More than 150 years ago, Dr. Ignaz Semmelweis demonstrated that hand-washing prior to treating patients dramatically reduced fatal infections. His contemporaries could not accept his results, and they did not adopt hand-washing as a standard of care. One in four women continued to die after childbirth from overwhelming infection. Semmelweis sunk into a depression, was committed to an asylum and then was beaten to death by the asylum guards just two weeks later.
Today, hand-washing is a known prerequisite for safe clinical care. Yet, recent reports demonstrated that clinicians still failed to wash their hands more than one-third of the time, and patients suffered as a result. With Medicare funds at stake, health care organizations are even going so far as to enact video monitoring programs to ensure clinicians are washing their hands with soap or properly sanitizing with alcohol-based sanitizers. Standard hand-washing for clinical care literally takes just 15 seconds, it saves lives, and everyone knows they should do it. So what's the problem?
The problem is that making changes in clinical workflow -- no matter how quick, easy, obvious and valuable the changes might seem -- is really tough when clinicians have more work than the time needed for it. Fifteen seconds multiplied by hundreds of patient interactions per shift equals a lot of time. With time as a zero-sum game (every new task steals time from an existing one), the harried clinician becomes forced into a devil's choice: Administer overdue medications to the heart attack patients or wash hands before entering each room and delay things even further. There is no right choice in this situation, only a guess as to which is less bad.
After 150 years, we finally have made headway on the hand-washing issue (sort of) by mounting anti-microbial hand gel bottles outside each patient room. Hands are cleaned without even a break in stride by the clinician. Success came only when hand-washing no longer stole time from other patient care activities.
The Secret to Success: Time
Every change introduced into medical practice -- no matter how valuable -- must be at least time-neutral, or it will risk failure. After all, we couldn't get clinicians to reliably wash their hands for 150 years until we made it time-neutral. For any change that steals time from clinicians, it's probably appropriate to ask, "Is this change even more valuable to patients than hand-washing?"
I suspect very few workflow changes in medicine could meet this criterion.
Clinicians are people too, and we can force time-stealing processes into their workflow with threats of termination or financial penalties. That's what's happening today in health IT.
Government financial incentives and penalties have increased health IT adoption. These changes have created unintended consequences as clinicians stop doing other clinically valuable activities to accommodate the time-stealing health IT and government requirements for its use. Residents now sit in front of a computer three times longer than they spend seeing patients.
The most insidious and damaging victim of stolen clinical time is "thinking time" -- time spent developing a differential diagnosis and choosing the best course of action. As decision making time erodes, the clinician is forced into "shotgun mode" -- ordering a barrage of expensive tests to cover for the shortchanged thinking time. Scariest of all, we have convinced ourselves that the dim-witted "decision support" provided by our electronic health record systems ("Warning! Narcotics interact with Valium!") more than makes up for lost thinking time. Perhaps that's why diagnostic errors remain "the most common, costly and dangerous errors made by doctors in the U.S."
We can resolve this conundrum by making health IT time-neutral (or better, time-saving) for clinicians. A number of technologies can achieve this, but too often they are seen as "nice to have" rather than an absolute pre-requisite to success. That must change.
The Highest and Best Use of Clinical Time
The valuable tasks that only the clinician can do are the highest and best use of clinical time. When clinicians stop doing other tasks in favor of these more valuable ones, everyone wins -- clinicians, hospitals, payers and especially patients.
Clinicians should never spend time logging into health IT. Proximity radio frequency identification tags, gesture recognition and biometrics (voice, fingerprint, iris and face) offer faster and more secure alternatives.
Clinicians should never spend time typing unless they prefer it. Modern speech recognition -- particularly a dedicated medical system that continuously learns through cloud-based voice profiles -- is fast, accurate and reliable.
Clinicians should never see a wait cursor for common tasks. Great coding and architectural design, an emphasis on system speed over user interface flashiness, appropriate investments in ultra-fast hardware and software infrastructures and an avoidance of optimizing IT convenience (e.g., virtualized operating systems) over clinical performance can achieve that goal. Sub-second response times for common clinical tasks can be achieved in large-scale clinical systems.
Clinicians should never spend time translating the encounter into "computer-speak." For example, clinicians should not be forced to use massive pick lists of computer-coded terms navigable only by search. Computers using modern day natural language understanding should automatically translate the clinician's free text into the standardized codes that computers understand.
Clinicians should never do work that computers could do for them. For example, clinicians manually populate problem and medication lists with computer-readable codes. In return, why can't the computer take a provisional pass at medication reconciliation on behalf of the clinician? Google Translate can match a word in Haitian Creole to its counterpart in Czech. Surely the EHR can guess that metformin might be the replacement for glipizide in a diabetic patient. If computers demand work from us then we should demand even more from them in return.
Clinicians should never have to duplicate work. For example, many call the increased specificity required by ICD-10 "a documentation problem more than a coding problem." Is that true? A fracture's type, laterality, and healing state have already been documented in the radiologist's reading. Why must the treating physician re-document that, and then re-do that work again when selecting an ICD-10 code? Computers empowered with natural language understanding should do this work. Like most EHR documentation, ICD-10 seems less a documentation problem and more a data reuse problem. Computers should fix this, not make it worse.
IT solutions that speed up clinicians rather than slowing them down with overhead and duplicate work can be achieved. These systems will easily integrate into workflows and create more thinking time, more bedside time, and more satisfied and healthy patients.