(Projektbeschreibung nur in Englisch)
B. Sageder1,2, K. Boegl1, W. Koller3,
C. Chizzali-Bonfadin1, K.-P. Adlassnig1,
A. Hatvan2, A. Berger4, L. Unterasinger4
Background
Surveillance of nosocomial infection is one of the prominent tasks for infection control
teams in hospitals [1]. Efficient surveillance needs to address various data sources within
the hospital such as patient administration, laboratory and other diagnostics, clinical
patient data management, and patient care documentation. With such enormous data
loads epidemiological alerts tend to pass unrecognized if these data are not managed by
computer programs.
Objective
The goal of this system is a knowledge-based monitoring of incoming data [2,3]. Two
kinds of monitoring are performed in this system: Monitoring of major collectives of
patients (monitoring of germs and antibiotic sensitivity patterns, monitoring of
crossinfections) and patient-oriented monitoring of suspected nosocomial infections. For
the first kind only microbiological data are of concern, for the second kind clinical patient
data are essential as well.
Material and Methods
Monitoring of Germs and Antibiotic Sensitivity Patterns
The user also defines how to ignore repetitive reportings, in other words, he defines the
time period within which no warnings are released when an identical isolate is reported
from the same patient.
The second step is the definition of the “alert organism” to be observed. The user selects
the microbial species he is interested in and may or may not define an antibiotic
sensitivity pattern as well as restrictions concerning specimens and senders.
Possible entries for antibiotic sensitivity are: - (resistant), + (sensitive), ± (intermediate),
* (not indicated), # (not tested).
An antibiotic which is not part of the defined antibiotic sensitivity pattern may have any of
these values.
1
Department of Medical Computer Sciences,
Section on Medical Expert and Knowledge-Based Systems,
University of Vienna Medical School,
Spitalgasse 23, A-1090 Vienna, Austria
e-mail:
Barbara.Sageder@akh-wien.ac.at
2
Software Unlimited, Billrothstraße 31/18, 1190 Vienna,
Austria
e-mail: hatvan@swun.com
3
Department of Hospital Hygiene, University of Vienna
Medical School, Währinger Gürtel 18-20, A-1090 Vienna, Austria
4
Department of Pediatrics, University of Vienna
Medical School, Währinger Gürtel 18-20, A-1090 Vienna, Austria
Initially, the user must specify which organisms with which antibiotic sensitivity patterns
he is interested in. Whenever a corresponding isolate occurs, the computer gives a
message. The user defines for specific groups of organisms which types of warnings shall
be released. So far, the alternatives provided by the program are a window on the screen,
an entry in a file, a printed document, an e-mail, or combinations of these. All parameters
(texts, filenames, addresses) are specified by the user. Any term in a text field that is
surrounded by curly brackets is interpreted as a variable (Fig. 16). When a warning
actually occurs, these variables are replaced by the current values.
Monitoring of Crossinfections
A crossinfection is the transmission of a germ from one person to another. The isolation
of the same germ with the same or a very similar antibiotic sensitivity pattern from both
persons is strongly indicative for such occurrences. Classes of such matching
microbiological results are classes of possible crossinfections. A physician still has to
check the elements of the classes to verify real crossinfections.
For each incoming microbiological result the program recognizes whether there are already matching results from different persons in the database. If a matching case is found, a new class of possible crossinfections is created. If the new germ fits in an already existing crossinfection class, it is added to it. The germ only fits to the respective class, if it is sufficiently similar to each member of the class.
Two germs are taken as matching when they share the species name and their sensitivity patterns are identical or differ only slightly. That means, at least x% (which is to be defined by the user) of the antibiotics must be tested on both germs (= definition of maximum percentage of “holes”), and maximum in y% of the antibiotics the sensitivity may differ slightly. What is a slight difference is also to be determined by the user.
As soon as a new class is created or a microbiological result is added to an existing
class, the user gets a warning in the format he previously defines. This is similar to the
definition for monitoring of germs and antibiotic sensitivity patterns, but there are different
messages for the creation of a new class and the addition of a result to an existing class.
Some germs with specific antibiotic sensitivity patterns occur so often that the monitoring
would respond too often (e.g., Staphylococci resistant to penicilline and sensitive to all
other tested drugs). However, this occurrence rarely is of any epidemiological
significance. The program offers the possibility to exclude such germs from monitoring
(Fig. 16).
A list of all crossinfection classes and lists of all microbiological results belonging to them can be viewed (Fig. 17,18).
Monitoring of Nosocomial Infections
Clinical data are transferred from the patient data management system (HP CareVue™)
that is in routine clinical use at the Department of Pediatrics, whereas microbiological
data are obtained from the MONI system. The knowledge-based analysis of the collected
data will be performed by the MedFrame system, that is currently under development at
the Department of Medical Computer Sciences. A specific definition and adaptation of
monitoring rules will be possible by using the MedFrame/KBuilder Toolkit (Fig. 19).
Results
All prototypes have been tested by the hospital infection control department for several
months. The retrospective studies have shown that a routine use of the system for active
monitoring could save much time for the members of the hospital infection unit. It reports
immediately all the defined alerts from the microbiological lab. So urgent actions can be
taken without loss of time. All other results are stored for further utilization, but the user
does not need to check them. At present the prototypes are being further developed and
reimplemented for routine use.
For the monitoring of nosocomial infections a pilot study has been started at the
Department of Pediatrics of the General Hospital of Vienna (Division of Neonatology and
Intensive Care) to demonstrate the feasibility and the clinical impact of the system in
monitoring of infected newborns.
Technical Specification
All current prototypes were implemented in Java (Symantec Visual Cafe) using ODBC for
database access. They are designed to monitor the incoming data, but can also be used
for retrospective studies. The second is especially useful during the evaluation period.
Conclusion
Retrospective studies have demonstrated the utility of the current prototypes. Together
with the users these will be further developed and other prototypes will be designed.
References