Two recent studies are shedding light on an important phenomenon in U.S. healthcare—the growing use by physicians in clinical practice, of medical scribes—and the issues around scribe use. Most medical scribes are medical students who are hired by physicians working in their offices and clinics, with the employment of those scribes designed to alleviate some of the documentation burden that is adding significantly to physicians’ workdays—and to their stress.
The two studies—one whose results were published this summer, and the other whose results were published earlier this month—underscore important concerns around the use of scribes. They were executed by Oregon Health & Science University (OHSU). The first study, done in collaboration with The Doctors Company (based in Napa, Calif.), the nation’s largest physician-owned medical malpractice insurer, revealed a lack of standardized training, and variability in experience among scribes, posing major risks to data accuracy and delivery of care.
Further, the first study, based on a survey of 355 physicians, found that:
> 44 percent of scribes have had no prior experience.
> Only 55 percent of scribes are trained by the doctor.
> Only 22 percent of scribes have had any form of certification.
> Around 24 percent of practices that use scribes hire them as employees.
> Nearly 13 percent of practices use scribe staffing agencies.
Meanwhile the second study, published under the title “Use of Simulation Based on an Electronic Health Records Environment to Evaluate the Structure and Accuracy of Notes Generated by Medical Scribes: Proof-of-Concept Study” in the publication JMIR Medical Informatics, involved the video-recording of the work of five scribes, all of whom had at least six months’ work experience working inside electronic health records (EHRs), and whose work transcribing an identical simulated physician-patient encounter was studied. That study found that “There was significant interscribe variability in note structure and content. Overall, only 26 percent of all data elements were unique to the scribe writing them,” the study found. What’s more, it found that, “Overall, there was a wide inter- and intrascribe variation in accuracy for each section of the notes with ranges from 50% to 76%, resulting in an overall positive predictive value for each note between 38% and 81%.”
The lead author of both studies is Jeffrey A. Gold, M.D., a critical care pulmonologist practicing at Oregon Health & Sciences University, in Portland, Ore., and a researcher.
Healthcare Informatics Editor-in-Chief Mark Hagland spoke with Dr. Gold twice about the implications of the two studies, as well as about the broader questions around the use of medical scribes in U.S. healthcare—first after the publication of the first study; and again, after the publication of the second study. Below are excerpts from those two interviews.
Dr. Gold, one broad question that comes to mind is whether there are any formal mechanisms for certifying medical scribes.
It’s really interesting; scribes fall into two categories. Medical students and others, who work full-time as scribes. Then there are people who work part-time as scribes, MAs and Pas [medical assistants and physician assistants]. And there are scribe service companies that certify them, and a lot of people don’t like to go through those companies. One company that does certify its scribes is ScribeAmerica, which requires that individuals go through 40 hours of online training, combined with 200 hours of work in practice. And there’s really no national set of guidelines over what scribes should be doing; it’s all over the place.
How does this strike you as a physician?
If I were a physician in clinic, I would say, I can’t keep up anyway, so at least I can be happy at work. That’s how the average physician feels about it, and rightfully so. EHRs are the single biggest cause of burnout, and anything you can do to untether physicians from the EHR will improve their satisfaction. But as a health IT safety person—the problem is that nobody knows how to use the EHR well at all. Every professional group studied has trouble using the EHR. And the trouble is that they’re the ones doing the training. It really is a challenge that most scribes have no significant medical knowledge at all.
And if you believe the literature, which finds that there is a big problem with selective data-gathering or selective data-processing—meaning that you don’t collect all the information, or it is incomplete or incorrect—then, theoretically, that phenomenon is contributing to delayed or incorrect diagnoses. That means that adding a middle-person into the process, to get the information in, or out, adds latent patient safety issues, especially if the people aren’t trained. If a scribe were just a pure Dictaphone—if all they do is regurgitate exactly, like a court stenographer, then that’s probably not a major issue; but as soon as the scribes are interpreting what’s going on, or are asked to find information or help physicians navigate the EHR—well, we know from our previous research that physicians themselves struggle to navigate the EHR. In addition, it’s the ones who struggle with navigating the EHR who turn to scribes, which is like asking a blind person to drive and to teach you how to drive. And that’s what scares me the most.
So what’s the high-level solution to this?
I think there are a few solutions to this. One is that there needs to be more specialty-specific, context-specific, training for the scribes. Second, there needs to be training for physicians on how to use and train scribes. Most physicians aren’t teachers. And if you’re going to be responsible for training a scribe…
But doctors don’t have the time to take courses to train scribes, correct?
We’re about to get a five-year grant from AHRQ [the federal Agency for Healthcare Research and Quality] to create an all-encompassing training program, one that is video-based, for scribes. It would video-based training, and a grading sheet at the end. I’ve applied for the grant; that’s in the public record.
How do you see this evolving over the next few years?
I don’t know. Some people believe that the EHRs will get useful enough that we will no longer need scribes in five to ten years; that the EHR vendors will take into account usability and workflow so that scribes become superfluous. I don’t think that’s true; I think as scribes become more and more prevalent, that the EHR vendors will give up on their efforts, because we will have found that workaround through this. The biggest problem with electronic health records is that we have taken our workflow in medicine, based on a paper-based workflow, and assumed that introducing a computer into the process wouldn’t change things. But the reality is that adding computerization into this, and then assuming that you can do the same number of things in the same amount of time, is ludicrous. The real answer would be to make patient visits longer, but that’s impossible because of payment and reimbursement issues.
So one thing we want to look at is, how much are scribe notes proofread and looked at by the physicians? The idea is that scribes are useful because they’re doing the inputting and the physician doesn’t need to be involved in the computer. The thing is, if the physician never looks at the notes, that’s a patient safety problem. On the other hand, if the physician has to spend two hours reviewing the scribe’s notes, then that actually will make the problem worse.
The issue with scribes is that there are multiple competing issues around why scribes exist in the first place. Some say physician notes already are not accurate. The question is whether there’s anyone arguing for patient safety, with regard to completeness of notes and completeness of billing within the notes.
Looking at the results of this second study, Dr. Gold, clearly, you found tremendous variability in note structure and content, correct?
Yes, that’s right. We don’t know how significant that is. It’s no different from physicians; five physicians recording the same patient encounter would describe things differently. Point number one is that scribes themselves are not like court stenographers; that’s the first thing. Some things might be described differently, and some things might not even make it into the note. And as an attending, you can’t just have someone write the note, and keep going. The physician will still end up being responsible. I think the most powerful thing here is that while there is value in scribes—and they are valuable, because physicians are getting burned out on documentation—scribes tend also to write things more consistently. And now, through a proof-of-concept study, we’ve been able to demonstrate that you can assess how good a scribe is at communicating what happened in an encounter, and can then improve it.
And not 100 percent of what happens in a visit is communicated in a note. And why is that? One element is when scribes just can’t keep up, and some random filtering occurs. Why would that be? One, the conversation may go to quickly; two, the EHR may be slowing down the scribe, just as it slows the physician down; and third, maybe the scribe just can’t type fast enough—maybe the scribe is just a terrible typist. You’d assume they could type well, just as you’d assume a taxi driver knows how to driver, right? But if scribes either are not good for any of those reasons, or because of the unique templates of a clinic—for whatever reason they can’t keep up—lack of medical knowledge or terminology, or whatever—that means that information is being filtered out.
When you’re a physician, you understand things based on your medical knowledge. And for the scribe, that filtering is not going to be determined based on medical knowledge, but based on random reasons. So this allows us to assess, and therefore, improve, how scribes perform. And this also allows you to see—with the grant we’re going to do, eventually, you’re going to be able to say, as a provider, I’ve had my scribes do these exercises, and I was able to see where their weaknesses were. To assume every scribe is perfect, is ludicrous, and to assume that they are all the same, is also ludicrous.
What would you say to laypeople who might be discomfited by the idea of non-perfect scribes being used?
What I would say is that, until the EHR gets better or our structure for how we see patients gets better, this may make things safer in the meantime. And so I would ask, how does your doctor do things now? They’re either typing into the computer all the time and trying to multitask. Or they would say, I don’t know when they do their documentation. And I would say, they’re doing it after you’ve left. And how much do they actually remember later on? The fact is that the physician is talking and documenting and thinking, and trying to process information at the same time, or documenting after the fact. And one could argue that a natural filtering takes place that actually is useful; that would be great if that filter happened right after the patient left the room. But the reality is that a lot of the documentation is actually not happening until the end of the day, after the physician has seen 20 patients.
So this is a case where it may not necessarily be making things more dangerous at all; it may just be shifting the error from one professional group to another. And because we have a professional group that has a singular role, to interface with the EHR, that it becomes easier to train for one purpose than to train a physician, who has to be trained on documentation, on meaningful use, on outcomes, on everything else. So if a person only needs to be trained on one thing, that’s an advantage. However, the problem is, that sounds all fine and dandy when we say, look, here’s someone who has only one job, to interface the EHR. However, we know that EHR training has never been great to begin with. And with this group, we need it more than ever, because that’s the only thing they do.
So I would argue that it can be made to be just as safe, if done the right way. The real issues will be things like, one, it’s a transient workforce, that’s a real problem. A large number of scribes are kids going to medical school, so they’re only doing it for a year. And it’s difficult to invest in people knowing you’ll only have them for a year, or for two at most. The second thing is that physicians are always going to have to be worried, as we do more and more coding of diagnoses. If the scribe is just writing the note, that’s one thing. The issue will be if scribes are also expanding and modifying the problem list, and are ordering tests and associating those tests with diagnoses, then the physician is also responsible for things that could potentially be billing fraud or other issues.
What will the landscape around this be like five years from now?
I honestly don’t know. Here’s the thing: if you’re an EHR vendor, you’re supposed to improve billing effectiveness, efficiency, clinical outcomes, everything. And I’m not sure what the vendors’ interest will be to dramatically change processes, if a workaround is already in place. So if you’re one of the biggest vendors, and you say, well, they’ve figured it out with scribes anyway; and the reality is that you go to the issues that are the most pressing need. And you have this ready supply of medical students. So I’m not sure scribes will go away. But when we look at EHR use, it’s not just efficient use, it’s also effective use.
And it’s not just about whether you can get through your day as quickly as possible, but whether you’re making good clinical decisions, and that’s based on what tests you order, and other considerations. And I do not believe for one second that when HER vendors created their products, that they intentionally created things that they thought would result in patient harm. I think they truly are trying to do the best job they can. The problem is not a lack of good intentions, but a lack of good processes.
This is about patient care scenarios that happen to involve the EHR. So it’s got to be patient-centric. And the problem now is that everybody and their brother is going to be creating their own patient-facing app. So if you don’t like the testing that the vendors are doing, what about the app developer at some random think tank or vendor out in La Jolla? So the elephant in the room is that what we need are standardized means for assessing whether we can deliver safe, efficient, and effective care for patients, using electronic devices. And this paper and our grant started with a focus on patient interaction, and can you improve that process? It’s not EHR-centric, it’s patient care-centric. So we need everyone to take care of complex patients across all specialties and professional groups, and make sure you don’t affect efficiency, but also that you don’t affect patient care.