Problems with EMRs: The Physicians' Perspective
We have been working with medical systems for several years at Human Factors International (HFI). Two of our prominent clients are the Indian Health Service (IHS), a division of Health and Human Services within the U.S. Government, and the Mayo Clinic. IHS and the Indian Tribes provide medical care to 1.9 million American Indians and Alaska Natives through a network of 45 hospitals and 288 health centers. The Mayo Clinic is a world-renowned medical practice based in Rochester, MN.
Our observations on EMRs are based on visits to several IHS facilities with dozens of health care providers using a single EMR system that has been locally customized. We also base our insights on an extensive cross-facility usability review for Mayo Clinic. For this review, we visited four different clinics in Rochester, Boston, Jacksonville, and Madison and observed approximately 20 physicians at work with four different EMRs.
Overall, our observations support insights from the literature. Physicians, indeed, find that EMRs take a long time to learn and often make them less productive. Physicians experience specific usability problems when working with EMRs that cause long training times and loss of productivity.
Long Training Times
EMRs provide an enormous range of functionality. This complexity can be bewildering for beginning users, and it simply takes time for them to understand what can be done and how to do it. This is typically a problem with the navigational structure of the EMR system. A typical EMR system contains hundreds and hundreds of screens that need to be accessed through the system's navigational scheme using tabs, buttons, and hyperlinks. Learning the right paths takes time.
Loss of Productivity
Even after a physician has learned to use the EMR, it often causes him or her to be less productive. The causes are many:
- Difficulty finding important information, e.g., Which note is the last full note? When was the patient's last mammogram? Does the patient need a tetanus shot?
- Typing the note is slower than writing it out.
- Some tasks are tedious to complete with the EMR, e.g., reordering chronic medications.
- The system response time can be long.
- Changing exam rooms requires logging off one workstation and enduring a lengthy login to a new one.
- Lack of integration places information needed for one task on multiple screens, e.g., reviewing lab results as affected by a change in medications.
- Searching for a procedure or diagnosis can produce useless results. Physicians speak one language, but the procedures and diagnoses are sometimes based on a different set of codes.
- Some screens contain information that is not needed, but missing information that is.
- Some screens are densely packed with information, making information hard to find.
This is only a partial list, and the list for each EMR system is different.
The Fundamental Problem: Skilled Users, Complex Tasks, and Complex Functionality
Fundamentally, usability problems arise in EMRs because of the interaction of three very complex components: physicians, their tasks, and the EMR. Physicians are highly trained professionals with very large knowledge bases and deep problem solving skills. In addition, physicians specialize in one of dozens of specialties. The tasks they complete, diagnosing and treating illnesses, are very complex and high risk. Finally, to support physicians as they complete their tasks, they need a very complex but efficient information system. Making complex systems usable is extremely difficult. It is no surprise that most EMRs struggle with usability.
Our focus in this paper is on the users' tasks. We certainly cannot cover all aspects of medicine known to physicians. Nor can we dive into all the usability problems of EMRs. Instead, we will focus on the tasks that physicians complete in the course of a patient encounter and highlight the corresponding usability problems that arise from variances in the way physicians work.
Variances in User Tasks
Overall there is enormous variability in the way that physicians work. Some examples include variations in the pace of work, the use of nurses, mode and timing of data entry, timing of patient load, task flow issues, and variations in needed functionality. Each of these deserves a brief explanation along with the corresponding usability issues that arise.
Pace of Work
The pace of work across physicians is highly variable. Some physicians see one patient per hour; some see four patients per hour. Logging in and out of a workstation as the physician rapidly changes exam rooms can be both annoying and unproductive. It gets worse for the physician if he or she is seeing more than one patient at a time, as is typical for many specialists.
Delegation of Work
The delegation of work varies across physicians. Some physicians do everything from taking vitals to ordering tests to reviewing systems. Others rely on assistants or nursing staff to complete work, either directly (e.g., instructing the nurse to order an MR scan of the patient's sinuses) or indirectly (e.g., nurses often have standing orders to administer routine tests and immunizations). Unfortunately, one system we observed requires that the nurses log into the EMR as themselves but order labs and procedures via proxy. This is a tricky process that will send the lab results and procedure results not to the ordering nurse but to the physician who needs to see them. If it is not done correctly, the physician will never see the results.
Getting data into the system can be done in different ways, and each provider has his or her preferences. (See Figure 1 for a sample note.) Younger physicians tend to be touch typists, while older physicians tend to prefer dictation. However, many physicians still prefer to write, and this can be done with a tablet. Finally, some physicians prefer combinations, e.g., importing boilerplate data, like medications and labs, while typing in their plans. One major EMR system we saw did not support dictation, forcing non-typists to hunt and peck. In another case where dictation was available, the department head mandated that all physicians type in their notes so that they would be immediately available. (Dictation typically requires a 24 hour turnaround.) This was a significant burden to non-typists.
Figure 1. Sample Note
Timing of Data Entry
The timing of data entry can be highly variable. Some physicians enter the note with the patient present, while others wait until the patient has left the exam room. Those with busy schedules tend to wait until the end of the day to enter their notes for all patients. Waiting until the end of the day to enter notes is potentially risky, as it relies on faulty human memory to reconstruct what occurred hours earlier.
Task Flow Variability
The physician's task flow can be highly variable and frequently interrupted. One physician may always review the patient's chart, then interact with the patient, then order labs, medications, and consultations with specialists, before composing the note and recording billing. Another physician may first interact with the patient, then order labs, then review the patient's chart, then order medications, and finally wait until the end of the day before composing the note and recording billing. (See Figure 2.) Both physicians, though, are constantly interrupted by their beepers, their phones, and the paging system. When this happens, they often lose track of where they are in their task flow because the EMR interface does not change to show which subtasks have been started or completed. This forces the physician to lose time by reviewing current work or face potential safety issues from incomplete work. (See Figure 3 for sample navigation structure from an EMR.)
Figure 2. Two different task flows for the same task
Figure 3. EMR navigation with no feedback on completed steps
Also, many of the EMR systems we saw did not support rapid switching from one patient to another and back again. This situation is surprisingly typical where physicians are interrupted at least once per hour by pages to attend to another patient's record. Switching between patients then becomes a frequent but time-consuming process.
Need for Different Functionality
Not surprisingly, different specialties need different functionalities. Obstetricians need a special set of vitals on the developing child. Ophthalmologists need to record drawings of eyes. Radiologists work with digitized images. Each specialty typically needs a specialized interface to support their unique needs. Unfortunately, not all EMRs provide this level of customization. Most systems we saw rely on a text-centric interface for recording physicians' notes; no drawings are possible. Another system at a remote clinic expected the local tech staff to customize screens for each of the specialties. However, the differences between specialists do not end with functionality.
Different Perspectives on Health
Different specialists have different perspectives on patient health. Certainly all physicians wish to "heal the lame." Orthopedists fix broken bones and surgeons repair damaged organs. However, in family medicine, and often in internal medicine, they are more focused on prevention, particularly when seeing a patient for a regular check-up. Getting a timely flu shot, getting regular mammograms, and controlling cholesterol levels can have enormous long-term health benefits. As a result, different physicians often need to see different information about the same patient. Unfortunately, the interfaces we saw in popular EMRs did not take these differences into account. In general, prevention-oriented family doctors saw the same user interface as orthopedists.
All these task differences—in pace of work, in delegation, in data entry, in functionality, and in perspective—contribute to the challenge of designing a usable EMR. However, there are yet more challenges that usability specialists face.