What we lose with data-driven medicine

The business of medicine in United States is about extracting reimbursements from payers while keeping regulators happy. Well-being and satisfaction of the patient is important only incidentally. Since most doctors would prefer to be in the business of helping patients, finding themselves in the business of billing insurance causes burnout.

Blaming the automation software for this fundamental misalignment is silly. Software is just a tool that optimizes for the needs of the business and the customer. And in American healthcare the patient is not the customer. Insurance companies and the government are the customer.

So I'm a medical student, as you can probably tell from my comment history. I once helped write a proposal for the American Medical Association related to medical scribes - individuals that record all the data from physicians into the EHR. The idea for these scribes were to allow physicians to focus on the patient care aspect rather than the documentation and billing.

Our proposal was approved almost unanimously by the student section of the meeting, but was argued extensively by the physician section, not on the proposals actual merit, but on the usefulness or lack thereof of scribes. Some delegates thought that they led to errors in documentation (our proposal was to study this), others thought they would be leapfrogged by newer technologies and thus unnecessary.

Long story short, I have two comments. The first, that having more scribes would be beneficial in letting doctors being doctors, perhaps being a simple solution to the problems demonstrated in the article.

The second is that doctors from all kinds of specialties probably agree on many issues affecting patient care and physicians at large. However, all I see are disagreements.

These EHR systems sound difficult and time-consuming to use. If they can't/won't be directly improved any time soon, why don't nurses and doctors have data entry specialist partners to do the clicking and typing? Why not take some people with a 2-year medical data entry certification and train them to use $INSTITUTION's customized EHR system? It seems crazy to make people who are qualified to actually diagnose and treat patients spend half their time on clerical work using cumbersome software.

Maybe people are worried that EHR quality would suffer if people with fewer credentials entered the data. According to the article EHR systems are already full of garbage data entered by physicians. Does it actually get worse if data entry specialists do the entry instead? This appears to be readily testable. You can have them do it in parallel with the physicians who enter their own EHR data now, and compare after a year.

Lamenting documentation requirements is as old as Medicaid itself. With more and more oversight by governing bodies and the burdens of "quality-metrics", the personal connection with carers will be increasingly distant.

I love electronic records because I don't have to handwrite anymore, and all the data I need is just a few clicks away. The problem isn't the electronic nature of the records, but rather the reporting and documentation requirements that come with it. Why are there increased amounts of documentation required? Because it is a form of rationing: make patient care so time consuming that fewer patients can be seen, and less can be billed long term. It really is a pernicious way to decrease healthcare expenses, but it comes at the cost of depersonalizing patients, physicians, and nurses.

That, and also legal protections because lawyers exploit any "bad outcome" as negligence, even though no negligence occurred.

Yes, doctors should focus on the patient and the family, not the computer. Technology should be scrutinized to be more supportive.

Yet a significant problem with this article is that it provides no comparison to other professions.

"A clinician will make roughly 4,000 keyboard clicks during a busy 10-hour emergency-room shift" -- I doubt that 4,000 "keyboard clicks" is unusual for any profession at present, even one where attention is moving across interactions with patients, colleagues, machines, and the computer. A page of text is about 3,000 characters.

(Aside: I'm actually not quite sure what is meant by a "keyboard click" is -- maybe the author is talking about mouse clicks or toggling checkboxes; elsewhere in the article the phrase is "4,000-key-clicks-a-day" -- if it's just 4,000 taps to keyboard keys, that's a pretty low number. I don't think that 4,000 checkbox or mouse clicks is even really such a big number. Answers here (https://www.quora.com/What-is-the-avarage-number-of-mouse-cl...) suggest between 5000 and 7000 mouse clicks/day. Another resources says a doc does about 2500 clicks/day -- http://www.healthcareitnews.com/infographic/infographic-one-... -- and that's for a 16 hour shift.)

In any case, is there something distinct about clinician work with regard to the use of computers, compared to what we're all doing? A claim that there is would strike me as special pleading. If there is an argument to be made, it must be made comparatively.

The author also claims that "Even if the E.H.R. is not the sole cause of what ails us, believe me, it has become the symbol of burnout" -- do doctors _really_ suffer from more burnout than other professions? _Doctor-authored_ studies may say so, but we need to have unbiased studies of burnout across the professions, and understand generally how being lassoed to a computer affects morale.

I realized the other day we find ourselves increasingly influenced by algorithms. Intentionally or not people are forming their lives around arbitrary algorithms. The YouTube guy who reacted outrageously in a Japanese suicide forest was forming his life decisions according to YouTube's algorithms to generate views. Now doctors are making medical choices based on algos, what's next?

At what point does the hivemind merge with algorithmic ML to form a sort of distributed AI?

This is one of the really sad thing about medicine today. For those of us who remember how going to the doctor used to be, the experience these days is so disheartening. All that slavish tending to to the computers and data entry instead of spending time with the patients. Nurses and even nursing assistance have also been greatly impacted by this.

I wonder what the actual split of those two hours of EHR per one hour of patient care is actually targeted for _medical_ information vs cost accounting and risk management?

My experience (and resulting skepticism) in data collection for business suggests that “the business” side is the driver behind many or most of the inefficiencies.

There are so many ideas here for how computer systems could be improved. It reads like an argument against the computer in a medical setting (or at least a move away from current, perceived, over-reliance) and yet systems could so easily fix 99% of these problems.

If anyone in this thread is a physician, and is looking for more autonomy in the way that they practice medicine my email is in my profile. We're working on fixing these problems at our clinic and what we've built could be of interest to you.

It's not like the doctors themselves are going to drive down the cost of healthcare. Machine driven medicine is the only way we can replace overpaid doctors who aren't doing anything much more intellectually challenging than car mechanics and yet are exploiting the American health care system for all it's got. At first the machines are going to be used for those without good health insurance, as those machines get smarter eventually even rich people will use them, and we can finally start to lower health care costs for everyone.

Some thoughts on why the system is broken. 1) The ONC certification process. (https://www.healthit.gov/topic/certification-ehrs/about-onc-...). It is way too complex to explain here, but it started with the incentive program CMS established in 2011 to push doctors to electronic health care records. Like a lot of CMS programs, the intention was there, but so was the opportunity for fraud and abuse. I can't remember the exact numbers, but the amount of certified EHR vendors dropped dramatically between the 2011 and 2014 certifications. The next round (currently 2017 stage 3) will further reduce that number. The 2014 round of certification definitely weeded out a lot of crappy EHR systems that were thrown together and sold to clinics. The problem now is twofold. First, it's becoming so burdensome and expensive to keep an EHR system certified, that only the well-financed (EPIC, Cerner, etc.) can afford to stay in the game. Second, it's extremely expensive for a clinic/hospital to switch EHR systems. Even if you have developed the most amazing EHR system known to man, getting a big hospital or clinic group to switch systems again will be next to impossible. 2) CQM and PQRS/Meaningful Use reporting. It's complicated, time consuming, and expensive. Prior to 2017, doctors were required to report to two different systems PQRS and CQM, both administered by CMS. PQRS has since been replaced by MIPS. Measures that appear to be identical between the two systems (i.e. CMS 69 and PQRS 128) sometimes have slightly different parameters. The measures themselves have versions and can change year to year. The entire system puts a huge financial burden on doctors. I get the intent. It’s the implementation that sucks. 3) Imagine if, on your job, you had to use a system of 70,000 different codes to identify each and every thing you did at work, and you had to justify each and every thing you did with up to 6 different reasons (out of a possible 70,000 reasons), and you had to submit this report each and every day, and if you made any mistakes, or if your reasons weren’t sufficient to justify your work, you didn’t get paid, and you then had to file an appeal to fight it, and it might be several months before you finally got paid. That’s the health care claim system. With the adoption of ICD-10 in 2015, the number of available diagnosis codes went from ~14k to ~70k. The number of Procedure codes went from ~4k to over 70k. If you’re into data analytics you probably had an orgasm. If you are a doctor, trying to get a heath care claim paid, your life got a lot worse. Does it really matter whether a patient got hurt because of a collision with a roller skater (ICD-10 V00.01). Guess what? Your doctor doesn’t get paid if he uses ICD-10 code V00.01. That’s because he has to indicate whether the collision was an initial (V00.01XA), subsequent (V00.01XD), or sequela (V00.01XS). The more complex the system, the more ways insurers can deny claims. It’s easy to get frustrated because your doctor/nurse spends all their time staring at their tablet/laptop clicking away instead of talking to you. Don’t get mad at them. It wasn’t their idea. To the people who are quick to paint the doctors as greedy, overpaid clerks who can and should be replaced by computer and AI, read this https://www.nytimes.com/2018/05/18/opinion/artificial-intell... . Keep on dreaming about your Elysium/Expanse fantasy where patients are hooked up to a machine and diagnosed/cured. It’s not happening now, it’s not going to happen any time soon. In the meantime, the most effective way for a doctor to treat a patient is to look at the patient, talk to the patient. It’s not reading tea leaves. There are subtle, non-verbal cues that are impossible to pick up if a doctor has his back to you because he/she is forced into a clerical role (see points 1,2,3 above). We are decades away from a computer and AI being able to do this. I hope I am wrong. I watch the Expanse too but it is just science fiction, the key word being fiction.

To the commenters claiming “scribes” can do this. It’s just not that easy. I was involved in developing a scribe system. It worked for a couple doctors. We thought we had the next big thing. The doctors were able to go in and actually talk to and look at the patient while a scribe sat in another room and listened to the conversation and watched a mirror of the doctor’s tablet (all with patients consent). By the time the doctor left the exam room, the progress notes were completed and prescriptions, follow-ups etc. were ready for order. The doctor just had to review and complete. System was great. Doctors were able to go home at a decent hour instead of spending 2 hours in the evening going over each patients encounter. When we tried to expand it failed. In hindsight it was easy to understand why. The scribes we initially used were CAs who had been working with these doctors for years and they could pick up on subtle verbal cues generating complete encounter notes just based on a few comments. In our experiment, it just didn’t work once we brought in scribes who had never worked with or around the doctors. I’m sure there’s a way to make this work and hopefully someone will one day. It would be nice to have a conversation with my doctor again. Anyway, just some thoughts. I see a lot of posts in here discounting the article and claiming the world is ready for computers to replace doctors, and it’s just the greed/ignorance of doctors holding us back. If you’re ready to put your life in the hands of AI and computers, I wish you the best of luck. I just want a system where doctors can actually be doctors again. The greedy people in our health care system don’t wear white coats. They run around with titles like “Hospital Administrator”, “Pharmaceutical Rep”, “Health Care Lobbyist”, “Senator/Congressperson” and “Insurer”.

Nothing we lose nothing. It's so frustrating...people LOVE to drum up fear about technology. But technology is an enhancer! It doesn't prevent anyone from doing anything they did before. If you want to use a horse and cart go for it. If you want to use a old fashioned rolodex instead of Facebook go for it. Use the tech or don't. The problem is people resist change even if it's for the objective scientifically proven better.

It sounds like the problem is in the people or procress then if they cut corners and aren’t thorough.

I didn’t see the computer or data failing in my hastey read through of the piece. It always contained what data the people entered.