Mr. Fitzpatrick has made an appointment for a general physical exam. At 31 years of age, Mr. Fitzpatrick remains active in his work in information technology, but he has not seen Dr. Anderson since his early 20s. In other words, during today's visit the 2 have a lot to talk about.
A thorough discussion of Mr. Fitzpatrick's family history does not reveal any significant conditions or diseases occurring at abnormally young ages in his relatives (eg, heart disease or cancer). The physical exam is likewise unremarkable, except for slightly elevated blood pressure on 2 separate readings (138/89 mm Hg and 140/89 mm Hg). Dr. Anderson counsels Mr. Fitzpatrick on several lifestyle changes that could affect his blood pressure, including reducing his salt intake and increasing exercise. Dr. Anderson frequently sees patients with elevated cholesterol in his family practice clinic, so he is quite cognizant of the need to monitor HDL- and LDL-cholesterol levels as well as triglycerides. He asks Mr. Fitzpatrick to return for a fasting blood test to check his total cholesterol, HDL, and triglycerides. Though reluctant because of work demands, Mr. Fitzpatrick agrees.
Later that afternoon, Dr. Anderson relates his actions to Dr. Monde, one of his practice partners. She is surprised to learn that he routinely screens patients in this age range for elevated cholesterol. In response, Dr. Anderson cites guidelines published in 2001 by the National Heart, Lung and Blood Institute's National Cholesterol Education Program (http://www.nhlbi.nih.gov/guidelines/cholesterol/index.htm). Showing her these guidelines online, he notes that they recommend screening with fasting lipid profiles every 5 years, beginning at age 20, for both men and women. Somewhat perplexed, Dr. Monde recalls that her last reading of the United States Preventive Services Task Force guidelines (issued in 2001) implied that screening for patients who, like Mr. Fitzpatrick, do not have any major risk factors for heart disease, need not begin until 35 years of age for men or 45 years for women. She also wonders whether the test must be a fasting one.
To help settle their debate, they turn to the National Guideline Clearinghouse, which offers a synthesis of clinical guidelines regarding lipid screening in adults. Their disagreement is confirmed, but they are dismayed to find several other organizations offering their own, sometimes conflicting, guidelines. Both physicians wish that a better method existed to keep them up-to-date with these clinical guidelines—after all, both guidelines are now 3 years old—and more importantly, to know what to do in these complex, conflicting situations.
Unaware of any conflict, Mr. Fitzpatrick arrives the next morning for his fasting blood test.
Clinical practice guidelines can be thought of as systems that support clinical decision making. The issues raised in this case by the use of traditional clinical guidelines could apply as well to the use of other, more novel clinical decision support systems (CDSS), such as computerized tools designed to assist clinician decision making. CDSS usually link patient-specific information with a knowledge base to provide alerts for dangerous orders, suggest diagnoses, recommend laboratory tests, or critique treatment plans. Some of these systems are integrated into an electronic medical record, and others are stand-alone systems that can be consulted by the physician.
In this particular case, Dr. Anderson made the decision to have Mr. Fitzpatrick return for a fasting cholesterol test based on his recollection of a particular guideline. Should he have relied as he did on his "unaided memory" when there was an information source available? Does the physician have an ethical obligation to the patient to use such systems, especially if they could be accessed conveniently at the point of care? Conversely, are there risks to the patient in using these systems, especially, as in the example in the case, when there may be conflicting information about the appropriate choice of action?
Rationale for Using CDSS
Reviewing the guidelines as Dr. Anderson did after the patient left the office was more effective than spending time surfing the Web while the patient was in the office. A good CDSS, however, might be able to display a synthesis of evidence in a format that would make it easier to apply, even in the short time available for an office visit. PDAs are gaining popularity among physicians and are easily used in an outpatient setting to access drug databases, clinical prediction rules, clinical guidelines, and other forms of decision support. While there is clear evidence that CDSS can reduce medical errors,1 at least in an inpatient setting, there are also studies showing that the CDSS advice is frequently ignored. Eccles and his colleagues in England integrated evidence-based guidelines into physician office computer systems.2 Their research showed no change in the quality of care after this effort, but they also found that the guidelines were rarely read, much less actively used. Hsieh et al studied the use of allergy-drug interaction alerts and found that 80 percent were overridden by the clinicians who received them.3
Part of the reluctance to use available systems may lie in the dilemma faced by the physicians in this case—that the evidence is conflicting or that the unique clinical situation justifies overriding a generic guideline. But there are other reasons the advice is not followed—reasons that may not be as clinically or ethically justifiable. Tamblyn et al found that when CDSS recommended a change in medication prescribed by another physician, physicians were reluctant to follow CDSS advice, but they were more comfortable changing their own treatment plans on that advice.4 As the evidence base for clinical decision making improves, and the technology to make that evidence easily accessible during the clinical encounter becomes available, it is likely that the standard of care may require use of these systems. There is even legal precedent relating to the use of technology in maritime law that, if found applicable to health care, could hold physicians liable for failure to use available technology, even if such use is not yet part of the standard of care.5
Risks of Using CDSS
The evidence for the benefits of CDSS in reducing harm to patients has led the Leapfrog Group to make clinical decision support in physician order entry systems a cornerstone of its patient safety agenda.6 Yet there are also risks in using this technology. If Drs Anderson and Monde decided to use automated guidelines, which guidelines should they use? When guidelines are fully or mostly integrated into an order entry system, the source of the recommendations may not be known to the user. The invisibility of the source may be particularly problematic when the data conflict, as in this case.
Knowledge base maintenance may also be a problem in at least 2 respects. First, in this case, some of the guidelines were up to 4 years old, and it was not always clear whether the guidelines were not updated because the evidence behind them had not changed (and hence, the guidelines were still valid) or whether the evidence had changed and the guidelines had not been revised to reflect that evidence. Second, while accurate physician documentation is an ethical and legal imperative, documentation in clinical charts is often far from complete and may not be accurate. When this incomplete information forms the input for clinical decision support systems, even a well validated system may not produce accurate results, potentially putting patients at risk for harm. Hsieh et al found that many clinician overrides of CDSS drug-allergy alerts were attributable to inaccurate allergy lists in the paper chart as well as in the automated system.3 One growing concern is over-reliance on a faulty system because the clinician may fail to recognize when the system is incorrect. Tsai et al found that residents' interpretations of electrocardiogram (ECG) data when they had access to an automated decision support system that gave an incorrect interpretation were worse than their unaided performance.7 Because legally and ethically the physician—and not the computer—is still the decision maker, clinicians must be able to judge how to identify appropriate sources for clinical decision support.
Guidelines for Use of CDSS
The Health on the Net Foundation has developed a code of ethical practices for Web sites that provide health information . Many of these guidelines apply to clinical decision support systems as well, eg, the knowledge base should be credible, based on best evidence, and updated regularly, and the source of the data should be clear to the user. To prevent users from relying on obsolete information, at least one of the drug databases available for PDAs provides regular updates and disables its system if the user does not access the updates. The systems should be well validated, and the American Medical Informatics Association and other organizations have advocated local software review committees to monitor performance within each organization to assure that systems are performing properly in the local environment.9 There should be a mechanism for physicians to override the system, but they should also be required to document the reason for the override, so that the reasons can be reviewed, discussed if necessary, and ultimately be used to improve the system itself.
Finally, the CDSS must be integrated into the workflow, or else even the best system will not get used. In this case, the Web site that compared the guidelines was useful for the doctors to study, but was probably too cumbersome, with too much information, to integrate at the point of decision making. CDSS fail to be used routinely if it takes too much time to enter data or to read and digest the information they provide. Also, too frequent alerts for inappropriate reasons can result in the important suggestions getting ignored.
In summary, Drs Anderson and Monde can be optimistic that there are technologies that can help keep them up-to-date, but they will need to be aware of the strengths and limitations of these systems to properly avail themselves of these potentially life-saving sources of information.
- Bates DW, Gawande AA. Improving safety with information technology. N Engl J Med. 2003;348(25):2526-2534.
- Eccles M, McColl E, Steen N, et al. Effect of computerised evidence based guidelines on management of asthma and angina in adults in primary care: cluster randomized controlled trial. BMJ. 2002;325(7370):941-944.
- Hsieh TC, Kuperman GJ, Jaggi T, et al. Characteristics and consequences of drug-allergy alert overrides in a computerized physician order entry system. J Am Med Inform Assoc. 2004;11(6):482-491.
- Tamblyn R, Huang A, Perreault R, et al. The medical office of the 21st century (MOXXI): effectiveness of computerized decision-making support in reducing inappropriate prescribing in primary care. CMAJ. 2003;169(6):549-556.
The T.J. Hooper, 60 F 2nd 737 (2d Cir 1932).
The Leapfrog Group. Leapfrog Fact Sheet. Available at: http://www.leapfroggroup.org/about_us/leapfrog-factsheet. Accessed October 18, 2004.
- Tsai TL, Fridsma DB, Gatti G. Computer decision support as a source of interpretation error: the case of electrocardiograms. J Am Med Inform Assoc. 2003;10(5):478-483.
Health on the Net. HON Code of Conduct (HONcode) for medical and health Web sites. Available at: http://www.hon.ch/HONcode/Conduct.html. Accessed October 18, 2004.
- Miller RA, Gardner RM. Summary recommendations for responsible monitoring and regulation of clinical software systems. Ann Intern Med. 1997;127(9):842-845.