Dear Editor

I share much of Dr Morris' enthusiasm for decision support. While involved in an evaluation of a decision support targeting vancomycin ordering practice,[1] I had the opportunity to observe the potential impact of this approach. However, as a clinician using the same computerized order entry system in daily practice, I also recognized the major limitation of this approach: users will not tolerate many such interventions at any given time.

Too many alarms can contribute to a sense of general noise, so that they lose their value.[2] Similarly, too many triggers for guidelines, or even alerts such as potential drug-drug interactions, will result in users clicking past all such screens, so that in the end, none of the screens will work. Forcing users to follow a given guideline or respond to a particular alert runs the risk of adding an intolerable time burden to frequent users of the system – e.g., interns and residents who write multiple orders a day.

Consider the admission orders for even a routine medical admission. Possible guidelines might relate to many medications (e.g., choices of antibiotics), diagnostic orders ('does this patient really need a KUB – the yield of plain abdominal radiographs is known to be low in most clinical situations'[3]), various prophylactic strategies (“do you want to order DVT prophylaxis?' '...stress ulcer prophylaxis?' etc.)

Elderly patients will trigger even more guidelines - does the patient need fall precautions? How about a soft matteress or other decubitus ulcer precautions? Pneumococcal vaccination prior to discharge?[4] Vlu vaccine? [4] Does the patient have an advanced directive?[5]

The list goes on, and this does not even include guidelines triggered by specific admitting diagnoses e.g., guidelines for treatment of community acquired pneumonia, acute coronary syndrome, hip fracture, gastrointestinal bleeding, stroke, etc. etc, not to mention important secondary diagnoses – 'This patient has diabetes: do you want to add an angiotensin converting enzyme inhibitor.' 'This patient has a diagnosis of congestive heart failure, but there is no record of an echocardiogram or other assessment of ejection fraction.' 'This patient is on prednisone; would you like to add a bisphosphonate to protect against osteoporosis?'

Thus, computerized systems offer a greater chance of success for implementation of a single guideline, it is unlikely that this benefit will generalize to more than a handful of such protocols at any given time. Further research will need determine optimal strategies for harnessing the potential of computerized decision support. Currently, though, it is unrealistic to think that an institution acquiring an order entry system could expect to impact practice in more than a few areas using this approach.

References

(1) Shojania KG, Yokoe D, Platt R, Fiskio J, Ma'luf N, Bates DW. Reducing vancomycin use utilizing a computer guideline: results of a randomized controlled trial. J Am Med Inform Assoc 1998;5:554-562.

(2) Cropp AJ, Woods LA, Raney D, Bredle DL. Name that tone. The proliferation of alarms in the intensive care unit. Chest 1994;105:1217- 1220.

(3) Harpole LH, Khorasani R, Fiskio J, Kuperman GJ, Bates DW. Automated evidence-based critiquing of orders for abdominal radiographs: impact on utilization and appropriateness. J Am Med Inform Assoc 1997;4:511-521.

(4) Dexter PR, Perkins S, Overhage JM, Maharry K, Kohler RB, McDonald CJ. A computerized reminder system to increase the use of preventive care for hospitalized patients. N Engl J Med 2001;345:965-970.

(5) Heffner JE, Barbieri C, Fracica P, Brown LK. Communicating do-not- resuscitate orders with a computer-based system. Arch Intern Med 1998;158:1090-1095.