To Combat Physician Burnout and Improve Care, Fix the Electronic Health Record

Performing several tasks, badly. The EHR is a lot more than merely an electronic version of the patient’s chart. It has also become the control panel for managing the clinical encounter through clinician order entry. Moreover, through billing and regulatory compliance, it has also become a focal point of quality-improvement efforts. While some of these efforts actually have improved quality and patient safety, many others served merely to “buff up the note” to make the clinician look good on “process” measures, and simply maximize billing.

Mashing up all these functions — charting, clinical ordering, billing/compliance and quality improvement — inside the EHR has been a disaster for the clinical user, in large part because the billing/compliance function has dominated. The pressure from angry physician users has produced a medieval solution: Hospital and clinics have hired tens of thousands of scribes literally to follow clinicians around and record their notes and orders into the EHR. Only in health care, it seems, could we find a way to “automate” that ended up adding staff and costs!

INSIGHT CENTER

As bad as the regulatory and documentation requirements are, they are not the largest problem. The electronic systems hospitals have adopted at huge expense are fronted by user interfaces out of the mid-1990s: Windows 95-style screens and dropdown menus, data input by typing and navigation by point and click. These antiquated user interfaces are astonishingly difficult to navigate. Clinical information vital for care decisions is sometimes entombed dozens of clicks beneath the user-facing pages of the patient’s chart.

Paint a picture of the patient. For EHRs to become truly useful tools and liberate clinicians from the busywork, a revolution in usability is required. Care of the patient must become the EHR’s central function. At its center should be a portrait of the patient’s medical situation at the moment, including the diagnosis, major clinical risks and trajectory, and the specific problems the clinical team must resolve. This “uber-assessment” should be written in plain English and have a discrete character limit like those imposed by Twitter, forcing clinicians to tighten their assessment.

The patient portrait should be updated frequently, such as at a change in clinical shifts. Decision rules determining precisely who has responsibility for painting this portrait will be essential. In the inpatient setting, the main author may be a hospitalist, primary surgeon, or senior resident. In the outpatient setting, it’s likely to be the primary care physician or non-physician provider. While one individual should take the lead, this assessment should be curated collaboratively, a la Wikipedia.

This clinical portrait must become the rallying point of the team caring for the patient. To accomplish this, the EHR needs to become “groupware” for the clinical team, enabling continuous communication among team members. The patient portrait should function as the “wall” on which team members add their own observations of changes in the patient’s condition, actions they have taken, and questions they are trying to address. This group effort should convey an accurate picture (portrait plus updates) for new clinicians starting their shifts or joining the team as consultants.

The tests, medications or procedures ordered, and test results and monitoring system readings should all be added (automatically) to the patient’s chart. But here, too, major redesign is needed. In reimagining the patient’s chart, we need to modify today’s importing function, which encourages users indiscriminately to overwhelm the clinical narrative with mountains of extraneous data. The minute-by-minute team comments on the wall should erase within a day or two, like images in SnapChat, and not enter and complicate the permanent record.

Typing and point and click must go. Voice and gesture-based interfaces must replace the unsanitary and clunky keyboard and mouse as the method of building and interacting with the record. Both documenting the clinical encounter and ordering should be done by voice command, confirmed by screen touch. Orders should display both the major risks and cost of the tests or procedures ordered before the order can be confirmed. Several companies, including Google and Microsoft, are already piloting “digital” scribes that convert the core conversation between doctor and patient into a digital clinical note.

Moreover, interactive data visualization must replace the time-wasting click storm presently required to unearth patient data. Results of voice searches of the patient’s record should be available for display in the nursing station and the physicians’ ready room. It should also be presentable to patients on interactive white boards in patient rooms. Physicians should be able to say things like: “Show me Jeff’s glucose and creatinine values graphed back to the beginning of this hospital stay” or “Show me all of Bob’s abdominal CT scans performed pre- and postoperatively.” The physician should also be able to prescribe by voice command everything from a new medication to a programmed reminder to be delivered to the patient’s iPhone at regular intervals.

Population health data and research findings should also be available by voice command. For example, a doctor should be able to say: “Show me all the published data on the side-effect risks associated with use of pembrolizumab in lung cancer patients, ranked from highest to lowest,” or “Show me the prevalence of postoperative complications by type of complication in the past thousand patients who have had knee replacements in our health system, stratified by patient age.”

AI must make the clinical system smarter. EHRs already have rudimentary artificial intelligence (AI) systems to help with billing, coding, and regulatory compliance. But the primitive state of AI in EHRs is a major barrier to efficient care. Clinical record systems must become a lot smarter if clinical care is to predominate, in particular by reducing needless and duplicative documentation requirements. Revisiting Medicare payment policy, beginning with the absurdly detailed data requirements for Evaluation and Management visits (E&M), would be a great place to start.

The patient’s role should also be enhanced by the EHR and associated tools. Patients should be able to enter their history, medications, and family history remotely, reducing demands on the care team and its supporting cast. Patient data should also flow automatically from clinical laboratories, as well as data from instrumentation attached to the patient, directly to the record, without the need for human data entry.

Of course, a new clinical workflow will be needed to curate all of this patient-generated data and respond accordingly. It cannot be permitted to clutter the wall or be “mainlined” to the primary clinical team; rather, it must be prioritized according to patient risk/benefit and delivered via a workflow designed expressly for this purpose. AI algorithms must also be used to scrape from the EHR the information needed to assign acuity scores and suggest diagnoses that accurately reflect the patient’s current state.

Given how today’s clinical alert systems inundate frontline caregivers, it is unsurprising that most alerts are ignored. It is crucial that the EHR be able to prioritize alerts that address only immediate threats to the patient’s health in real time. Health care can learn a lot from the sensible rigor and discipline of the alert process in the airline cockpit. Clinical alerts should be presented in an easy-to-read, hard-to-ignore color-coded format. Similarly, hard stops — system-driven halts in medication or other therapies — must be intelligent; that is, they must be related to the present reality of the patient’s condition and limited to clinical actions that truly threaten the health or life of the patient.

From prisoners to advocates. The failure of EHRs thus far to achieve the goals of improving health care productivity, outcomes, and clinician satisfaction is the result both of immature technology and the failure of their architects to fully respect the complexity of converting the massive health care system from one way of doing work to another. Today, one can see a path to turning the EHR into a well-designed and useful partner to clinicians and patients in the care process. To do this, we must use AI, vastly improved data visualization, and modern interface design to improve usability. When this has been accomplished, we believe that clinicians will be converted from surly prisoners of poorly realized technology to advocates of the systems themselves and enthusiastic leaders of efforts to further improve them.