ABSTRACT

DEAN F. SITTIG, JONATHAN M. TEICH, JEROME A. OSHEROFF, and HARDEEP SINGH

State-of-the-art electronic health record systems with advanced clinical decision support (CDS) capabilities can fundamentally improve quality and reduce costs of health care. [1,2] However, these outcomes have not been universally achieved.[3,4] As the study by Fiks et al [5] in this issue of Pediatrics demonstrates, providing CDS in the form of “alerts” to encourage desired health care activities may not be sufficient to make a substantial impact. [6] Maximizing the potential of CDS for improving quality and safety of care requires attention to several factors, not all of which are related to the computer system.[7]

The goal for the study by Fiks et al was to increase vaccination rates in asthmatic children, so in examining the results one must fi rst consider what caused the low vaccination rate in their population. Several factors could account for the low initial vaccination rates and, hence, could explain the minimal improvements with alerting. Without knowledge about these factors, it may be too much to expect alerts alone to fi x the problem. Alerts are helpful when an unusual occurrence must come to a physician's attention or when a necessary process might be overlooked in a busy encounter. When other underlying problems lead to low vaccination rates, such as poor patient acceptance, difference of opinion about vaccinating patients late in the season, or low priority of vaccination when a patient has an acute problem, they must be addressed before the alert can be

successful. Indeed, studies of infl uenza vaccination reminders in adults have had varying results, and in some cases these results were directly attributable to such noncomputable factors. [8,9] It would have been enlightening if the decision support used in this study also captured the reasons for failure of the providers to act on the alert by having them select or enter a reason for nonvaccination. [10]

In addition, one should also consider whether presentation of the vaccination alert as soon as the patient encounter was opened within the electronic health record was the best CDS intervention to achieve the desired objectives, compared with other intervention types such as facesheet displays, order sets, patient education handouts, and end-of-visit forms. In a guide to CDS implementation that we published in 2005, [11] we suggested that different types of CDS presentation, applied at different parts of the visit workfl ow, can have very different effects depending on what it is that one is trying to encourage the physician to do. Moreover, communication through group academic detailing (used in this study) may not be the best strategy to educate and change the behavior of clinicians regarding the concepts behind clinical alerts. [12,13]

To achieve a specifi c clinical objective by using a CDS intervention, one must consider whether the communication and acceptance groundwork has been laid to maximize the intervention's impact, and also consider what type of CDS, applied when in the encounter, is likely to have the greatest impact. [14] From the aforementioned CDS guidebooks [11,14] and other published reviews of CDS effectiveness factors, [15,16] we support the following list of questions to consider before the implementation of any real-time, point-of-care CDS intervention designed to interrupt clinicians during their work.