Graphic Element Markup   Table of contents   Indexes   Authoring Tools and the Expanding Radius of Deployment
 

Meta-analysis of clinical documents for principled markup of Medical Records
 Angelo   Rossi Mori
  Dr.
  CNR-ITBM  Viale Marx 15
Rome   RM  Italy  00137
Phone: +39 06 827 7101
Fax: +39 06 827 3665
Fax: +39 06 86090 289
Email: rossi@color.irmkant.rm.cnr.it
 
Biographical notice:
Consorti, F.
IV Clinica Chirurgica, Université La Sapienza, Roma
 Italy  
 Rome  
 
Angelo Rossi Mori graduated in Physics (1974) and he is a researcher at the Institute for Biomedical Technology of the National Research Council, in Rome. His major reaserch topic is on structures for terminologies and medical records. He is Deputy Convenor of WG II on Terminology CEN/TC251 on Healthcare Informatics and Project Team Leader of PT27 on "Domain Termlist" for Communication of Electronic Healthcare Record.
 F.   Consorti
  MD
  IV Clinica Chirurgica, Université La Sapienza, Roma  Rome   RM  Italy
 
Biographical notice:
 
Fabrizio Consorti graduated in Medicine (1980) and specialized in General Surgery (1985). Since 1986 researcher at the 4th Inst. of General Surgery of the Univ. of Rome. His main research areas are: clinical use of tumoral markers, clinical nutrition, immunity and surgery, colorectal surgery, medical informatics. He is a member of PT27.
 
ABSTRACT:
 
In order to prepare an European Prestandard on "Communication of Electronic Health Care Record - Part 2: Domain Termlist" we carried out an analytical study about names of clinical documents, titles of generic sections, names of data elements, according to our terminological methods.
 

Introduction
 
We describe the structures used to organize clinical information in patient records, that can be the basis for a principled markup and advanced computer processing. The patient record is a set of nested "chinese boxes", with the following properties:
  •  a record folder may be made of documents
  •  some documents may be divided into sections
  •  some sections may be organized into purpose-related clusters
  •  some clusters may be made of a list of homogeneous statements
  •  some statements may be articulated into explicit, processable details
At each level, the usage of an explicit construct is "in competition" with free-text narrative, that presents an unlimited expressive power and is suitable for human presentation but not for further processing. Actual usage and granularity of the above constructs are in fact adapted to each item of clinical information, considering also its topic and the purpose for storing and transmitting it. For example, details on drug prescriptions tend to be fully explicit and systematic in most record systems (and therefore individually processable by computer), while the report of a diagnostic investigation tends to be expressed as narrative separated into broad sections (i.e. suitable for presentation to humans, not for further processing by computer) A good markup system should consider this variability, and provide an adequate set of descriptors for the actual constructs used in a given record.
 

Background
 
The Technical Committee TC251 "Healthcare Informatics" of CEN (the European Standardization Body) is preparing a standard, in 4 parts, about "Communication of Electronic Health Care Record":
  •  part 1 : Architecture (basic attributes of nested components - i.e. complexes and record items)
  •  part 2 : Domain Termlist (tables to characterize the components)
  •  part 3 : Distribution rules (how to attach constraints to each component to regulate access)
  •  part 4 : Messages (how to send the clinical information and the related data)
Each part is prepared by a Project Team (PT), that writes the drafts and processes the comments. Part 2 - Domain Termlist - is under the responsibility of PT27. Part 2 should provide a set of tables to characterize the components, to allow a receiving application to process some coarse grained information from received records, without preliminary agreements on: record organization, data elements, coding systems. A standard cannot impose a particular organization to a record system. Nevertheless, it should facilitate communication of clinical information and its processing by the receiving record system, by providing an appropriate markup to preserve the structure of the information existing in the sending system. The topic is innovative and needs extensive field tests. CEN procedures allow to approve a provisional version of this standard (called "European Prestandard"), as the basis for three years of distributed testing. After revision, the final version (EN) for this standard is planned for the year 2003. In order to prepare the First Working Document (FWD) of "Part 2: Domain Termlist" of the Prestandard, we carried out an analytical study of
  •  several lists of document names, e.g. Notes (progress notes), Requests (diagnostic request), Reports (discharge summary, report of diagnostic procedure), Notifications (infectious disease notification), Charts (observation chart), Study products (images, signals);
  •  700+ titles from 7 authoritative sources of section titles, e.g. Reason for contact, Family history, Treatments Plans;
  •  several large lists of names for data items, e.g. ocular reflexes, birth weight, drinking habits.
We used the approach for terminological modelling adopted by CEN/TC251 and the methodology developed during the GALEN project. The need for integration between the terminological perspective and the information modelling perspective within the patient record structures was discussed in the three workshops of the European Project ToMeLo in 1997 and 1998.
 

Results
 
We defined three major layers for structuring clinical information within a record system - corresponding to major constructs in the other standards of the series - , with a number of related tables and lists.
 

Layer 1. structuring the record into documents and sections (record complexes)
 
A record may be organized in nested complexes, i.e. documents or sections. Complexes convey to the receiver the original organization and may contain structured and/or unstructured clinical data (e.g. free text or multimedia not directly processable from a semantic point of view). Even without using further structures, their names (i.e. document names or titles of sections) may be useful, e.g.
  1.  to build a cumulative table of content from different record systems - i.e. to browse data on the same patient that are written according to different record structures - , for example dealing with records from different healthcare services in shared care; or
  2.  to limit the search within complexes with a particular name in text-based information retrieval, for example when searching for patients with particular features in a large hospital database.
We prepared two classifications (one with 20 documentation categories and one with 30 categories of sections), a taxonomy of 200 generic section titles, a system of 75 descriptors to annotate the above titles (41 descriptors in 8 primary annotation types, 34 descriptors in 7 secondary types).
 

Layer 2. structuring a complex into clinical statements (data items)
 
Certain sections may be organized into a list of smaller homoneneous processable units, called here clinical statements, each representing a "situation", e.g.
  •  a list of symptoms,
  •  a list of laboratory results,
  •  a list of planned activities.
Clinical statements can be rearranged in different ways for different purposes, i.e. they can individually handled according to task-dependent criteria, e.g. to create dynamic views. But each item - when moved - should be always interpreted according to the original context. To obtain this performance, a clinical statement must be made of two explicit parts:
  •  the "clinical concept" regarding abstract record-independent features, i.e. the nature of the situation (e.g. surgical procedure) and its essential details (e.g. body part, approach, extent, instruments);
  •  the "status concept" regarding factual circumstances in a particular patient record, e.g. information about the subject of information (e.g. "patient" or "relative") or the actual status (e.g. an activity may be "done" or "planned", a finding may be a goal or the result of observation).
We defined 14 relevant subtypes of situation:
  •  situation
  •  
    •  condition or state
    •  
      •  allergy
    •  events and accidents
    •  life features
    •  healthcare activity
    •  
      •  physical examination
      •  investigation
      •  recording images and signals
      •  decision-making activity
      •  therapeutic activity
      •  
        •  drug-related activity
        •  immunization activity
      •  information activity
      •  administrative activity
We also prepared a list of status concepts. Among them, a particular subset was identified as "safety critical" and should be carefully handled and precisely marked because they heavily affect the interpretation of the clinical concepts (default values as they are usually understood by a record system are underlined):
 
SUBJECT OF INFORMATION patient , relative of patient, fetus/neonate, mother, donor, other entity
LIFE CYCLE done , in process, planned, exceptionally not done
POTENTIALITY actual , goal, predicted, at risk
PROCESS STATUS new , ongoing, former
CERTAINTY certain , uncertain
NEGATION affirmative , negated
 
A suitable amount of context can be then conveyed into status concepts, in order to achieve the required independence from the original record organization. With this kind of markup it could be possible for example to process different record systems - using various coding systems - in order to extract a health summary with lists of homogeneous items, as allergies, diseases in family history, immunizations, planned activities. Of course, the actual markup must be safely assigned to clinical information items. It should considered an unsafe procedure to assign markup to a list of free text statements just on the basis of the section names under which they are presented. Structured data entry is instead suggested as the substrate to control the proper markup. In this way, the system of descriptors developed for layer 1 can be also used to safely annotate a structured statement.
 

Layer 3. structuring the content of a statement
 
Certain statements may have an explicit internal structure, i.e. a record system may be able to handle individual details within particular kinds of statement. For example, within the statements on laboratory tests, a system can handle explicitly test name, sampling date, numeric value and units. This set of explicit details can be used, for example, to draw a chart that shows the trend of different variables. Drug prescriptions are another example of highly structured statements. We defined the idea of "archetype", to describe the details that may appear in the internal structure of the various kinds of statements, e.g.:
 
habits >
.is a life feature
.has habit type kind-of-habit (e.g alcohol, smoking, non-medical drug)
..has quantity number
..has units unit of measurement (e.g. cigarettes/day)
..has duration time interval
..has stop-time date, time interval (e.g. stopped in 1994, 3 years ago)
 
Archetypes set requirements on the content of attributes and on the coding systems. It is anticipated that the terminological requirements about well-formed statements are satisfied by the most common coding systems. Developers and users can use archetypes to increase their awareness on the overlaps among details expressed within coded values, data element names, and section titles. Archetypes are also useful to design and use terminological services to extract specific details from entries of advanced terminological systems and to prepare uniform structured data input, so that structured data can be easily transmitted and interpreted by different systems. If there is a messaging standard on a topic, it is expected that details of statements in the record correspond to the fields of the message, otherwise the message cannot be easily implemented. Therefore the archetypes and the models behind the messages should be fully harmonized too. We prepared a provisional list of archetypes for each of the 14 subtypes of situations.
 

The links between the above constructs
 
Finally we studied the relationships among the previous constructs, as derived from the corpora. No exhaustive list of relationships was available as reference, so our results should be considered as a preliminary attempt to allow for experimenting with this mechanism. We produced a tentative list of 30+ link names, e.g
 
healthcare activity is documented by note
situation is assigned to problem
condition is interpretation of findings, report
informing has topic record component
 
Links are divided into 6 categories:
 
  •  generic links
  •  documenting and reporting links (between information and documents)
  •  organizational links (between an activity and its context, e.g. a contact)
  •  clinical links (established by professional judgement)
  •  circumstantial links (e.g. between informing activity and its target or topic)
  •  view management links (between items and views)
 
We expect that an extensive usage of links among record components will be an important feature enabled by a proper markup of electronic record systems, i.e. one of the major benefits from the structured record on computer.
 

Conclusions
 
With our experimental studies on a reasonably wide and differentiated set of corpora, from different countries, we prepared a basic European Prestandard, to allow to process coarse grained information from different record systems without preliminary agreements on record organization, coding systems, data elements.It can be applied in data transmission from a record system to another, as well as for homogeneous browsing of different records of the same patient in different locations (i.e. the "virtual patient record", the "life-long patient record").Our achievements are the starting material for a principled markup of clinical information in the patient records, that can be enforced by further standards by different organizations (e.g. HL7, CEN/TC251, ISO/TC215).
 
Acknowledgments
 
Work undertaken by CEN/TC251/PT27 on "Health Care Informatics - Electronic Health Care Record Communication - Part 2: Domain Termlist" under mandate of European Union to CEN, with the contribution of the EU Projects GALEN-IN-USE and ToMeLo.
 
Bibliography
CEN/ TC251 "Health Informatics", at http://www.centc251.org
GALEN and GALEN-IN-USE. documentation (1992-99), available from AL Rector, Medical Informatics Group, Dept. Computer Science, Univ. Manchester, Manchester M13 9 PL, UK e-mail galen@cs.ac.man.uk ; see also http://www.cs.man.ac.uk/mig/galen and http://www.cs.man.ac.uk/mig/giu
ToMeLo (an European Project) see documentation at http://www.ehm.kun.nl/tomelo/#workshops
Rossi Mori A, Consorti F, Galeazzi E. Standards to support development of terminological systems for healthcare telematics. Meth Inform Med 1998 (in press)
Galeazzi E, Rossi Mori A, Consorti F, Errera A. A cooperative methodology to build conceptual models in medicine. In: MIE 97, IOS Press 1997
Galeazzi E, Rossi Mori A, Consorti F, Errera A. MEDITA: the reference methodology for cooperative production of dissections by terminological analysis. GALEN-IN-USE, internal report MET01/97, May 1997
Rossi Mori A, Galeazzi E, Consorti F, Bidgood WD Conceptual schemata for terminology: a continuum from headings to values in patient records and messages JAMIA symp. suppl., 1997:650-4
Rossi Mori A, Consorti F. Integration of clinical information across patient records: a comparison of mechanisms used to enforce semantic coherence. IEEE TITB, 1998 (in press)
Rossi Mori A, Consorti F. Exploiting the terminological approach from CEN/TC251 and GALEN to support semantic interoperability of healthcare record systems. Int J Med Inform 1998; 48:111-124
Alschuler L, Sokolowski R, Rossi Mori A, Dudeck J. SGML/XML in healthcare information exchange standards, in: Proceedings Conf. SGML/XML Europe, Paris, May 98
HL7 SGML/XML Special Interest Group. The Kona architecture proposal. Draft available at http://www.mcis.duke.edu/standards/hl7/committees/sgml/kona.htm
Rector AL, Nowlan WA, Kay S, Goble CA, Howkins TJ. A framework for modelling the Electronic Medical Record. Meth Inform Med 1993; 32:109-19
Rector AL, Nowlan WA, Kay S. Foundations for an Electronic Medical Record. Meth Inform Med 1991;30:179-86
ASTM E1384-91. Standard Guide for Description for Content and Structure of an Automated Primary Record of Care
Severs M (on behalf of the Multi-professional Working Group). Headings for Communicating Clinical Information from the Personal Health Record, a position paper. IM&T training unit, Solihull (UK), September 1997
AMA/HCFA, Documentation Guidelines for Evaluation and Management Services, 1998, available at http://www.hcfa.gov/medicare/mcarpti.htm
Servers M. EPR Programme, Occasional Papers - paper 1: Headings for Communicating Clinical Information from the Personal Health Record. NHS Executive (UK), April 1998, IMG ref. C3231
Severs M (on behalf of the Multi-professional Working Group). Headings for Communicating Clinical Information from the Personal Health Record, a position paper, revision 1998
Pollock D. (ed) Data Elements for Emergency Department Systems (DEEDS). US Department of Health and Human Services, National center for injury prevention and control, Atlanta, Georgia, release 1.0 1997
National Committee on Vital and Health Statistics (NCVHS). Core Health Data Elements, US Department of Health and Human Services, National Center for Health Statistics, August 1996
SNOMED DICOM Microglossary (SDM), see http://www.snomed.org/sdm/sdm.htm
Vaughan NJA, Home PD, Vermeij D, Sando SH, Cashman SJ (eds). The Development of a Generic Message for Use in Diabetes Care. from the "Fifth Gubbio Workshop of the EASD Study Group DO IT" held jointly with the European Union BIOMED Concerted Action project MFIT, GUBBIO, (PG-Italy) 16 -18 May 1997.
Graphic Element Markup   Table of contents   Indexes   Authoring Tools and the Expanding Radius of Deployment