One of the givens of EHR life is that users, especially physicians, spend excessive time keying into EHRs. The implication is that much keyboarding is due to excessive data demands, poor usability or general app cussedness. There’s no end of studies that support this. For example, a recent study at the University of Wisconsin-Madison’s Department of Family Medicine and Community Health in the Annals of Family Medicine found that:
Primary care physicians spend more than one-half of their workday, nearly 6 hours, interacting with the EHR during and after clinic hours. The study broke out times spent on various tasks and found, unsurprisingly, that documentation and chart review took up almost half the time.
Figure 1. Percent Physician’s Time on EHR
This study is unique among those looking at practitioners and EHRs. They note:
Although others have suggested work task categories for primary care,13 ours is the first taxonomy proposed to capture routine clinical work in EHR systems.
They also make the point that they captured physician EHR use not total time spent with patients. Other studies have reached similar EHR use conclusions. The consensus is there too much time keyboarding and not enough time spent one to one with the patient. So, what can be done? Here, I think, are the choices:
- Do Nothing. Assume that this is a new world and tough it out.
- Use Scribes. Hire scribes to do the keyboarding for physicians.
- Make EHRs Easier. Improve EHRs’ usability.
- Make EHRs Smarter. Adapt EHRs to physician’s needs through artificial intelligence (AI) solutions.
- Offload to Patients. Use patient apps to input data, rather than physician keyboarding.
Examining the Alternatives
1. Do Nothing. Making no change in either the systems or practioners’ approach means accepting the current state as the new normal. It doesn’t mean that no changes will occur. Rather, that they will continue at an incremental, perhaps glacial, pace. What this says more broadly is that the focus on keyboard, per se, is wrong. The question is not what’s going in so much as what is coming out compared to old, manual systems. For example, when PCs first became office standards, the amount of keyboarding vs. pen and paper notations went viral. PCs produced great increases in both the volume and quality of office work. This quickly became the new norm. That hasn’t happened with EHRs. There’s an assumption that the old days were better. Doing nothing acknowledges that you can’t go back. Instead, it takes a stoic approach, and assumes things will get better eventually, so just hang in there.
2. Scribes. The idea of using a scribe is simple. As a doctor examines a patient, the scribe enters the details. Scribes allow the physician to off load the keyboarding to someone with medical knowledge who understands their documentation style. There is no question that scribes can decrease physician keyboarding. This approach is gaining in popularity and is marketed by various medical societies and scribe services companies.
However, using scribes brings a host of questions. How are the implemented? I think the most important question is how a scribe fits into a system’s workflow. For example, how does an attending review a scribe’s notes to determine they convey the attending’s clinical findings, etc. The attending is the responsible party and anything that degrades or muddies that oversight is a danger to patient safety. Then, there are questions of patient privacy and just how passive an actor is a scribe?
If you’re looking for dispositive answers, you’ll have to wait. There are many studies showing scribes improve physician productivity, but few about the quality of the product.
3. Make EHRs Easier. Improving EHR usability is the holy grail of health IT and about as hard to find. ONC’s usability failings are well known and on going, but it isn’t alone. Vendors know that usability is something they can claim without having to prove. That doesn’t mean that usability and its good friend productivity aren’t important and are grossly overdue. As AHRQ recently found:
In a review of EHR safety and usability, investigators found that the switch from paper records to EHRs led to decreases in medication errors, improved guideline adherence, and (after initial implementation) enhanced safety attitudes and job satisfaction among physicians. However, the investigators found a number of problems as well.
These included usability issues, such as poor information display, complicated screen sequences and navigation, and mismatch between user workflow in the EHR and clinical workflow. The latter problems resulted in interruptions and distraction, which can contribute to medical error.
Additional safety hazards included data entry errors created by the use of copy-forward, copy-and-paste, and electronic signatures, lack of clarity in sources and date of information presented, alert fatigue, and other usability problems that can contribute to error. Similar findings were reported in a review of nurses’ experiences with EHR use, which highlighted the altered workflow and communication patterns created by the implementation of EHRs.
Improving EHR usability is not a metaphysical undertaking. What’s wrong and what works have been known for years. What’s lacking is both the regulatory and corporate will to do so. If all EHRs had to show their practical usability users would rejoice. Your best bet here may be to become active in your EHR vendor’s user group. You may not get direct relief, but you’ll have a place, albeit small, at the table. Otherwise, given vendor and regulatory resistance to usability improvements, you’re better off pushing for a new EHR or writing your own front EHR front end.
4. Make EHRs Smarter. If Watson can outsmart Kent Jennings, can’t artificial Intelligence make EHRs smarter? As one of my old friends used to tell our city council, “The answer is a qualified yes and a qualified no.”
AI takes on many, many forms and EHRs can and do use it. Primarily, these are dictation – transcription assistant systems. They’re known as Natural Language Processing (NLP). Sort of scribes without bodies. NLP takes a text stream, either live or from a recording, parses it and puts it in the EHR in its proper place. These systems combine the freedom of dictation with AI’s ability to create clinical notes. That allows, the theory maintains, a user to maintain patient contact while creating the note, thus solving the keyboarding dilemma.
The best-known NLP system Nuance’s Dragon Medical One, etc. Several EHR vendors have integrated Dragon or similar systems into their offerings. As with most complex, technical systems, though, NLP implementation requires a full scale tech effort. Potential barriers are implementation or training shortcuts, workflow integration and staff commitment. NLP’s ability to quickly gather information and place it is a given. What’s not so certain is its cost effectiveness or its product quality. In those respects, its quality and efficacy is similar to scribes and subject to much the same scrutiny.
One interesting and wholly unexpected NLP system result occurred in a study by the University of Washington Researchers. The study group used an Android app NLP dictation system, VGEENS, that captured notes at bedside. Here’s what startled the researchers:
….Intern and resident physicians were averse to creating notes using VGEENS. When asked why this is, their answers were that they have not had experience with dictation and are reluctant to learn a new skill during their busy clinical rotations. They also commented that they are very familiar with creating notes using typing, templates, and copy paste.
The researchers forgot that medical dictation skills are just that, a skill and don’t come without training and practice. It’s a skill of older generations and that keyboarding is today’s given.
5. Offload to Patients. I hadn’t thought of this one until I saw an article in the Harvard Business Review. In a wide ranging review, the authors saw physicians as victims of medical overconsumption and information overload:
In our recent studies of how patients responded to the introduction of a portal allowing them to e-mail health concerns to their care team, we found that the e-mail system that was expected to substitute for face-to-face visits actually increased them. Once patients began using the portal, many started sharing health updates and personal news with their care teams.
One of their solutions is to off load data collection and monitoring to patient apps:
Mightn’t we delegate some of the screening work to patients themselves? Empowering customers with easy-to-use tools transformed the tax reporting and travel industries. While we don’t expect patients to select what blood-pressure medications to be on, we probably can offload considerable amounts of the monitoring and perhaps even some of the treatment adjustment to them. Diabetes has long been managed this way, using forms of self-care that have advanced as self-monitoring technology has improved.
This may be where we are going; however, it ignores the already crowded app field. Moreover, every app seems to have its own data protocol. Health apps are a good way to capture and incorporate health data. They may be a good way to offload physicians’ keyboarding, but health apps are a tower of protocol Babel right now. This solution is as practical as saying that the way to curb double entering data in EHRs is to just make them interoperable.
What’s an EHR User to Do?
If each current approach to reducing keyboarding has problems, they are not fatal. I think that physician keyboarding is a problem and that it is subject to amelioration, if not solution.
For example, here’s Nordic’s Joel Martin on EHR usability:
… In reality, much of this extra work is a result of expanded documentation and quality measure requirements, security needs, and staffing changes. As the healthcare industry shifts its focus to value-based reimbursement and doing more with less, physician work is increasing. That work often takes place in the EHR, but it isn’t caused by the EHR’s existence.
Blaming the EHR without optimizing its use won’t solve the problem. Instead, we should take a holistic view of the issues causing provider burnout and use the system to create efficiencies, as it’s designed to do.
The good news is that optimizing the EHR is very doable. There are many things that can be done to make it easier for providers to complete tasks in the EHR, and thereby lower the time spent in the system.
Broadly speaking, these opportunities fall into two categories.
First, many organizations have not implemented all the time-saving features that EHR vendors have created. There are features that dramatically lower the time required to complete EHR tasks for common, simple visits (for instance, upper respiratory infections). We rarely see organizations that have implemented these features at the time of our assessments, and we’re now working with many to implement them.
In addition, individual providers are often not taking advantage of features that could save them time. When we look at provider-level data, we typically see fewer than half of providers using speed and personalization features, such as features that let them rapidly reply to messages. These features could save 20 to 30 minutes a day on their own, but we see fewer than 50 percent of providers using them.
Optimization helps physicians use the EHR the way it was intended – in real-time, alongside patient care, to drive better care, fewer mistakes, and higher engagement. Ultimately, we envision a care environment where the EHR isn’t separate from patient care, but rather another tool to provide it.
What does that mean for scribes or NLP? Recognize they are not panaceas, but tools. The field is constantly changing. Any effort to address keyboarding should look at a range of independent studies to identify their strengths and pitfalls. Note not only the major findings, but also what skills, apps, etc., they required. Then, recognize the level of effort a good implementation always requires. Finally, as UW’s researchers found,surprises are always lurking in major shake ups.