Alert Fatigue

In July we witnessed the last of the NASA space shuttle flights as Atlantis blasted into orbit.  I recall being in high school as the shuttle program was in its stages of final development and writing a report on the new technology.  I still vividly remember the National Geographic article (from which I scalped most of my information—cheating from Wikipedia was not yet invented) that promised a future of cheap and frequent trips into space.  The original specs of the shuttle program estimated that we’d see around 50 flights a year and that each would cost taxpayers a paltry $15 million.  Instead, the reusable orbiter made it into space about 4 times annually at an average cost of $1.5 billion per launch.  In defending the shuttle program and its apparent shortcomings, the original program director Bob Thompson clarifies that it did exactly what it was designed to do:

"The shuttle gave the country a very capable work vehicle to go into low Earth orbit and do everything from carrying experiments there to launching satellites from there, to building space stations, to building telescopes like Hubble. It was a multipurpose work vehicle."

The space shuttle provided the basic service—Fed Ex service to earth’s orbit—but failed to live up to its vaunted potential.

The discussion surrounding the winding down of the shuttle program reminds me of our experience with our newly initiated electronic medical record (EMR).  For about a year now we’ve had this system up and running and at this point the general consensus among my colleagues is that it does what it was designed to do: provide a computer-based repository for our medical records.

What it ideally could do is an entirely different matter.  In my mind it has yet to live up to my expectations of a system that actually improves the care I provide my patients.  A vivid example of this is a concept known as alert fatigue, a modern take on the “boy who cried wolf” motif.  If you test the high school fire alarms only once a year you’ll get a pretty brisk response from the impassive teenagers who diligently hustle toward the exits.  Test the alarm once a week, however, and when a real fire sparks up in the boys’ bathroom or the chemistry lab you’ll find that no one pays attention to the alert.

Weather forecasters recently began debating the relative dangers of alert fatigue when it comes to tornado warnings especially after the recent spate of twisters across the country.   Sirens, they note, are heard so often that they no longer elicit the appropriate response from the public.

In the world of medicine we’ve begun to experience our own form of alert fatigue thanks to EMR.  In theory, EMR should be our best friend when it comes to prescribing medications for our patients.  As the world of pharmacology becomes more diverse and complex, and as doctors find themselves increasingly practicing in narrower niches of specialty, we are more likely than ever to prescribe medications that can potentially interact with other drugs or have adverse side effects that could easily be avoided.  Prior to EMR we relied on our pharmacy colleagues to catch our occasional prescribing mistakes and give us a phone call to tactfully remind us of our slip-up and suggest a safer alternative.  But pharmacists aren’t perfect either and can feasibly miss a troublesome interaction here or there.

Not so with the computer.  It has the capability to draw upon massive amounts of information in order to immediately assess the safety and appropriateness of any drug I try to prescribe and provide a warning when necessary.  It never grows tired and it never forgets the most rare side effect or interaction.  In theory it should be the perfect companion in our attempts to minimize drug-related adverse events.

What really happens, as with the case of the well-intended space shuttle program, barely mirrors the preconceived ideal.  Our current EMR produces so many warnings about nearly every single medication—from toxic chemotherapy drugs to the most innocuous lozenge—that doctors very quickly learn to ignore all warnings coming from the computer system.

For example, just this week my schedule was quite heavy with office work and I subsequently visited with dozens of patients.  With each I offered new prescriptions for their ongoing cardiac medications—the EMR actually makes prescription refill quite easy.  I found that I wasn’t able to refill a single medication without at least one dire warning arising about each drug.  The system seemed to demonstrate no qualitative bias, throwing up as many red flags about aspirin, for example, as it did for medications such as warfarin, amiodarone, and digoxin—all drugs that require active diligence to avoid interactions.

One particular example is illustrative of my frustration.  I had a patient on amlodipine (a blood pressure pill) and wanted to ratchet up the patient’s medications to improve control of hypertension.  An easy way to do this is to provide a medication that combines amlodipine with another agent.  I entered into the EMR the new drug I wanted to add, but did so prior to deleting the previous prescription for amlodipine.  Alerts went up, as expected, but what surprised me is that the computer program failed to detect my apparent duplication of the drug amlodipine and instead brought to my attention a cautionary warning about the use of blood pressure drugs in elderly patients (he was 76).  Thus, it failed to catch what could have been a legitimate, problematic error and only succeeded in reminding me that my patient is old.

We see alerts every time we prescribe clopidogrel (Plavix) and aspirin together despite the fact that Plavix is almost never prescribed in the absence of aspirin.  Whenever we combine cholesterol or blood pressure medications—a practice common in a majority of cardiac patients—we get alerts.  If ever I try to add a drug from the ACE-inhibitor class I get no fewer than 3 red flags even when there is no real reason not to use these medications (which, by the way, serve as the backbone of care in many cardiac and renal patients).  In the end I get so many warnings that I find myself reading none of them.

I’m not alone in this frustration.  A 2009 study published in the Annals of Internal Medicine showed that I’m not the only doctor blowing through the deluge of unhelpful warnings.  An evaluation of three and a half million electronic prescriptions among nearly 3000 doctors showed that prescribers overrode the alerts more than 90% of the time.

My objection to our current system might be just the ranting of a frustrated blogger if not for one particular problem.  The belief that an omniscient computer database has reviewed our new prescriptions and an experienced physician has read and consciously overridden each alert lulls us into a sense of security that belies the reality that we doctors have become numb to all the alerts and are simply clicking through them without paying much heed.  We are inviting danger by believing in our system’s ability to catch drug errors when we know that the real dangers are needles in a haystack of false alarms.

I can only hope that our pharmacists recognize these realities and continue to provide the service we’ve come to rely on for years prior to the advent of EMR.  As doctors we still need to remain well versed on the subject of drug interactions and side effects since I’m sure we won’t be able to rely on our EMR to fill in the gaps as we’d hoped.  Perhaps at some point our EMR will live up to its potential as a tool that allows us to really improve the care we provide to our patients.  Until then, the best we can say is that it does no more than what it was designed for—turning our paper charts into ones and zeros for storage on a hard drive.