Computer Decision Support for Antibiotic Prescribing
February 15, 2002 11 minutes read
Computer Decision Support for Antibiotic Prescribing
Conference Coverage
Synopsis: This meeting, held in Deer Valley, Utah, last summer, was organized by Matthew Samore (University of Utah) and Neil Fishman (University of Pennsylvania) to discuss the status and future of computer applications for decision making in antibiotic use. — Joseph F. John, Jr., MD, and Alan Tice, MD, FACP
Overview
Neil Fishman reviewed the background of antibiotic control, management and, eventually, stewardship. The current program at the University of Pennsylvania Hospital features a Handbook of Antibiotic Therapy with a rapid web site (www.med.upenn.edu/bugdrug). The Fishman program at HUP projects to realize huge savings for the hospital at a fraction of the cost to run the program.
Matt Samore reviewed where we are with respect to computer decision support systems. He raised several questions: Can the computer represent group knowledge? Do we need computerized order entry? Can computer systems capture the art of medicine? Why are there not more successful computerized programs of antibiotic use?
The Problems with Antibiotic Use
Julie Gerberding, who outlined the need for change and measures of antibiotic use, reviewed the problems of antibiotic use and the plan that the Centers for Disease Control and Prevention is planning to put into place. These include improving information systems, limiting antibiotic use, shortening duration of therapy, narrowing spectrum, surveillance of antimicrobial resistance, and reducing use of devices that predispose to infection. The CDC program should be ready for marketing to the physician community and the public soon, but most of the elements are presently available on the web at www.cdc.gov/drugresistance.
Alan Tice pointed out the need for changes in antimicrobial use in the outpatient setting, as most antibiotics are used outside the hospital, and the potential reservoir for resistant organisms is far greater. The control over antibiotic use is much more difficult in ambulatory care than in the hospital but there are already remarkable systems in place that profile and monitor even individual prescribing for the pharmaceutical industry. The potential for gathering more information about adverse effects of antimicrobial therapy through the OPAT Outcomes Registry was also noted with now more than 10,000 cases available for analysis.
Victor Yu presented the problems he has faced with overuse of antibiotics in the intensive care units and the success he has had with dramatically reducing antibiotic use through tracking outcomes of low-risk patients randomized to a 3-day course of 1 antibiotic vs. the usual multiple drug regimen for an indefinite period.1
Computer Decision Support Systems
John Bartlett noted he had evaluated computer support systems through focus groups of primary care physicians who do most of the prescribing. The consensus seemed to be that for an information system to be helpful and practical, it must have validity and credibility as well as being fast (20 seconds or less), unbiased, easy to navigate, and free. It is a tough challenge.
A variety of computer programs have been developed for support in antimicrobial prescribing. These include the Latter Day Saints Hospital program reported by John Burke, David Klassen, and Stanley Pestotnik. That program is still in place and functioning in the Intermountain Health Care system. Some of the initial investigators have also developed the TheraDoc Antibiotic Assistant program which gathers and integrates patient-specific information (age, weight, history of allergies, other diseases), laboratory information (culture results, renal function, other laboratory values) and integrates them to assist with specific recommendations for antibiotics and dosing. It is being developed as part of integrated packages at hospitals. Stan Pestotnik also pointed out that the most valuable tool is one that can be used at the time of decision making, and potentially at the bedside.
John Bartlett reported the Johns Hopkins ABX program was developed as a disease-specific model with recommendations for infections and pathogens but does not provide patient-specific information. It is available at no cost through the Internet. More than 17,000 people have already downloaded the program—most within the last 2 months. An HIV decision-making module is in the works.
Dale Gerding reported icanMD is also developing programs for antibiotic decision making. The focus will be on quick and simple answers for clinical problems with the ability to "drill down" if more information is needed and time is available. There will be a basic 1-page format for comparing antibiotics that will include evidence-based medicine ratings, toxicity, and cost factors. He noted the great efforts and cost involved in developing such programs and keeping them updated through a variety of experts and advisors. Sources of funding are also a problem as most of the money available is through pharmaceutical companies, who may not have the same objectives and goals as physicians.
Victor Yu also presented information on the antibiotic information system he is developing for the internet. It will bring a remarkable depth of references and information to one’s fingertips and be continually updated. It will also have a search mechanism with the ability to focus on the specific question or problems far more easily than Medline or any other information system available. It may well replace textbooks because of their inflexibility.
Therapeutics and Guidelines
Scott Fridkin of the CDC reviewed the essential elements for surveillance and decision support. Resistance varies by the type of ICU. Several surveillance systems in the United States collect and publish useful data, both at the hospital level and regional level. The ABC system example for invasive pneumococcal disease can be found in the literature.2,3 The audience raised a question of information overload and how to deal with it.
David Schwartz at Cook County/Rush Medical Center in Chicago reported data from some of the work emerging from the Chicago Antimicrobial Resistance Project (CARP). Dr. Schwartz and his contingent are proving that antibiotic usage can be tracked on a PC. Still, for example, costs to assemble data, on prescription errors, are high and could not be done in Cook County/Chicago example without grant support.
Ron Polk at the Medical College of Virginia discussed ways to rank recommendations and guidelines based on the level of evidence. One problem is that infectious diseases physicians are more comfortable assuming similar outcomes for different antimicrobials. We have not yet moved to using pharmacokinetic/pharmacodynamic parameters as generally reliable predictors of clinical response. Nevertheless, this area needs much more study to determine the value of measuring outcomes based on antibiotic choice.
Joseph John and Philip Mathew were charged with discussing how to measure quality of antimicrobial therapy. They proposed a model whose outcome measures include efficacy (cure, improvement), clinical response (mortality, morbidity), adverse reactions, costs, and antimicrobial resistance. The challenge remains to validate such an approach.
Hand Held Applications
Hand held computers were noted to offer tremendous opportunities for data collection as well as decision-making. Many physicians and most house staff already own Palm Pilots or comparable devices with or without wireless web access. There are not only a variety of new Palm device models available that are smarter and prettier than ever before, but there is increasing competition from Compaq (ipaq), Handspring (Visor), HP (Journada), and others. It is also theorized that hand helds will actually get physicians back to the bedside so they can interface more with patients.
There are already free programs available through the internet that provide antibiotic information and advice for treating specific infections and pathogens. These include the qID program from ePocractes and the ABX program from Johns Hopkins. There are rumors a Sanford edition will also be available in a hand held program but the booklet remains the gold standard for others to be compared with in terms of access, speed of access, and format.
Outcomes Measures and Control
Marc Lipstich tried to answer the question of whether we should try to maximize diversity of antibiotic selection or use standardized selections. Lipstich uses mathematical models to predict the success or failure of a given strategy but he emphasizes that the first issue is deciding upon the aim of antibiotic policy. For example, a medical center may choose to wait until resistant organisms emerge before changing to a different regimen, a policy different from one that attempts to minimize the use of empiric therapy with ineffective antibiotics. Mathematical models can be applied to these situations as well as to antibiotic cycling. Statistical analysis of the theoretical likelihood of bacteria developing resistance on antibiotic therapy suggests it is more likely with sequential switching rather than random use and that the least chance is with combination therapy.
While the published considerations consider development of resistance to be independent, there may well be useful applications of modeling in antibiotic use that will be useful.4
Then Terry Clemmer at LDS Hospital discussed how computers could replace paper, but emphasized that a trusting environment was needed in hospitals. He emphasized that the field needed more evidence since only 15% of the approach used at LDS was based on good evidence. He had 3 provocative propositions:
- Standardize at the local level: one size does not fit all;
- Develop protocols to build a collaborative environment to bring change to a local setting;
- Deliver protocols to unleash creativity and novelty on the front lines.
John Jerrigan at the CDC gave a compassionate view of how hard it is to prioritize antibiotic resistance to the clinical outcome of individual patients. Physicians remain under extreme pressure to use antibiotics, sensing a narrow margin for therapeutic error.
Charles Mullet gave a pediatric view of successfully implementing a computerized antibiotic support program. There is a need to know literally what the work flow is like on a ward to understand the process of computerized order entry. If a system is to succeed in the pediatric setting, many tactics have to be changed from the adult approach.
Charles Huskinds from the Mayo Clinic raised the issue of whether clinical trials can answer some of the questions about computerized support of antibiotic use. Perhaps this can be done with specific issues like the value of antibiotic cycling, however, Huskinds felt that randomized, controlled trials (RCTs) are preferable, but not mandatory, and are best used when the magnitude of the effect is likely to be small and when confounding becomes a major problem. RCTs suffer from the high cost of performing such studies.
Stephen Harbarth made the final presentation on whether antibiotic data assessed at the group level needs the addition of individual subject level measurements. For instance, in the Lancet data were recently presented with industry’s input regarding antibiotic use in Europe.5 From 1993 to 1997 there was not much change. Group analysis can show that improved antibiotic use may reduce resistance and costs and, perhaps, improve patient outcome. What antibiotics should not do is preclude the physician from entertaining a differential diagnosis. In order for such data to have application to the individual patient, there seems to be more need for validation of process for, as David Schwartz emphasized as commentary to the presentation, "Antibiotics are used one patient at a time."
Surveillance Systems
Trish Perl presented her experience with gathering surveillance information for infection control on a Palm Pilot with a program that can automatically download lab and microbiology information. She found that nurses could find relevant data far more efficiently than going through charts, could record information at the bedside, and spent only 1.5 hours in the average ICU with the Palm Pilot compared with their usual of 8 hours per week. The data collected can also be more easily stored and analyzed than with prior methods.
Comment by Joseph F. John, MD, & Alan D. Tice, MD, FACP
This meeting comes at a time when, in the current medical environment, there are multiple incentives to apply computers. Not only is there an explosion of medical information, but appropriate treatments are becoming increasingly complex. The Institute of Medicine has recently highlighted the potential for inappropriate therapy—one of many "medical error" classes. While "to err is human," there is, nevertheless, an increasingly fertile field for attorneys looking for medical errors. What is attractive is the concept that computers can bring organization and even safety to one specific area of prescribing where it is obviously needed. Computers in medicine offer true solutions to complex problems, but they also engender fear, particularly in those hospitals that appreciate the theoretical value but lack the magnitude of dollars to implement computer systems. The fact is that many, if not most, hospitals still do not have interactive computer systems among billing, laboratory, radiology, administration, and patient care. Aside from trying to practice medicine in this Tower of Babel, it is a good time for physicians to find ways to use computers and seek cost-effective solutions to some of our most challenging problems.
One participant in the conference, Stan Pestotnik, commented that medicine is "drowning in data yet starving for information." Tools will be needed to measure and manage, but there is hope with the incentives of the IOM and the nearly $1 billion set aside for research funding by the department of Health and Human Services. Although it was disappointing to hear how little progress has actually been made in solving some of the riddles of antibiotic use, it was obvious there has been progress and that there are a lot of creative minds working on the problems.
Much of the present efforts seem to be directed toward improving compliance with guidelines although we are not sure what they should be. While education and compliance are important, additional efforts could focus on true outcomes with indicators that can evaluate appropriate antibiotic use and provide feedback to physicians on an ongoing basis. The evolving systems have the potential of gathering data with courses of treatment, tracking standardized outcomes indicators, and continually updating a database that can be used to compare results and help in therapeutic decisions. This could provide a way of learning through experience similar to what physicians do, although in a more structured format.
Adverse drug effects were not emphasized at the conference yet remain an important aspect of decision making with antibiotics. Toxicity in many cases becomes as good a reason as antibiotic resistance to limit inappropriate usage. Little has been reported about antibiotic selection based on adverse effects although it is an integral part of a risk-benefit decision that must be made with each order. More information about antibiotic side effects can easily be gleaned from electronic medical records and computer applications as well.
It was interesting to note the progress and effect the pharmaceutical industry has on antibiotic prescribing. Not only are their marketing efforts far more effective than any academic or governmental response so far, but they also have been applying computers to antibiotic prescribing for years and have developed sophisticated systems that academic antibiotic use monitors can only dream of. Obviously a cooperative relationship would be best, but the goals and incentives differ. To what extent physicians will be able to match their systems or even channel them toward the public interest remains to be determined, but the track record is not good and the time to respond is now.
Although much of the information presented was about influencing physician-prescribing behavior, most of it was in a negative way with formulary restrictions, obstacles to ordering, and requiring permission to use.
It seemed clear that whatever computer applications are adopted, they will need to include local factors such as the antibiotic susceptibility pattern in the community, the season of the year, the location of the patient, and local formularies. There is also the factor of local resources, which may not be able to support a computer system or integrate results. The only widely applied computerized system—one that works amazingly well—is sponsored and supported by the government in our Department of Veterans Affairs Medical Centers.
Future meetings of the group and/or a conference on Computer Decision Support for Antibiotic Decision Making were discussed—as well as workshops or symposia at IDSA, ICAAC, and possibly other conferences.
References
1. Singh N, et al. Short-course empiric therapy for patients with pulmonary infiltrates in the intensive care unit. Am J Respir Crit Care Med. 2000;162: 505-511.
2. MMWR. Morb Mortal Wkly Rep. 2001;44:656-661.
3. N Engl J Med. 2000;343:917-924.
4. Bonhoeffer S, et al. Evaluating treatment protocols to prevent antibiotic resistance. Proc Nat Acad Sci. 1997;94:12106-12111.
5. Cars O, et al. Lancet. 2001;357:1851-1853.
Dr. John, Professor of Medicine and Microbiology, University of Medicine & Dentistry—New Jersey, Robert Wood-Johnson Medical School, is Co-Editor of Infectious Disease Alert. Dr. Tice, Infections Limited, PS, Tacoma, WA; Infectious Disease Consultant, John A. Burns School of Medicine, University of Hawaii, Honolulu, HI, is Associate Editor of Infectious Disease Alert.
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content