Christopher Bonafide
1. One of your research interests has been alarm fatigue and reducing unnecessary physiologic monitoring. In a discussion of one of your recent articles (https://jamanetwork.com/learning/audio-player/14307072), you report that half of the alarms you studied were invalid and <1% were actionable or required an intervention. Can you give us the bottom line on how hospitals (and we as pediatric hospitalists) can do a better job with this? What sorts of QI measures might our readers begin implementing at their own institutions?
Yes – as a resident I was completely oblivious to the idea that there could be a downside to physiologic monitoring. I probably put 90% of my patients on continuous ECG and pulse ox monitoring without a second thought. I truly thought that when alarms went off, nurses dropped everything and responded. Monitoring felt like a safety net. Then I had this a-ha moment as a new attending where I was having a conversation with a nurse about a patient who was getting worse, and her phone (which was being sent a text every time a monitor alarm sounded) just kept beeping and vibrating over and over. Finally, she set it down on a counter, walked away, and said “let’s go talk over here, these alarms are driving me nuts.” I realized that the entire system of continuous monitoring was broken – it was not working as it was intended to work, and it had the potential to fail the patients who were truly deteriorating and would actually benefit from a noisy alarm and an interruptive text to a nurse. So my team started thinking about how we could measure alarm fatigue. We threw around lots of ideas. Follow nurses around with stopwatches and time how long it takes them to respond? Put fancy tracking devices on nurses and get data that way? Maybe just ask them how long it takes them to respond? None of those made sense, so we ended up using video cameras – little GoPro cameras actually – to measure alarm fatigue. In an early pilot, we even had nurses wear GoPro cameras while caring for patients in the PICU. That was super interesting but the nurses hated it, so eventually we landed on just mounting lots of cameras in patient rooms and at the nurses’ station, which is how we performed the study that you mentioned. The main findings were of course that the vast majority of alarms were not actionable on a pediatric ward, and response time was slow.
So, what can we do? If you want to be part of a large collaborative effort, think about joining the PRIS Network (https://www.prisnetwork.org) and taking part in the EMO Study, which stands for “Eliminating Monitor Overuse.” Our goal over the next 2 years is to study the overuse of pulse oximetry in bronchiolitis across a large group of hospitals in the US and Canada. We will first measure overuse across 40+ hospitals and then use methods from the field of implementation science to collaboratively design a strategy to safely de-implement the unnecessary use of pulse oximetry in children under 2 with bronchiolitis who are not requiring supplemental oxygen.
At a local QI level, there are a number of different approaches you can take to reducing unnecessary alarms and hopefully thereby accelerating responses to alarms and alleviating alarm fatigue. Data helps – a lot. If you are able to get a sense of how many alarms are occurring, even if it’s just on 1 unit for 1 day, that can be very powerful in convincing others of the need for improvement. You could get that from the biomedical engineering department if you’re lucky and they have systems in place to track alarms (never hurts to ask – you might be surprised) or you could engage an eager student to count alarms from a central monitor station for a few hours at a time. Then, armed with some data, you can decide if you want to first address the problem of who is being monitored with strategies like implementing clinical pathways, order sets, or using audit-and-feedback, or address the problem of how patients are being monitored (for example, what settings are being used). Using classic QI methods like driver diagrams and impact-effort matrices can be really helpful in scoping out the problem, linking to key drivers, and mapping to interventions.
At CHOP we’ve found that one of the easiest ways to reduce alarms is to widen the default alarm parameters on the monitors – in other words, look at the settings for SpO2 low, HR high, etc that are automatically set when a patient is admitted. Are they age-based, or the same of all patients in the hospital? Are they appropriate? When we did that at CHOP, we realized that our default parameters were really conservative, much more so than most of our peer hospitals. So we made some modest changes, and reduced alarms by about 40%.
2. Last year, you wrote a piece in JAMA entitled, “The Emerging Market of Smartphone-Integrated Infant Physiologic Monitors.” In it, you talk about the growing number of parents scared into purchasing home vital sign monitors such as the “smart sock” monitor (Owlet Baby Care) at $250 each. Can you tell us how you address the worried parent of a bronchiolitis or BRUE patient who wants to purchase the Owlet or another home monitoring alarm?
Sure, this is a really interesting area. My main issues with the currently marketed monitors are: (a) none are FDA-cleared, so the companies have not had to demonstrate that they are safe or accurate, (b) the marketing of these products builds upon and benefits from the fear and anxiety of SIDS experienced by new parents and parents to be, and (c) the American Academy of Pediatrics Task Force on SIDS has issued recommendations that include “Do not use home cardiorespiratory monitors as a strategy to reduce the risk of SIDS,” so there isn’t any medical reason to gather medical data on healthy babies.
When I have parents in the hospital ask me about monitoring their healthy baby at home, I sit down with them and we discuss their very real fears and worries. We talk about the ABCs of safe sleep – the AAP recommendations that are based on evidence. We discuss what I’ve observed in testing these monitors – that they can fail to detect true problems (false negatives), and that they can also generate false alarms (false positives) that are frightening and might land them in the emergency room unnecessarily. Since, ultimately, my relationship with a hospitalized patient and their family is temporary and ends at discharge, if they are still interested in using a monitor despite my recommendations, I tell them that if they do decide to use one, they need to make a plan in collaboration with their pediatrician of what to do when the alarm goes off in the middle of the night. And then I’ll often give the pediatrician a heads-up about this discussion so they are prepared for it.
I think I’ve gotten a bit of a bad rap of being anti-technology or anti-innovation for my views on this. To be clear, I am incredibly supportive of companies that are innovating in the pediatric medical device space. It’s an underserved, relatively small market and we need all the attention we can get. I also think that in the future, everyone will be monitored – adults, children, all of us. The differences will be that the devices will be more accurate (less susceptible to false negatives and false positives) and will be smarter – in other words, we will know how to process and provide decision support to parents and clinicians for readings that are out of “normal ranges” but do not represent risks to the patient, to avoid overdiagnosis and overtreatment. For example, we know that many infants who are home recovering from bronchiolitis have occasional desaturations, and infants with these desats have the same outcomes as other infants without desats (https://jamanetwork.com/journals/jamapediatrics/fullarticle/2498407). But in our current environment, if a parent calls their PCP with an oxygen saturation of 89% on a home baby monitor, I’d bet 10 out of 10 PCPs would send the child to an emergency room.
In general, across hospital and home monitoring, I consider where we are right now to be the “awkward adolescence” of monitoring. We are in a situation where we have to avoid monitoring large groups of patients because we have not yet perfected how to use the data to benefit patients who are well or who have mild illnesses.
3. I’m a huge fan of your article, “Optimization of drug–drug interaction alert rules in a pediatric hospital's electronic health record system using a visual analytics dashboard.” In it, you describe how you used a group of pediatric clinical pharmacists to deactivate clinically irrelevant drug-drug interactive alerts, reducing alert rates from ~59 per 100 orders to 25 per 100 orders for pharmacists (also reducing it for providers). How difficult was it to get buy-in from front line providers and pharmacists for this project? How do you plan to maintain this project and reassess these alerts moving forward?
Thanks – that was really fun work to do, spearheaded by my friend and hospitalist colleague Eric Shelov, MD, MBI. The keys to the success of this project were taking a data-driven approach, partnering with pharmacy from the very beginning of the project, and broadly engaging stakeholders across the organization, including front line providers. Engagement from the front-lines was relatively easy since the project was going to decrease their alert burden and improve their experience with the EHR. Our clinical pharmacists did the “heavy lift” of the project, comprehensively reviewing drug interaction information from a variety of sources, but since their group was most impacted by the burden of false-positives, they were eager to contribute to this thoughtful reduction approach. The more daunting buy-in to get was patient safety leaders on board. They were concerned about the idea of turning off alerts. In the end, we were successful in developing an evidence-based proposal suggesting that the safety risk posed by alert fatigue was greater than that of deactivating carefully selected alerts. The project was operationalized within the Clinical Decision Support Committee at CHOP, which provided a forum for weekly review and discussion of alerts and multidisciplinary, data-driven decisions about alerts that could be safely deactivated in the system to reduce alert burden. I’m a bit removed from the project now, but Dr. Shelov now co-chairs the Clinical Decision Support Committee that oversees monitoring of these alerts. The group has recently expanded their focus to monitoring and optimization of other decision support tools, such as order sets and more sophisticated alerts that can evaluate multiple factors simultaneously.
4. You were the senior author on an article discussing implementation of a chest pain pathway in the Emergency Room (published Sep 2016 in Pediatr Emerg Care). With the pathway, there was a reduction in the rate of chest X-rays ordered, but an increase in outpatient follow-up (15 to 29%) with no visits resulting in diagnosis of a new cardiac condition. Does this mean that we should be “Choosing Wisely” and begin to do less with pediatric chest pain? How did the outpatient providers feel about the increased referrals/follow up?
This was an interesting project to be a part of. I view this project as a mostly negative study; the most interesting aspects to me were the outcomes that did not change after implementation. Here’s how the project started - a group of cardiologists embarked on a QI project because their perception was that there was a ton of excess cardiac-related resource utilization in the ED for chest pain. They were specifically concerned about excess cardiology consults, excess lab testing (e.g. troponins), ECGs, and echocardiograms. The problem was that they did not have baseline data when they started – their understanding of the issue was initially based on their Division’s perceptions only. They designed the clinical pathway and integrated it into the electronic health record. After implementation, they approached me to help mentor them, which I was glad to do. Together, we gathered the baseline data (which we of course should have gathered before even designing the intervention) and the post-intervention data. The most interesting findings to me were that (1) baseline resource utilization was already quite low before the intervention was implemented – lower than the perceptions of the cardiologists suggested, and (2) we did not observe changes in rates of ECG, blood tests, or echocardiograms. I think the reason for the lack of changes is that utilization of most of those diagnostic tools was already quite appropriate in the baseline period, although we couldn’t prove that point with the data we had. The lesson from this project is that it is critically important to obtain good baseline data to confirm your perceptions of a problem before designing and implementing an intervention- whether it’s QI or research. I don’t think the data from this project are strong enough to support a Choosing Wisely statement, but there is definitely room for more QI and research in this area.
5. You are Board-certified through the ABPM-Clinical Informatics Subspecialty Board and are a faculty member associated with Children’s Hospital of Philadelphia’s Clinical Informatics fellowship. How have you utilized this advanced training in your role as a hospitalist and what advice do you have for those interested in additional informatics training?
Clinical informatics is a great “niche” for hospitalists. When residents or fellows interested in PHM careers come to me for career advice, one of the first things I tell them is that they need to develop their niche – a relatively narrow area of specialization/expertise that gives them unique value within their organization and keeps them from getting burnt out from doing 100% clinical care all the time. Over time, if your niche truly does provide value to your hospital or department, it may be possible to negotiate for some protected time to devote to further developing your niche or sharing your expertise with others. As I have grown up as a hospitalist, and EHRs with comprehensive clinical decision support capabilities have become nearly ubiquitous, clinical informatics has developed into a really viable and rewarding niche. I have used my informatics background primarily to support QI and research endeavors. For example, a clinical pathway can be much more powerful if it has a robust order set behind it, and especially if that order set presents itself prominently when a patient with the right characteristics shows up on your service as a new admission. In research, if you know what the clinical data look like and where to find them in the hospital’s data warehouse, the data become less mysterious and your research efforts become much more nimble.
6. Jack Percelay often ends his list-serve commentary with the phrase “that’s just my 2 pennies.” What are your two cents?
Don’t be afraid to occasionally take a risk if it could help you develop your niche and achieve your goals. If you’re interested in informatics, volunteer to help make your EHR better. If you’re interested in research, sign up to be a site PI for a multicenter PRIS project. If you’re interested in QI, take a course and volunteer to lead the development of a clinical pathway at your hospital. Over time, small sacrifices like these can bring more joy to your work day and get you closer to actually living your dream job.
Yes – as a resident I was completely oblivious to the idea that there could be a downside to physiologic monitoring. I probably put 90% of my patients on continuous ECG and pulse ox monitoring without a second thought. I truly thought that when alarms went off, nurses dropped everything and responded. Monitoring felt like a safety net. Then I had this a-ha moment as a new attending where I was having a conversation with a nurse about a patient who was getting worse, and her phone (which was being sent a text every time a monitor alarm sounded) just kept beeping and vibrating over and over. Finally, she set it down on a counter, walked away, and said “let’s go talk over here, these alarms are driving me nuts.” I realized that the entire system of continuous monitoring was broken – it was not working as it was intended to work, and it had the potential to fail the patients who were truly deteriorating and would actually benefit from a noisy alarm and an interruptive text to a nurse. So my team started thinking about how we could measure alarm fatigue. We threw around lots of ideas. Follow nurses around with stopwatches and time how long it takes them to respond? Put fancy tracking devices on nurses and get data that way? Maybe just ask them how long it takes them to respond? None of those made sense, so we ended up using video cameras – little GoPro cameras actually – to measure alarm fatigue. In an early pilot, we even had nurses wear GoPro cameras while caring for patients in the PICU. That was super interesting but the nurses hated it, so eventually we landed on just mounting lots of cameras in patient rooms and at the nurses’ station, which is how we performed the study that you mentioned. The main findings were of course that the vast majority of alarms were not actionable on a pediatric ward, and response time was slow.
So, what can we do? If you want to be part of a large collaborative effort, think about joining the PRIS Network (https://www.prisnetwork.org) and taking part in the EMO Study, which stands for “Eliminating Monitor Overuse.” Our goal over the next 2 years is to study the overuse of pulse oximetry in bronchiolitis across a large group of hospitals in the US and Canada. We will first measure overuse across 40+ hospitals and then use methods from the field of implementation science to collaboratively design a strategy to safely de-implement the unnecessary use of pulse oximetry in children under 2 with bronchiolitis who are not requiring supplemental oxygen.
At a local QI level, there are a number of different approaches you can take to reducing unnecessary alarms and hopefully thereby accelerating responses to alarms and alleviating alarm fatigue. Data helps – a lot. If you are able to get a sense of how many alarms are occurring, even if it’s just on 1 unit for 1 day, that can be very powerful in convincing others of the need for improvement. You could get that from the biomedical engineering department if you’re lucky and they have systems in place to track alarms (never hurts to ask – you might be surprised) or you could engage an eager student to count alarms from a central monitor station for a few hours at a time. Then, armed with some data, you can decide if you want to first address the problem of who is being monitored with strategies like implementing clinical pathways, order sets, or using audit-and-feedback, or address the problem of how patients are being monitored (for example, what settings are being used). Using classic QI methods like driver diagrams and impact-effort matrices can be really helpful in scoping out the problem, linking to key drivers, and mapping to interventions.
At CHOP we’ve found that one of the easiest ways to reduce alarms is to widen the default alarm parameters on the monitors – in other words, look at the settings for SpO2 low, HR high, etc that are automatically set when a patient is admitted. Are they age-based, or the same of all patients in the hospital? Are they appropriate? When we did that at CHOP, we realized that our default parameters were really conservative, much more so than most of our peer hospitals. So we made some modest changes, and reduced alarms by about 40%.
2. Last year, you wrote a piece in JAMA entitled, “The Emerging Market of Smartphone-Integrated Infant Physiologic Monitors.” In it, you talk about the growing number of parents scared into purchasing home vital sign monitors such as the “smart sock” monitor (Owlet Baby Care) at $250 each. Can you tell us how you address the worried parent of a bronchiolitis or BRUE patient who wants to purchase the Owlet or another home monitoring alarm?
Sure, this is a really interesting area. My main issues with the currently marketed monitors are: (a) none are FDA-cleared, so the companies have not had to demonstrate that they are safe or accurate, (b) the marketing of these products builds upon and benefits from the fear and anxiety of SIDS experienced by new parents and parents to be, and (c) the American Academy of Pediatrics Task Force on SIDS has issued recommendations that include “Do not use home cardiorespiratory monitors as a strategy to reduce the risk of SIDS,” so there isn’t any medical reason to gather medical data on healthy babies.
When I have parents in the hospital ask me about monitoring their healthy baby at home, I sit down with them and we discuss their very real fears and worries. We talk about the ABCs of safe sleep – the AAP recommendations that are based on evidence. We discuss what I’ve observed in testing these monitors – that they can fail to detect true problems (false negatives), and that they can also generate false alarms (false positives) that are frightening and might land them in the emergency room unnecessarily. Since, ultimately, my relationship with a hospitalized patient and their family is temporary and ends at discharge, if they are still interested in using a monitor despite my recommendations, I tell them that if they do decide to use one, they need to make a plan in collaboration with their pediatrician of what to do when the alarm goes off in the middle of the night. And then I’ll often give the pediatrician a heads-up about this discussion so they are prepared for it.
I think I’ve gotten a bit of a bad rap of being anti-technology or anti-innovation for my views on this. To be clear, I am incredibly supportive of companies that are innovating in the pediatric medical device space. It’s an underserved, relatively small market and we need all the attention we can get. I also think that in the future, everyone will be monitored – adults, children, all of us. The differences will be that the devices will be more accurate (less susceptible to false negatives and false positives) and will be smarter – in other words, we will know how to process and provide decision support to parents and clinicians for readings that are out of “normal ranges” but do not represent risks to the patient, to avoid overdiagnosis and overtreatment. For example, we know that many infants who are home recovering from bronchiolitis have occasional desaturations, and infants with these desats have the same outcomes as other infants without desats (https://jamanetwork.com/journals/jamapediatrics/fullarticle/2498407). But in our current environment, if a parent calls their PCP with an oxygen saturation of 89% on a home baby monitor, I’d bet 10 out of 10 PCPs would send the child to an emergency room.
In general, across hospital and home monitoring, I consider where we are right now to be the “awkward adolescence” of monitoring. We are in a situation where we have to avoid monitoring large groups of patients because we have not yet perfected how to use the data to benefit patients who are well or who have mild illnesses.
3. I’m a huge fan of your article, “Optimization of drug–drug interaction alert rules in a pediatric hospital's electronic health record system using a visual analytics dashboard.” In it, you describe how you used a group of pediatric clinical pharmacists to deactivate clinically irrelevant drug-drug interactive alerts, reducing alert rates from ~59 per 100 orders to 25 per 100 orders for pharmacists (also reducing it for providers). How difficult was it to get buy-in from front line providers and pharmacists for this project? How do you plan to maintain this project and reassess these alerts moving forward?
Thanks – that was really fun work to do, spearheaded by my friend and hospitalist colleague Eric Shelov, MD, MBI. The keys to the success of this project were taking a data-driven approach, partnering with pharmacy from the very beginning of the project, and broadly engaging stakeholders across the organization, including front line providers. Engagement from the front-lines was relatively easy since the project was going to decrease their alert burden and improve their experience with the EHR. Our clinical pharmacists did the “heavy lift” of the project, comprehensively reviewing drug interaction information from a variety of sources, but since their group was most impacted by the burden of false-positives, they were eager to contribute to this thoughtful reduction approach. The more daunting buy-in to get was patient safety leaders on board. They were concerned about the idea of turning off alerts. In the end, we were successful in developing an evidence-based proposal suggesting that the safety risk posed by alert fatigue was greater than that of deactivating carefully selected alerts. The project was operationalized within the Clinical Decision Support Committee at CHOP, which provided a forum for weekly review and discussion of alerts and multidisciplinary, data-driven decisions about alerts that could be safely deactivated in the system to reduce alert burden. I’m a bit removed from the project now, but Dr. Shelov now co-chairs the Clinical Decision Support Committee that oversees monitoring of these alerts. The group has recently expanded their focus to monitoring and optimization of other decision support tools, such as order sets and more sophisticated alerts that can evaluate multiple factors simultaneously.
4. You were the senior author on an article discussing implementation of a chest pain pathway in the Emergency Room (published Sep 2016 in Pediatr Emerg Care). With the pathway, there was a reduction in the rate of chest X-rays ordered, but an increase in outpatient follow-up (15 to 29%) with no visits resulting in diagnosis of a new cardiac condition. Does this mean that we should be “Choosing Wisely” and begin to do less with pediatric chest pain? How did the outpatient providers feel about the increased referrals/follow up?
This was an interesting project to be a part of. I view this project as a mostly negative study; the most interesting aspects to me were the outcomes that did not change after implementation. Here’s how the project started - a group of cardiologists embarked on a QI project because their perception was that there was a ton of excess cardiac-related resource utilization in the ED for chest pain. They were specifically concerned about excess cardiology consults, excess lab testing (e.g. troponins), ECGs, and echocardiograms. The problem was that they did not have baseline data when they started – their understanding of the issue was initially based on their Division’s perceptions only. They designed the clinical pathway and integrated it into the electronic health record. After implementation, they approached me to help mentor them, which I was glad to do. Together, we gathered the baseline data (which we of course should have gathered before even designing the intervention) and the post-intervention data. The most interesting findings to me were that (1) baseline resource utilization was already quite low before the intervention was implemented – lower than the perceptions of the cardiologists suggested, and (2) we did not observe changes in rates of ECG, blood tests, or echocardiograms. I think the reason for the lack of changes is that utilization of most of those diagnostic tools was already quite appropriate in the baseline period, although we couldn’t prove that point with the data we had. The lesson from this project is that it is critically important to obtain good baseline data to confirm your perceptions of a problem before designing and implementing an intervention- whether it’s QI or research. I don’t think the data from this project are strong enough to support a Choosing Wisely statement, but there is definitely room for more QI and research in this area.
5. You are Board-certified through the ABPM-Clinical Informatics Subspecialty Board and are a faculty member associated with Children’s Hospital of Philadelphia’s Clinical Informatics fellowship. How have you utilized this advanced training in your role as a hospitalist and what advice do you have for those interested in additional informatics training?
Clinical informatics is a great “niche” for hospitalists. When residents or fellows interested in PHM careers come to me for career advice, one of the first things I tell them is that they need to develop their niche – a relatively narrow area of specialization/expertise that gives them unique value within their organization and keeps them from getting burnt out from doing 100% clinical care all the time. Over time, if your niche truly does provide value to your hospital or department, it may be possible to negotiate for some protected time to devote to further developing your niche or sharing your expertise with others. As I have grown up as a hospitalist, and EHRs with comprehensive clinical decision support capabilities have become nearly ubiquitous, clinical informatics has developed into a really viable and rewarding niche. I have used my informatics background primarily to support QI and research endeavors. For example, a clinical pathway can be much more powerful if it has a robust order set behind it, and especially if that order set presents itself prominently when a patient with the right characteristics shows up on your service as a new admission. In research, if you know what the clinical data look like and where to find them in the hospital’s data warehouse, the data become less mysterious and your research efforts become much more nimble.
6. Jack Percelay often ends his list-serve commentary with the phrase “that’s just my 2 pennies.” What are your two cents?
Don’t be afraid to occasionally take a risk if it could help you develop your niche and achieve your goals. If you’re interested in informatics, volunteer to help make your EHR better. If you’re interested in research, sign up to be a site PI for a multicenter PRIS project. If you’re interested in QI, take a course and volunteer to lead the development of a clinical pathway at your hospital. Over time, small sacrifices like these can bring more joy to your work day and get you closer to actually living your dream job.