GermWatcher/GermAlert
Washington University Medical Informatics Lab / BJC Health System
MEDICINE

GermWatcher and Germ Alert are part of a  family of medical expert systems designed to support Infection Control specialists in detecting, tracking, and investigating hospitalized patients with significant contagious infections so that new outbreaks can be quickly identified and contained before additional patients are exposed or infected.

Please describe your application and the information technology used in conjunction with it. Please keep your language simple and your explanations non-technical.

The Medical Informatics Laboratory at Washington University Medical Center has one principal goal - to improve the quality of medical care by providing health care professionals with the right information at the right time. One of the principal activities of our group is to minimize the occurrence of complications from inappropriate medication administration. Because many different providers can prescribe medications without full knowledge of the consequences of interactions from medications already received, the Laboratory has developed a group of software systems that integrate pharmacy and clinical information. This new software provides pharmacists and clinicians with an additional "safety net" that assures a higher likelihood of safe prescriptions even in elderly patients receiving multiple medications.

The most deadly complication of any hospital admission is an infection. Hospital-acquired (nosocomial) infections are frequently fatal because infections caused by organisms present in hospital environments are more likely to be resistant to most common antibiotic treatments. The U.S. Centers for Disease Control and Prevention (CDC) estimates that nosocomial infections involve more than 2 million patients, directly or indirectly contribute to an additional 58,000 deaths, and cost the nation more than $4.5 billion annually. The CDC also estimates that 36% to 54% of these adverse outcomes are preventable with aggressive Infection Control services.

GermWatcher/GermAlert is a family of expert systems designed to support Infection Control specialists in detecting, tracking, and investigating infections in hospitalized patients. Three basic activities by Infection Control professionals form the first line of defense in controlling the spread of infections in hospitalized patients:

• Continuous surveillance of all infections to detect unexpected increases in existing infection rates, infections with common organisms having high resistance to the usual antibiotics, and new infections with organisms not previously seen in this hospital;
• Investigation of unusual infections detected by the surveillance system to locate and contain a common source or cause of the infections; and
• Aggressive preventive programs to eliminate substandard practices or ineffective methods to prevent or reduce the risk of further infections to additional patients and healthcare workers.

By combining expert systems and database management technologies, the GermWatcher family empowers Infection Control professionals to attack a serious multifactorial problem with a variety of tools and approaches.

GermWatcher is an expert system that models the CDC's National Nosocomial Infection Surveillance System (NISS) culture-based definitions for nosocomial infections. NNIS is a set of national standards for nosocomial infections created by the CDC so that national infection rates can be monitored. Following an extensive validation process, GermWatcher has successfully automated the basic surveillance tasks at Barnes and Jewish Hospitals in St. Louis, Missouri.

GermAlert is an expert system that contains rules for "infectious disease emergencies." To develop GermAlert, we asked a simple question, "What infections would cause you to stop what you are doing and get on the telephone to initiate new procedures or investigations?" Because there are no national guidelines for infectious disease emergencies, GermAlert's rules model the concerns of our Infection Control specialists.

Suspect is an expert system that contains rules for organisms resistant to usual antibiotics. These organisms appear for one of three reasons: (1) the antibiotic resistance pattern was mistyped by the technician, (2) the resistance pattern has been described in the literature but is rarely seen at our institution, and (3) the resistance pattern has never been seen at our institution. Since clinicians select antibiotic therapies based on antibiotic resistance information, it is critical to detect unusual resistance profiles so that proper therapy is delivered.

Reportable Diseases Monitor is an expert system that encodes the State of Missouri rules for infections which have significant public health risks. This system automatically generates the State's Reportable Diseases form which includes all relevant patient and clinical data required by Missouri's Department of Health. A pilot program to automate the transfer of this data to local, county, and state epidemiology offices is in progress.

The collection of expert systems that comprise the GermWatcher family has dramatically changed the way infection control is performed at Barnes and Jewish Hospitals. By combining expert systems and database technologies, Infection Control specialists can concentrate on the tasks they do best Ð infection outbreak investigation, outbreak intervention and healthcare professional education. In response to GermWatcher's success, this technology is being expanded to additional medical centers in metropolitan St. Louis. As we continue our multi-faceted attack on one of medicine's most feared complications in hospitalized patients, the impact of this technology on the risk of infections in our patients will be immeasurable.

These programs are in daily use by infection control specialists throughout our medical center and we believe they enhance patient care and possibly even save lives. Our experience suggests that appropriately applied information technology does not increase the cost of medical care - it lowers these costs and improves as well.

 Has your application helped those it was designed to help? How has it affected them? What are its most important benefits? What positive impact will your application have beyond its immediate users? Will it change how others live And / Or work? How will it impact society?

Although infections are one of the most deadly complications in hospitalized patients, multiple studies by the U.S. Centers for Disease Control and Prevention have shown that diligent multi-faceted Infection Control programs can reduce infection rates by 32% to 54%. The GermWatcher family of expert systems is designed to support Infection Control specialists in detecting, tracking, and investigating infections so that new outbreaks can be recognized and contained before additional patients are infected. GermWatcher automates a number of critical, yet time-consuming, daily tasks so that Infection Control specialists can concentrate on the tasks they do best Ð infection outbreak investigation, outbreak intervention and healthcare professional education.

Prior to the development of the GermWatcher system, highly trained Infection Control specialists would manually inspect every positive microbiology culture result looking for unusual infections. At Barnes Hospital (now called Barnes-Jewish South), over 900 cultures are processed daily. The manual review of these cultures consumed no less than three (3) hours per day Ð time which was not spent investigating worrisome new infections or rising infection rates.

Consistent interpretation of the CDC's NNIS criteria is essential for detecting true changes in long-term historical infection rates. The NNIS defintions require over 100 pages of detailed descriptions to ensure this consistence. We have documented that specialists without expert system support do not apply the NNIS criteria uniformly. GermWatcher enforces the consistent application of NNIS definitions so that comparable infection rates are obtained.

Since its February 1992 deployment at Barnes Hospital, GermWatcher automates the task of reviewing microbiology culture results and applying the NNIS definitions. GermAlert automates an early-alerting system for "infectious emergencies" Ð infections which require immediate intervention due to their extremely contagious nature or due to their risk to hospital personnel or other patients. Suspect detects organisms that are highly resistant to usual antibiotics, which then must be treated with very potent "last-resort" antibiotics or complex multi-drug therapies. Reportable Diseases Monitor automates the mandated notification of reportable infectious diseases to the Missouri Department of Health. Collectively, the GermWatcher family is a multi- faceted set of tools which support a multi-disciplinary attack on a deadly complication of hospitalization.

In July 1995, the GermWatcher systems expanded to Jewish Hospital (now called Barnes-Jewish North). In the coming year, the programs will be extended to two additional large non-academic community medical centers. Our goal is to make this advanced technology accessible to smaller medical centers as well as large academic sites to reduce the spread of life-threatening infections in all clinical settings.

 How did information technology make this application possible? Describe any new technology used and/or cite innovative uses of existing technology. For example, did you adapt an old software program to solve a new problem? Or did you define a problem and develop new technology to solve it?

The GermWatcher family of expert systems leverages a number of information technologies:

• The use of internetworking standards to share data between mainframe databases and Unix-based systems.
• The coupling of a commercial relational database system with a widely available expert system shell developed by the National Aeronautics and Space Administration (NASA). We have released the source code for this expert system / database coupling to the public domain and have received correspondence acknowledging the use of this code throughout the world.
• The reuse of the same expert system shell, called the GermWatcher Engine, for all expert systems in the GermWatcher family. Extensive software reuse has enabled the rapid development of subsequent members of the GermWatcher expert systems.
• The use of familar graphical user interfaces to hide sophisticated information technologies from nontechnology-literate end-users. From the end-user's perspective, the GermWatcher family appears as a set of simple PC-based programs. The existence of a large Unix-based server which executes multiple expert systems and houses a complex relational database system is completely hidden.

A key challenge has been to design the GermWatcher system so that these expert systems can be made available to other medical centers. Standards-based designs and the use of dictionary tables has resulted in a separation of hospital-specific naming and local formatting conventions from national coding standards and expert system rules. Because of this design, we have reimplemented GermWatcher in a second academic center which has a completely different hospital information system infrastructure. Based in this initial successful port, we are expanding the technology into two additional community-based medical centers.

 Is your application unique? Is it original? How? Is it the first, the only, the best or most effective application of its kind? How did your application evolve? What is its background?

The GermWatcher project was conceived during a series of inter-disciplinary meetings between the Division of Medical Informatics, the Division of Infectious Diseases, and the Barnes Hospital Quality Assessment Department. The first expert system, GermWatcher, was created to unburden the Infection Control specialists from the tedious yet critical task of manual review of positive microbiology culture results. The project was of particular interest because the success of the reasoning system was based on nationally recognized and validated explicit definitions promoted by the U.S. Centers for Disease Control and Prevention. Because of the existence of the CDC rules, GermWatcher has applicability beyond a single medical institution.

The GermWatcher concept was first described and implemented in a system developed by Scott Evans, David Classen and others from the Latter Day Saints (LDS) Hospital at the University of Utah. Our contribution has been to develop the GermWatcher family of tools using readily available commercial products and public-domain technologies rather than the self-developed, nonportable technology present at LDS Hospital. In addition, the GermWatcher family embodies a number of unique features:

• A simple yet effective method for sharing data between a public-domain expert system shell's internal memory and a commercial relational database system. We have released the code for this linkage into the public domain and have received positive feedback from investigators around the world.
• A generalized expert system shell, called the GermWatcher Engine, used to implement all of the expert systems in the GermWatcher family. Thus, the GermWatcher Engine represents a successful example of substantial software code reuse to solve different but closely related tasks in Infection Control. Most recently, we have demonstrated the versatility of the GermWatcher Engine by using it to develop a set of expert systems outside the domain of Infection Control Ð the monitoring of medications that require complex dosing adjustments based on patient-specific characteristics such as age, sex, and kidney function.

 Has your application achieved or exceeded its goals? It it fully operational? How many people benefit from it? Describe future plans for the project.

We have assessed the success of the GermWatcher family using a number of direct and indirect metrics:

• Formal blinded evaluations of the GermWatcher program have demonstrated that the program's accuracy, relative to a reference standard, exceeds the performance of individual Infection Control specialists.
• As a by-product of the GermWatcher evaluations, we have demonstrated that the unassisted performance of the Infection Control specialists has improved substantially. The specialists attribute this improvement to their daily use of GermWatcher, which provides constant educational exposure to the CDC's NNIS criteria.
• The Infection Control specialists at Jewish Hospital actively pursued the deployment of the GermWatcher family in their hospital, even though the program was developed and deployed at another hospital.

Barnes and Jewish Hospitals have now merged into a single integrated hospital. The combined Infection Control department has experienced substantial turnover and downsizing. Objective studies have shown that the GermWatcher systems have enabled fewer specialists to provide Infection Control surveillance services to more patients without decreasing accuracy or quality of service. This outcome provides the strongest evidence that the GermWatcher systems have markedly improved the efficiency and effectiveness of the combined Barnes-Jewish Infection Control Department.

Future plans for the GermWatcher family include:

• Expansion into two large community-based hospitals in St. Louis where Infection Control services have requested the functionality.
• 
  The addition of data visualization capabilities, using the underlying relational database infrastructure. Infections have both a temporal and a spatial component Ð it takes time for a patient to manifest an infection which he/she may have acquired while in a different location in the hospital. We are creating 3-D data animations to visualize outbreaks as a clustering of infections that appear in the same location or at the same time during the data animation.

 What were the most important obstacles that had to be overcome in order for your work to be successful? Technical problems? Resources? Expertise? Organizational difficulties?

The GermWatcher project has been successful because an interdisciplinary group of dedicated innovative professionals looked beyond organizational boundaries to eliminate obstacles. The most significant barrier early in the GermWatcher project was the transfer of patient-specific data from hospital mainframe computers to a university database server. Issues of patient confidentiality and data ownership required cross-institutional trust and rigid adherence to policies, procedures, and safeguards. Technology provided solutions to implement data security policies. Individual commitment to excellence and professional conduct provided the leadership to overcome organizational concerns.

The existence of an interdisciplinary team which encompasses Computer Science knowledge in expert systems and relational database systems, Infection Control knowledge of surveillance and outbreak investigations, and Infectious Diseases knowledge of hospital and clinical epidemiology resulted in the creation of a set of systems which delivers high quality expert-level clinical decision support.

Exhibits

• The Opponent - Hospital Infections
• The GermWatcher Family
• The Defense - Advanced Technology
• The Advantage - Containment / Control
• The GermWatcher Team - 1995

For More Information:

Contact Thomas Bailey, MD
Washington University School of Medicine
314-454-8321
tbailey@im.wustl.edu