Editor's note: Dr. Chopra is Chief of the Division of Hospital Medicine and Associate Professor of Medicine at the University of Michigan Medical School. His research efforts focus on improving the safety of hospitalized patients by preventing hospital-acquired complications—particularly those associated with peripherally inserted central catheters. We spoke with him about safety of peripherally inserted central catheters and central lines.

Dr. Robert Wachter: What got you interested in the problems with lines?

Dr. Vineet Chopra: A lot of work was being done on central lines in the intensive care unit (ICU). But I didn't see a lot of research or interest on IV devices outside the ICU, particularly on the wards—which is where I was most encountering them. It was around this time that peripherally inserted central catheters (PICCs) were becoming more prevalent in the ward setting, and I began to notice a lot of hospitalists varied in how they used them. For example, I could come on service and a person would have a PICC placed for difficult access, and the next week someone would have a triple lumen device for no clear reason. I would try to understand what led my colleagues to make those decisions about the PICC, and often there was no clear reason for their choices. Because I wanted to focus my career on a subject we as hospitalists deal with every day that was underdeveloped, this seemed like the perfect opportunity.

RW: Was the initial impetus the appropriateness of use, complications of, cost of, variations in, or all of the above?

VC: All of the above, but I started to begin with understanding epidemiology and variation with the use of PICCs. I didn't know this would fly as a research career, because the first question that came up was who "owns" PICCs? Is it the clinician that orders them? Is it the inserter that places them? Insertion was quickly moving to nurse-led teams, but I didn't see any substantive research from inserters when it came to variation in use of PICCs. Similarly, interventional radiologists often inserted these lines, but there was little data or evidence regarding indications for use or appropriateness in their literature. The same was true for end-users of the device. For example, surgeons and oncologists often order these devices, but a consistent approach to their use was lacking. As I looked more, I began to realize that there was no clear owner of the device or the issues related to it. So my first task was to explore the scope of the ownership and outcomes issues; that's where we started looking at indications for use and complications.

RW: How central of a problem was diffuse ownership as you tried to define best practices? It seems like the analogy is with central lines, and it's a little clearer who owns that because most of them live in the ICU space. What did you come to think about this issue of ownership?

VC: The first epiphany for me was when I realized that there was no consensus about who is best suited to insert these devices or who owned them after they were placed. When I started to perform systematic reviews to understand the literature, I began to see evidence in the nursing , radiology , oncology, surgery literature, etc. Then, there were data that came from very specific patient populations like cystic fibrosis or sickle cell, where vascular access represented a lifeline. As I started looking at these populations, these diseases, these providers, I understood that they had differing views and approaches about PICCs. Consequently, many of the recommendations from these groups were unique and tailored to their context—so much so that it was hard to take these data and apply them to hospitalist practice. 

Not surprisingly, then, you also had substantial variation in rates of complications, process metrics with respect to dwell time, how many lumens your catheter should have, and how complications should be managed. As I went around looking for one source of truth, I began to realize there wasn't one. The irony was that hospitalists had begun ordering the insertion of a lot of these devices—so we were squarely in the middle of the issue. I recognized that we in hospital medicine were ideally positioned to lead in this space. And so, my research journey began. As part of my early career development award, I put together a group of 15 experts from across the world (including a patient) to come up with recommendations for what IV device you should use to help decision-making with PICCs. The panel included experts from disciplines such as radiology, cancer, infectious diseases, pharmacy, surgeons, and vascular access nurses as well as hospitalists. Every stakeholder has their own way of doing it. My goal was to try to get us all on the same page. 

RW: You said one thing that was important was that the complications and hazards had been underappreciated. What did you come to learn?

VC: When I was a trainee, many viewed PICCs as safer than other IV devices. In large part, this was because of their unique insertion route. In adult patients, PICCs are usually placed in the veins of the upper extremity. Unlike central lines, you don't have to go anywhere near the great vessels of the neck or the chest, so you avoid injury to those vessels, which can be life-threatening. You avoid injury to the lung—that dreaded complication of pneumothorax. As well, if people start to bleed during insertion, with a PICC it's easy to manage. You just put pressure at that site in the arm and compress the vessel against the underlying humerus. Compare that to the subclavian vessels or a lacerated internal jugular vein—those are hard to compress. The unique aspect of their site of insertion fueled this perception that they were safer than central lines. 

Some early work also found that PICCs were associated with a lower rate of infections than central venous catheters. So many hospitals (and hospitalists) had adopted a strategy of preferentially placing PICCs (given their insertion safety and perceived lower risk of infection) over central lines. That was the state of the science then, all of which ultimately changed as we recognized that these early data regarding PICCs came from studies of patients who were ready to discharge home or never touched the hospital. There is, after all, a fundamental difference between patients that needed to be in the hospital and still needed central venous access compared to patients who could go home and care for the line. A lot of data on infection risk came from patients in the outpatient setting or those transitioning to the outpatient setting. It was a classical fallacy in terms of inference—we looked at the data without considering the context. Much of my earlier work was focused on unraveling this to understand what the true rates of infectious and noninfectious complications were.

RW: Is it that we were missing complications of patients that occurred after they left? Or the patient populations are so different that you have to adjust for that to understand the relative safety of the two approaches?

VC: First, it was retrospective data, so you were missing events that were just not well documented, if documented at all. Second, the patients getting PICCs were fundamentally different from hospitalized patients in that they were healthier. To show this, we performed a systematic review and meta-analyses, where we organized studies based on where patients were when they got their PICC. We showed that if a patient was in the outpatient setting, the risk of infection with a PICC compared to any other line was 28% lower. There were certainly aspects of PICCs that helped ensure safety—because the site of insertion, for instance led to easier and timelier care as opposed to the neck or the groin. But if studies included patients that were hospitalized when they got their PICC, there was no difference in the risk of infection between a PICC and a central venous catheter. 

It's said that it takes 17 years before evidence changes practice. But we published that systematic review in 2012, and in 2014 the Society for Healthcare Epidemiology of America and the Infectious Disease Society of America published an update to the strategy for preventing central line–associated bloodstream infection (CLABSI), where they specifically instructed hospitals to "not use PICCs as a strategy to reduce the risk of bloodstream infection." This recommendation was likely carefully worded because the guideline writers knew that is what hospitals and many clinicians were doing. A lot of the studies that they used to support that recommendation came from the work that we and others had done, so it was nice to see our work being used to update guidelines and change practice.

RW: Recognizing that the infection was one of only a few complications of central lines, if you had a choice of a PICC versus a central line, even if the infection rates are the same, wouldn't you still choose the PICC?

VC: That's exactly what the ICU doctors asked me with respect to our work. Clearly, when you put PICCs in, these are safer. You think of a patient with unstable spine where you cannot safely access the neck to put a line in, or a patient who is coagulopathic—surely the dangers of placing something in the neck or chest outweighed anything else. What we discovered to help inform this debate was that while the insertion risks may be lower, the thrombotic risk associated with PICCs was much higher than previously recognized. 

In 2013, we did a systematic review of 62 studies and published a paper in The Lancet. We found that the average rate of PICC DVT [deep vein thrombosis] in studies where PICCs were placed in patients on the wards that were not critically ill and did not have cancer was about 3%. If the study included patients with malignancy or the indication for the PICC was cancer-related, the rate of DVT doubled to 6%. If patients were in the ICU when the PICC was placed as the central line, the thrombosis rate was double that of cancer patients—about 13%. Overall, it's about a 5% risk of DVT and while that doesn't sound terrible, remember that the average risk of DVT in a general medical patient without a malignancy or critical illness is less than 1%. So, using a PICC substantially increases thrombotic risk. When making a decision about a central line or PICC, I always emphasize looking at both infectious and noninfectious complications—and DVT is front and center when thinking about the noninfectious harms.

It turns out what makes PICCs so attractive for insertion (the arm veins) is dangerous with respect to thrombosis. Because PICCs are inserted into smaller veins in the arm, they are more likely to provoke DVT. Think about it: when you insert a PICC, your needle causes endothelial injury, in a patient that's hypercoaguable. Your device is placed in a smaller vessel (than the IJ [internal jugular] or the subclavian) so that it occupies most of the cross-sectional diameter of that vein, so you have no space around the catheter for blood to flow and get venous stasis. So you have the perfect setup for Virchow's triad. When we compared PICCs head-to-head with central lines, we found that DVT rates were 2.5 times higher with PICCs. That was shocking. And that's where our work has had some of the most impact in terms of informing the risk–benefit discussion of these devices in ICUs, in patients with malignancy, and for patients with difficult access where they may not be the best choice.

RW: And are the DVTs as consequential? You would imagine if they're happening in smaller vessels, they might not be as clinically important?

VC: Well, it's true that they don't embolize to the lungs as much as lower extremity clots. But there are consequences. One is once you have a DVT in the arm, that vein essentially becomes inflamed and ultimately occludes and thromboses. If you think about venous access for that patient, you're essentially depleting limited venous real estate in the arm. This is extremely relevant for patients with chronic kidney disease, where the use of a PICC has been one of the strongest predictors of failure of an arterial venous fistula. A PICC can be disastrous because the clot can essentially distort your venous anatomy and essentially leave you with either no target veins for fistula creation or a nonworking dialysis access.
Second, thrombosis can often cause a lot of patient discomfort. Arm edema, arm swelling, the whole problem with venous outflow can be quite troublesome for patients. Deep vein thrombosis in the arm also has important consequences in terms of postphlebitic syndrome. Essentially, pain and discoloration or disfiguration of the arm because of DVT can be quite challenging to manage. Finally, remember that the treatment is not benign. Systemic anticoagulation for at least 3 months has risks and costs. But the biggest impact, in my opinion, is the loss of functioning arm veins. This loss can have huge impact to overall care and management of patients with chronic disease.

RW: When you're trying to educate people or systems about the complications of PICCs, what are the alternatives that you're beginning to steer people to or at least getting them to consider?

VC: The bigger picture here—even beyond PICCs—is how we approach venous access in general. Think about it—it's the most commonly performed procedure in the world today. Every patient gets an IV device, regardless of age, gender, or other demographics. Yet, the evidence that we have to figure out what the best device for that patient is next to nonexistent. My dream is that for every patient, for every hospitalization, there's one device that lasts them the entire stay. Not multiple peripheral IV catheters or a PICC on every admission because they are a "difficult stick." It's getting it right and getting it right for their entire vascular access journey. So when we started thinking about alternatives to PICCs, we started thinking about this bigger concept: how can we create guidance for clinicians to inform decisions in IV access? That's where we decided to create appropriateness criteria and where the Michigan Appropriateness Guide for Intravenous Catheters, or MAGIC, was born.

MAGIC focused on not just the PICC, but also on what other options are available. There are three categories of peripheral devices, and there are three categories of central devices. On the peripheral side, you have the mighty but small peripheral intravenous catheter. You have a device that has resurfaced in the past decade, the midline catheter. The third device that sits in-between the two is an ultrasound-guided peripheral, long peripheral or extended-dwell catheter. These are like peripherals IVs, but they are longer and placed using ultrasound into deeper veins of the arm. Because their tip ends in deeper veins, they can last longer than typical peripheral IV catheters.

And then you have the central line category, which includes the acute nontunneled central lines in the ICU (e.g., the IJs, femoral, and subclavian catheters). You have tunneled catheters, where the inserter creates a space between the site of catheter entry into the vein and exits into the skin. The idea is that space prevents skin pathogens from getting into the bloodstream and causing infections. Then you have totally implanted ports, which are devices that feature a reservoir and a catheter. They reservoir sit beneath the surface of the skin and you access them when you need to and when you're done, you deaccess the port so that the skin becomes a natural barrier to infection. So, the PICC is just one of those six choices. The question then becomes, which one do you choose and for what?

The goal of MAGIC was to use appropriateness to direct clinicians to the options they should consider. What we wanted to do was become the one-stop shop for thinking about the vascular access journey for patients. MAGIC is focused on the PICC, but it does so knowing that you have to give clinicians options and evidence to inform their decision-making. The MAGIC algorithm asks for the proposed duration of therapy, what devices are available, and whether you are trying to infuse irritants or vesicants. 

Using inputs from this three-dimensional cube, we devised MAGIC to walk people through the decision of what IV device was most appropriate. The joy for me in doing this work was I didn't anticipate the impact it would have. Today, MAGIC is used in hundreds of hospitals across the US and the world. 

Another thing we recognized when we published this paper (which was close to 50 pages long) was that no one was going to read a document that lengthy to make a decision that needed to happen expediently. It was a moment in my career where I realized that the academic products are important, but in order to have clinical impact—a 50-page paper may not be the best approach. So we did what a Silicon Valley entrepreneur may do—we created a free mobile application for iOS and Android devices. Essentially, the app enables the MAGIC recommendations, but in a way that allows decisions to be made with four quick questions. It walks people through the indication for insertion, what devices may be available, how long they need venous access, and what they're trying to administer or infuse. It's a good lesson in implementation science: our paper has been read about 2000 or 3000 times, but the app has been downloaded over 10,000 times. I go to places and talk to folks, and while they often will know about the app, they won't know about the paper—which to me is a job well done. It's informing clinical care at the bedside, which is precisely where we wanted to have impact.

RW: What are some of the lessons you've learned about evidence to practice? How do you move from we now know what we think is right in different circumstances to scaling that and making sure that people are always doing the right thing?

VC: That's the Holy Grail of quality improvement, isn't it? The first moment of insight was that education by itself is a weak intervention. It's simply not enough to provide people with evidence. Yet, the quality improvement work that we do often has a very heavy—if not monocentric—educational component. It's not that people don't want to learn. It's that time is limited, people forget and people change. You might see initial improvements with education, but you don't see sustainability. 

The second lesson is to learn that culture will eat strategy for lunch. To have impact with respect to changing quality and decision making, you must understand organizational culture and embed interventions within it. It doesn't take more than a few people in a key space to say "we're not going to do this" or "this makes no sense" for your best-laid plans to fail. One of the key strengths of MAGIC is that it is readily applicable to a number of settings and organizational cultures.

We have tackled culture by making the right thing to do the easy thing to do. We do that in several ways with MAGIC. First, we do it via the EHR. For example, we have changed our entire order set for IV devices at Michigan to be completely informed by MAGIC. And because physicians are burnt out with hard stops and other alerts, we didn't want it to be cumbersome. So, we changed the default options in the order set to be the ideal recommendations. Second, we created the app to allow for decision-making at the bedside. In this way, the most important principles are shared up front.

For instance, a key risk factor for infections and thrombosis is the number of lumens a catheter has. For DVT, the more lumens you have, the greater the gauge of the catheter, which means the more space it occupies in a vein and higher risk for venous stasis. Ordering a single lumen PICC should thus lower risk of thrombosis. We set that as the default in the app and the orderset. We also built in evidence for use of PICCs in chronic kidney disease. We want to encourage avoidance of PICC use in chronic kidney disease, so our Epic orderset pulls in GFR [glomerular filtration rate] data directly into the clinicians line of sight and provides warnings regarding PICC use if the GFR is below 45 mL/min. It makes the right thing to do the easy thing to do. 

Although we focused a lot on medicine providers, on hospitalists—it's worth noting that some of the biggest improvements in our institution came from outside of medicine. They came from specialties we didn't touch. For example, I never talked to surgery or pediatrics or some of the other specialties when we launched the new orderset. Yet, when you look at ordering practices over time, they often had the greatest improvements in terms of appropriateness. You could argue that's because they had more improvement to make. But that's where the culture comes in. People often don't think about these decisions, they do them on autopilot or don't know the evidence behind the recommendations. Creating a streamlined path that allows colleagues to reflect on why you need a catheter and what aspects you should consider changes how we approach these choices.

The other advantage of an electronic orderset was that it created an audit trail. If providers were repeat offenders, we are able to track them because every order has a provider signature attached to it. For example, when we first launched, we had some providers who would select the triple lumen catheters, and the reason would be free-text words like "FFPPGGHH." They would still be able to order that device, but there was a paper trail left behind that allowed us to interface with them. In many ways, PICCs are the epitome for collaborative science because multiple providers touch the device from ordering, inserting, and caring for them. As one of our physician champions here for this work, I had the opportunity to engage with folks who often were not ordering devices appropriately or were not aware of risks and alternatives. Having that voice and the technology built in to hardwire this process has allowed us to be leaders when it comes to using PICCs at Michigan.

RW: You mentioned that one of the things it does is create an audit trail to show people who are not following the guidelines, and you said you're the champion—the one who "engages" folks. Tell me what that looks like in real life. Are there circumstances where someone's ignoring the app, doing it all wrong, and they get a call from Dr. Chopra to talk about your PICC ordering behavior?

VC: Haha! No, that's not it. I fully recognize that there's not a Dr. Chopra in every hospital, so let me share how we do this across Michigan within a collaborative qualitive improvement project funded by Blue Cross/Blue Shield of Michigan. We've asked participating hospitals to develop a vascular access committee that comes together every month and looks at their data including aspects, such as indications for use and outcomes. We collect detailed data from the medical record at each hospital and provide directions as to what areas to improve on. Hospital committees have representatives from a number of specialties—infectious diseases, nephrology, surgery, the emergency department—and they all look at the data and discuss improvement opportunities. In this way, there is no "police" that responds to offenders, but a group of clinicians who bring unique perspectives to indications for use and outcomes that can help improve hospital-wide practice.

Our committees at each of our 40-plus member hospitals gives members an opportunity to look at the data and ask, "does this make sense" and "what was the issue/what went wrong here?" Every institution has clinicians who just don't want to conform or believe their way is the right way, and some of those people may get an email or phone call from leaders. Yes, I've had to have several of these discussions at Michigan, but I'll say that most of the times, there is genuine surprise regarding what is learned and shared. This has really been the case in the Hematology–Oncology space, where a lot of PICC use occurs. Some oncologists prefer PICCs because if their patients get infected, they can pull these catheters very easily to help treat the infection. It turns out the data may not support this fear, and there is now good Level 1 evidence suggesting that ports do have lower risks of complications than PICCs in patients with solid cancers. The committees can often be the key way to share that knowledge.

For example, I tell our oncologists, "I understand you're concerned about safety, but the data doesn't support some of your concerns, at least from what we're seeing so far." And whether that's a face-to-face, phone call, a friendly nudge, or an email—all of those apply with respect to helping improve patient safety.

RW: It reminds me a little bit about IVC [inferior vena cava] filters, that in some ways you've come out with what appears to be an easier alternative, and it gets overused and the evidence is not there. There's some logic to why it might make sense, but it takes a while to accumulate the evidence about the downsides. Then, it takes even longer to disseminate that and get past people's instincts about how easy and safe the thing is.

VC: If you think about percutaneous coronary intervention or a CT scan or an MRI, it's the same story. The Gartner Foundation hype cycle is a nifty framework that defines how a technology diffuses and is relevant here. And it's true for PICCs; it's true for IVC filters. First, there's a technology trigger, a new shiny toy that everyone wants to jump all in on. Then comes the Peak of Inflated Expectations, where overuse dominates. After every peak comes a valley—the Trough of Disillusionment in the hype cycle, where evidence about risks and complications changes perceptions of the technology from beneficial to potentially harmful. Then, (my favorite phase of the cycle) is the Slope of Enlightenment, where the balance between risks and benefits gets better delineated so appropriate use can emerge. Finally, we get to the Plateau of Productivity, where use of technology occurs in a balanced way. PICCs are a great metaphor in the past 20 years of having gone through that cycle, just like IVC filters as you mentioned.

RW: What's next for you in this journey?

VC: We've done a lot of work with our EHR tools, but there's more work to be done. We've been working with Epic to create a standardized order set that I'm hoping will be implemented in every hospital that has this platform. We are actively looking to diffuse that so everyone can experience the Michigan MAGIC! We are also in the middle of revamping our app and including a patient-facing component. As of now, the app only walks you through the MAGIC algorithm and is clinician-centric. The new app will begin by asking whether you are a patient, a researcher, or a clinician. And if you choose patient, information on what patients need to know about their PICCs—what questions they should ask their clinician before getting one, for example—are featured. We're also building videos for patients on how they should care for the device—things like flushing, dressing, maintenance, etc.—that will be within the app. We also want them to know when they should seek help. The app is thus going to be an important way for us to activate patients, who so far have been very much the passive recipients of these devices. 

Finally, as I said earlier, I would love for there to be a day where every patient has one device for their entire hospital stay. But to get there, we need to truly empower and enable vascular access teams and nurses. Much of what I've learned about PICCs, about vascular catheters, I have learned from that group. They do this better than any of us. They have knowledge about things like tip position, about ECG to track the catheter to the right place, about number of lumens, about catheter-to-vein ratios, things that you and I never learned in medical school. What's next for me is empowering and enabling these providers so that we move the field from ordering a line to consulting vascular access teams to help determine what device is most appropriate. The provider will do an assessment, figure out what the right choice is given what the clinician is trying to do in terms of infusion, access, drugs, and then place that device in for you. That's the future I'm trying to march toward. I think we have a long way to go, but that's where we're going.