Editor's note: Eric G. Poon, MD, MPH, is Director of Clinical Informatics at Brigham and Women's Hospital and Assistant Professor of Medicine at Harvard Medical School. Dr. Poon's research has focused on using health information technology to improve patient safety. He oversees the development and implementation of clinical applications including computerized physician order entry (CPOE) and barcode-assisted electronic medication administration record, and was lead author on the first rigorous study demonstrating the impact of a bar coding system in a hospital pharmacy. We asked him to speak with us about how such technology can augment medication safety.

Dr. Robert Wachter, Editor, AHRQ WebM&M: What got you interested in looking at bar coding as a research endeavor?

Dr. Eric G. Poon: Many patient safety technologies are available to prevent medication errors, and CPOE certainly has captured the imagination of health care workers and policy workers alike. Research has shown that CPOE is capable of reducing the incidence of serious medication errors by about 50%; there are still many other opportunities to prevent them. Specifically, according to Leape's 1995 study, only about a third to a half of serious medication errors are committed during the physician prescribing stage, while up to two-thirds of serious medication errors occur during the transcription, dispensing, and administration stages. Other studies have looked at the incidence of administration errors in community-based settings. Barker found that almost one in five medication administrations were associated with an error, and a good number of them actually had the potential to harm patients.

Another way to look at it is that, in the health care industry, we are already pretty good at giving the correct medication at the correct time to the correct patient. Our accuracy rate is probably greater than 95% in most instances. However, the catch is that we give so many medications so many times a day—even a very small error rate can constitute a lot of opportunities for harm. We can do a lot through education and through improved ergonomics in the health care setting, but those strategies can only go so far when the reliability is already at the 95% to 99% stage. To really get to a very high reliability state, you may need a more precise type of technology that can actually verify without fail every time a medication has been given that you are giving it to the right patient, at the right dose, and at the right time. This is why our hospital invested in bar coding technology as the next quantum leap for improving medication safety.

As for my interest in this technology from a research standpoint, there is still not a lot of information out there to show how effective this technology is, whether its benefits justify its costs, and how to design and implement it in the best possible way. As our hospital began to invest in this technology, our research team thought that we would be remiss if we did not take advantage of the opportunity to answer these important questions.

RW: Given that such a high percentage of errors in the medication process happen at a point that begins beyond computerized order entry, it's interesting that so much of the research has been weighted towards CPOE, and there has been so little study of bar coding or other systems designed to address these other aspects. Why do you think that's been?

EP: For a variety of reasons, there has been more investment in CPOE in the health care IT industry. And the places in the academic hospitals that have implemented this technology are naturally places where the technology can be studied. There is some thought in the field that if an error is committed early in the medication use process during the order-writing stage, it perpetuates itself throughout the system. And if you catch those errors early, maybe you can do a little bit more good there. But I think we realize, at least locally, that while CPOE can do a lot to reduce errors, before we had bar coding in our hospital, a lot of incident reports reflected failure points in the transcription, dispensing, and administration stages. That also fits with the research data. So I'm hopeful that over the next couple of years we'll be able to have more definitive studies that describe the impact of this technology on patient safety.

RW: How would you characterize the evidence on barcode technology?

EP: I think the evidence is still mediocre at this point. There is some early evidence from the VA Hospital showing that, over the years as barcode technology has been implemented, the rate of errors did drop. But that particular study had some significant limitations in that they were mostly looking at incident reports, and we know that incident reports can significantly underestimate the incidence of errors. We do have fairly good evidence in terms of how well barcode technology can reduce the incidence of dispensing errors. We actually did a study that shows that we reduced the incidence of medication errors with the potential to harm patients by about 60%. That result is something that perhaps health care executives can hang their hats on, at least based on the experience of one hospital.

RW: From your work in looking at bar coding systems, what are the elements of an effective system?

EP: At the Brigham and Women's Hospital, we have learned that, for a barcode system to be successful, you really need to make sure that it fits into the workflow of the clinician. The design of the device with which you actually scan the barcode and scan the patient identifier (perhaps on the wristband) needs to be well accepted by the clinicians—because if the clinicians want to bypass the technology, it's always possible to do that. If the clinician doesn't believe that scanning all these barcodes before administration has any impact, then even the best designed technology is not going to exert any impact. So getting the form factor really well sorted out is important. Getting good clinician buy-in that this technology is going to work and partnering with a software vendor that is capable of delivering something that fits into a clinician's workflow obviously are going to be key to success. Other things that we have learned at our hospital are that in order for barcode medication administration technology to be successful, you really need to think about linking up the bedside point-of-care administration with both the CPOE system and the pharmacy system. In some ways, I think of the medication administration process as a triangle, and the three vertices of the triangle represent ordering, dispensing, and administering. Once the physician orders are entered electronically in the CPOE system, they need to be transmitted into the pharmacy system. Then the pharmacy needs to approve the medications and set the right schedules electronically for the nurses. All the information needs to be pushed over to the electronic medication administration records, so that the nurse can be prompted to give the medications at the right time. So the three corners of the triangle need to be able to communicate with each other effectively. One of the things that we're very proud of at our hospital is that we were able to design and implement a system that really is very tightly integrated. Some of the early efforts were able to leverage the barcode technology, but those systems tended in the early days to be fairly isolated. So errors could actually be created from time to time because those systems existed in a silo.

RW: Can you give any anecdotes that illustrate the form factor not being right, and then what nurses or doctors would do to work around or violate the intent of the system?

EP: Sure. The organizations that have actually had the most experience with barcode technology are the VA Hospitals. In their early incarnations of the technology, they had barcode scanners that were tethered to really large medication administration carts, so nurses often had to wheel around these very unwieldy carts that contained the medications for half the floor to each patient's room. Because the scanners were tethered by a wire to the medication administration carts, nurses often had to go into all kinds of contortions to scan the barcode on the patient's wristband. You can imagine this type of setup wasn't really designed with the best ergonomics in mind. Of course the VA has improved this technology, and we at the Brigham and Women's Hospital actually have learned a lot from what the VA has done. But as you can imagine, in the early days, nurses may not have accepted it very well, and when nurses are in a rush to finish the medication passes, they might just bypass the scanning and take away the benefit of the technology. When you have technology that's supposed to improve medication safety and people don't use it, the clinicians might be lulled into a false sense of security and they might be even less careful than they used to be. So I think these are some of the issues that early adopters of the technology have had to overcome.

RW: We've spoken so far about barcode medication administration. What other things, including laboratories, pathology, patient transport, or right side surgery, can barcode technology be used for?

EP: I think that there are definitely lots of opportunities to use barcode technology to make sure that lab specimens are correctly identified. Our hospital is also in the middle of implementing barcode scanning at the point of specimen collection to make sure that we can leverage barcode technology to its fullest. Other areas where the technology could definitely be useful may include other testing types. For example, making sure when patients have their EKGs done that the correct patient's identifier is tagged along with the EKG. Beyond that, there is the emergence of radiofrequency identification (RFID) capable of tracking patients and equipment throughout the hospital. That may ultimately be an important tool to make sure that providers know where patients are, and the organization can actually manage the flow of the patients more effectively. Some of the uses of barcode technology may not have any direct patient safety impact but definitely could increase the efficiency of the hospital if it's implemented correctly and perhaps increase satisfaction, and ultimately I think patients could benefit from that type of technology.

RW: Bar coding seems like very old technology—we had it in the supermarket 20 years ago. Is there a role for RFID or newer technologies in the medication management process, or is bar coding the best that we'll be able to do?

EP: The field has been asking that very question for some time. While RFID is becoming less and less expensive, it is difficult to compete with the cost of generating a barcode. With the FDA ruling that was put in place a few years ago that each medication used in the hospital needs to have a barcode at the unit dose level, the cost of putting a barcode on every medication dose will come down steadily. And RFIDs, as wonderful as they are, still can cost quite a bit more than bar coding. For example, when you're talking about an aspirin that costs a few pennies, tagging it with an RFID that might be three or four times the actual cost of the medication does not make a lot of economic sense. Where proponents of RFID may have an easier time justifying a business case or a return on investment may be around patient tracking and equipment tracking in the hospital.

RW: So, you mentioned the business case. Is there a business case for bar coding today?

EP: We have done a pretty good cost–benefit analysis looking at the role of barcode technology in the pharmacy dispensing process. The results have actually been very encouraging. Using an internally developed system such as ours, we were able to reach the break-even point within about 1 year of going live within the pharmacy. One of the reasons we were able to achieve these financial benefits is because the pharmacy is such a high-volume place and it dispenses so many medications, barcode technology is actually very good at reducing the last 0.25% of the errors. So that's why the cost–benefit analysis came up so favorably.

RW: Where do the savings come from, and do you account for the extra nurses' time that it might take to do the scanning?

EP: We don't yet have the data to account for the patient safety impact of barcode scanning at the bedside. We have data at the pharmacy. But I will point out that we have pretty good data showing that at least with our implementation of barcode technology, there was no increase in the amount of time nurses spent on medication administration. That may be somewhat counterintuitive. When we looked at our data more closely, we did find that after the implementation of barcode technology, nurses were spending a fraction more time on the mechanics of scanning the patient's wristband and interacting with the computer. But one thing that we found was that with this technology you can actually help organize the nurses' workflow better. For example, with the technology we can actually help nurses organize their day better, so that the nurses will be prompted to give the medications, and the computer will actually keep track of when the patients are due for medications. With this technology, the nurse will not have to hunt around the unit looking for the paper medication administration records. So there is some time savings there. And because our nurses are on the computers all the time, they have actually better access to electronic resources to look up medication administration protocols. Which is why, although they actually have to spend a little bit more time with the mechanics of scanning, the amount of time spent on medication administration overall came out about even.

RW: Ten years from now, what will the medication administration process look like?

EP: Ten years from now, I really believe that we will have very commonplace business solutions that will take care of the three vertices of the triangle that I talked about. I think that we'll have better products that provide integrated systems to take care of physician ordering, pharmacy approval and dispensing, and medication administration at the point of care. And within the pharmacy and at the point of care, the medication use process will be guided at various times by the barcode scanning technology. I can also see how this technology will begin to move into the higher risk areas such as oncology, or in neonatal intensive care units. These are some of the areas that we are actively working on. For example, as we rolled out this technology initially, with the complexity of chemotherapy protocols in the oncology units, we found that we needed an additional layer of sophistication in our software. So when we rolled out barcode technology, we decided to not roll it out on the oncology unit, in order to have time to actually work on this technology to make sure that we don't expose our patients to any additional risk or harm.

RW: A final question: if a hospital CEO today wanted to get into the IT game with an eye toward patient safety, and the choice is to focus on building electronic health record, computerized order entry, bar coding, or something else first, how would you advise that person in terms of prioritization?

EP: This is a really difficult question, ripe for some good research. There is a plethora of choices that each organization can make. One of the things that I may suggest would be for each organization to consider the capacity for change. Each organization may come to a different conclusion. As we talked about earlier, the evidence base for CPOE, while not completely airtight, is still better than that for barcode technology. Although I'm hopeful that will probably change. And as evidence comes out, I think there will be some comparative studies looking at whether one technology might be more cost effective than another. I would say based on our experience that there seems to be a pretty good case, both in terms of patient safety and return on investment, in investing in a good pharmacy system. While the pharmacy does not produce the lion's share of medication errors, the cost associated with redoing the technology in the centralized pharmacy system probably is less, mainly because most pharmacies tend to be run more centrally. If you're talking about CPOE and barcode medication administration, both technologies tend to touch more people within the hospital, and therefore converting from a paper-based system to an electronic system comes at a higher cost. So there may be a case for concentrating on the pharmacy, particularly if resources are limited. But I think the jury is still out as to which technology should definitively be the first place to start.