| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Proceedings of the 24th Arnold O. Beckman Conference |
1 MDS Laboratory Services, US, 5217 Maryland Way, Suite 303, Brentwood, TN 37027. Fax 615-661-7922; e-mail raller{at}mdslabsus.com.
Abstract
Recent advances in electronic connectivity technology make conceivable almost instantaneous movement of data from the patient or laboratory to any other point in the healthcare system. The Internet, combined with new standards for wireless data transmission, has erased many of the previous physical barriers. Interface engines and formatting/content standards have facilitated the connection of multiple disparate systems in medicine. However, major logical barriers persist, and overcoming these barriers is key to achieving functional connectivity. These include fundamental issues of patient safety (such as reliably identifying the patient), capture of the most clinically meaningful data (patient history, physical examination findings, physician diagnostic impressions, and full range of orders), unambiguous identification of data elements, and synchronization of control files among multiple different systems within the healthcare enterprise. Fortunately, tools exist to address each of these areas. This report cites illustrative examples of such tools.
In an ideally connected medical practice environment, all clinical data would be unambiguously captured electronically at their primary sources and instantly transmitted to wherever in the healthcare system they were needed for clinical care. Data would be presented in context, and actionable items would be flagged for intervention. Finally, the system would facilitate assimilation and comprehension by the caregiver.
Remarkably, we now have electronic and computer tools available that make feasible the transmission of data to the physician’s alphanumeric pager as soon as they have been analyzed in the laboratory or produced by a point-of-care testing device. Thus, within 2 min of analysis, Susan Donahue’s abnormal "LAP" is displayed on the physician’s pager (Fig. 1
).
|
Capture of the most important clinical data is the weakest link in our connectivity scenario. The technological tools available to electronically transmit those data have far outstripped actual usage in medicine. Placing data into context and flagging of actionable events have been the foci of other presentations and reports. Likewise, it is well recognized that how information is presented to the physician has a direct effect on how well it is assimilated and has a major influence on whether appropriate action is taken.
Technology of Electronic Connectivity
Modern local and wide-area networks utilizing fiber optics and advanced protocols such as asynchronous transfer mode have made possible the ultra-high-bandwidth transmission of data. Likewise, infrared and radio technologies have continued to advance the speed of local wireless connections. The Institute of Electrical and Electronic Engineers 802.11b radio frequency interoperability standard is capable of transferring 11 megabits/s with a direct sequence radio operating at 2.4 gigahertz (1). At the same time, wide-area wireless networks are expanding, both those based on digital-cellular technology (2) and dedicated wireless internet providers (3).
All of these technologies are ephemeral and rapidly evolving. There are, however, certain enduring requirements that will be crucial for patient care no matter what technological medium is used to transfer the data. The network must be highly reliable and available at all times for patient care. Transmissions must be highly secure. As reinforced by the recently published Health Insurance Portability and Accountability Act regulations (4), all communications must be encrypted, and devices such as digital certificates are advocated to ensure that communication is with a duly authorized and authenticated user.
Another tool, the interface engine, is essential for maintaining reliable communications between systems. Such engines transform protocols and formats as well as adjudicate proprietary inconsistencies between systems (Fig. 2
) (5).
|
In addition to the Health Level 7 (HL7) 1 healthcare industry standard format for information transmission (6), specific record and content standards are being developed for certain areas within laboratory medicine, such as the Connectivity Industry Consortium for Point-of-Care Testing (7). In addition, general standards are being promulgated to specify not only format, but also record content, such as the Extensible Markup Language (XML) and version 3 of HL7.
With these tools, new capabilities are now theoretically possible. The entire patient database can be available during data capture to improve relevance of data collection and during information display to the physician to ensure that data are presented in an appropriate context.
But No Panacea
Unfortunately, this marvelous physical connectivity is not the panacea. In the scenario described above, there are serious problems. The phlebotomist was tired, neglected to check the wristband, and the blood specimen was actually drawn from Sally Smith. The history, physical exam, and clinical evaluation are far more critical to making a diagnosis than is an isolated laboratory result. And what "LAP" laboratory test are we actually talking about? Leucine amino peptidase? Leukocyte alkaline phosphatase?
What Are the Most Important Missing Links in Connectivity of Clinical Data?
The most egregious error we continue to make in patient care is failure to reliably and robustly identify the patient. Despite laboratorians’ claims that "laboratory data represent 80% of the objective data in the medical record", it is well accepted among clinicians that the most important information in patient diagnosis and management is a well-taken patient history—and after that, the physical examination. Direct capture of the entire gamut of physician orders is key to the construction of physician-alerting systems. We will never be able to build effective alerting rules in the absence of unambiguous naming of clinical data elements, not only laboratory results, but also clinical observations. Finally, diverse systems-control dictionaries must be kept in synchronization. Fortunately, tools are now readily available to address each of these shortcomings.
robust identification of the patient
This issue may be considered in two stages. The first stage involves the following question: how do we achieve reliable identification of a patient presenting to the health system? A health identification card is helpful, but uninsured people commonly use their siblings’ insurance cards. Biometrics are a more reliable identification tool. Fingerprint patterns are readily recognized by automated equipment and are widely used as a password substitute for computer access control (8). A new biometric technology is now available that can evaluate the vast complexity of the human iris, enabling query from a large database without need for manual entry of a claimed identity (9).
The second stage occurs once a patient is actively involved in a healthcare encounter. How does one ensure accurate identification at each intervention or sample collection? The most widely used technology is a barcoded wristband applied to the patient. This was prototyped during the 1980s, and a full-scale demonstration was mounted at the American Hospital Association meeting in 1988. However, it is only recently coming into widespread use. For example, the Veterans Health Administration (VA) is now implementing the use of barcoded wristbands in all 160+ VA facilities (10)(11). Radio frequency identification tags have recently been introduced and are easier to use, being readable through blankets and without requiring positioning of the wristband (12). For either technology, handheld computer units can scan the identification and record information about specimen identification (or medication or blood administration) (13).
capture of patient history
As noted earlier, the most important information for clinical decision-making is the patient history. In most healthcare environments, this is still captured in handwritten form and is therefore unusable for communication and computation.
Several tools have been developed for automated, electronic collection of patient history. One such example is the Problem Knowledge Coupler (14). These tools are designed to discover, elucidate, and identify problems and also address and manage these problems. Each coupler is focused on a specific type of patient chief complaint and does not presume any specific disease process.
To use the Problem Knowledge Coupler, the user first identifies the chief complaint (Fig. 3
). On the basis of this chief complaint, the system asks about many specific symptoms, which may be entered as positive, negative, or uncertain. Many more questions are asked than would be included in a typical physician-patient interview, allowing the elucidation of less common types of conditions. Findings may be annotated. After history responses and physical findings are entered (Fig. 4
), the coupler indicates the relative likelihood of various conditions based on the information at hand (Fig. 5
). The physician can then make a judgment of the likely diagnosis and proceed with management decisions.
|
|
|
unambiguous naming of data elements
The Laboratory Observation Identifier Names and Codes (LOINC) (15)(16) were first published in 1995 for use in overcoming the Babel of laboratory test names. Subsequently, LOINC was extended for use in clinical medicine (e.g., blood pressure) and renamed "Logical Observation Identifier Names and Codes". The objective of the LOINC project has been to provide a universal identifier for laboratory and clinical observations so that information about observations in electronic messages can be pooled in electronic medical record systems and in research and management databases. Chaired by Clem McDonald, MD (Lab LOINC), and Stanley Huff, MD (clinical LOINC), these codes have come a long way toward avoiding clinical Babel. Patient data from many sources can be merged for clinical care, patient data can be pooled for outcomes management, and HL7 observation reports can be standardized.
LOINC has been endorsed by the American Clinical Laboratory Association, was used by Quest to standardize 21 diverse regional laboratory nomenclatures, and is also being implemented by LabCorp, ARUP, Mayo, Microbiology Reference Lab, and others. All US veterinary laboratories use the system. Managed care companies (such as Aetna and Kaiser) require laboratories to provide LOINC codes for tests reported on managed care patients. Integrated delivery systems, such as Intermountain Health Care (Utah), Partners (Boston), Clarian/Indiana University, and Columbia Presbyterian use LOINC to add meaning to clinical databases. The 3M clinical database requires all inflowing data to be LOINC encoded. The Centers for Disease Control is using LOINC for electronic reporting of infectious disease surveillance and tumor registry data. Other governmental medical records efforts, including the Department of Defense and the VA, use it as well.
LOINC structures each observation name in six parts (Table 1
). The component, or analyte (such as potassium or systolic blood pressure), may also include challenge information (such as 1-h postprandial). The property measured, such as a substance concentration or pressure, is the second part. The third part is a time aspect: was this an observation at a point in time, or a 24-h specimen/averaging? The fourth part is the system (specimen source or organ), such as urine, whole blood, plasma, or cerebrospinal fluid. The fifth part is precision [quantitative, ordinal (1+, 3+), or qualitative/nominative ("yellow")]. Finally, if methodology causes results to be significantly different from one method to another, the methodology is specified.
|
Some examples of fully specified LOINC names are illustrated in Table 2
. Again, it is important to emphasize that the LOINC name is far more than a simple analyte name. It specifies the observation specifically enough that one should see close to identical results, regardless of what laboratory/healthcare organization produced the result.
|
LOINC is copyrighted, with the stipulation that use is free for all purposes. The copyright exists to prevent development of variants. Any distribution of the database must include the copyright notice, all six parts of the name, and the related names.
The LOINC database has grown steadily in size, and as of February 2001 includes >28 000 terms. The database is distributed with RELMA (the Regenstrief LOINC Mapping Assistant), which facilitates the mapping of an observation dictionary to the LOINC lexicon. Table 3
lists several web sites of interest.
|
LOINC assists laboratorians with standardization, provides knowledge about laboratory tests, and helps ensure accurate billing (and is required by several managed care organizations), productivity benchmarking, management tools, and better comprehension of referral and esoteric testing.
Determining the LOINC codes for one’s laboratory database can be approached in two segments. Although a typical laboratory may report results for >3000 distinct tests, it is likely that only a few hundred of these are performed in house. One should concentrate on encoding those assays performed in house and ask the esoteric laboratory to provide LOINC codes for everything referred out.
It is quite appropriate that we ask esoteric laboratories to give us an unambiguous definition of the tests they report. The LOINC code is such a standardized definition. Only the laboratory performing an assay can accurately LOINC code the result.
synchronization of dictionaries/master files
A final hurdle in connectivity is to ensure that all dictionary/control files within our various interacting systems have compatible sets of entries. LOINC helps accomplish this for test result files, but many other files require synchronization. Fortunately, tools to accomplish this task have become commercially available. One such tool is HealthPatterns’ Master File Manager (17). This system creates a metarecord that incorporates all data elements and fields that may be contained in various incarnations of a particular master file; in essence, the metarecord is a superset of all of the master files of synchronized systems. For example, the physician master file might contain a Drug Enforcement Administration number in one system and a beeper number in another. The metarecord would contain both of these and would be used to drive synchronous changes in all systems.
Conclusions
Physical connectivity of systems is rapidly becoming ubiquitous through the Internet and a variety of wired and wireless technologies. In medicine, we can simply use what has been developed and refined in other industries. However, we still have fundamental issues that physical connectivity does not resolve. We must deal with the correct patient. We must capture the most important clinical information (history, physical, and physician observations) that currently is consigned to handwritten oblivion. Observations must be identified in unambiguous fashion, rather than by a vague and misinterpretable string of text, and multiple systems dictionary/control files must be kept in synchronization. Fortunately, robust tools exist to deal with each of these issues.
Footnotes
1 Nonstandard abbreviations: HL7, Health Level Seven; VA, Veterans’ Health Administration; and LOINC, Logical Observation Identifier Names and Codes. 
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
