What will it take to make the health quality revolution succeed?

The US health system has pervasive quality of care problems. Aaron showed this in comprehensive detail (start here). The famous report To Err is Human estimated that between 44 and 98 thousand Americans were killed each year by medical errors. Elizabeth McGlynn found that US adult patients receive only about 55% of recommended medical services and Rita Mangione-Smith found that children get only 47%.

So how do we make things better? There have been successful quality revolutions in auto manufacturing and military flight safety. I used to employ a former naval officer as a software engineer. He told me that the Navy taught him that

You get the results you inspect, not the results you expect.

This is exactly right (and it’s true even if you work for yourself). Motivation helps, but quality doesn’t improve unless you also inspect what gets done and hold people accountable. Atul Gawande popularized the idea that checklists could standardize medical practice and make risky medical procedures safer in just the way that they made air travel safer.

So what we need to do is implement lots of checklists in our electronic health records, so that healthcare providers can document that they have followed best practices. Right?

I believe in checklists, but there are fundamental limits on what they can accomplish. I see two problems.

First, clicking boxes isn’t the only kind of documentation that providers have to create. Providers also need to write a narrative about the case, so that other members of the team understand the state of the game and what they need to do next. A lot of important data about care goes into that “free text” but machines can’t read it. And if machines can’t read that data, you can’t efficiently inspect that care. So the provider has to document the care twice, once by dictating a note and again by clicking boxes. This is a bad use of provider time.

Second and more important, the safety problem is a lot harder in medicine. The body is far more complex than an airplane. Airplanes, moreover, are designed to make inspection quick and efficient, while bodies aren’t designed at all. One upshot is that there are many more things in medicine that we ought to document but that we aren’t. Unfortunately, we are hitting a limit in health care providers’ available time:

American Journal of Emergency Medicine. 2013 Sep 20.

4000 Clicks: a productivity analysis of electronic medical records in a community hospital ED.

Hill RG Jr, Sears LM, Melanson SW.


OBJECTIVE: We evaluate physician productivity using electronic medical records in a community hospital emergency department.

METHODS: Physician time usage per hour was observed and tabulated in the categories of direct patient contact, data and order entry, interaction with colleagues, and review of test results and old records.

RESULTS: The mean percentage of time spent on data entry was 43% (95% confidence interval, 39%-47%). The mean percentage of time spent in direct contact with patients was 28%. The pooled weighted average time allocations were 44% on data entry, 28% in direct patient care, 12% reviewing test results and records, 13% in discussion with colleagues, and 3% on other activities. Tabulation was made of the number of mouse clicks necessary for several common emergency department charting functions and for selected patient encounters. Total mouse clicks approach 4000 during a busy 10-hour shift.

CONCLUSION: Emergency department physicians spend significantly more time entering data into electronic medical records than on any other activity, including direct patient care. Improved efficiency in data entry would allow emergency physicians to devote more time to patient care, thus increasing hospital revenue.

Got that? These ER docs are spending 60% more time on data entry than they are on direct patient care. We can’t add more documentation because you can’t squeeze blood from a stone.

So what’s the solution? Maybe we can make documentation more efficient, although what’s easier than a mouse click? Electronic health records will get smarter and acquire the ability to read and comprehend providers’ free text narratives. Then providers won’t need to dictate and redundantly click. But these are just stopgaps.

The quality revolution will occur through the roboticization of routine healthcare. That is, care will get a lot better when we have machines that can interact intelligently with patients, including understanding what patients tell them. These will be machines that access enormous volumes of clinical data and use continuously improving algorithms to personalize care to the patient’s preferences and physiology. Robots are awesome at consistently following best practices and they will happily generate limitless machine-readable documentation.

This won’t happen soon, but look around: you can see the foundations. This is, for example, where Google is headed. The company is not so much about search or cell phone operating systems as about solving fundamental problems of artificial intelligence, such as natural language understanding. Of course, the evolution of machine intelligence is a bigger event than a mere revolution in health care. We live in interesting times.


Hidden information below


Email Address*