Review Paper c.J. McDonald, P.R. Review Paper Dexter, B. Takesue, J.M Overhage Regenstrieflnstitute for Health Care, Health Informatics Standards: Indiana University Medical Center, Indianapolis, Indiana, USA A View From Mid-America 1. Perspective behavior. Highway speed limits are a cently, everyone used their own idio­ familiar example. Most health syncratic codes to identify laboratory 'Those who develop standards face informatics message standards are tests and clinical measurements. Fur­ a number of problems and confusions. enabling. They provide a way to trans­ ther, health-care also deals with free For voluntary health informatics stan­ mit whatever information has been text, containing information which we dards efforts, the real problem might collected (e.g., a chestX-:ray report or cannot yet resolve into discrete items. be described as a reverse tragedy of a set of vital signs) to another medical A further challenge for medical data the commons[!]. In contrast to farin­ facility. By themselves they will not standards is the strict need to maintain ers on English commons, having more induce a particular behavior, such as privacy while simultaneously allowing information systems (cows) browsing the mandatory collection of a particu­ access by many health-care profes­ on the same informational commons lar set ofclinical variables wanted by a sionals, and the accompanying neces­ provides the most economic gain for regulatory agency. As a conse­ sity of carrying patient identifying in­ all. The tragedy is that for early adopt­ quence, regulatory agencies may dis­ formation along with the medical data. miss a voluntary informatics stan­ Banks only need to know account ers of standards the cost of putting their systems into the standard com­ dard and write their own very spe­ numbers, not individual persons. mons can exceed their gain, and this cialized flat file standard instead. It Finally, Nathan Myhrvold's asser­ inhibits the adoption of standards. The would be better for all of us if regu­ tion that there will "always be a soft­ challenge is to get enough information lators wrote their requirements (as­ ware crisis because ambition absorbs providers onto the standardized com­ suming they are reasonable) as an­ all advances" applies equally to stan~ mons to create the critical mass to other "layer" to the enabling stan­ dards. As soon as we sniff success in draw in all ofthe other potential brows­ dard, for example, as a table listing one domain of standards we extend ers. the variables required for a particu- our expectations. We master stan­ We also face a confusion between . lar transmission. dards for patient registration and struc­ standards that enable and those that Misconceptions also exist about tured test results, and then we want to control. An enabling standard says: the difficulty ofhealth-care standard­ standardize the entire work flow of the "Here is a vehicle. This is how it ization efforts. Banking is fully stan­ health-care system. operates. You are free to use it as you dardized and health-care should be no Despite our laments, standards de­ will." The most successful technical more difficult, or so the argument goes. velopers have made substantial progress standards have been enabling. They Note that banking transactions all deal toward more automated sharing of typically provide an interface or enve­ with one completely fungible quantity: information among health-care orga­ lope to carry, or process, some kind of money. Health-care deals with tens of nizations and providers. In the follow­ tforrnation. They do not prescribe thousands ofdifferent quantities-- test ing we will document those areas about hat information must be included in values, drug doses, clinical measure­ which we are most familiar and which the envelope. The CD-ROM music ments, etc. Health-care deals with lit­ we hope will be of interest to the ~nnat and MIDI music format are erally hundreds of thousands of de­ reader. Table 1 provides a summary of ~d ellamples of enabling standards. scriptors. Count the code/vocabulary these standards, including the respon­ Regulatory organizations are more entries in SNOMED or the Read Codes sible organization and contact infor­ terested in standards that control to confirm these numbers. Until re- mation. fbook ofMedical Informatics 1997 67 Revi.ew Paper Table l: Summary of standards/contact information AMIA American Medical Informatics Association http://www .amia.org/ or ami a- [email protected] ASTM American Society for Testing and Materials http://www .mcis.duke.edu/standards/ ASTM/astm.htm ATCC American Type Culture Collection sales@ atcc.org or [email protected] CAS Chemical Abstract Society [email protected] CORBA Common Object Request Broker Architecture info@ omg.org CPRI Computerized Patient Record Institute http://www .cpri .org/ CPT Current Procedural Terminology http://www .mcis.duke.edu/standards/ termcode/cpt4.htm DICOM Digital Imaging and Communications in Medicine http://www .mcis.duke.edu/standards/ DICOM/dicom.htm HIMSS Healthcare Information and Management Systems http://www .himss.org/ Society HL7 liealth Level Seven http://www.mcis.duke.edu/standards/ HL7d/hl7.htm or [email protected] ICDlO-PCS International Classification of Diseases www.who.org/programmes/mnh.ems/ icdl 0/icdl O.htm IEEE The Institute of Electrical and Electronics http://www .mcis.duke.edu/standards/1 Engineers EEE/ieee.htm IETF Internet Engineering Task Force ftp://ietf.org/ or ietf-ediint@ imc.org IUPAC Intern.ational Union or Pure and Applied http://www.mcis.duke.edu/standards/ Chemistry termcode/iupac.htm standards @regenstrief.iupui.edu or LOINC Laboratory Observation Identifier Names and http://www.incis.duke.edu/standards/ Codes termcode/loinc.htm SNOMED Systematized Nomenclature of Human and http://www .mcis.duke.edu/standards/ Veterinary Medicine termcode/snomed.htm SSL Secure Sockets Layer http://ds.internic. net/internet- drafts/draft-ietf-tls-ssl-version 3-00. txt UMDNS Universal Medical Device Nomenclature System http://www.mcis.duke.edu/standards/ termcode/ecri.htm http://www.mcis.duke.edu/standards/ UMLS Unified Medical Language System termcode/umls.htm or wth@ nlm.nih.gov 2. Message Standards ing, admission discharge and transfers, the capture of clinical data from auto­ financial transactions and master file mated bed-side instruments, clinical HL7 has progressed since we last exchange. It also includes specifica­ trial management, immunization report· reported on its status. It has completed tions for entirely new areas, including ing, and adverse product experience balloting of version 2.3 [2]. This ver­ appointment scheduling, problem list reportng. In printed form, including sion includes enhancements to existing maintenance, nursing goal mainte­ cross refernces, it is now 863 pages. areas: order entry, observation report- nance, referral notices, the US UB92, The electronic version of the penulti- 68 Year book ofMedical Informatics 1997, Review Paper mate draft of the vs. 2.3 document is even now. As we increase the scope The goal of the vocabulary SIG is to available at no cost from the Duke and the detailed workings of the health­ specify the codes and vocabulary sys­ HL7 server. (http://www.mcis.duke. care world, how will health systems tems that should be used for each edu/standards/ HL7 /hl7 .htm) developers be able to cope with such coded field in HL7, not to create new The HL7 SIGOBT group has been large and detailed models and how will vocabularies per se. The vocabulary working on using CORBA [3] and, they accommodate the variations and SIG is collaborating with the National Microsoft OLE (ActiveX) as alterna­ evolution of real health systems? One Library of Medicine (NLM) to incor­ tive mechanisms for delivering HL7 possibility would be to create more porate HL7 vocabulary lists in the message content [4]. This approach abstract models with fewer parts, that NLM's Unified Medical Language has the advantage that the information is, fewer objects with more System(UMLS) [10]. If this effort is content of a message is identical expressivity. Huff presents a good successful, and other U.S. standards whether delivered via ASCII charac­ example of model abstraction by re­ groups follow a similar strategy, the ters, OLE, or CORBA. Two succes­ ducing a medical record model down UMLS could provide a mechanism for sive OLE prototypes have been built. to about six heavy-duty objects [6]. In unifying the contents of various mes­ Fourteen different parties demonstrated his proposal, all patient attributes are sages standards. The goal ofthe clini­ OLE interconnections at the HIMSS modeled as time-varying observations. cal alert SIG is to define a way to '96 meeting in San Diego. Ittook three This is also the modeling approach of convey alert messages from decision days to prepare this demonstration, one large U.S. pharmaceutical manu­ support system using the Arden Syn­ one day of which was consumed by facturer, and is reminiscent ofthe PEN tax [11,12] torepresentmedicalrecord the physical network installation. This & PAD model which is, in many ways, and clinical care systems. relatively fast setup contrasts with the an even more compact and exquisite HL7 now has affiliates in New six-week marathon required to pre­ model [7]. Zealand, Australia (where HL7 is be­ pare for the first HL7 demo atthe 1990 The advantage of greater abstrac­ ing adopted as a national standard), New Orleans HIMSS Conference. tion is ease of implementation, flexibil­ Canada, Germany, and the Nether­ Hewlett-Packard