Examples of
Integration
The CPMC
Clinical Repository is furnished with the controlled medical
vocabularies. It provides information about department-based vocabularies
to application developers (through browsing tools) and their applications
(through query facilities). For example, someone writing a program to display
laboratory data would need access to the controlled vocabulary used by
each laboratory system providing data to the central database in order
to understand how those data were structured. The resulting application
would also need access to the vocabulary in order to translate the data
properly for display to users. The vocabulary repository has been designed
in such a way that the applications can function independently of changes
to the vocabularies of the departmental systems.
Clinical Database Design
Laboratory Results Review
Confidential
Data Display
Automated Decision Support
The
Medline Button
One of the biggest beneficiaries of the availability of the MED has
been the group charged with the expansion of the clinical database. With
the addition of each new departmental system, the first step is to model
its controlled vocabulary in a way that is commensurate with previous vocabularies.
If the new vocabulary can be accommodated by the existing model, then the
database applications can handle the new data using the same methods as
are used for existing data. For example, as data about medical procedures
are provided by a new system, they can be stored and displayed using the
same tables and software that works with older procedure data.
The initial application for laboratory results review functioned by displaying one set of test results (for each battery of tests) at a time, in simple reverse chronological order. A later modification allowed all results of a particular type of test or battery to be displayed sequentially. However, the original laboratory system often provided the same or similar clinical information under different codes. For example, there were twelve different tests which reported blood glucose. To make matters more complicated, in 1994 a new laboratory system was added which supplies an entirely new set of tests and codes. Unfortunately, the laboratory systems do not provide any means for knowing which tests were similar.
However, since the MED includes information about specimens and substances,
it is a simple matter for the MED editing tools, using an automated classification
technique, to collect similar concepts together (e.g., "all tests which
measure glucose in a blood specimen"). Furthermore, even if concepts are
organized in one way (for example, according to the laboratory which performs
the test), this does not prevent the MED from adding additional classification
schemes (such as, according to similar clinical significance). Thus, a
program which displays all chemistry tests one way and all microbiology
tests another way, can still determine that a test (such as "hepatitis
B antigen") is to be displayed as a chemistry test, even if (as in this
case at CPMC) it is performed in the microbiology laboratory.
Another application which takes advantage of MED information is one
which is closely linked to the results review application, called Confidential
Data Display. This application displays highly sensitive clinical information
only after reverifying the user's credentials. The challenge for the builders
of the application was to determine which clinical information was to be
considered by the application to be sensitive. A test result may become
sensitive depending on changing criteria (such as political pressure) and
a new laboratory can add a sensitive test without notifying the application
developer. We therefore created a class of concepts in the MED called "confidential
data item". Tests which fall into this category are added as descendants
of the concept as they are added to the system or as their politics change.
This new classification does nothing to disrupt the old classifications
that the test may have had, but now the display application simply needs
to know if the test in question is a descendant of the concept "confidential
data item". Thus, even though the vocabulary changes, the application remains
unchanged.
The clinical alerting system at CPMC is integrated with the clinical
information system through the coding of clinical information provided
in the MED. For example, when examining the results of tuberculosis reports,
it is designed to send an alert when a positive culture result. As the
laboratory adds new result terms, the alerting system needs a way to tell
the positive from the negative, for example, to know that "M avium
intracellulare"
is a positive result and "no growth to date" is negative. The laboratory
system does not provide this additional classification information; however,
the MED does. Thus, the alerting system can send a message when a culture
report comes back with a new type of tuberculosis and is able to ignore
creative new ways to say "negative".
As part of the efforts to link library and clinical information sources, experiments have been conducted in which one or more patient data items are used to initiate literature retrieval. Part of this effort involves the translation of terms in clinical vocabularies into those used to index the medical literature (i.e., MeSH - the Medical Subject Headings). Information drawn from the UMLS has been included in the MED to support the translation process. As a result, users looking at a laboratory test can have the system offer a set of possible questions of interest, customized to be relevant to the particular test, and then select a question and have the information search conducted automatically, using the MED to translate from laboratory terms to MeSH. The additional information in the MED can be used for generating questions and strategies as well. For example, if the user is looking at a digoxin level, the system can generate the question "What are the side effects of digoxin toxicity?". The is possible because the MED links digoxin tests and the chemical digoxin.