In Conversation With... Barbara Drew, RN, PhD
Editor's note: Dr. Drew, a nurse researcher, is the David Mortara Distinguished Professor of Physiological Nursing and Clinical Professor of Medicine in Cardiology at the University of California, San Francisco. She also founded the ECG Monitoring Research Lab in the School of Nursing and mentored numerous graduate students pursuing studies in the field of electrocardiology. We spoke with her about the perils and prevalence of alert fatigue.
Dr. Robert M. Wachter: What got you interested in alarms and alerts?
Dr. Barbara Drew: The wakeup call for me was the incident in 2010 where a patient was being monitored at a prestigious medical center on a telemetry ward. He was walking around the unit the day of the event with his wife and waiting for a heart pacemaker to be put in. A bunch of low heart rate alarms sounded, but the nurses who worked that day—10 in fact—never recall hearing those alarms. Then the patient's crisis level alarm went off because his heart rate went below 40, and no one saw or heard that either. The patient was discovered dead. So when the Centers for Medicare & Medicaid Services came in to investigate, they said that the nurses not having heard these alarms were an indication of alarm fatigue and that it contributed to the patient's death.
That provoked me to contact our nursing administrator at UCSF to ask, "Do you think we have the same problem here at UCSF?" And she said, "Let me come to your office. I'll be there in 5 minutes." And in the 5 minutes that it took for her to walk to my office, which is adjacent to the Medical Center, she snapped three smartphone pictures, all indicative of alarm fatigue. Things like sound speakers turned to the wall to muffle the alarm sounds. An alert on a central station that said, "Contact biomed immediately." The alarm speakers had been unplugged; things like that.
RW: When you first began investigating this, did you have a theory as to what the problem was, and did that shift as you came to understand it more deeply?
BD: When I say that I was alerted to it in 2010, I already was aware of what I call "alarm creep." I first started with cardiac monitoring clinically way back in the late 1960s, when I was a student nurse and then a new nurse—it's when hospital monitoring began. Every year the monitor manufacturers have been eager to add more parameters, just because they could add more things to monitor. And of course they're in fierce competition with each other—these companies so often use these as selling points. But every added vital sign, parameter, or value that they were adding added another layer of alarm. Before you knew it, we had this myriad of alarms—many of which are only nuisance alarms because they're not clinically actionable.
So it's been slowly coming on for a long time. It made sense to me that this would happen. Over the years, engineers for industry have come to me frequently and said things like, "Hey guess what, we can now digitally record blah, blah, blah, or x." I would think about it and ask, "Well why would you want to do that? That would just be irritating. We don't need to know that." But just because they could, they often added these features without real thought by anybody about what is this all going to result in.
RW: Interesting. So it sounds like the old story about the frog and the hot water. It crept up on us and by the time people woke up to the danger, we were already pretty far in the boiling water.
RW: You've done a lot of research on this issue and found some startling statistics. Can you highlight some of the key statistics that gave you an understanding of the breadth and depth of the problem?
BD: I have been in this cardiac monitoring business for over 40 years. But I had absolutely no clue of the staggering number of alarms that we would find in our intensive care units (ICUs) at UCSF. We collected every unique alarm in a 31-day period in our 5 adult ICUs. We had an average census of 66 patients per day during those 31 days, and we captured a total of more than 2.5 million unique alarms in that period. When I saw that number, I thought there was a mistake—but there wasn't. I remember talking to our chief medical officer the same day. I said, "Hey, guess how many alarms we had in a 31-day period?" And he said, "I'm sure there were a lot. Were there 100,000?" And I said, "No, up." "Were there 200,000?" But none of us had any idea that there were so many alarms. Nurses had told us that there were and that it was becoming background noise to them, but this number really was shocking, even to me.
RW: I use that number in my book and periodically with audiences, and 100,000 does seem to be the number that people say. They think they're being kind of wacky when they say 100,000. They think: It can't be that high. Then when you say 2.5 million, they're flabbergasted. When the data came out around the real numbers, did nurses find that validating in some way—we've been complaining about this for years and no one has taken us seriously. Did you hear any of that kind of feedback?
BD: All the time. Nurses have struggled with this. They have complained about this. They often interface with industry, with the sales people. Once the sale has been made, there's very little incentive for anything to be done about it from the company's point of view. But nurses have been living with this, and they often say to me, "And those are just the cardiac monitor alarms. We have infusion pump alarms, bed alarms, and ventilator alarms, and on and on and on." And they're absolutely right. This is just one cause of the cacophony of sounds in a patient's room that not only interrupt the patient's rest and interfere with conversations doctors and nurses have with patients, but it's also a real safety concern.
RW: In part because of your work, we've thought more about the impact on the nurses of these alarms going off. What does it feel like to be a patient or a family member in a room where these are going off every 7 minutes?
BD: It's interesting that you ask that because this topic really resonates with patients. When my niece's newborn was in a neonatal ICU shortly after birth, she called me in the middle of the night just so stressed because the baby's monitor was sounding all the time. She thought the baby was dying and nobody was coming. But I've talked to other patients who say they had to go home to get a good night's sleep because of all the environmental noise in the room. Patients see their monitors in the ICU because they're big colored flat-screened devices. They see the flashing numbers and hear the alarm sounds, and often the families who are visiting also see that. I talked with a husband whose wife was in the ICU and he said, "I literally went out in the hall and did jumping jacks to try to get the attention of the nurse down at the other end of the hall to come and see what was wrong because the cardiac monitor was alarming." This topic resonates with anybody who has had a loved one or has been in the hospital. On our patient satisfaction questionnaire, we always struggle to get a good rating on noise.
RW: Could you recount that story of you asking that nurse about what would make her worried about a patient?
BD: I asked the nurse, "With all these level of alarms—we have crisis, warning, advisory, textual—what would really catch your attention that something was wrong?" She thought about it for a second and she said, "If I came in the unit and everything was quiet—if there were no alarms, no sounds, I would know something was wrong."
RW: That's just such a wonderful story that captures the essence of the problem. Whose fault is this?
BD: The monitor manufacturers have obviously erred on the safe side and tried to make an alarm for every little change so that they never miss an event. They're worried about the legal consequences. They have such a sensitive device, but then that puts it all on the user's back to filter through the false alarms before they find one true alarm. The device manufacturers bear some of the responsibility here, and the FDA does too in their postmarket surveillance. I have been to the FDA to talk to them about this. They do have an alarm fatigue task force now looking into the problem and trying to formulate an effective response.
The other thing is that clinicians want to find one size fits all. Alarms have default settings for what is too high and what is too low for every vital sign measurement and every alarm parameter. But everybody knows one size doesn't fit all, and clinical staff need to tailor the alarms for their individual patient when their patient is an outlier and doesn't fit into the hospital default. That would save a lot of false alarms.
Then sometimes the patients are at fault. Not knowingly, but in our study we found a few outlier patients who drive almost all of the arrhythmia alarms. These might be patients who are agitated and have a lot of motion artifact that triggers a lot of false arrhythmia alarms. We have to look at these outlier patients and weigh the benefits and risks of monitoring them. Maybe it would be better for all the other patients to take that patient off the monitor if it isn't critically important for that individual—there wouldn't be so many false alarms and other patients who have true alarms would be noticed. We need to look at this issue in many different ways to solve it.
RW: It strikes me that the incentive for manufacturers, and maybe even for the individual hospital, from a risk management standpoint is to make sure the alarm parameters stay pretty tight. We're so worried about missing something that the idea that people are going to get 100 false alarms for every real one is acceptable. If hospitals or the manufacturers were held harmless from a malpractice perspective and then made thoughtful decisions about alarm parameters, they would be protected from liability if a patient did have something go wrong. Does the malpractice system need to change on this?
BD: Probably. Although the case that I talked about earlier is one in which the company might be sued because that had so many false alarms that they went unnoticed. In the 31-day period that we analyzed, we had 17 cardiopulmonary arrests in these ICU patients during that period, and all 17 cases were heralded by numerous arrhythmia, heart rate, systolic and diastolic blood pressure, respiratory—all kinds of alarms. Not one of them was missed. We could eliminate a lot of alarms before we would reach the point where we would worry about missing a real important event.
RW: With the current technologies, through the interventions you've already talked about—tailoring the alarms to the individual patient parameters so you don't use the default settings and looking for patients whose physiology or current issues are causing the alarms to get set off a lot—what percentage of the alarms do you think you could get rid of safely?
BD: Probably half of them.
RW: That's not that encouraging, if you're at 1.2 million. I was hoping you'd say 90%, but you don't think so?
BD: No, but I think that there are other things that we could do. Our lab at UCSF is interested in creating smarter alarms that combine multiple measurements. Currently the way all devices work is that separate measurements (EKG, blood pressure, SpO2, respiration) all have separate alarms, but they don't talk to each other. They don't say that this is not motion artifact, but it's really truly VT [ventricular tachycardia] because the blood pressure fell off, the SpO2 went down, or other things changed. If some of these alarm parameters were fused together, we could make a much smarter device. For example, atrial fibrillation is a very common arrhythmia in older people in the hospital, and there is an atrial fibrillation alarm. But for somebody who's in chronic atrial fib, we don't need to know every 5 minutes that the patient is still in atrial fib. So, while you wouldn't want to silence that alarm for other people who are not in atrial fib, these things could be improved.
There's no reason why alarms couldn't be more intuitive, like smartphones. When you buy a smartphone you leave the store and you figure out how to use it pretty much without any in-service or any orientation. Yet it's a very complex device that does a lot of different things. These monitors are computer devices, so there could be prompts that would come up and help the user—the nurse or staff. For example, a prompt could come up indicating that this patient's heart rate has averaged 135 for the last 5 minutes, and ask, "Do you want to increase the high heart rate alarm limit because our hospital default is 130?" We are a generation who knows how to use that; we have smartphones and interact with this stuff all the time. I'm trying to encourage the cardiac monitoring companies to build more intuitive devices because we have all levels of staff. We have travelers [nurses from other sites] and new graduates who are unfamiliar with this equipment in our hospital. So the more that these devices can be user-friendly and interactive, we could really improve the safety of our care.
RW: You just brought up a potential conundrum: part of the challenge is young grads, travelers working in unfamiliar circumstances. In some ways, the argument centers on whether you want to increase the parameter to 135, or should the nurse have the authority without having to contact the doctor to do that. Is there a tension of professional boundaries—particularly whether the nurse has both the training and then the appropriate authority to change the parameters in ways that would ease alarm fatigue?
BD: Yes, that's an interesting question because I've often wondered if we shouldn't have a contextual type of alarm, in which the nurse starts her shift and there is a prompt asking, "Who are you?" And, "Please rate yourself how much you know about the monitor." If you're a novice, then 1, and up to 3 if you're an expert. And then have the monitor behave differently if you're a novice than if you're an expert. I've talked to monitor manufacturers about this idea of making it contextual. They do this in other areas of industry. That could be thought of in the future whether that would ever be something that we could consider doing.
RW: When I met with the IBM Watson people, their vision for the future encapsulates a lot of what you're talking about, but even goes beyond that contextually. I want to see whether you think this is realistic in any of our lifetimes. That the IT system knows who you are, knows that this is a patient in the cardiac ICU who has this disease and has been having these problems, and knows that in the last 1000 cases an alarm has fired in this type of patient—in this situation, at this time of day—it's always been a false alarm and therefore it should not fire in this patient. In other words, it's taking into account a huge amount of big data, the context of the patient, the context of the provider, time of day, geography, and then making a decision about the test characteristics of the alarm at that particular moment and either firing or not firing. Does that sound like science fiction, or can we get there in 5 to 7 years?
BD: Well there's a lot of work on prediction algorithms. One of my colleagues at UCSF, Xiao Hu, has been funded by NIH to look at what he calls a super alarm, which takes into account various parameters from the electronic health record and lab work and tries to create a predictive algorithm about whether you as a patient are stable or going in the right direction or in the wrong direction. So the nurses would know who's in trouble ahead of time, before the actual code blue happens. Other centers are looking into these predictive algorithms too. Now with everything being electronic, there's no reason why there couldn't be some of these helpful predictive algorithms. We're monitoring so much data that if we could figure out what to look at, it's reasonable to think that in 5 years we could have much more clinically meaningful alarms.
RW: How does the issue of alert and alarm fatigue play out outside of the ICU?
BD: Alarm fatigue is probably a much bigger problem on a general medical floor than in the ICU. It's easier for us to study it in the ICU, so we chose to do that. However, this study really needs to be replicated on general medical floors. Many of those patients are also on cardiac monitors, and there are fewer nurses per patient. The patients in the ICU are often sedated, and they're quieter because they're on mechanical ventilation or they have invasive lines and they lie more quietly in bed, so there's less motion artifact. But patients on the general floor walk in the halls; they go to the bathroom. They can trigger a lot more alarms. So the real elephant in the room is what happens in terms of alarm fatigue on the general medical floors. That needs to be the next thing to be talked about. Our study, while shocking, with more than two and a half million alarms in a month, is just the tip of the iceberg of the problem that we would see if we studied it in other environments.