- Email: [email protected]
The HELP system and its application to infection control
jSx~rna1 of Hospital Infection
The HELP
system
J. P. Burke,
Division
(1991)
18 (Supplement
A), 424431
and its application control
D. C. Classen, S. L. Pestotnik, -L. E. Stevens
to infection
R. S. Evans
and
of Infectious Diseases, LDS Hospital and University of Utah School of Medicine, Salt Lake City, Utah 84143, USA Summary: The HELP system is a comprehensive hospital information system that is linked to an allied financial data base. The clinical data base integrates information from areas such as admitting, pharmacy, radiology, surgery, pathology, nursing, respiratory therapy, and the clinical laboratories, including microbiology. This allows for the creation of an electronic medical record that contains all the clinical and financial data for each patient. The HELP system combines both communication and advice features through the use of data- and time-driven algorithms. We have used the HELP system to automate the surveillance and analysis of hospital-acquired infections and to identify patients at high risk for nosocomial infection. The expert system features have also been used to suggest alternatives for patients receiving inappropriate antimicrobial therapy, to improve the timing of antibiotic prophylaxis in surgery, and to curtail unnecessarily prolonged prophylaxis. Automated hospital information systems such as HELP can facilitate the investigation of a broad range of infection control, quality improvement, and cost-containment issues. Keywords:
Computers;
expert
systems;
infection
control;
surveillance.
Introduction Hospital infection control units on both sides of the Atlantic have made extensive use of computers to analyse data collected by surveillance personnel.’ In the future, linkage of desk-top personal computers, laboratory minicomputers and hospital mainframe computers will be commonplace. The improved communication capability between computers in hospitals has obvious potential benefits to improve the efficiency of the surveillance process as well as to prompt the application of infection control measures.2 Computerized hospital information systems are still in their infancy. The medical information systems that are now emerging can be grouped into two categories: ‘communication’ systems that store, retrieve and transmit information, and ‘advice’ systems that can supply stored information to the solution of clinical problems. 3 HELP (an acronym for Health Evaluation Correspondence to: Dr John P. Burke, Division Street, Salt Lake City, Utah 84143, USA. 01956701/91/06A424+08
of Infectious
Diseases, LDS Hospital, Q 1991 The Hospital
$03.00/O
424
8th Ave. and C Infection
Society
The HELP hospital
information
system
425
through Logical Processing) is one of the first integrated hospital information systems that combines the functions of both communication and advice systems. The HELP system was designed to meet the clinical, of hospitals and to provide administrative and research needs decision-support capability. 4 This computer system has been developed over the past 20 years at the 1LDS Hospital in Salt Lake City, Utah, USA. The purpose of this paper is to describe some of the infection-control applications of the HELP system over the past two decades and, further, to emphasize that these applications were made possible in a stepwise fashion as the HELP system evolved gradually into a comprehensive automated hospital data base. An historical approach will be used to describe applications that were enabled by existing data elements at each stage in the process of bringing various h.ospital services online. It should be clear at thle outset that the development of such an information system represents an extensive long-term commitment by the institution. It should become clear that, as comprehensive and sophisticated as it is, the system is sti.11 incomplete and that the most exciting epidemiological applications remain in the future. Indeed, one of the most desirable attributes of the HELP system is its flexibility, improvability and receptivity to new technology and advances in medical science itself. The existing HELP system may be viewed, in one sense, as a prototype that holds important lessons for dlesigners of future medical computer systems. Features
of the HELP
system
The initial application of the IHELP system was in the intensive care units of LDS Hospital to provide haemodynamic and electrocardiographic monitoring and to capture data from the clinical laboratory. Warner and his associates used a Control Data Corporation 3300 computer in these early efforts and adopted design criteria that included the need for modularity to accommodate an ever-expanding data base and the need for data-driven processing of medical decision algorithms in order to minimize the data entry required by medical personnel. By 1978 the system had been installed hospital-wide with terminals in every nursing unit. From 1979 to 1981 the system was re-developed in order to better meet the administrative needs of the hospital and to prevent downtime. The current Tandem system has 10 central processing units (CPUs) with seven gigabytes of online disk storage, more than 500 terminals, and 100 laser printers. The system is interfaced with minicomputers that assist in collecting and reviewing data from the intensive care units, the clinical laboratory, the heart catheterization laboratory, and the pulmonary function laboratory. The HELP system now creates a computerized medical record for each patient that contains coded data in a hierarchical structure rather than in free text. All patient data are stored permanently in two formats: a
426
J. P. Burke
et al.
long-term abstract of coded demographic and diagnostic information that is kept on-line, and a comprehensive collection of all data on removable disk packs, which is available for clinical and research purposes. The interfaces with other computers in the hospital allow for the automatic storage of test results from the laboratory computer system and of financial information that includes cost, charge and reimbursement data. Information from the HELP system can also be downloaded to personal computers and accessed by modems from physicians’ offices and homes to generate daily or 7-day ‘rounds reports’. In addition to this integrated data base, one of the most distinctive features of HELP is its decision-support function that has been strengthened and expanded at each stage in its growth. Simple algorithms and more complicated decision trees are used to generate patient-specific advice or ‘alerts’. These algorithms can be modified or added to as needed, and the decisions themselves become a permanent part of the data base. This knowledge base of programmed medical logic is automatically activated when certain key information is entered into the mainframe computer (data-driven) or specific modules can be set to run at a certain time of day (time-driven). The great appeal of such a system is its ability to give assistance to clinicians without requiring laborious data entry by the themselves’ and the fact that protocol-generated physicians recommendations can reduce the rate of medical errors due to the intrinsic limits of physicians.6 Infection
control
applications
Urinary catheter studies, 1970-75 The HELP system permitted a growing number of infection-control applications as the system became more fully developed and comprehensive. The first applications were in the form of research projects in patients with indwelling urinary catheters. Because microbiology test results were not yet available to the HELP system, we began a monitoring programme that required the collection of a catheter urine specimen each day from every patient with an indwelling catheter. The results of these urine cultures were entered into the mainframe computer and became the basis for a series of protocol-generated randomized controlled trials of urinary catheter equipment and catheter care practices.‘-i4 The computer was used in these studies to capture routine clinical data, to direct the collection of appropriate bedside observations by study personnel using algorithms, and to randomize the patients to various interventions. In addition, the urinary catheter monitoring became an automated tool for the surveillance of nosocomial urinary tract infections, relieving the infection control practitioners of this task and creating an archive for research issues. One of the intriguing aspects of these data derives from the more accurate classification of hospital-acquired urinary infections that was
The
HELP
hospital
information
system
427
achieved; for example, nearly one-half of the nosocomial urinary infections discovered by traditional surveillance were shown to represent bacteriuria present at the time of catheter placement, and a nearly equal number of ‘missed’ cases were uncovered by the intensive daily culturing.15 Antibiotic monitoring, 1975-80 In 1975 the hospital pharmacy was linked to the HELP system and a computerized monitoring service became clinically useful as algorithms were designed to alert and warn of potential adverse drug reactions and interactions.” Because of the broad nature of the data base even at this early stage in its development, the drug monitoring service was able to consider patient allergies, concomitant drugs likely to cause interactions, renal and hepatic dysfunction likely to affect drug excretion and clearance, and underlying diseases associated with specific drug reactions. Antibiotics accounted for 40% of all the drug warnings and therefore received special attention. We developed algorithms, for example, to identify patients who were receiving potentially nephrotoxic antibiotics and who were not being appropriately monitored with tests of renal function and patients who were receiving inappropriate dosing of antibiotics based on their renal or hepatic function. We observed that physicians complied with more than 90% of these computer-generated reminders, in part, because of the enhanced role of the pharmacists, who assessed each alert and notified the physicians directly. Also, we observed an educational effect of these reminders, with improved drug ordering and monitoring and an associated decline in the freq.uency of certain alerts.*’ Computerized infectious disea.se monitor, 1980-85 In the early 1980s microbiology test results were added to the system through an interface with thle laboratory computer system. Microbiology data are now integrated with medical information from other clinical areas, importantly with admission-discharge-transfer (ADT) data, most pharmacy, radiology, surgery and respiratory therapy. With these advances, the elements were in place to develop a computerized infectious disease monitoring system.18 We created a knowledge base of algorithms that is automatically activated when specific microbiology data are entered into a patient’s computer file, and the decisions are stored until needed. The computerized infectious disease monitor has enabled us to automate the surveillance of nosocomial infections. This surveillance mechanism has some of the same limitations that apply to any laboratory-based system; non-cultured infections, infections at some sites and viral infections are missed.” However, we have shown that infection control practitioners found more hospital-acquired infections using computer screening when compared with traditional surveillance methods while requiring only 35% of the time.” More importantly, false-positive diagnoses of
428
J. P. Burke
et al.
hospital-acquired infections were made with nearly the same frequency by the computer and by the infection control practitioners. Computer screening has several other advantages over traditional retrospective chart reviews. This automated surveillance results in more timely alerting of infection control staff to patients likely to have a hospital-acquired infection. For example, clusters of infection can be more rapidly detected by threshold analysis to permit prompt application of control measures, and downloading of data to personal computers facilitates outbreak investigations and statistical analyses. Methods are now being developed to activate computer algorithms by sources of information in addition to microbiology test results. Quality improvement, 1985-90 The computerized infectious disease monitor enables other clinical applications in addition to screening for nosocomial infections. An immediately useful monitoring involves linking antibiotic susceptibility reports with antibiotic prescriptions. When inconsistencies occur between antibiotic therapy and in-vitro susceptibility data, computer algorithms generate alerts, which are assessed by a clinical pharmacist who then notifies attending physicians. In nearly one-half of such alerts, physicians were simply not yet aware of the test results, but a variety of other circumstances has been appreciated. We have presented results that confirm the value of this method to identify and correct errors in antimicrobial prescribing and to assure the appropriate use of therapeutic antibiotics.21 These methods can also be used to further improve antibiotic therapy through algorithms that use susceptibility test results and a programmed knowledge base to recommend less costly or less toxic antibiotics than those actually prescribed. Prophylactic antibiotic use can also be defined and improved using the interfaces between the surgery operating room schedule, the pharmacy and the microbiology laboratory. The elements essential to the computer decision analysis with respect to surgical prophylaxis include the patient’s name and location, the surgeon and planned surgical procedure, the patient’s drug allergy history and drug use during hospitalization, and a list of all surgical procedures with an assessment of the clinical utility of perioperative antibiotic prophylaxis. In a 2-year study we were able to establish the magnitude of misapplication of perioperative prophylaxis and to develop computer-generated reminders to improve usage. These reminders were associated with improved timing of antibiotic use and with a concurrent decline in postoperative wound infection rates.” We also identified patients receiving antibiotic prophylaxis for prolonged durations, and pharmacists used the alerts to place ‘stop orders’ with the consent of surgeons. This programme was shown to reduce the average charges for antibiotic prophylaxis in surgery by US $42 without compromising patient care.23
The HELP hospital
information
system
429
The HELP system has also been linked to the IBM-based financial computer system that contains cost, charge and reimbursement data for all patients. This includes a sophisticated comprehensive microcosting system for all aspects of patient care. The data base can now be used to address a wide variety of issues, such as the effect of various strategies for antibiotic selection on total hospital costs, on rates of hospital-acquired infections, and on the use of other hospital services.
The
future
The computer developments that will allow future applications are well under way. First, bedside terminals have been installed in patient rooms and a computer-based nurse charting system is fully functional in several nursing care units and all intensive care units. This new charting system includes bedside entry of vital signs, measurements such as fluid intake and output, daily weights, and oblservations such as the presence of a rash or diarrhoea. With this new source of data integrated with the system, an adverse drug event monitoring program has been developed. Patients likely to be experiencing an adverse drug event are ‘flagged’ based on new orders for drugs likely to be prescribed to treat adverse drug reactions (e.g. anticonvulsants, antipruritics, antidiarrhoeals), sudden termination or change in dosage of a drug, laboratory abnormalities, or nurse charting of a rash, seizure, diarrhoea, etc. A. clinical pharmacist then uses these signals to investigate and to assign a probability score for adverse drug events based on standard algorithms.24 In the US, involvement of infection control with expanded programmes of hospital epidemiology that include quality assessment and pharmacoepidemiology appears to be growing.25 Infection control units already have the responsibility for monitoring antibiotic use, and automated systems, such as the one we have developed using HELP, can fulfil the broader needs of epidemiologists. The second ongoing development that has important implications for the future is the installation of HELP in other hospitals owned by Intermountain Health Care (IHC), the non-profit corporate parent of LDS Hospital. IHC owns and operates 23 hospitals in the intermountain region of the US and has begun the installation of HELP in all of these hospitals. There is growing recognition of the value of a large automated data base drawn from a regional network of hospitals. In infection control there will be increased needs to stratify patients by various risk factors, severity-of-illness scores, and underlying illnesses. HELP has already been used to develop a logistic regression model for predicting hospital-acquired infections,26 and the installation of HELP in other hospitals will provide a method to validate these predictors and to evaluate prevention strategies.
J. P. Burke
430
et al.
Discussion
Shortliffe’ has commented on the paradox of the persisting notion that the application of computers to medical practice has had limited success despite ample evidence to the contrary. While it is true that attempts to use the computer in medical diagnosis have been unsatisfactory, there is a steady increase in the use of computers as clinical decision-support tools. At the LDS Hospital, physicians have enthusiastically embraced the computer and have found the HELP system to be essential for patient care. The HELP system has replaced the paper medical record except for physicians’ notes. The success of the system seems to be based, in part, on data entry at its source by the appropriate clerical, nursing or laboratory personnel. In some instances, as in the laboratory, data entry is entirely automated. The automated use of a knowledge base of algorithms to analyse patient data means that health care professionals are not required to enter data or request evaluation of the data to receive decisions. The fear that computers may replace physicians seems quaint in retrospect. Computers will not replace epidemiologists either, but they can and will assist in epidemiological decision making. Data should not be collected to serve only one purpose, and hospital epidemiologists are embracing broader areas of concern in quality assessment and cost-containment concurrently with the rapid growth in automated hospital data bases. These systems will become commonplace and will offer greatly expanded applications to hospital epidemiology. For the full potential to be realized, microbiologists, pathologists and epidemiologists must be involved in planning these systems and make the system developers aware of their special needs. Moreover, the development of an automated hospital data base takes place incrementally and requires many parallel changes in the daily operation of the institution. The experience of the infection control unit at the LDS Hospital illustrates how new applications can be incorporated as the system grows. This
work
was supported,
in part, by U.S.P.H.S.
grants
HS 06028 and AI 15655.
References and use of computers in hospital infection control. 1. Wenzel RP, Streed SA. Surveillance J Hosp Infect 1989; 13: 217-229. WR. Information, computers and infection control. J Hasp Infect 1990; 15: 2. Gransden l-5. 3. Rennels GD, Shortliffe EH. Advanced computing for medicine. Sci Am 1987; 257: 154-161. RM, Clayton PD, Warner HR. The HELP system. r Med Syst 4. Pryor TA, Gardner 1983; 7: 87-102. EH. Computer programs to support clinical decision making. JAMA 1987; 5. Shortliffe 258: 61-66. 6. McDonald CJ. Protocol-based computer reminders, the quality of care and the non-perfectibility of man. N EnglJ Med 1976; 295: 1351-1355. RA, Burke JP, Dickman ML, Smith CB. Factors predisposing to bacteriuria 7. Garibaldi during indwelling urethral catheterization. N Engl J Med 1974; 291: 215-219.
The
HELP
hospital
information
system
431
8. Britt MR, Garibaldi RA, Miller WA, Hebertson RM, Burke JP. Antimicrobial prophylaxis for catheter-associated bacteriuria. Antimicrob Agents Chemother 1977; 11: 240-243. 9. Garibaldi RA, Burke JP, Britt MR, Miller WA, Smith CB. Meatal colonization and catheter-associated bacteriuria. N Engl J Med 1980; 303: 3 166318. 10. Burke JP, Garibaldi RA, Britt :MR, Jacobson JA, Conti M, Alling DW. Prevention of catheter-associated urinary tract infections: Efficacy of daily meatal care regimens. AmJ Med 1981; 70: 655-658. 11. Jacobson JA, Burke JP, Kasworm E. Effect of bacteriologic monitoring of urinary catheters on recognition and treatment of hospital-acquired urinary tract infections. Infect Control 1981; 2: 227-232. 12. Burke JP, Jacobson JA, Garibaldi RA, Conti MT, Alling DW. Evaluation of daily meatal care with poly-antibiotic ointment in prevention of urinary catheter-associated bacteriuria. J Ural 1983; 129: 331-334. 13. Burke JP, Larsen RA, Stevens LE. Nosocomial bacteriuria: Estimating the potential for prevention by closed sterile urinary drainage. Infect Control 1986; 7 (Suppl): 96699. 14. Larsen RA, Burke JP. ThLe epidemiology and risk factors for nosocomial catheter-associated bacteriuria caused by coagulase-negative staphylococci. Infect Control 1986; 7: 212-215. 15. Britt MR, Burke JP, Epstein B, Mooney B, Krall B, Miller WA. Monitoring of nosocomial urinary tract infection: A comparison of methods for surveillance of nosocomial urinary tract infection. Abstract 422. Program and Abstracts of the 17th Interscience Conference on Antimicrobial Agents and Chemotherapy, New York, 12-14 October 1977. 16. Hulse RK, Clark SJ, Jackson CJ, Warner HR, Gardner RM. Computerized medication monitoring system. Am J Hasp Pharm 1976; 33: 1061-1064. 17. Burke JP. The influence of computerized antibiotic monitoring on physician prescribing patterns. Symposium: Influencing Physician Usage of Antimicrobials. Program and Abstracts of the 17th Interscience Conference on Antimicrobial Agents and Chemotherapy, New York, 12-14 October 1977. 18. Evans RS, Gardner RM, Bush AR et al. Development of a computerized infectious disease monitor. Comput Biomed Res 1985; 18: 103-113. 19. Nystrom B. Surveillance of hospital-associated infections. Infection 1989; 17: 43-45. 20. Evans RS, Larsen RA, Burke JP et al. Computer surveillance of hospital-acquired infections and antibiotic use. JAMA 1986; 256: 1007-l 011. 21. Pestotnik SL, Evans RS, Burke JP, Gardner RM, Classen DC. Therapeutic antibiotic monitoring: Surveillance using a computerized expert system. Am J Med 1990; 88: 4348. 22. Larsen RA, Evans RS, Burke JP, Pestotnik SL, Gardner RM, Classen DC. Improved perioperative antibiotic use and reduced surgical wound infections through use of computer decision analysis. Infect Control Hosp Epidemiol 1989; 10: 316-320. 23. Evans RS, Pestotnik SL, Burke JP, Gardner RM, Larsen RA, Classen DC. Reducing the duration of prophylactic antibiotic use through computer monitoring of surgical patients. DICP, Annals of Pharmacotherapy 1990; 24: 351-354. 24. Classen DC, Pestotnik SL, Evans RS, Stevens LE, Bass SB, Burke JP. Concurrent monitoring of adverse drug events through the use of a hospital information system. 6th International Conference on Pharmacoepidemiology, Anaheim, California, 10-l 3 August 1990. 25. Burke JP, Tilson HH, Platt R. Expanding roles of hospital epidemiology: Pharmacoepidemiology. Infect fcontrol Hosp Epidemiol 1989; 10: 253-254. 26. Evans RS, Burke JP, Classen DC, Stevens LE, Goodrich KM, Pestotnik SL. Computerized identification o!F hospital-acquired infections and high-risk patients. Third Decennial International Conference on Nosocomial Infections, Atlanta, Georgia, 31 July-3 August 1990.
The HELP
system
J. P. Burke,
Division
(1991)
18 (Supplement
A), 424431
and its application control
D. C. Classen, S. L. Pestotnik, -L. E. Stevens
to infection
R. S. Evans
and
of Infectious Diseases, LDS Hospital and University of Utah School of Medicine, Salt Lake City, Utah 84143, USA Summary: The HELP system is a comprehensive hospital information system that is linked to an allied financial data base. The clinical data base integrates information from areas such as admitting, pharmacy, radiology, surgery, pathology, nursing, respiratory therapy, and the clinical laboratories, including microbiology. This allows for the creation of an electronic medical record that contains all the clinical and financial data for each patient. The HELP system combines both communication and advice features through the use of data- and time-driven algorithms. We have used the HELP system to automate the surveillance and analysis of hospital-acquired infections and to identify patients at high risk for nosocomial infection. The expert system features have also been used to suggest alternatives for patients receiving inappropriate antimicrobial therapy, to improve the timing of antibiotic prophylaxis in surgery, and to curtail unnecessarily prolonged prophylaxis. Automated hospital information systems such as HELP can facilitate the investigation of a broad range of infection control, quality improvement, and cost-containment issues. Keywords:
Computers;
expert
systems;
infection
control;
surveillance.
Introduction Hospital infection control units on both sides of the Atlantic have made extensive use of computers to analyse data collected by surveillance personnel.’ In the future, linkage of desk-top personal computers, laboratory minicomputers and hospital mainframe computers will be commonplace. The improved communication capability between computers in hospitals has obvious potential benefits to improve the efficiency of the surveillance process as well as to prompt the application of infection control measures.2 Computerized hospital information systems are still in their infancy. The medical information systems that are now emerging can be grouped into two categories: ‘communication’ systems that store, retrieve and transmit information, and ‘advice’ systems that can supply stored information to the solution of clinical problems. 3 HELP (an acronym for Health Evaluation Correspondence to: Dr John P. Burke, Division Street, Salt Lake City, Utah 84143, USA. 01956701/91/06A424+08
of Infectious
Diseases, LDS Hospital, Q 1991 The Hospital
$03.00/O
424
8th Ave. and C Infection
Society
The HELP hospital
information
system
425
through Logical Processing) is one of the first integrated hospital information systems that combines the functions of both communication and advice systems. The HELP system was designed to meet the clinical, of hospitals and to provide administrative and research needs decision-support capability. 4 This computer system has been developed over the past 20 years at the 1LDS Hospital in Salt Lake City, Utah, USA. The purpose of this paper is to describe some of the infection-control applications of the HELP system over the past two decades and, further, to emphasize that these applications were made possible in a stepwise fashion as the HELP system evolved gradually into a comprehensive automated hospital data base. An historical approach will be used to describe applications that were enabled by existing data elements at each stage in the process of bringing various h.ospital services online. It should be clear at thle outset that the development of such an information system represents an extensive long-term commitment by the institution. It should become clear that, as comprehensive and sophisticated as it is, the system is sti.11 incomplete and that the most exciting epidemiological applications remain in the future. Indeed, one of the most desirable attributes of the HELP system is its flexibility, improvability and receptivity to new technology and advances in medical science itself. The existing HELP system may be viewed, in one sense, as a prototype that holds important lessons for dlesigners of future medical computer systems. Features
of the HELP
system
The initial application of the IHELP system was in the intensive care units of LDS Hospital to provide haemodynamic and electrocardiographic monitoring and to capture data from the clinical laboratory. Warner and his associates used a Control Data Corporation 3300 computer in these early efforts and adopted design criteria that included the need for modularity to accommodate an ever-expanding data base and the need for data-driven processing of medical decision algorithms in order to minimize the data entry required by medical personnel. By 1978 the system had been installed hospital-wide with terminals in every nursing unit. From 1979 to 1981 the system was re-developed in order to better meet the administrative needs of the hospital and to prevent downtime. The current Tandem system has 10 central processing units (CPUs) with seven gigabytes of online disk storage, more than 500 terminals, and 100 laser printers. The system is interfaced with minicomputers that assist in collecting and reviewing data from the intensive care units, the clinical laboratory, the heart catheterization laboratory, and the pulmonary function laboratory. The HELP system now creates a computerized medical record for each patient that contains coded data in a hierarchical structure rather than in free text. All patient data are stored permanently in two formats: a
426
J. P. Burke
et al.
long-term abstract of coded demographic and diagnostic information that is kept on-line, and a comprehensive collection of all data on removable disk packs, which is available for clinical and research purposes. The interfaces with other computers in the hospital allow for the automatic storage of test results from the laboratory computer system and of financial information that includes cost, charge and reimbursement data. Information from the HELP system can also be downloaded to personal computers and accessed by modems from physicians’ offices and homes to generate daily or 7-day ‘rounds reports’. In addition to this integrated data base, one of the most distinctive features of HELP is its decision-support function that has been strengthened and expanded at each stage in its growth. Simple algorithms and more complicated decision trees are used to generate patient-specific advice or ‘alerts’. These algorithms can be modified or added to as needed, and the decisions themselves become a permanent part of the data base. This knowledge base of programmed medical logic is automatically activated when certain key information is entered into the mainframe computer (data-driven) or specific modules can be set to run at a certain time of day (time-driven). The great appeal of such a system is its ability to give assistance to clinicians without requiring laborious data entry by the themselves’ and the fact that protocol-generated physicians recommendations can reduce the rate of medical errors due to the intrinsic limits of physicians.6 Infection
control
applications
Urinary catheter studies, 1970-75 The HELP system permitted a growing number of infection-control applications as the system became more fully developed and comprehensive. The first applications were in the form of research projects in patients with indwelling urinary catheters. Because microbiology test results were not yet available to the HELP system, we began a monitoring programme that required the collection of a catheter urine specimen each day from every patient with an indwelling catheter. The results of these urine cultures were entered into the mainframe computer and became the basis for a series of protocol-generated randomized controlled trials of urinary catheter equipment and catheter care practices.‘-i4 The computer was used in these studies to capture routine clinical data, to direct the collection of appropriate bedside observations by study personnel using algorithms, and to randomize the patients to various interventions. In addition, the urinary catheter monitoring became an automated tool for the surveillance of nosocomial urinary tract infections, relieving the infection control practitioners of this task and creating an archive for research issues. One of the intriguing aspects of these data derives from the more accurate classification of hospital-acquired urinary infections that was
The
HELP
hospital
information
system
427
achieved; for example, nearly one-half of the nosocomial urinary infections discovered by traditional surveillance were shown to represent bacteriuria present at the time of catheter placement, and a nearly equal number of ‘missed’ cases were uncovered by the intensive daily culturing.15 Antibiotic monitoring, 1975-80 In 1975 the hospital pharmacy was linked to the HELP system and a computerized monitoring service became clinically useful as algorithms were designed to alert and warn of potential adverse drug reactions and interactions.” Because of the broad nature of the data base even at this early stage in its development, the drug monitoring service was able to consider patient allergies, concomitant drugs likely to cause interactions, renal and hepatic dysfunction likely to affect drug excretion and clearance, and underlying diseases associated with specific drug reactions. Antibiotics accounted for 40% of all the drug warnings and therefore received special attention. We developed algorithms, for example, to identify patients who were receiving potentially nephrotoxic antibiotics and who were not being appropriately monitored with tests of renal function and patients who were receiving inappropriate dosing of antibiotics based on their renal or hepatic function. We observed that physicians complied with more than 90% of these computer-generated reminders, in part, because of the enhanced role of the pharmacists, who assessed each alert and notified the physicians directly. Also, we observed an educational effect of these reminders, with improved drug ordering and monitoring and an associated decline in the freq.uency of certain alerts.*’ Computerized infectious disea.se monitor, 1980-85 In the early 1980s microbiology test results were added to the system through an interface with thle laboratory computer system. Microbiology data are now integrated with medical information from other clinical areas, importantly with admission-discharge-transfer (ADT) data, most pharmacy, radiology, surgery and respiratory therapy. With these advances, the elements were in place to develop a computerized infectious disease monitoring system.18 We created a knowledge base of algorithms that is automatically activated when specific microbiology data are entered into a patient’s computer file, and the decisions are stored until needed. The computerized infectious disease monitor has enabled us to automate the surveillance of nosocomial infections. This surveillance mechanism has some of the same limitations that apply to any laboratory-based system; non-cultured infections, infections at some sites and viral infections are missed.” However, we have shown that infection control practitioners found more hospital-acquired infections using computer screening when compared with traditional surveillance methods while requiring only 35% of the time.” More importantly, false-positive diagnoses of
428
J. P. Burke
et al.
hospital-acquired infections were made with nearly the same frequency by the computer and by the infection control practitioners. Computer screening has several other advantages over traditional retrospective chart reviews. This automated surveillance results in more timely alerting of infection control staff to patients likely to have a hospital-acquired infection. For example, clusters of infection can be more rapidly detected by threshold analysis to permit prompt application of control measures, and downloading of data to personal computers facilitates outbreak investigations and statistical analyses. Methods are now being developed to activate computer algorithms by sources of information in addition to microbiology test results. Quality improvement, 1985-90 The computerized infectious disease monitor enables other clinical applications in addition to screening for nosocomial infections. An immediately useful monitoring involves linking antibiotic susceptibility reports with antibiotic prescriptions. When inconsistencies occur between antibiotic therapy and in-vitro susceptibility data, computer algorithms generate alerts, which are assessed by a clinical pharmacist who then notifies attending physicians. In nearly one-half of such alerts, physicians were simply not yet aware of the test results, but a variety of other circumstances has been appreciated. We have presented results that confirm the value of this method to identify and correct errors in antimicrobial prescribing and to assure the appropriate use of therapeutic antibiotics.21 These methods can also be used to further improve antibiotic therapy through algorithms that use susceptibility test results and a programmed knowledge base to recommend less costly or less toxic antibiotics than those actually prescribed. Prophylactic antibiotic use can also be defined and improved using the interfaces between the surgery operating room schedule, the pharmacy and the microbiology laboratory. The elements essential to the computer decision analysis with respect to surgical prophylaxis include the patient’s name and location, the surgeon and planned surgical procedure, the patient’s drug allergy history and drug use during hospitalization, and a list of all surgical procedures with an assessment of the clinical utility of perioperative antibiotic prophylaxis. In a 2-year study we were able to establish the magnitude of misapplication of perioperative prophylaxis and to develop computer-generated reminders to improve usage. These reminders were associated with improved timing of antibiotic use and with a concurrent decline in postoperative wound infection rates.” We also identified patients receiving antibiotic prophylaxis for prolonged durations, and pharmacists used the alerts to place ‘stop orders’ with the consent of surgeons. This programme was shown to reduce the average charges for antibiotic prophylaxis in surgery by US $42 without compromising patient care.23
The HELP hospital
information
system
429
The HELP system has also been linked to the IBM-based financial computer system that contains cost, charge and reimbursement data for all patients. This includes a sophisticated comprehensive microcosting system for all aspects of patient care. The data base can now be used to address a wide variety of issues, such as the effect of various strategies for antibiotic selection on total hospital costs, on rates of hospital-acquired infections, and on the use of other hospital services.
The
future
The computer developments that will allow future applications are well under way. First, bedside terminals have been installed in patient rooms and a computer-based nurse charting system is fully functional in several nursing care units and all intensive care units. This new charting system includes bedside entry of vital signs, measurements such as fluid intake and output, daily weights, and oblservations such as the presence of a rash or diarrhoea. With this new source of data integrated with the system, an adverse drug event monitoring program has been developed. Patients likely to be experiencing an adverse drug event are ‘flagged’ based on new orders for drugs likely to be prescribed to treat adverse drug reactions (e.g. anticonvulsants, antipruritics, antidiarrhoeals), sudden termination or change in dosage of a drug, laboratory abnormalities, or nurse charting of a rash, seizure, diarrhoea, etc. A. clinical pharmacist then uses these signals to investigate and to assign a probability score for adverse drug events based on standard algorithms.24 In the US, involvement of infection control with expanded programmes of hospital epidemiology that include quality assessment and pharmacoepidemiology appears to be growing.25 Infection control units already have the responsibility for monitoring antibiotic use, and automated systems, such as the one we have developed using HELP, can fulfil the broader needs of epidemiologists. The second ongoing development that has important implications for the future is the installation of HELP in other hospitals owned by Intermountain Health Care (IHC), the non-profit corporate parent of LDS Hospital. IHC owns and operates 23 hospitals in the intermountain region of the US and has begun the installation of HELP in all of these hospitals. There is growing recognition of the value of a large automated data base drawn from a regional network of hospitals. In infection control there will be increased needs to stratify patients by various risk factors, severity-of-illness scores, and underlying illnesses. HELP has already been used to develop a logistic regression model for predicting hospital-acquired infections,26 and the installation of HELP in other hospitals will provide a method to validate these predictors and to evaluate prevention strategies.
J. P. Burke
430
et al.
Discussion
Shortliffe’ has commented on the paradox of the persisting notion that the application of computers to medical practice has had limited success despite ample evidence to the contrary. While it is true that attempts to use the computer in medical diagnosis have been unsatisfactory, there is a steady increase in the use of computers as clinical decision-support tools. At the LDS Hospital, physicians have enthusiastically embraced the computer and have found the HELP system to be essential for patient care. The HELP system has replaced the paper medical record except for physicians’ notes. The success of the system seems to be based, in part, on data entry at its source by the appropriate clerical, nursing or laboratory personnel. In some instances, as in the laboratory, data entry is entirely automated. The automated use of a knowledge base of algorithms to analyse patient data means that health care professionals are not required to enter data or request evaluation of the data to receive decisions. The fear that computers may replace physicians seems quaint in retrospect. Computers will not replace epidemiologists either, but they can and will assist in epidemiological decision making. Data should not be collected to serve only one purpose, and hospital epidemiologists are embracing broader areas of concern in quality assessment and cost-containment concurrently with the rapid growth in automated hospital data bases. These systems will become commonplace and will offer greatly expanded applications to hospital epidemiology. For the full potential to be realized, microbiologists, pathologists and epidemiologists must be involved in planning these systems and make the system developers aware of their special needs. Moreover, the development of an automated hospital data base takes place incrementally and requires many parallel changes in the daily operation of the institution. The experience of the infection control unit at the LDS Hospital illustrates how new applications can be incorporated as the system grows. This
work
was supported,
in part, by U.S.P.H.S.
grants
HS 06028 and AI 15655.
References and use of computers in hospital infection control. 1. Wenzel RP, Streed SA. Surveillance J Hosp Infect 1989; 13: 217-229. WR. Information, computers and infection control. J Hasp Infect 1990; 15: 2. Gransden l-5. 3. Rennels GD, Shortliffe EH. Advanced computing for medicine. Sci Am 1987; 257: 154-161. RM, Clayton PD, Warner HR. The HELP system. r Med Syst 4. Pryor TA, Gardner 1983; 7: 87-102. EH. Computer programs to support clinical decision making. JAMA 1987; 5. Shortliffe 258: 61-66. 6. McDonald CJ. Protocol-based computer reminders, the quality of care and the non-perfectibility of man. N EnglJ Med 1976; 295: 1351-1355. RA, Burke JP, Dickman ML, Smith CB. Factors predisposing to bacteriuria 7. Garibaldi during indwelling urethral catheterization. N Engl J Med 1974; 291: 215-219.
The
HELP
hospital
information
system
431
8. Britt MR, Garibaldi RA, Miller WA, Hebertson RM, Burke JP. Antimicrobial prophylaxis for catheter-associated bacteriuria. Antimicrob Agents Chemother 1977; 11: 240-243. 9. Garibaldi RA, Burke JP, Britt MR, Miller WA, Smith CB. Meatal colonization and catheter-associated bacteriuria. N Engl J Med 1980; 303: 3 166318. 10. Burke JP, Garibaldi RA, Britt :MR, Jacobson JA, Conti M, Alling DW. Prevention of catheter-associated urinary tract infections: Efficacy of daily meatal care regimens. AmJ Med 1981; 70: 655-658. 11. Jacobson JA, Burke JP, Kasworm E. Effect of bacteriologic monitoring of urinary catheters on recognition and treatment of hospital-acquired urinary tract infections. Infect Control 1981; 2: 227-232. 12. Burke JP, Jacobson JA, Garibaldi RA, Conti MT, Alling DW. Evaluation of daily meatal care with poly-antibiotic ointment in prevention of urinary catheter-associated bacteriuria. J Ural 1983; 129: 331-334. 13. Burke JP, Larsen RA, Stevens LE. Nosocomial bacteriuria: Estimating the potential for prevention by closed sterile urinary drainage. Infect Control 1986; 7 (Suppl): 96699. 14. Larsen RA, Burke JP. ThLe epidemiology and risk factors for nosocomial catheter-associated bacteriuria caused by coagulase-negative staphylococci. Infect Control 1986; 7: 212-215. 15. Britt MR, Burke JP, Epstein B, Mooney B, Krall B, Miller WA. Monitoring of nosocomial urinary tract infection: A comparison of methods for surveillance of nosocomial urinary tract infection. Abstract 422. Program and Abstracts of the 17th Interscience Conference on Antimicrobial Agents and Chemotherapy, New York, 12-14 October 1977. 16. Hulse RK, Clark SJ, Jackson CJ, Warner HR, Gardner RM. Computerized medication monitoring system. Am J Hasp Pharm 1976; 33: 1061-1064. 17. Burke JP. The influence of computerized antibiotic monitoring on physician prescribing patterns. Symposium: Influencing Physician Usage of Antimicrobials. Program and Abstracts of the 17th Interscience Conference on Antimicrobial Agents and Chemotherapy, New York, 12-14 October 1977. 18. Evans RS, Gardner RM, Bush AR et al. Development of a computerized infectious disease monitor. Comput Biomed Res 1985; 18: 103-113. 19. Nystrom B. Surveillance of hospital-associated infections. Infection 1989; 17: 43-45. 20. Evans RS, Larsen RA, Burke JP et al. Computer surveillance of hospital-acquired infections and antibiotic use. JAMA 1986; 256: 1007-l 011. 21. Pestotnik SL, Evans RS, Burke JP, Gardner RM, Classen DC. Therapeutic antibiotic monitoring: Surveillance using a computerized expert system. Am J Med 1990; 88: 4348. 22. Larsen RA, Evans RS, Burke JP, Pestotnik SL, Gardner RM, Classen DC. Improved perioperative antibiotic use and reduced surgical wound infections through use of computer decision analysis. Infect Control Hosp Epidemiol 1989; 10: 316-320. 23. Evans RS, Pestotnik SL, Burke JP, Gardner RM, Larsen RA, Classen DC. Reducing the duration of prophylactic antibiotic use through computer monitoring of surgical patients. DICP, Annals of Pharmacotherapy 1990; 24: 351-354. 24. Classen DC, Pestotnik SL, Evans RS, Stevens LE, Bass SB, Burke JP. Concurrent monitoring of adverse drug events through the use of a hospital information system. 6th International Conference on Pharmacoepidemiology, Anaheim, California, 10-l 3 August 1990. 25. Burke JP, Tilson HH, Platt R. Expanding roles of hospital epidemiology: Pharmacoepidemiology. Infect fcontrol Hosp Epidemiol 1989; 10: 253-254. 26. Evans RS, Burke JP, Classen DC, Stevens LE, Goodrich KM, Pestotnik SL. Computerized identification o!F hospital-acquired infections and high-risk patients. Third Decennial International Conference on Nosocomial Infections, Atlanta, Georgia, 31 July-3 August 1990.