01bates(299-308)

Journal of the American Medical Informatics Association Volume 8 Number 4 Jul / Aug 2001 Focus on Quality Improvement
Reducing the Frequency ofErrors in Medicine UsingInformation Technology DAVID W. BATES, MD, MSC, MICHAEL COHEN, MS, RPH, LUCIAN L. LEAPE, MD, J. MARC OVERHAGE, MD, PHD, M. MICHAEL SHABOT, MD, THOMAS SHERIDAN, SCD A b s t r a c t Background: Increasing data suggest that error in medicine is frequent and
results in substantial harm. The recent Institute of Medicine report (LT Kohn, JM Corrigan,
MS Donaldson, eds: To Err Is Human: Building a Safer Health System. Washington, DC:
National Academy Press, 1999) described the magnitude of the problem, and the public
interest in this issue, which was already large, has grown.
Goal: The goal of this white paper is to describe how the frequency and consequences of
errors in medical care can be reduced (although in some instances they are potentiated) by
the use of information technology in the provision of care, and to make general and specific
recommendations regarding error reduction through the use of information technology.
Results: General recommendations are to implement clinical decision support judiciously;
to consider consequent actions when designing systems; to test existing systems to ensure
they actually catch errors that injure patients; to promote adoption of standards for data and
systems; to develop systems that communicate with each other; to use systems in new ways;
to measure and prevent adverse consequences; to make existing quality structures meaningful;
and to improve regulation and remove disincentives for vendors to provide clinical decision
support. Specific recommendations are to implement provider order entry systems, especially
computerized prescribing; to implement bar-coding for medications, blood, devices, and
patients; and to utilize modern electronic systems to communicate key pieces of asynchronous
data such as markedly abnormal laboratory values.
Conclusions: Appropriate increases in the use of information technology in health care—
especially the introduction of clinical decision support and better linkages in and among systems,
resulting in process simplification—could result in substantial improvement in patient safety.
■ J Am Med Inform Assoc. 2001;8:299–308.
Affiliations of the authors: Harvard Medical School, Boston, This work is based on discussions at the AMIA 2000 Spring Massachusetts (DWB); Institute for Safe Medication Practices, Congress; May 23–25, 2000; Boston, Massachusetts.
Huntingdon Valley, Pennsylvania (MC); Harvard School of Public Correspondence and reprints: David W. Bates, MD, MSc, Division of Health, Boston (LLL); Indiana University of Medicine, General Medicine and Primary Care, Brigham and Women’s Hospital, Indianapolis, Indiana (JMO); University of California–Los Angeles 75 Francis Street, Boston, MA 02115; e-mail: <[email protected]>.
School of Medicine, Los Angeles, California (MMS); MassachusettsInstitute of Technology, Cambridge, Massachusetts (TS).
Received for publication: 10/09/00; accepted for publication: 3/16/01.
Our goal in this manuscript is to describe how infor- recommended by the IOM involves all stakeholders— mation technology can be used to reduce the frequen- professionals, health care organizations, regulators, cy and consequences of errors in health care. We begin professional societies, and purchasers. Health care by discussing the Institute of Medicine report and the organizations are called on to work with their profes- evidence that errors and iatrogenic injury are a prob- sionals to implement known safe practices and set up lem in medicine, and also briefly mention the issue of meaningful safety programs in their institutions, in- inefficiency. We then define our scope of discussion cluding blame-free reporting and analysis of serious (in particular, what we are considering an error) and errors. External organizations—regulators, profession- then discuss the theory of error as it applies to infor- al societies, and purchasers—are called on to help mation technology, and the importance of systems establish standards and best practice for safety and to improvement. We then discuss the effects of clinical hold health care organizations accountable for imple- decision support, and errors generated by information technology. That is followed by management issues, Some of the best available data on the epidemiology the value proposition, barriers, and recent develop- of medical injury come from the Harvard Medical ments on the national front. We conclude by making a Practice Study.4 In that study, drug complications number of evidence-based general and specific recom- were the most common adverse event (19 percent), mendations regarding the use of information technol- followed by wound infections (14 percent) and tech- ogy for error prevention in health care.
nical complications (13 percent). Nearly half theevents were associated with an operation. Most work The Institute of Medicine Report and
on prevention to date has focused on adverse drug Iatrogenesis
events and wound infections. Compared with thedata on inpatients, relatively few data on errors and Errors in medicine are frequent, as they are in all injuries outside the hospital are available, although domains in life. While most errors have little poten- errors in follow-up5 and diagnosis are probably espe- tial for harm, some do result in injury, and the cumu- cially important in non-hospital settings. lative consequences of error in medicine are huge. While the IOM report and Harvard Medical Practice When the Institute of Medicine (IOM) released its Study deal primarily with injuries associated with report To Err is Human: Building a Safer Health System in errors in health care, the costs of inefficiencies related November 1999,1 the public response surprised most to errors that do not result in injury are also great.
people in the health care community. Although the One example is the effort associated with “missed report’s estimates of more than a million injuries and dose” medication errors, when a medication dose is nearly 100,000 deaths attributable to medical errors not available for a nurse to administer and a delay of annually were based on figures from a study pub- at least two hours occurs or the dose is not given at lished in 1991, they were news to many. The mortality all.6 Nurses spend a great deal of time tracking down figures in particular have been a matter of some pub- such medications. Although such costs are harder to lic debate2,3 although most agree that whatever the assess than the costs of injuries, they may be even The report galvanized an enormous reaction from Scope of Discussion
both government and health care representatives.
Within two weeks, Congress began hearings and thePresident ordered a government-wide feasibility In this paper, we are discussing only clear-cut errors study, followed in February by a directive to govern- in medical care and not suboptimal practice (such as mental agencies to implement the IOM recommenda- failure to follow a guideline). Clearly, this is not a tions. During this time, professional societies and dichotomous distinction, and some examples may be health care organizations have begun to re-assess helpful. We would consider a sponge left in the patient after surgery an error, whereas an inappro-priate indication for surgery would be suboptimal The IOM report made four major points—the extent of practice. We would consider it an error if no postop- harm that results from medical errors is great; errors erative anticoagulation were used in patients in result from system failures, not people failures; achiev- whom its benefit has clearly been demonstrated (for ing acceptable levels of patient safety will require example, patients who have just had hip surgery).
major systems changes; and a concerted national effort However, we would not consider it an error if a is needed to improve patient safety. The national effort physician failed to follow a pneumonia guideline and Journal of the American Medical Informatics Association Volume 8 Number 4 Jul / Aug 2001 prescribed a commonly used but suboptimal antibi- resourceful and inventive, and they can recover from otic, even though adherence to such guidelines will both their own and the equipment’s errors in creative almost certainly improve outcomes. Although we ways. In comparison, machines are more depend- believe that information technology can play a major able, which means they are dependably stupid when role in both domains, we are not addressing subopti- a minor behavior change would prevent a failure in a neighboring component from propagating. The intel-ligent machine can be made to adjust to an identified Theory of Error
variable whose importance and relation to other vari-ables are sufficiently well understood. The intelligent Although human error in health care systems has human operator still has usefulness, however, for he only recently received great attention, human factors or she can respond to what at the design stage may engineering has been concerned with error for sever- be termed an “unknown unknown” (a variable al decades. Following the accident at Three Mile which was never anticipated, so that there was never Island in the late 1970s, the nuclear power industry any basis for equations to predict it or computers and was particularly interested in human error as part of human factors concerns, and has produced a number Finally, we seek to reduce the undesirable conse- of reports on the subject.7 The U.S. commercial avia- quences of error, not error itself. Senders and tion sector is also very interested in human error at Moray10 provide some relevant comments that relate present, because of massive overhaul of the air traffic to information technology: “The less often errors control network. A few excellent books on human occur, the less likely we are to expect them, and the more we come to believe that they cannot happen… .
While it is easy and common to blame operators for It is something of a paradox that the more errors we accidents, investigation often indicates that an opera- make, the better we will be able to deal with them.” tor “erred” because the system was poorly designed.
They comment further that, “eliminating errors local- Testimony of an operator of the Three Mile Island ly may not improve a system and might cause worse nuclear power plant in a 1979 Congressional hear- ing11 makes the point, “ If you go beyond what the A medical example relating to these issues comes from designers think might happen, then the indications the work of Macklis et al. in radiation therapy12; this are insufficient, and they may lead you to make the group has used and evaluated the safety record of a wrong inferences. …[H]ardly any of the measure- record-and-verify linear accelerator system that dou- ments that we have are direct indications of what is ble-checks radiation treatments. This system has an error rate of only 0.18 percent, with all detected errorsbeing of low severity. However, 15 percent of the The consensus among man–machine system engi- errors that did occur related to use of the system, pri- neers is that we should be designing our control marily because when an error in the checking system rooms, cockpits, intensive care units, and operating occurred, the human operators assumed the machine rooms so that they are more “transparent”—that is, “had to be right,” even in the face of important con- so that the operator can more easily “see through” flicting data. Thus, the Macklis group expressed con- the displays to the actual working system, or “what cern that over-reliance on the system could result in an is going on.” Situational awareness is the term used accident. This example illustrates why it will be vital to in the aviation sector. Often the operator is locked measure to determine how systems changes affect the into the dilemma of selecting and slavishly following overall rate of not only errors but accidents. one or another written procedure, each based on ananticipated causality. The operator may not be sure Systems Improvement and Error Prevention
what procedure, if any, fits the current not-yet-understood situation.
Although the traditional approach in medicine has Machines can also produce errors. It is commonly been to identify the persons making the errors and appreciated that humans and machines are rather punish them in some way, it has become increasing- different and that the combination of both thus has ly clear that it is more productive to focus on the sys- greater potential reliability than either alone. How- tems by which care is provided.13 If these systems ever, it is not commonly understood how best to could be set up in ways that would both make errors make this synthesis. Humans are erratic, and err in less likely and catch those that do occur, safety might surprising and unexpected ways. Yet they are also A system analysis of a large series of serious medica- in a controlled trial that computerized physician tion errors (those that either might have or did cause order entry systems resulted in a 55 percent reduction harm)13 identified 16 major types of system failures in serious medication errors. In another time series associated with these errors. Of these system failures, study,22 this group found an 83 percent reduction in all of the top eight could have been addressed by bet- the overall medication error rate, and a 64 percent reduction even with a simple system. Evans et al.23have also demonstrated that clinical decision support Currently, the clinical systems in routine use in can result in major improvements in rates of antibiot- health care in the United States leave a great deal to ic-associated adverse drug events and can decrease be desired. The health care industry spends less on costs. Classen et al.24 have also demonstrated in a information technology than do most other informa- series of studies that nosocomial infection rates can be tion-intensive industries; in part as a result, the dream of system integration been realized in feworganizations. For example, laboratory systems do Another class of clinical decision support is comput- not communicate directly with pharmacy systems.
erized alerting systems, which can notify physicians Even within medication systems, electronic links be- about problems that occur asynchronously. A grow- tween parts of the system—prescribing, dispensing, ing body of evidence suggests that such systems and administering—typically do not exist today.
may decrease error rates and improve therapy, Nonetheless, real and difficult issues are present in thereby improving outcomes, including survival, the the implementation of information technology in length of time patients spend in dangerous condi- health care, and simply writing a large check does tions, hospital length of stay, and costs.25–27 While an not mean that an organization will necessarily get an increasing number of clinical information systems outstanding information system, as many organiza- contain data worthy of generating an alert message, tions have learned to their chagrin.
delivering the message to caregivers in a timely wayhas been problematic. For example, Kuperman et Evaluation is also an important issue. Data on the al.28 documented significant delays in treatment effects of information technology on error and even when critical laboratory results were phoned to adverse event rates are remarkably sparse, and many caregivers. Computer-generated terminal messages, more such studies are needed. Although such evalu- e-mail, and even flashing lights on hospital wards ations are challenging, tools to assess the frequencyof errors and adverse events in a number of domainsare now available.14–19 Errors are much more fre-quent than actual adverse events (for medicationerrors, for example, the ratio in one study6 was100:1). As a result, it is attractive from the sample sizeperspective to track error rates, although it is impor-tant to recognize that errors vary substantially intheir likelihood of causing injury.20 Clinical Decision Support
While many errors can be detected and corrected byuse of human knowledge and inspection, these rep-resent weak error reduction strategies. In 1995, Leapeet al.13 demonstrated that almost half of all medica-tion errors were intimately linked with insufficientinformation about the patient and drug. Similarly,when people are asked to detect errors by inspection,they routinely miss many.21 It has recently been demonstrated that computerizedphysician order entry systems that incorporate clini- Alert detection system. Three major forms of cal decision support can substantially reduce medica- critical event detection occur—critical laboratory alerts, tion error rates as well as improve the quality and effi- physiologic “exception condition” alerts, and medication ciency of medication use. In 1998, Bates et al.20 found Journal of the American Medical Informatics Association Volume 8 Number 4 Jul / Aug 2001 2 Wireless alerting system. In the Cedars-Sinai system, alerts are initially detected by the clinical system, then
sent to a server, then via the Internet, then sent over a PageNet transmitter to a two-way wireless device.
have been tried.29–32 A new system, which transmits includes appropriate patient identification informa- real-time alert messages to clinicians carrying alphanumeric pagers or cell phones, promises to Alerts are a crucial part of a clinical decision support eliminate the delivery problem.33,34 It is now possible system,35 and their value has been demonstrated in to integrate laboratory, medication, and physiologic controlled trials.27,35 In one study, Rind et al.27 alerted data alerts into a comprehensive real-time wireless physicians via e-mail to increases in serum creatinine in patients receiving nephrotoxic medications or Shabot et al.33,34 have developed such a comprehen- renally excreted drugs. Rind et al. reported that when sive system for patients in intensive care units. A e-mail alerts were delivered, medications were adjust- software system detect alerts and then sends them to ed or discontinued an average of 21.6 hours earlier caregivers. The alert detection system monitors data than when no e-mail alerts were delivered. In another flowing into a clinical information system. The detec- study, Kuperman et al.35 found that when clinicians tor contains a rules engine to determine when alerts were paged about “panic” laboratory values, time to therapy decreased 11 percent and mean time to reso-lution of an abnormality was 29 percent shorter. For some kinds of alert detection, prior or relateddata are needed. When the necessary data have been As more and different kinds of clinical data become collected, alerting algorithms are executed and a available electronically, the ability to perform more decision is made as to whether an alert has occurred sophisticated alerts and other types of decision sup- (Figure 1). The three major forms of critical event port will grow. For example, medication-related, lab- detection are critical laboratory alerts, physiologic oratory, physiologic data can be combined to create a “exception condition” alerts, and medication alerts.
variety of automated alerts. (Table 1 shows a sample When an alert condition is detected, an application of those currently included in the system used at formats a message and transmits it to the alphanu- Cedars-Sinai Medical Center, Los Angeles, meric pagers of various recipients, on the basis of a California.) Furthermore, computerization offers table of recipients by message type, patient service many tools for decision support, but because of space type, and call schedule. The message is sent as an e- limitations we have discussed only some of these; mail to the coded PIN (personal identification num- Among the others are algorithms, guidelines, order ber) of individual caregivers’ pagers or cell phones.
sets, trend monitors, and co-sign forcers. Most The message then appears on the device’s screen and sophisticated systems include an array of these tools. Sample of Wireless Alerts Currently in Use at Cedars-Sinai Medical Center, Los Angeles, California PEEP (positive end-respiratory pressure) > 15 cm H Systolic BP < 80 mm Hg and no pulmonary artery catheter Systolic BP < 80 mm Hg and pulmonary wedge pressure Urine output < 0.3 cc/kg/hour and patient not admitted in Re-admission to intensive care unit < 48 hours after discharge Heparin flush ≥ 500 unitsHeparin injection ≥ 5000 units Alert if urine output is low (< 0.3 cc/kg/hour for 3 hours) and the patient is receiving gentamicin, tobramycin, vancomycin, penicillin, ampicillin, Augmentin (amoxicillin/clavulanic acid), piperacillin, Zosyn (piperacillin/tazobactam), oxacillin, Primaxin (imipenem/ cilastatin), or Unasyn (ampicillin/sulbactam).
Medication–laboratory data trend alerts: Alert if serum creatinine level increases by > 0.5 mg/dL in 24 hours and the patient is receiving any of the following drugs: gentamicin, tobramycin, amikacin, vancomycin, amphotericin, digoxin, procainamide, Prograf (tacrolimus), cyclosporin, or ganciclovir.
Errors Generated By Information Technology
cian order entry systems are separate from the phar-macy system, which requires double entry of all Although information technology can help reduce orders. This may result in electronic/computer-gener- error and accident rates, it can also cause errors. For ated medication administration records (MARs) that example, if two medications that are spelled similar- are derived from the order entry system database, not ly are displayed next to each other, substitution the pharmacy database, which can result in discrepan- errors can occur. Also, clinicians may write an order cies and extra work for nurses and pharmacists. Furthermore, many computerized physician order In particular, early adopters of vendor-developed entry systems lack even basic screening capabilities order entry have reported significant barriers to suc- to alert practitioners to unsafe orders relating to over- cessful implementation, new sources of error, and ly high doses, allergies, and drug–drug interactions.
major infrastructure changes that have been necessary While visiting hospitals in 1998, representatives of to accommodate the technology. The order entry the Institute for Safe Medication Practices (ISMP) process with many computerized physician order tested pharmacy computers and were alarmed to dis- entry systems currently on the market is error-prone cover that many failed to detect unsafe drug orders.
and time-consuming. As a result, prescribers may Subsequently, ISMP asked directors of pharmacy in bypass the order entry process totally and encourage U.S. hospitals to perform a nationwide field test to nurses, pharmacists, or unit secretaries to enter written assess the capability of their systems to intercept or verbal drug orders. Also, most computerized physi- common or serious prescribing errors.36 To partici- Journal of the American Medical Informatics Association Volume 8 Number 4 Jul / Aug 2001 pate, pharmacists set up a test patient in their com- puter system, then entered actual physician prescrip- Percentage of Pharmacy Computer Systems That tion errors that had actually led to a patient’s death or serious injury during 1998 (Table 2). Only a smallnumber of even fatal errors were detected by current These anecdotal data suggest that current systems may be inadequate and that simply implementing the current off-the-shelf vendor products may not have the same effect on medication errors that hasbeen reported in research studies. Improvement of vendor-based systems and evaluation of their effectsis crucial, since these are the systems that will be Management Issues
* All these orders are unsafe and have resulted in at least one fatal- A major problem in creating the will to reduce errors ity in the United States. However, most pharmacy systems did notdetect them, and even among those that did, a large percentage has been that administrators have not been aware of allowed an override without a note. Data reprinted, with permis- the magnitude of problem. For example, one survey sion, from ISMP Medication Safety Alert! Feb 10, 1999.36 showed that, while 92 percent of hospital CEOs Copyright Institute for Safe Medication Practices.
reported that they were knowledgeable about the fre-quency of medication errors in their facility, only 8 result in a 12.7 percent decrease in total charges and percent said they had more than 20 per month, when a 0.9 day decrease in length of stay.41 Even without in fact all probably had more than this.37 Probably in full computerization of ordering, substantial savings part as a result, the Advisory Board Company found can be realized: data from LDS Hospital23 demon- that reducing clinical error and adverse events strated that a program that assisted with antibiotic ranked 133rd when CEOs were asked to rank items management resulted in a fivefold decrease in the on a priority list.38 A number of efforts are currently frequency of excess drug dosages and a tenfold under way to increase the visibility of the issue. For decrease in antibiotic-susceptibility mismatches, with example, a video about this issue, which was devel- substantially lower total costs and lengths of stay.
oped by the American Hospital Association and theInstitute for Healthcare Improvement, has been sent Barriers
to all hospital CEOs in the United States, and a num-ber of indicators suggest that such efforts may be Despite these demonstrated benefits, only a handful of organizations have successfully implemented clin-ical decision support systems. A number of barriers The Value Proposition
have prevented implementation. Among these arethe tendency of health care organizations to invest in For information technology to be implemented, it administrative rather than clinical systems; the issue must be clear that the return on investment is suffi- of “silo accounting,” so that benefits that accrue cient, and far too few data are available regarding across a system do not show up in one budget and this in health care. Furthermore, there are many hor- thus do not get credit; the current financial crisis in ror stories of huge investments in information tech- health care, which has been exacerbated by the Balanced Budget Amendment and has made it veryhard for hospitals to invest; the lack, at many sites, of Positive examples relate to computer order entry. At leaders in information technology; and the lack of one large academic hospital, the savings were esti- mated to be $5 million to $10 million annually on a$500 million budget.39 Another community hospital One of the greatest barriers to providing outstanding predicts even larger savings, with expected annual decision support, however, has been the need for an savings of $21 million to $26 million, representing extensive electronic medical record system infra- about a tenth of its budget.40 In addition, in a ran- structure. Although much of the data required to domized controlled trial, order entry was found to implement significant clinical decision support is already available in electronic form at many institu- ment safe practices.42 One of the first of these practices tions, the data are either not accessible or cannot be will be the implementation of computerized physi- brought together to be used in clinical decision sup- cian order entry systems. Similarly, a recent Medicare port because of format and interface issues. Existing Patient Advisory Commission report suggested that and evolving standards for exchange of information that the Health Care Financing Administration con- (HL7) and coding of this data are simplifying this sider providing financial incentives to hospitals that task. Correct and consistent identification of patients, adopt physician order entry systems.43 The Agency doctors, and locations is another area in which stan- for Healthcare Research and Quality has received $50 dards are needed. Approaches to choosing which million in funding to support error reduction information should be coded and how to record a research, including information technology–related mixture of structured coded information and strategies. California recently passed a law mandating that non-rural hospitals implement computerizedphysician order entry or another application like it by Some organizations have moved ahead with adopt- 2005.44 Clearly, many look to automation to play a ing such standards on their own, and this can have major role in the redesign of our systems.
great benefits. For example, a technology architectureguide was developed at Cedars-Sinai Medical Center Recommendations
to help ensure that its internal systems and databasesoperate in a coherent manner. This has allowed them Recommendations for using information technology to develop what they call their “Web viewing sys- to reduce errors fall into two categories—general tem,” which allows clinicians to see nearly all results suggestions that are relevant across domains, and on an Internet platform. Many health care organiza- very specific recommendations. It is important to rec- tions are hamstrung, because they have implemented ognize that these lists are not exhaustive, but they do so many different technologies and databases that contain many of the most important and best-docu- mented precepts. Although many of these relate to A second major hurdle is choosing the appropriate the medication domain, this is because the best cur- rules or guidelines to implement. Many organizations rent evidence is available for this area; we anticipate have not developed processes for developing and that information technology will eventually be implementing consensus choices in their physician shown to be important for error reduction across a groups. Once the focus has been determined, the wide variety of domains, and some evidence is organization must determine exactly what should be already available for blood products, for example.45,46 done about the selected problem. Regulatory and The strength of these recommendations is based on a legal issues have also prevented vendors from pro- standard set of criteria for levels of evidence.47 For viding this type of content. Finally, despite good therapy and prevention, evidence level 1a represents precedents for delivering feedback to clinicians for multiple randomized trials, level 1b is an individual simple decision support, changing provider behavior randomized trial, level 4 is case series, and level 5 for more complex aspects of care remains challenging.
The National Picture
General Recommendations
A national commitment to safer health care is devel- Implement clinical decision support judiciously oping. Although it is too soon to determine how it (evidence level 1a). Clinical decision support can will “play out” (the initial fixation on mandatory clearly improve care,48 but it must be used in ways reporting has been an unwelcome diversion, for that help users, and the false-positive rate of active example), it seems clear that many stakeholders have suggestions should not be overly high. Such deci- a real interest in improving safety. Doctors and other sion support should be usable by physicians. professionals are in the interesting position of being Consider consequent actions when designing sys- expected to be both leaders in this movement and the tems (evidence level 1b). Many times, one action recipients of its attention. Already a national coalition implies another, and systems that prompt regard- involving many of the leading purchasers, the ing this can dramatically decrease the likelihood of Leapfrog Group, which includes such companies as General Motors and General Electric, have an-nounced their intention to provide incentives to hos- Test existing systems to ensure that they actually pitals and other health care organizations to imple- catch errors that injure patients (evidence level 5).
Journal of the American Medical Informatics Association Volume 8 Number 4 Jul / Aug 2001 The match between the errors that systems detect Specific Recommendations
and the actual frequency of important errors isoften suboptimal.
Implement provider order entry systems, especial-ly computerizing prescribing (evidence level 1b).
Promote adoption of standards for data and sys- Provider order entry has been shown to reduce the tems (evidence level 5). Adoption of standards is serious medication error rate by 55 percent.20 critical if we are to realize the potential of infor-mation technology for error prevention. Standards Implement bar-coding for, for example, medica- for constructs such as drugs and allergies are espe- tions, blood, devices, and patients (evidence level 4). In other industries, bar-coding has dramaticallyreduced error rates. Although fewer data are avail- Develop systems that communicate with each able for this recommendation in medicine, it is like- other (evidence level 5). One of the greatest barri- ly that bar-coding will have a major impact.52 ers to providing clinicians with meaningful infor-mation has been the inability of systems, such as Use modern electronic systems to communicate medication and laboratory systems, to readily key pieces of asynchronous data (evidence level exchange data. Such communication should be 1b). Timely communication of markedly abnormal seamless. Adopting enterprise database standards laboratory tests can decrease time to therapy and the time patients spend in life-threatening condi-tions.
Use systems in new ways (evidence level 5). Elec-tronic records will soon facilitate new, sophisticated Our hope is that these recommendations will be use- prevention approaches, such as risk factor profiling ful for a variety of audiences. Error in health care is a and pharmacogenomics, in which a patient’s med- pressing problem, which is best addressed by chang- ications are profiled against their genetic makeup.
ing our systems of care—most of which involveinformation technology. Although information tech- Measure and prevent adverse consequences (evi- nology is not a panacea for this problem, which is dence level 5). Information technology in general highly complex and will demand the attention of and clinical decision support in particular can cer- many, it can play a key role. The informatics commu- tainly have perverse and opposite consequences; nity should make it a high priority to assess the continuous monitoring is essential.50 However, effects of information technology on patient safety. such monitoring has often not been carried out. Itshould also be routine to measure how often rec- ommendations are presented and how often sug- 1. Kohn LT, Corrigan JM, Donaldson MS (eds). To Err Is Human: gestions are accepted and to have some measures Building a Safer Health System. Washington, DC: National 2. McDonald CJ, Weiner M, Hui SL. Deaths due to medical errors Make existing quality structures meaningful (evi- are exaggerated in Institute of Medicine report. JAMA.
dence level 5). Quality measurement and improve- ment groups are often suboptimally effective.
3. Leape LL. Institute of Medicine medical error figures are not Increasing the use of computerization should exaggerated [comment]. JAMA. 2000;284:95–7.
4. Leape LL, Brennan TA, Laird NM, et al. The nature of adverse make it dramatically easier to measure quality events in hospitalized patients: results from the Harvard continually. Such information must then be used Medical Practice Study II. N Engl J Med. 1991;324:377–84.
5. Gurwitz JH, Field T, Avorn J, et al. Incidence and preventability of adverse drug events in nursing homes. Am J Med. 2000;109:87–94.
Improve regulation and remove disincentives for 6. Bates DW, Boyle DL, Vander Vliet MB, Schneider J, Leape LL.
vendors to provide clinical decision support (evi- Relationship between medication errors and adverse drug dence level 5). The regulation relating to informa- events. J Gen Intern Med. 1995;10:199–205.
7. Rasmussen J. Human errors: a taxonomy for describing tion technology is hopelessly outdated and is cur- human malfunction in industrial installations. J Occup Accid.
rently being revised to address such issues as pri- vacy in the electronic world.51 One issue that 8. Reason J. Human Error. Cambridge, UK: Cambridge relates to error in particular is that vendors, with some cause, fear being sued if they provide action- 9. Norman DA. The Design of Everyday Things. New York: oriented clinical decision support. Thus, the sup- 10. Senders J, Moray N. Human Error: Cause, Prediction and port either is not provided or is watered down.
Reduction. Mahwah, NJ: Lawrence Erlbaum, 1991.
11. Testimony of the Three Mile Island Operators. United States President's Commission on the Accident at Three Mile Island, vol 34. Shabot M, LoBue M, Chen J. Wireless clinical alerts for critical 1. Washington, DC: U.S. Government Printing Office, 1979:138.
medication, laboratory and physiologic data. Proceedings of 12. Macklis RM, Meier T, Weinhous MS. Error rates in clinical the 33rd Hawaii International Conference on System radiotherapy. J Clin Oncol. 1998;16:551–6.
Sciences (HICSS); Jan 4–7, 2000; Maui, Hawaii [CD-ROM].
13. Leape LL, Bates DW, Cullen DJ, et al. Systems analysis of Washington, DC: IEEE Computer Society, 2000.
adverse drug events. ADE Prevention Study Group. JAMA.
35. Kuperman G, Sittig DF, Shabot M, Teich J. Clinical decision support for hospital and critical care. J HIMSS. 1999;13:81–96.
14. Brennan TA, Leape LL, Laird N, et al. Incidence of adverse 36. Institute for Safe Medication Practices. Over-reliance on com- events and negligence in hospitalized patients: results from the puter systems may place patients at great risk. ISMP Medication Harvard Medical Practice Study I. N Engl J Med. 1991;324:370–6.
Safety Alert, Feb 10, 1999. Huntingdon Valley, Pa.: ISMP, 1999.
15. Barker KN, Allan EL. Research on drug-use-system errors. Am 37. Bruskin Goldring Research. A Study of Medication Errors and J Health Syst Pharm. 1995;52:400–3.
Specimen Collection Errors. Commissioned by BD (Becton- 16. Classen DC, Pestotnik SL, Evans RS, Burke JP. Computerized Dickinson) and College of American Pathologists. Feb 1999.
surveillance of adverse drug events in hospital patients.
Available at http://www.bd.com/bdid/whats_new/survey.
17. Jha AK, Kuperman GJ, Teich JM, et al. Identifying adverse 38. The Advisory Board Company. Prescription for change: drug events: development of a computer-based monitor and toward a higher standard in medication management. Wash- comparison to chart review and stimulated voluntary report. J Am Med Inform Assoc. 1998;5(3):305–14.
39. Glaser J, Teich JM, Kuperman G. Impact of information events 18. Gandhi TK, Seger DL, Bates DW. Identifying drug safety issues— on medical care. Proceedings of the 1996 HIMSS Annual from research to practice. Int J Qual Health Care 2000;12:69–76.
Conference. Chicago, Ill.: Healthcare Information and 19. Karson AS, Bates DW. Screening for adverse events. J Eval Management Systems Society, 1996:1–9.
40. Sarasota Memorial Hospital documents millions in expected 20. Bates DW, Leape LL, Cullen DJ, et al. Effect of computerized cost savings, reduced LOS through use of Eclipsys’ Sunrise physician order entry and a team intervention on prevention Clinical Manager [press release]. Delray Beach, Fla.: Eclipsys of serious medication errors. JAMA. 1998;280(15):1311–6.
Corporation; Oct 11, 1999. Available at: http://www.
21. Bates DW, Cullen D, Laird N, et al. Incidence of adverse drug events and potential adverse drug events: implications for 41. Tierney WM, Miller ME, Overhage JM, McDonald CJ.
Physician inpatient order writing on microcomputer worksta- 22. Bates DW, Miller EB, Cullen DJ, et al. Patient risk factors for tions: effects on resource utilization. JAMA. 1993;269:379–83.
adverse drug events in hospitalized patients. Arch Intern 42. Fischman J. Industry Preaches Safety in Pittsburgh. U.S. News 23. Evans RS, Pestotnik SL, Classen DC, et al. A computer-assist- 43. Medicare Payment Advisory Commission. Report to the ed management program for antibiotics and other anti-infec- Congress: Selected Medicare Issues, Jun 1999. Available at: tive agents. N Engl J Med. 1998;338:232–8.
http://www.medpac.gov/html/body_june_report.html.
24. Classen DC, Evans RS, Pestotnik SL, et al. The timing of pro- phylactic administration of antibiotics and the risk of surgical- 44. California Senate Bill No. 1875. Chapter 816, Statutes of 2000. wound infection. N Engl J Med 1992;326:281–6.
45. Lau FY, Wong R, Chui CH, Ng E, Cheng G. Improvement in 25. Tate K, Gardner RM, Scherting K. Nurses, pagers and patient transfusion safety using a specially designed transfusion specific criteria: three keys to improved critical value report- wristband. Transfus Med.2000;10(2):121–4.
ing. Proc Annu Symp Comput Appl Med Care. 1995;19:164–8.
46. Blood-error reporting system tracks medical mistakes [press 26. Tate K, Gardner RM, Weaver LK. A computerized laboratory release]. Dallas, Tex: The University of Texas Southwestern alerting system. MD Comput. 1990;7:296–301.
Medical Center at Dallas; Nov 22, 1999. Available at 27. Rind D, Safran C, Phillips RS, et al. Effect of computer-based http://irweb.swmed.edu/newspub. Accessed Mar 12, 2001. alerts on the treatment and outcomes of hospitalized patients.
47. Centre for Evidence-based Medicine. Levels of evidence and grades of recommendation; Sep 8, 2000. Available at: 28. Kuperman G, Boyle D, Jha AK, et al. How promptly are inpa- http://cebm.jr2.ox.ac.uk/docs/levels.html. Accessed Mar 12, tients treated for critical laboratory results? J Am Med Inform 48. Haynes RB, Hayward RS, Lomas J. Bridges between health 29. Bradshaw K. Computerized alerting system warns of life-threatening care research evidence and clinical practice. J Am Med Inform events. Proc Annu Symp Comput Appl Med Care. 1986;10:403.
30. Bradshaw K. Development of a computerized laboratory alert- 49. Overhage JM, Tierney WM, Zhou X, McDonald CJ. A ran- ing system. Comput Biomed Res. 1989;22:575–87.
domized trial of “corollary orders” to prevent errors of omis- 31. Shabot M, LoBue M, Leyerle B. Inferencing strategies for auto- sion. J Am Med Inform Assoc. 1997;4:364–75.
mated alerts on critically abnormal laboratory and blood gas 50. Miller R, Gardner RM. Summary recommendations for data. Proc Annu Symp Comput Appl Med Care. 1989;13:54–7.
responsible monitoring and regulation of clinical software sys- 32. Shabot M, LoBue M, Leyerle B. Decision support alerts for tems. Ann Intern Med. 1997;127:842–5.
clinical laboratory and blood gas data. Int J Clin Monit 51. Gostin LO, Lazzarini Z, Neslund VS, Osterholm MT. The pub- lic health information infrastructure: a national review of the 33. Shabot M, LoBue M. Real-time wireless decision support alerts law on health information privacy. JAMA.1996;275:1921–7.
on a palmtop PDA. Proc Annu Symp Comput Appl Med Care.
52. Bates DW. Using information technology to reduce rates of medication errors in hospitals. BMJ. 2000;320:788–91.

Source: http://sce.umkc.edu/~leeyu/MEDSS/References/mde-1.pdf