Reply to Atreja et al.

We thank Atreja et al. for their response to our article, “Countrywide Computer Alerts to Community Physicians Improve Potassium Testing in Patients Receiving Diuretics”,1 which appeared in JAMIA Nov/Dec 2003. Their comments raise some interesting points, which we would like to address.

Atreja et al. affirm that no evidence-based data exist supporting annual testing of serum potassium levels in patients receiving diuretics. This issue has, in fact, been addressed, and periodic potassium testing has been recommended outside of the pharmaceutical industry.2 Clearly, in the absence of specific testing frequency guidelines, clinical judgment should remain the main determinant of testing frequency. This is not to say that no testing is required. The computer-generated potassium reminder was introduced in Maccabi Healthcare Services due to the ubiquity of prescriptions for diuretics in our middle-aged and older population combined with an unexpectedly large proportion of patients not having a potassium level determined either before or after the institution of diuretic therapy.3 Atreja et al. suggest that some physicians might not want to test their patients for hypokalemia because their levels have been stable for years, while others test just to avoid further reminders.

We agree that physicians who did not order blood tests could have a valid clinical reason for ignoring the computer-generated reminder, and this response is completely valid. Our reminder uses carefully neutral language to alert the physician that a blood test relevant to patients taking diuretics has not been performed for one year. The text, as reproduced in our article,1 states the name and date of the most recently purchased diuretic and notes that no potassium blood test was performed during the previous 12 months. We certainly don't say: “Do a potassium test and do it now!” The recommendation to test may be implicit but certainly is not mandatory. Anecdotal physician response to these computer-generated reminders has been favorable, and no complaints about the reminder were received by our department from physicians.

The implementation of our alert system was “cheap and simple” due to its seamless inclusion into the physician's workflow. We consider it unlikely that many physicians ordered laboratory tests “against their clinical judgment” just to avoid repeated reminders. The proportion of reminder screens in the laboratory results flow was low for most physicians. Reminders that vanish from the computer screen with a click of the mouse button cannot be considered onerous, and no great effort is required to ignore them.

Although no cost–benefit analysis is planned, it is reasonable to assume that there were increased costs beyond the cost of potassium blood tests themselves. However, we view any medication changes, follow-up tests, or other expenses, resulting from our reminders as an acceptable part of medical practice based on the clinical judgment and responsibility of the prescribing physician. In fact, we assumed that physicians would take the opportunity afforded by the reminder to review and perhaps order other blood tests where comorbidity exists.

Certainly, an analysis of potassium blood levels before and after the intervention would have been welcome. However, technical hurdles prevented the inclusion of these data, putting explicit outcome measures beyond the scope of this study.

This study represents a pilot intervention, and we will be refining our targeting of alerts using the existing medical information system. Focusing on subpopulations and modifying reminder frequency using risk stratification,4,5 such as age, previous blood chemistries, or recent drug changes, should enhance the efficacy of these interventions.