Skip to content Skip to navigation

Doing Alerts Right

January 31, 2008
by James Feldbaum, M.D.
| Reprints
Implemented correctly, alerts can greatly enhance a CPOE system, but done wrong, they can send physicians back to pens and paper

James feldbaum, m.d.

James Feldbaum, M.D.

The electronic medical alert arguably represents one of the best opportunities to improve the quality of patient care as part of computer-assisted decision support. In its most common, albeit primitive form, it is the allergy alert. In its most advanced manifestation, it is the unsolicited life-saving knowledge transfer from the electronic medical record to the practitioner in real time at the point of care.

Despite its promise, medical alerts are often despised as interruptions to medical work and thought flow; they are considered superfluous and ignored, or become the genesis of what has become widely known as “alert fatigue.”

As part of their exuberance implementing electronic medical records, physicians and pharmacists often rush to create electronic medical alerts in hope of harnessing the real clinical power of the new electronic systems. Unfortunately, it is this unbridled and well-meaning enthusiasm that often leads them astray. In this commentary, I would like to outline some rules, steps, and processes that can align the desire to implement alerts with the likelihood of succeeding in the effort.

The first question is most frequently about timing. When is it best to begin thinking about and crafting alerts? Clearly, some alerts, like allergy and drug interaction checking, come with our software right out-of-the-box. Alerts which would seem innocuous are most often overly inclusive and instigate over-alerting and alert fatigue.

In an article published in the “Archives of Internal Medicine” physicians overrode 91.2 percent of drug allergy and 89.4 percent of “high-severity” drug interaction alerts. There was not a statistically significant number of adverse drug events (ADEs) between the two groups. So, for openers physicians will find themselves in need of a process and a governance to deal with all of the alerts and the perception of over-alerting.

Let's begin by defining a step-by-step thought process:

Why are we even thinking about an alert? Is there really a problem? Have we performed any metrics that clearly define this issue as a problem? The need to comply with core measures, regulatory deficiencies, or documented incidents makes a clear case for action, but more often our issues are those of perception that often fail to reach problem status in the face of closer analysis.

Will this particular alert improve care? Once we have clearly documented that a problem exists, will a well-designed alert solve it? Is an alert the best way to achieve our goal? Have other modalities of intervention (education) been tried? Have we established metrics to assess the impact of the corrective modalities that we implement?

Who should be the target for this alert? Are some medication alerts better or more efficiently screened by non-physicians or do they need to appear synchronous with ordering?

What is the urgency for this alert? Is the alert so critical that it warrants interrupting an order session?

How best do we display the warning? Many studies have documented the ineffectiveness of non-obtrusive warnings. How provocative do you want your alert display to be? A medical center in Jerusalem employed a flashing screen visible whenever an individual patient's chart or entire ward's results was accessed on any hospital computer. It was effective.

What steps must a practitioner make to comply with an alert?

Next, assuming there are significant improvements in care to be achieved, let's describe six qualities of a well-designed alert:

An alert must access only accurate high quality data. An alert based on patient weight or home medications, both notoriously unreliable, will create false alerts. Advanced alerts need to rank, calculate, and compare data in order to determine when parameters for alerting are met. It is garbage in, garbage out all over again.

It must present information to the right person at the right point in the ordering workflow. Do not underestimate the cost in time and effort for a practitioner to respond to an alert.

It must improve care or outcomes. Do not assume that an alert has had the anticipated effect. Measure it. If it doesn't work, rethink it, redesign it or trash it. Avoid alerting for the sake of alerting.

It must be based on sound and compelling science. Is it controversial? There is much national debate about the standard of what we refer to as “best practice” or “evidence-based medicine.” Define how your organization obtains consensus on clinical decision making before crafting alerts.

It should make the act of compliance simple and fast. Many programs employ “actionable-alerts” affording the provider the opportunity to take recommended action within the alert window itself. Conversely, an opportunity must be given for a practitioner to document the rationale for bypassing an alert.

It must consolidate data that either by its volume or subtlety a practitioner might miss when caring for the patient.

Personally, I like alerts conceptually and I like the thought exercise that crafting them entails. But with all of the caveats described above, how can an organization get the courage to fire one up? The answer is simple. Run them in the background. Imagine them, discuss them, research them, craft them, monitor them and refine them. Just don't alert anyone. When an electronic alert seems ready for prime time, choose carefully the target for the alert.