 WI 130.1.1:2003 a histology report may be divided into sections describing the: macroscopic appearance, microscopic features and the conclusion of the service provider based on these findings. Each of these sections may consist of free-text, measurements (e.g. size, weight etc.) and code values representing the findings. Different healthcare parties may be responsible for different parts of a report. Furthermore, overall responsibility of reviewing and signing-off the reports may be assured by another supervisory healthcare party.
    • According to “evidence-based pathology”, only features that are reproducible and relevant – with demonstrated diagnostic or prognostic significance – should be reported in description [Fleming 02]

Sub-optimal implementation of these templates in Structured Data Entry (SDE) systems

Standardization efforts have resulted in structured reporting from a display perspective only, with the data elements being unstructured from an information system perspective. Some vendor tools have facilitated the capture of report items in an information system structure, but these systems lack widespread acceptance because many vendors use methods that are difficult to use and/or create unacceptably formatted reports; others simply do not offer them at all. In many cases the vendor products have been counterproductive as they often produce a poorly formatted report with many redundancies (as many labs simply append them to their usual reports) or do not support innovative formatting requirements that many labs require.

Hypothesis

It is possible to define an information model for structuring and coding AP findings and to provide selection processes in order to render the relevant information and clinical decision support to the different healthcare providers in different contexts, essentially for patient care (in or outside hospitals) and also for clinical research & epidemiology (Cancer registries & Centers for Diseases Control), tissue banking, biorepositories, training/education and others... Front-end (for point of care capture and viewing of report data) and back-end (for database storage and manipulation of the data) software applications are required. The elements can be represented by form controls or in other innovative ways to prompt and facilitated data entry. Individual items populate the controls - pick lists, combo boxes, check boxes, radio buttons, etc. Obviously for this to work well a large, structured specialty medical knowledge base is needed also. Once a user-friendly front end and well-designed back end are coupled, the advantages of SDE/SD systems outline previously become reality. Given the current disarray in pathology documentation, SD need be the most important goal of pathology informatics. All of the other informatics technologies are hampered by the lack of SDE/SD systems and will not reach their potentials without them. And as noted above the lack of such systems stymie efforts to avoid preventable medical errors, improve the efficiency of care delivery and financing, and to generate accurate healthcare data for a variety of purposes. At this point it is appropriate to mention another form of error: error in aggregate data. Aggregate data, or data focused on certain medical conditions from numerous patients, are valuable for research and for establishing policy. Such data are extremely valuable; for example, for speeding up drug development, for guiding policy aimed at decreasing healthcare expenditures, and for many other purposes. However, much aggregate healthcare data currently being gathered is seriously flawed, as it is abstracted from illegible, incomplete, cluttered free-text charts often by poorly trained employees of hospitals and other agencies. Useful aggregate data will require that valid, complete data is validated at the point of care with SDE systems and retrieved by SD systems. Confidentiality, security issues and transmission transactions are out of the scope of this white paper.

Definitions & Background

Structured clinical document

  • Clinical Report Document (CRD. A clinical report is a clinical document reporting a medical event by a healthcare practitioner. Typically this document will be a part of the official medical record. The term “Clinical Report” is equivalent to the HL7 term “Clinical Document.” (need to reference this)
  • Structured Report Document (SRD)

The structure facilitates electronic representation, transmission and storage of the report. The primary focus of this paper will be on the information systems automation perspective.

    • Levels of structure

A structured report document is a clinical report whose content is structured at one or more levels of detail. The structure may consist at the lowest level of meaningful section headers. At a finer level, section content may be structured into sub-sections or even individual data items. By contrast, an example of an unstructured report is a pure narrative report without sections, headers or sub-sections. (this is in alignment with CDA definition). Structure is often expressed by means of metadata. There are different degrees of structuring information corresponding to different ‘‘levels’’ of machine readability of the report.

      • Level 1: Within a free text or narrative report, sections may be identified by human beings but they are not identified by “Observation identifiers” in a coding system (not readable by computers). An ‘‘unconstrained’’ message format (level 1) allows for free text in order to facilitate the transfer of unstructured report. This approach (“old style”) provides maximum compatibility with older systems and simplifies the implementation process from a technical standpoint.
      • Level 2 : The semi-structured report consists in textual information structured in sections identified with “Observation identifiers”. Possible coding systems include LOINC, SNOMED CT. Level 2 adds a specification for section constraints within the message in order to provide some structure while still allowing for unconstrained elements within the headings.

Examples of “Observation identifier” (NAACCR Pathology Laboratory Electronic Reporting version 3.0) : Final Diagnosis (LOINC code: 22637-3 Path report final diagnosis); Text Diagnosis (LOINC code: 33746-9 Pathologic findings); Clinical History (LOINC code: 22636-5 Path report relevant Hx); Gross Pathology (LOINC code: 22634-0 Path report gross description); Micro Pathology (LOINC code: 22635-7 Path report microscopic observation), etc.

      • Level 3 : The structured report consists in a list of structured observations (coded data or findings or items) based on templates (e.g. in the US CAP cancer check-lists and in France the SFP (French society of pathology) templates CRFS). Templates are identified by template identifiers and version. Not only the content structuring is important to facilitate the retrieval of information, but also the coding of these contents with standard codes. The most frequent used coding systems in anatomic pathology domain are SNOMED Clinical Terms® (SNOMED CT®), ICD-O-3 and ADICAP in France. Level 3 provides for fully structured ‘‘entry level templates’’ and is by far the most granular, allowing for maximum machine readability.

In essence, each increasing level allows for additional machine readability, but the clinical content of the notes should be identical in all three levels.

    • Purpose and scope

Depending on the activity (patient care, research, training) the features that are reproducible and relevant could be slightly different. It is important to analyze the workflow of anatomic pathology reporting and to adapt the reporting IT solution to their specific use (e.g cancer surveillance programs, for example, may utilize pathology reports to identify new cases and collect information on previously reported cases. As part of quality assurance initiatives, some items of the AP report are required for accreditation purposes by INCA (in France) or The American College of Surgeons’ Commission on Cancer (in US)).

  • Electronic Structured Report Document (ESRD): Structured electronic report is a SRD that is authored, transmitted or stored by electronic means. (Written this way so as not to exclude those who use structure in their document but don’t have an electronic means to use Anatomic Pathology Report (APR)

A CR of a pathology specimen examination (e.g. surgical pathology, cytopathology, autopsy).

  • Structured APR (SAPR): An APR subset of SR.
  • Structured data entry (SDE): An enabling technology for Structured Documentation. Structured data can be acquired in two ways. First, electronic user interfaces support primary capture in structured form, often in real time. Use of predefined templates (or forms, guides, patterns, outline, scheme, schema, etc.) in report creation (e.g. pick-lists, combo-boxes, radio buttons). Second, one can capture unstructured data, i.e., narrative text, and then attempt to structure essential information within it either manually or automatically.

Templates

  • Structure

In the context of structured data, metadata exists at two levels

    • Data element (or item) : name, description, data type, validity (range, value sets, etc)
    • Group of data elements (or grouped items): relative order of elements within the group, branching logic, computations and complex validation
  • Codes from clinical terminologies
    • Reference terminologies (RT) (e.g SNOMED CT or LOINC) are developed to provide exact and complete representations of a given domain’s knowledge including entities and their relationships. They are typically optimized to support storage, retrieval and classification of clinical data
    • Codes: Terms or expressions expressing data elements as well as groups are defined to support clinician’s entry of patient information in computer programs (interface terminology).

Standards & Systems

  • Clinical documents: There are many examples worldwide of implementation of structured reports usually using Web technologies (e.g such as the CAP electronic cancer check lists in XML). But, these formats are specific for each country which prevents their exchange universally.

Several studies have focused on defining an appropriate IT standard to comprise the structured and encoded clinical documents.

    • ISO/IEC 11179 standard

The ISO/IEC 11179 standard was originally developed for descriptive metadata repositories and concerned with issues such as provenance. ISO/IEC 11179 is now in use by several US government organizations. The National Cancer Institute’s cancer Data Standards Repository (caDSR) is also based upon ISO/IEC 11179. In this standard, a data element is defined as a combination of a concept and a value domain: the latter defines a set of properties such as data type, range, list of values, maximum length, minimum length and format (a limited type of pattern validation, such as YYYY-MM-DD for dates). ISO/IEC 11179 was not designed in the context of form-based SDE.

    • HL7’s Electronic Data Capture Instrument (eDCI)

HL7’s Electronic Data Capture Instrument (eDCI) is a draft standard. It is intended to support reusable templates for electronic case report forms, and is based in part on XForms, a W3C standard. XForms itself, originally intended to be a significant improvement over HTML 4.0 forms, is reasonably powerful. However, it does not, by itself, meet the needs of medical SDE; its ability to perform computations has been deliberately limited, and for the latter purpose it is intended to be embedded in a host environment/language that provides these abilities. While several commercial implementations or variations of XForms (Microsoft Office InfoPath, IBM Alphaworks, and the opensource Chiba project), the cross-platform variations in the host language make porting difficult.

    • HL7’s Clinical Document Architecture (CDA)

A generic structure for clinical documents (discharge summaries, progress notes, history and physical reports, prior lab results, etc.). HL7CDA is the most reliable standards that cover these needs [Dolin 06]. It allows the clinical data to be both machine and human-readable and provides a framework for the incremental growth in the amount and granularity of structured, codes-bound clinical information. HL7’s Continuity of Care Document standard (CCD) was created by making the The American Society for Testing and Materials’ (ASTM) Continuity of Care Record (CCR) compliant with HL7’s reference information model. (Get HL7 Structured Reporting WG input) (also look at xforms more)

    • CDISC (Clinical Data Interchange Standards Consortium) Operational Data Model

A standards body called CDISC (the Clinical Data Interchange Standards Consortium) working on defining standards for exchange of both data and metadata in the context of clinical studies. The CDISC standards deal mostly with definitions of structured data.

    • HL7 V3 SDA, R1
    • DICOM SR
    • HL7 v3 : Domains Laboratory, Specimen, Observation
  • Coding Systems
    • SNOMED (distinguish between CID and legacy codes), LOINC, ADICAP, ICD-O
  • Integration profiles
    • ITI Sharing Value Sets integration profile is a technical solution that aims to define, manage, maintain and share the value sets corresponding to coded items. For example, the value sets related to specimen identification and description, and to specific anatomic pathology observation and diagnostic codes.
      • HL7 v2.5 transactions (MSH, PID, PV1, OCR, OBR, OBX, SPM) : Placer Order Management, Filler Order management, Order Result Management, Procedure Scheduled and update, and SPM segment(Specimen Type, Specimen Additives, Specimen Roles, Specimen Availability).
      • DICOM transactions : Specimen module : Container Types, Container component Types, Anatomic Pathology Specimen Types, Breast Tissue Specimen Types, Specimen Collection Procedures, Specimen Sampling Procedures, Specimen Stains, Specimen Preparation Types, Specimen Fixative, Specimen Embedding Material.
      • CDA template must be harmonized…

Previous approaches and current status

  • Survey: International initiatives

We collect the different initiatives related to anatomic pathology reporting in different contexts (patient care (anatomic pathology reporting to clinicians, to health networks, to personal electronic healthcare record, etc), epidemiology (control disease centres, cancer registries, screening organisms, etc), biomedical research (clinical trials, research in anatomic pathology) and in different countries. We investigate for the regulatory aspects (security and confidentiality), organizational aspects, medical aspect (content, structure and coding of the anatomic pathology report) and technical aspects (architecture, formats and transactions). There are several international initiatives intending to define templates for structured anatomic pathology reports:

  • In the United States, the College of American Pathologists (CAP) has published approximately 72 protocols/checklists for reporting cancer. To further enhance implementation, these cancer checklists are released in an XML schema format. This distribution of the checklists also includes the appropriate SNOMED CT codes to allow vendors to implement and retrieve on discrete data items.
  • The Canadian Association of Pathologists endorsed the content of the CAP’s cancer protocols/checklists as standards to be used nationwide by Canadian pathologists in June 2009.
  • The Royal College of Pathologists of Australasia is currently collaborating with the CAP Cancer Committee on the cancer protocols.
  • In France (INca/FSP/InVS): 24 cancer check lists in text format, 9 of them will be available in CDA format (esophagus, stomach, pancreas, colon and rectum).
  • In Netherlands (AORTA Implementation guide HL7v3 Anatomic pathology version 1.02 – Nictiz)

We appreciate your participation:
APSR Survey
View responses
Modify responses

  • Survey: Pathology report coding details and foreseen evolution

In order to understand issues related to vocabularies used for coding reports, we are collecting information from countries about coding systems used for the various aspects of pathology report, about national policies about coding, about participation in IHSTDO and about translation of SNOMED-CT. Data is used only in aggregated form, to give maximum freedom in answers.

Survey is here: Pathology Report Coding Survey

Use Cases/contexts

Different activities

So, anatomic pathology reports (APR) may be designed for multiple uses, essentially in patient care (diagnostic and prognostic information, tumor board) and also in epidemiology (centers for diseases control) and clinical research (cancer registries), etc.
We have to carefully analyse these different contexts (workflow) of producing APR in order to define both templates and IT solution which support these different uses.

  • PATIENT CARE
    • Context 1 : anatomic pathology reporting for clinicians
    • Context 2 : anatomic pathology reporting for healthcare networks (e.g tumor board)
  • RESEARCH & EPIDEMIOLOGY
    • Context 1 : research in the anatomic pathology field (retrospective studies in AP laboratories)
    • Context 2 : anatomic pathology reporting for public health agencies (cancer registries, Centers for Diseases Control, etc.)
    • Context 3 : anatomic pathology reporting for tissue banking and biorepositories.
  • TRAINING & EDUCATION
  • Others...

Different specimen types

Moreover, for each context we should consider in synoptic/structured reports the different specimen types: operative specimens and biopsies in surgical pathology, fluid and FNA specimens in cytopathology, and tissue micro array in molecular biology.

Work plan

  • First step
    • Collect and analyze available structuring architectures & coding processes of APR in different countries
    • Analyze available technical formats (CDA, ...)
    • Analyze available coding systems (SNOMED, LOINC, ICD-O, ADICAP...).
  • Second step
    • Propose templates (information models) containing the value sets for coded items, for a subset of the available SFP CRFS & CAP synoptics (e.g esophagus, stomach, pancreas, colon, breast, lung and prostate cancers)
    • Validate these templates by pathologists and public health organizations(cancer registries, centers for diseases control, screening organizations, etc)
  • Final step
    • Specify the appropriate architectures, actors and transactions for implementing the reporting workflow in and outside AP lab
    • Identify the technical solution to store and share structured anatomic pathology reports (HL7 v2/v3 messages, XDS, CDA, DICOM SR...)
    • Define selection and filtering processes that regulate the deliverance of the relevant information into the different healthcare providers.

Open issues

  • How to handle the versioning issue of the CDA implementation guides (international & national extensions)? E.g. cancer protocol versions; need metadata that identifies version; need a query use case to illustrate this;
  • How should be reported cases with multiple cancers (2 different instances of the same checklist or even of different checklists)
    • Comes from NAACCR; e.g. both breast and colon specimen in same patient; also, multiple primaries in same specimen; might also combine multiple specimens in one report; need multiple use cases; IACR vs NCI rules for multiple primaries;
  • Will the SNOMED CT codes associated to the CAP electronic Cancer Checklists (CAP eCC) be available?
    • Will SCT be able to cover all concepts?
    • How will we combine other value sets?
    • Need better data input ways
  • How will local/state-specific controlled vocabularies be handled?
  • Shall derived specimens dedicated to biobanking be reported in the report?
    • “This derived specimen is sent to tissue banking” possible statement in checklists.
    • V3 CDA has balloted models for tissue banking
    • Check Goldsmith article – in gross description; Per Article, if done, must be documented in the report.
    • Usually tissue is de-identified when sent to tissue bank so wouldn’t be in the document; can say “all tissue has not been submitted”
    • Should it be in the laboratory tracking system – but not in report?
    • Need to find out what other HL7 groups are doing. Go to Project Insight homepage to see where all HL7 projects are.
  • How to identify/describe derived specimens (a block derived from a specimen would be identified/described with its own identifier but should also carry an identifier/description for the parent specimen)?
    • Include descriptions of all parent specimens to derived specimen, e.g. case, specimen(s), block, slide level, tma cores, tissue profiles on the same slide, microdissection tissue from slide(s).
  • Medical (semantic) consensus is not easy to achieve at regional/national/international level about important features that should be reported as well as the vocabulary and/or code system to use. How to reach the highest level of consensus?
    • Collaborative efforts between different groups of experts/organizations
    • Develop tools, e.g., online surveys, to assist in reaching consensus
    • Address modifiers/search tags for uncertainty and ambiguity and link to modified term
  • Methods to modify for uncertainty and ambiguity must be defined technically
  • How to link CDA documents to DICOM images?
    • CDA provides means to reference an external document from a level 3 entry
    • Document sharing infrastructure stores as documents the manifest to DICOM objects which may be stored externally. The infrastructure provides links between the CDA report and DICOM objects.
  • How to maintain consistency between templates dedicated to patient care and those dedicated to research?
    • Will monitor practices for possible future action
    • Methods for identifying confidential data elements?
  • Standards for data not intended for data not to be included in the patient report document

References

  • Error detection in anatomic pathology. Zarbo RJ, Meier FA, Raab SS. Arch Pathol Lab Med 2005;129:1237-1245.
  • Evidence-based cellular pathology. Fleming KA. Lancet 2002;359:1149-1150. HL7 Clinical Document Architecture R2, Dolin et al, J Am Med Inform Assoc. 2006;13:30–39. DOI 10.1197/jamia.M1888.
  • Interactive exploration of model-based automatically extracted data, Coden et al, Myo Clinic.
  • Patient safety and error reduction in surgical pathology. Nakhleh RE. Arch Pathol Lab Med 2008;132:181-185.
  • Recommendations for quality assurance and improvement in surgical and autopsy pathology. Association Of Directors Of Anatomic And Surgical Pathology, Nakhleh R, Coffin C, Cooper K. Hum Pathol 2006;37:985-988.
  • Reporting guidelines for clinical laboratory reports in surgical pathology, Goldsmith JD, Siegal GP, Suster S, Wheeler TM, Brown RW. Arch Pathol Lab Med. 2008 Oct;132(10):1608-16.

Kohn LT, Corrigan JM, ed, Donaldson M, ed. To Err Is Human: Building a Safer Health System. Washington, DC: Institute of Medicine; 1999

  • Clement J. McDonald, MD; Michael Weiner, MD, MPH; Siu L. Hui, PhD Deaths Due to Medical Errors Are Exaggerated in Institute of Medicine Report. JAMA 284:1, July 2001
  • Kronz JD, Westra WH, Epstein JI. Mandatory Second Opinion Surgical Pathology at a Large Referral Hospital. Cancer 1999; 86: 2426-35.
  • Gupta D, Layfield LJ. Prevalence of Inter-Institutional Anatomic Pathology Slide Review. Am J Surg Pathol 24: 280-284, 2000.
  • Lind AC, Bewtra C, Healy JC et. al. Prospective Peer Review in Surgical Pathology. Am J Clin Pathol 1995; 104: 560-6.
  • Foucar and Foucar, Error in Anatomic Pathology, AmJClinPathol, October 2000 http://www.ajcp.com/.
  • October 10 CAP Virtual Management College report by Dr. Elizabeth Hammond: “Reporting on Cancer Specimens: How to Use the Cancer Protocols and Case Summaries to Improve Quality and Simplify Practice” (http://www.cap.org/html/member/vmc.html#schedule).
  • Busam, KJ, Lack of Relevant Information for Tumor Staging in Pathology Reports of Primary Cutaneous Melanoma. AmJClinPathol August 2001.