The flaws with eCWs system aside, the larger problem with EHRs, and the fact that to date they have not improved outcomes in any significant way, lies with the fact that they are primarily designed to “optimize” revenue cycle management. Everything has to be documented correctly - or else it can’t be billed for or may get rejected on audit.
Hospital staff, including physicians, are responsible for the accuracy of this documentation, but that’s a big responsibility to bear.
The only way this may change is if we change how we pay for our care. This is not on the horizon at this time. Even the value based programs that CMS has implemented are really fee for service with a look back and adjustments to payments made based on savings that were encountered.
"Everything has to be documented correctly - or else it can’t be billed for or may get rejected on audit."
The clue is right here. This is how complexity in the private health insurance market drives costs and inefficient outcomes into every other aspect of our health care. A thousand different plans each with different billing standards and the software reflects the need to focus on making sure you can meet all the billing standards.
You think a government payment system is going to be less complex? Won't demand correct documentation? Have you ever dealt with a government records system?
Evidence from around the world is that government plans are likely to be cheaper and more effective. Are US private plans standardized regarding what will be covered, as well as the justification of a given procedure or drug? As far as I've had direct experience every plan is different, and even within the same plan there have been inconsistent approvals, doctors have to often get re-involved to add more or different justifications for procedures that are denied then covered. It's all unnecessary added overhead as exemplified by the relative costs of more than 2x other nations with government unified healthcare.
I'd much rather be paying for the time of my doctor to treat me instead of fighting my insurance company armed with a similarly high-overhead software package.
Uh medicare is more efficient than private insurance, despite having it's hands tied by current legislation keeping it from negotiating drug prices. Medicare also takes on the oldest in the population demographic, needing the most and highest cost care that the private insurance companies are then relieved from managing.
You’re completely right and wrong at the same time.
Since Medicare reimbursements are so much lower than private insurance - the consequence of moving to all Medicare would have immense consequences, and the dollars removed would Largely need to be replaced, or quality and access to care would be dramatically impacted
These rhetorical questions are disingenuous given that many examples of working government healthcare already exist. You’re asking if a problem that has been solved multiple times is solvable. The answer is yes.
That hasn’t been my experience with Medicare or Medicaid and I haven’t ever spoken with a healthcare provider that agrees with that statement. I also have family that has used Medicaid and the health care is abysmal.
The Administrator of the US Centers for Medicare and Medicaid Services has a salary of $165,000.
I don’t envy the person making that kind of money, but given that the government can do a similar task cheaper, I question the legitimacy of the boards decision to pay an excessive amount, which impacts the profitability of the firm.
> The only way this may change is if we change how we pay for our care.
Agree 100% on this. The incentives in a fee-for-service world are just so badly misaligned with the goals of functional healthcare system.
> Even the value based programs that CMS has implemented are really fee for service with a look back and adjustments to payments made based on savings that were encountered.
So CMS is acutely aware that merit-based adjustments on top of a fee-for-service model is an incremental improvement (and a very modest increment at that). So within the Medicare's Quality Payments Program, there is MIPS (Merit-based Incentive Payment System) which is exactly what you describe and where most providers default to right now. But there are also several APMs (Alternative Payment Models) that providers can elect to participate in, some of which have full or partially-capitated payment mechanisms where there is a fixed fee per beneficiary, rather than the fee-for-service model. Everything I've heard from CMS suggests that APMs are the future and fee-for-service needs to go away. That's not going to happen over night but MIPS is the narrow end of the wedge to get us there.
Are you familiar with the free market medical movement? It's definitely not a fix for the entire US healthcare system, but I think it has the potential to create some major positive improvement in outpatient care for people who don't live hours away from the nearest cashpay specialist and can afford to pay for small things out of pocket.
I work in inpatient now, but I previously worked for a radiology group that was cashpay focused. The total cost for a brain MRI at most locations was over 10x less than the technical component alone for the hospital I'm currently working for.
A related aspect is liability. Since EHRs are going to be looked at in any postmortem for a health intervention gone wrong (where lawsuits happen), providers have an incentive to not write anything that implies uncertainty. "I didn't see anything" is a hard position to get in trouble with - "I thought it might have been pneumonia but didn't do anything" can bite you.
The funny thing is, the early stimulus has lead to so much investment in US EHRs that they’re quite competitive internationally and being installed in other public-payer health systems.
What I would like to see is a comparison of truly foreign EHRs.
FTFA, non-US clinicians write up much shorter notes in the same EHR. So there’s something about the pressures on the US user that don’t exist elsewhere.
Private insurers are already moving toward value based care where providers bear some of the risk. For example it's common to have bundled payments for certain things like joint replacements covering the procedure itself and all necessary follow-up care. But the fee-for-service model will still be used on some claims for many years to come.
>Even the value based programs that CMS has implemented are really fee for service with a look back and adjustments to payments made based on savings that were encountered.
And let’s not forget one of the more troubling aspects of EHRs and Medicare patients is that physicians are now required to turn over all patient EHR data to the government (CMS) every year for a reimbursement bonus and failure to turn over patient records results in a penalty.
Government collecting patient digital health records...what could possibly go wrong?
The parent comment gets this a little wrong, but there is truth to the data collection aspect.
An increasing number of physicians have to be part of an integrated health network in order to get competitive reimbursement rates. Those networks upload individual patient data to state exchanges which in turn have started to upload to nationwide databases.
The rollout of this is staggered and uneven, as of a couple years ago when I stopped working at a company with its own electronic health record product.
As a sibling comments correctly notes, the CMS bonus (or absence of penalty) does not depend on direct uploading. It does, however, depend on using a system which has the capability of doing that uploading ('sharing').
Finally, I saw over the past 10 years a giant increase in the number of records audited by health insurers. Those audits involve giving the entire patient file to the insurer.
I'm waiting for the day one of these large databases gets compromised and everyone acts shocked that the data was being collected.
Physicians are in no way required to turn over all patient EHR data. CMS collects quality measures in the aggregate, numerators and denominators. There may be vendors in the middle who are collecting the data to calculate the measures, but the patient-level data does not go to CMS. (Obviously, their claims do, they're a payor.)
> He says Epic uses such insights to improve the client experience. But coming up with fixes is difficult because doctors “have different viewpoints on everything,” he says.
This suggests to me they are doing UX design based on "what specific interfaces and interface changes the customer asks for." While common in "enterprise" sales, I think we all know this simply _does not work_.
And it's not just an issue of the executives who make the decisions asking for different things than the end-users would. They say it's "doctors" who have "different viewpoints" we can charitably assume they really are getting UX feedback from the end-users not just from executives (I don't really believe it, but let's be charitable).
It's that UI/UX design is hard, and "just doing what the customer tells me to do" never works. The customer is not a UX designer. What the customer thinks will help may be a "local miminum" or may not actually help at all. even if it is end-users telling you what they want. end-users may be experts on the problems and frustrations they are having, but not on the design solutions.
> “We had all the right ideas that were discussed and hashed out by the committee,” says Mostashari, “but they were all of the right ideas.”
> "“The industry was like, ‘I’ve got this check dangling in front of me, and I have to check these boxes to get there, and so I’m going to do that.’ ”"
Yup, sounds like enterprise software alright.
> “It’s like asking nine women to have a baby in a month.”
That's even a paraphrase from Fred Brooks author of "The Mythical Man-Month", although the speaker may not have known it. (Brooks prob wasn't the first one to say it... but might have been about software?) https://en.wikipedia.org/wiki/Brooks%27s_law
It is and it is just like Salesforce or Oracle or SAP. It is the people who never have to use the software that buy it and force it on the people who do and they hate every minute of it.
I used to work at a hospital and some clinics in a previous life. The administration went forward with a new EMR system without actually getting the users to "test" it beyond a webinar.
I left before it started to be implemented. Not sure how it went, but it's been 4 years and they're still "working on it" according to some old co-workers still there.
Furthermore, the problem is that in these situations its the lack of iterations, and that workflows are hard to change, design is basically impossible. (Sure, an expert working group could spend time on it, but usually by the time these big projects reach the end-users and experts, it's already crunch time.)
> end-users may be experts on the problems and frustrations they are having, but not on the design solutions
This is the key here. We have continuously designed our healthcare system for “quick-fix” instead of the right solution that takes longer. Even the HITECH push for EHR adoption and Meaningful Use - twisted the arms of providers to find and use an EHR before they could be built to positively impact patient care.
Actually, what happens is they let some nurse administrator at the hospital design the UI. which leads to some horrible "kitchen sink" design w/ insane amounts of buttons and toolbars.
Part of the problem with UX is also the fact that large customizable systems like this push a lot of the "design" down to the IT staff responsible for maintaining the software.
* Lack of standard, OPEN, free to implement formats; for everything.
* That are required as a native format for all vendors considered for selection.
* Ironically, HIPPA, which correctly makes collecting data a liability.
However this also makes releasing that data to patients difficult.
It makes patients authorizing the release to other entities even more difficult.
It contributes to making critical data during emergency situations difficult to locate.
-
Some kind of standard civil contract singing PKI might be a technical solution, but
That's beyond the scope of healthcare solutions, it's a national ID / notary issue
* Lack of a central registry to at least annotate where patient records exist and can be located.
* A multitude of private healthcare cost management organizations (HMOs)**
* Artificially constrained supply in non-emergency contexts
* No shopping at all during emergency contexts (nor should there be)
(HMOs) which seem mostly to focus on for-profit results rather than actually improving quality of care, patient outcomes or actual medical costs; at least in most cases.
In fairness, even HN can't get people to stop using monospace for quoting or prose, so one can understand why it's hard to get providers to adhere to a consistent standard.
Indeed. I think I am going to start downvoting posts that quote by (or otherwise use) indention. It's unreadable on mobile. Why is this so hard to fix?
Mostly because HN has a custom markup solution and attempts to reformat single-line breaks, requiring the use of monospace for ANY level of control over formatting at all, and otherwise restricting everything to separated paragraphs that are difficult to read.
If I'm wrong and it allow either limited BBcode, white-listed HTML entities for markup, or some flavor of Markdown please let me know.
I would love to have ACTUAL bullet points and text that flow wraps per line, with any newline I enter becoming either a P block if it's a paragraph, or a <br> tag if it's not followed by a second newline.
I don't try to get real work done on mobile; it just isn't a good media for reading or composing anything complex, and a large threaded forum IS complex.
Mobile is good for getting and sending small updates, and consuming videos at lunch.
OP is talking about writing long-form comments on HN. I guess you could consider this "entertainment" too, but it doesn't change the fact that smartphones are bad for this task.
> Lack of standard, OPEN, free to implement formats; for everything.
I was surprised when I talked to a doctor family member awhile back and learned that this was not part of the ACA. A big sweeping re-write of the health care regulatory structure would have been a good time to require data interoperability.
Maybe the new plans both parties are supposedly interested in getting done this time around will include this. I won't hold my breath though.
A strong requirement for interoperability appears in the 21st Century Cures Act of 2016. The ONC is now in the process of writing regulations to implement that law.
HL7 has had open standards for moving data between systems for 30+ years. But actually building those interfaces takes a lot of work. Every EHR has its own propriety internal data model so there's always some data mapping and conversion needed.
Standards created before any software ecosystem exists to give feedback and evidence on what the standards should be, seldom turn out well.
If they had just created all of those standards and central systems _before_ rolling out EHR to everyone, I doubt it would have gone any better, honestly.
The articles does go into some of the social/political influences that led to a requirement or incentives for everyone to have EHS before anyone knew what it was going to look like, at a speed that prevented incremental piece-by-piece development of the ecosystem and adjustments to requirements, standards, and best practices based on what was found.
Without such an incremental process, I think it would be doomed no matter what.
At my institution, they developed their own EMR in house. And it works really well, presumably because 1) incentives are as well aligned as they could be, and 2) it allows incremental improvement. This latter point is crucial - the basic system went out, everyone could use it and make suggestions, and then other services were gradually integrated (radiology, pathology). The same people working on the EMR keep working on it over (thus far) 7 years. It also was dramatically less expensive than say getting a product from Epic, which is what the next iteration will be. Everyone is worried.
Worked for one of these companies and I’m honestly surprised they don’t kill more people. Most of the time I go to a doctor using the one I work on I apologize, and frequently have to help them with a bug or nonsense interface thing. I recently had to get an MRI and forgot to watch the screen while doc was filling out an order. I would have noticed the form not save the second half appropriately (was an existing bug from when I worked there that I wanted to fix but was told no). Ended up delaying my head MRI for a couple weeks. Good thing it turns out there was nothing wrong. Learned my lesson about not giving up arguments where the other person insists a bug is “intended behavior.”
Ah yes, the whole “working as designed” excuse. It’s like ISO9001 approval for your concrete parachute company. Yes, they’re consistent, but they don’t actually work.
I'll first say that I haven't read the full article, which frankly, is the same as most everyone who writes a comment.
The title and intro reminded me of this piece from the New Yorker a few months ago that written by a doctor and his experiences and others with the computers in the doctor field.
The summary is that health care, in the US and other countries too, is a giant system where even little changes take time to go through and become accepted. People who say there's an easy solution for electronic records are in no way correct.
It's a great read both in the information it gives, and also how good of a write the doctor is. I love his writing style.
I agree, not much more info than the New Yorker article.
Lots of anecdotes, but what I really care is: do outcomes improve or not? The computer can screw up, but so can the human, so which is better now?
Very US-centric too, other than the point about US doctor’s electronic notes being 3-4x as long as rest of world. Is the problem the EHR or the culture?
I don’t care about clicks. It’s more crude than using BMI for obesity. Replacing the mouse with a giant scroll wheel won’t improve things despite eliminating clicks. We use mice because they’re excellent human-machine interfaces for most users. (I like keyboard shortcuts, but you shouldn’t design for me).
How many minutes/seconds, or units of cognitive load, does it require to order that ibuprofen? And to get that ibuprofen to the from order to patient?
Are all of the clicks 10x10 pixel checkboxes on a 2 megapixel display?
Is it easy to do the right thing and hard to do the wrong thing?
What’s the full-cycle error rate of EHR vs paper requisition? And cost?
I noticed a lot of the complaints about "problems" with
the EHR is in reality mistakes made by humans that the software could theoretically have caught, like bad drug combinations, but it didn't either because the app just wasn't that smart, or because it generated alerts the humans ignored due to warnings overload (presumably itself caused by the risk of lawsuits).
This doesn't seem really like an issue with the software per se - paper can't catch mistakes in what you write on it.
I got the impression that people were complaining that such concerns are easily hidden in the confusion of the interface. With paper records, perhaps the hospital would have a system of making the summary page of the record (which is supposed to be updated before it gets put away) have important information about drug allergies, etc. In the EHR it may be 10 or 12 point text tucked into one line at the top or bottom of the screen and may not be visible on every page. Some physicians also have to deal with several different EHRs depending on where they're working on a given day.
Paper records are straightforward. You read through them completely and then go to work. If you don't do that, you're taking on a risk that you missed something important. EHRs can make that a little harder by obfuscating the history, only pulling up parts of the history at once, etc.
I think there's an interesting role to play for AI in creating useful, readable summaries of medical records that physicians can use. Probably too hard to get right to prevent accidentally censoring information that was important.
I'm glad this finally made it to the front page because it is an important story of how culture and true customers shape software development. Notice that government incentives spurred hospitals to quickly adopt software that physicians (and a whole host of other clinicians and supporting actors) use. Hospitals/Health Systems/etc != Physicians.
Some reactions to the article I sent to my team:
Bulletpoint summary: [1] “The article includes a brilliant comment from WellSpan SVP/CIO Hal Baker, MD: ‘Physicians have to cognitively switch between focusing on the record and focusing on the patient … I have yet to see the CEO who, while running a board meeting, takes minutes, and certainly I’ve never heard of a judge who, during the trial, would also be the court stenographer. But in medicine … we’ve asked the physician to move from writing in pen to [entering a computer] record, and it’s a pretty complicated interface.’”
EPtalk by Dr. Jayne 3/21/19: [2] “The piece hooks the reader by opening with a story that details a patient’s death from a brain aneurysm, with the lack of diagnosis being influenced by failure of the head scan order to be transmitted by her physician’s EClinicalWorks EHR.”
Reader Survey Results: How I Would Change EHRs: [3] “There is no perfect technology. Our ability to acknowledge data integration is key is tantamount.”
Not related, but also interesting: [4] HIStalk Interviews Grahame Grieve, FHIR Architect and Interoperability Consultant: “A lot of doctors I talk to think about this as a technology problem, but it’s not a technology problem. It’s an information problem, and so technology can’t solve it. It needs clinicians to make clinical agreements in order to get clinical interoperability.”
> Physicians have to cognitively switch between focusing on the record and focusing on the patient … I have yet to see the CEO who, while running a board meeting, takes minutes,
Aren't doctors usually assigned an assistant? My GP has one, dentists have one. (I'm in Hungary.)
An assistant to write my notes? Wow, that would be great. I've been practicing for 12 years and haven't had one yet. (Primary care/public health.)
There are certainly scribes out there and some docs use them, but the practice is hardly universal nor is it without cost. And, of course, some are better and more accurate than others.
When I had an accident in Dallas in 2012, the specialist at Baylor general had a foot pedal-activated recorder (one of those micro cassette tape deals), and then sent off his recorded notes to be transcribed. 1970s technology. Seemed pretty effective.
I used to dictate everything when I did hospitalist work. It seemed pretty efficient and notes were in the chart within hours. There's a cost there too, but it might be better borne by tertiary facilities and medical centers. It's the small offices where the expense can only be spread over a couple docs.
Yes and no. There are often lots of medical assistants of various sorts, ranging from someone titled "medical assistant" (who is qualified to take your temp, confirm you are the right person and little else) to "Physicians Assistants" or "Nurse Practitioners" who for a lay person are basically doctors with a lesser title.
An assistant specifically for documentation is known as a "scribe." This person follows a physician and does all the typing into the EHR so the doc can focus on the patient. My children's pediatrician recently recently started using one and it does allow the physician to focus more on the patient. My understanding is that they have to pay for the scribe out of their own pocket.
So the Hungarian healthcare system is in complete shambles, because of lack of funding, drastically severe power imbalances (doctors are scarce so they can do whatever they feel like, and they exploit their privileges, usually they simply ask for money to make up for the aforementioned lack of funds - but then nurses and other staff see almost nothing of these "funds"), and of course enormous lack of efficiency (due to lack of funds, patients wait a lot, travel a lot, and then wait some more, and then the doc doesn't have much time for each patient, so the results/outcomes are not great, many things get overlooked).
GPs get money from our NHS thing (it's called Country Healthcare Treasury), after patient visits and after assigned patients (patients can choose their GP), but they have to bear the fixed costs (leasing an office, scribe, etc). Since the GP system is sort of overloaded, there seems to be enough money for every GP to afford a scribe.
On a side note - there are a ton of EMR startups, but do any implement a doctor/provider-friendly UI w/ some sort of native or web interface, that meets some minimum amount of HHS/medicare bullet points? Something designed to make writing a H&P or SOAP note (a typical clinical document describing a visit) as easy as possible?
We use Allscripts and (when it's actually up and not crashing), it's painful to even type in the thing...the lag is so bad. Data support is terrible, and the thing suffers from "too much UI" syndrome in that to add any information requires too much button clicking to open dialog/selector boxes in a UI/architecture that clearly can't support it. (let's put it this way, they make you use Internet Explorer to launch some Active X app that fires up Remote Desktop into their app which runs on some underpowered instance of Windows Server 2003 or something)
At this point I'd settle for something that shows the letters on the screen right when I type them...
Enterprise software is often [a.] optimized for needs not in alignment with a diverse user population's needs, and [b.] written, configured and customized by different groups of people each with an incomplete understanding of the user populations needs. I think of the situation as that of the "Blind men and the elephant."
Plus "workflow" -- humans doing certain things in certain orders, entering them in a computer in a process which has components that also involve interacting with other software, non-software organizational systems, humans, and offline tasks -- is one of the hardest things to automate.
Because different people in different contexts will have different needs. Not just predilections. And not just because different individuals have idiosyncracies. But because they are inevitably doing _different kinds of work_ in _different organizational systems that are set up differently_, and you're trying to sell the same software to all of them, because it would not be cost effective to give each organization bespoke software. (And as we know, "customization" offers only limited escapes).
Not all "enterprise" software might be workflow-centered. But "ERP" systems like SAP are, and so are Electronic Health Records. In spades.
At the Chaos Communication Congress last year, there was this talk about electronic health record apps for patients and it shows how insecure they are and also how on an philosophical position they are a not so good idea:
We handled labs, notes, scripts, etc. (Our product didn't touch any of the financials, billing, etc. So I can't speak firsthand to the fraud stuff mentioned in the article.)
Everyone involved understood that data sharing between competitors would never be feasible. In other words, useful interchange requires someone with a big stick, such as single payer.
The push for ICD-10 (from insurers & consultants?) was ridiculous, is a work multiplier, and has zero benefit for patient care.
Our usability was pretty bad, but miles better than others (Epic, Cerner). We had the benefit of a clean slate, but too few resources to hire real, full time UI people. Also, working with (mid 2000s era) doctors was mostly brutal.
That said, healthcare is just about one of the most complicated things I've worked on. And rapidly changing. Of course the user experience sucks. I wouldn't expect any of it to improve until the domain settles down. For example, since lab reporting is industry standardized, we were able to do a pretty good job on those UIs/reports.
ICD-10 has some benefits for patient care in that it can be more specific than ICD-9. That helps when one provider sends a coded list of patient conditions to another, like for a specialist consultation.
By the time ICD-10 was mandated, ICD-9 had been obsolete for years and remaining on that version was no longer tenable. Going forward the whole industry will probably have to upgrade to ICD-11 at some point. It's just a cost of doing business. Like if you write an application targeting Windows XP, eventually Microsoft stops supporting that platform and you have to expect to upgrade occasionally.
ICD-9 had one code for animal bites. ICD-10 has too many codes. My favorite is alligator bites: https://www.icd10data.com/ICD10CM/Codes/V00-Y99/W50-W64/W58-...
For statistical purposes and public health data, knowing what animal bit people is useful, but from a fee-for-service side, having to determine and adjust the amounts paid for all these new codes seems like unnecessary complications.
It's a hierarchical code system so there's no additional complication for payers. They can just set a single coverage policy and reimbursement rate for all animal bites.
As a doctor I'm still waiting for the eMacs of EHRs/EMRs. Something with a bit a learning curve but where your hands rarely need to leave the keyboard. And where you have emmet like intellisense.
Unfortunately a lot of other countries do it better because they're in a completely different healthcare system. Which is to say that the focus of US EHR systems is billing, the focus in other countries is on patient care.
That might sound like a trivial difference, but medical staff at US healthcare providers are required to record significantly more information in order for it to be reimbursed. For example in other countries you aren't recording supply usage on a "by patient" basis, but by ward or even by department.
So patient charts are much simpler abroad. So you could copy a foreign EHR system but it would be completely unusable or require so much alteration you might have well start again. The issues are somewhat systemic.
Even Medicare which is essentially socialized medicine in the US, has substantially higher documentation requirements than equivalent programs abroad.
Fee-for-Service vs salary makes a difference too. In FFS, the quicker you finish, the quicker you go home for the same pay.
When you’re salaried by the health system, it’s an easier sell to enter in your own orders and sell the downstream savings as funding your next pay increase.
Doctor here - honestly, a paper record works great generally.
The really useful low hanging fruit of EHR is an electronic portal for radiology results, blood results, discharge summaries and letters. If you have a way to link family doctor/pharmacy prescriptions with hospitals, thats great too.
My ideal health IT system assuming perfect AI and everything else would be:
- Automatic recording of everything discussed and done
between nurses/doctors with no input from the doctor and producing a record which reads in a sensible way
- Automatic prompting for things the doctor/nurse was likely missed or should discuss during a consult (but only to a sensible level of detail)
- Automatic checking for common ommissions, or things that are being done that perhaps shouldn't be, so a patient who has a microcytic anaemia triggers a check to ask a doctor "do you want to order iron studies on this patient?"
That's the system we're building. We've been at it for over three years. Not long before we leave stealth mode and switch from R&D to commercialization. We're internally-funded, which means we don't answer to investors, we don't have to live in SV, SFO or NYC, we don't have to hire people the VC firms like/trust (we have senior executive experience at multi-nationals), and, most importantly, we can focus strictly on what's best for our colleagues and our patients.
Most of the innovation is happening in the FHIR space. And this is a very international space. As far as "interoperability" is concerned, we are all figuring this out together.
edit: I should add that US budgeting funds for adding interoperability requirements sure does help.
Mandate truly electronic record portability by making a common format, designed by a comitee of stake holders (doctors, EHR companies, etc). All EHR systems have to be able to massively import and export records from and to this format, by law.
Reducing the friction to swap providers should give a good jolt of energy to the industry (reducing moats encourages new competition)
They went wrong, but they don't have to stay that way! Commure is working to make better software for doctors, and if you might be interested in helping fix this - here's our whoishiring post https://news.ycombinator.com/reply?id=19544133
(disclaimer: I used to work on EHRs, hence the name...)
This article is melodramatic and terrible, and I feel sorry for anyone who gets their opinion on EHRs from it. It reads like some kind of expose, like it's uncovering some disturbing truths and showing the reader that our system is totally broken, or something. Not true at all, it's just that some people have bad experiences that get sensationalized and put into an attention-grabbing news piece. And then people read it and only hear about the bad stuff.
As someone who's worked in EHRs, I can tell you we're a lot better off with EHRs than ever before (particularly compared to hospitals that used to put everything on paper, in a freaking file cabinet...) and we're just improving with time. Granted, bad EHR software does exist out there, but as a whole we make patient care significantly better.
I put some real examples below. I think we have tons of anecdotal evidence pointing to the positive impact of EHRs, but I'll admit we don't have many statistics either (good or bad). I think you basically have to work in the EHR/healthcare space to see the positive impacts for yourself, but I hope this post helps people see the good things we do, too.
+ Catching medication administration and other procedural errors that would lead to patient harm and death
+ Allowing patient info to be accessible at other hospitals, when it's needed to adequately care for a patient (real example that has happened countless times: patient comes into the ER unconscious, but has a nametag; nurse uses it to look up the pt in the system, but he's never been to that hospital before. No problem; nurse requests his record from the hospital he has been to, and gets his whole history, allergies, etc so they can better determine a cause and what treatment is OK for the patient)
+ Allowing anonymized patient data to be used for population health and other research; a doctor used this to catch the Flint, MI water crisis
+ Nurses/docs can just look up patient info in one system instead of searching through an ever-growing file cabinet, or having to go get their paper record from another department, etc. It's just all in one place. Hospital admins can more easily run anonymized reports/statistics on their patients, departments, etc. to see how everything's going.
+ EHRs are able to suggest tests or screens for the patient depending on their symptoms and their whole medical history (which, again, is all in one place). In some cases, this is just a nice reminder for the provider, but in other cases can indicate something deeper is going on with the patient that really needs to be tested.
+ Sending prescriptions to pharmacies is often automatic instead of needing a call, which essentially removes another point of failure. Sending information about specimens, like blood tests or scans, can also be automatic depending on the EHR.
+ Patients may have a patient portal with the EHR that gives them a view of their past history/visits and provides an easy way to request prescription refills, schedule appointments (no call required!), and receive test results.
> >+ Catching medication administration and other procedural >errors that would lead to patient harm and death
Depends on the end user. EMR medication checking is only as good as the tech who put the bar codes on the bags. I've seen a number of med errors b/c nursing blithely scanned "correct" barcodes on wrong meds. Other meds are routinely ordered wrongly or the errors are so useless the ordering MD/RN just clicks through them to dismiss.
The best hedge against med errors are hiring dedicated pharmacists per floor double-checking the orders.
> + Allowing patient info to be accessible at other hospitals, when it's needed to adequately care for a patient (...)
This never happens. EMRS are not routinely interoperable with each other. Especially EPIC who prides itself on lock in and exclusivity.
We always get records printed out and faxed.
>+ Allowing anonymized patient data to be used for population health and other research; a doctor used this to catch the Flint, MI water crisis
Maybe if you're the head honcho w/ top level access and a bunch of EPIC minions to do this for you. For the average med student/resident/grad student....eh. Did you ever try and get broad clinical data out of EPIC or Cerner or anything else? It takes a huge amount of technical head scratching the way everything is implemented. Most pilot projects w/ academic centers still have a fellow or resident click through each record and gather data manually for their research. Otherwise a huge committee is required to wave their arms to...download data from a MUMPS database.
+ Nurses/docs can just look up patient info in one system instead of searching through an ever-growing file cabinet, or having to go get their paper record from another department, etc. It's just all in one place. Hospital admins can more easily run anonymized reports/statistics on their patients, departments, etc. to see how everything's going.
It's one system but the data is hidden away in a thousand clicks. There have been attempts to make "dashboard/portal" interfaces in EPIC but I suspect the API's and toolkits are not really there...no self respecting MD trusts those.
+ EHRs are able to suggest tests or screens for the patient depending on their symptoms and their whole medical history (which, again, is all in one place). In some cases, this is just a nice reminder for the provider, but in other cases can indicate something deeper is going on with the patient that really needs to be tested.
Again - depends on the implementation. Maybe in the ideal world, but most of the errors and suggestions that pop up are considered nuisances and not correct.
+ Sending prescriptions to pharmacies is often automatic instead of needing a call, which essentially removes another point of failure. Sending information about specimens, like blood tests or scans, can also be automatic depending on the EHR.
For e-scribe, this is gradually getting better - we still have to sometimes resort to fax or phone scripts for a lot of this.
For lab tests - it's still an issue, we never reliably get "e-results" that automagically pop into our EHR. It's still fax. A lot of it may rest of implementations of lab providers (Quest, LabCorp, etc) but whatever the reason, this still sucks.
+ Patients may have a patient portal with the EHR that gives them a view of their past history/visits and provides an easy way to request prescription refills, schedule appointments (no call required!), and receive test results.
We paid our megacorp EHR for this feature but it's "down" most of the time. Other companies may have better uptime. AT least for test results and records. YMMV also for scheduling...calls are still pretty much required in that we have to screen insurances for patients who are actually allowed to see us, rather than have them show up at appointment time and get denied by their HMO.
Fair warning: put on your flame-retardant suits; the following rant contains adult content and is for mature audiences only.
I started my career as a professional software engineer, where I worked on porting the UNIX kernel to an SMP machine, then I went to grad school in CS, with a focus on database theory and AI, back when AI was unfashionable and NOT funded by the NSF or DARPA, and then I went on to medical school.
I'm also an industry insider, having worked for one of the larger (marketshare-wise) EMR companies.
So I'm a doc. A real one. As in I spent 4 years busting my ass at a top college, where I graduated phi beta kappa, while all the privileged white frat boys spent most of their college time getting drunk and fucking all the cheerleaders. Then I spent another five sleep-deprived years as an underpaid resident.
Now that you know my credentials, let's get a few things straight:
1. Physicians HATE Epic. It's responsible for an epidemic of physician burnout. I personally know physicians who have refused to work for the companies that acquired their practices because they were told they'd have to use Epic after the acquisition was finalized. And I personally know physicians who have left the practice of medicine AFTER being forced to use Epic. It just wasn't worth it to them.
2. The non-disparagement clauses are true. I've seen then. You are not allowed to talk about the very real patient harm and threats to patient safety that Epic has caused. Epic's CareEverywhere merges the wrong patient data across health systems way too often. These are HIPAA violations. But the health systems refuse to report the HIPAA violations because of the civil and criminal fines they would face, not to mention the violation of the non-disparagement clauses. I've seen it first-hand. It's all too real. And patients have been harmed. But no one has ever outed this problem.
3. You remember the old trope: "No one gets fired for buying IBM"? The modern equivalent is "No one gets fired for buying Epic." In reality, the health systems' C-Suites are clueless; they don't listen to their physicians. Their only concern is their job security; not their patient's welfare, and not the mental health of the physicians that they employ. To them, physicians are replaceable commodities.
4. Health systems purchased Epic in the FALSE belief that it would pay for itself via Medicare bonus payments. Bullshit. It never did, and it never will. Magical thinking is too fucking prevalent in the health systems' C-Suites. And there's no sign of it ending.
5. It's an EMR stupid. Not an EHR. When we stared automating the paper chart (medical record), we called it an ELECTRONIC MEDICAL RECORD. Then some suit -- a non-clinically-trained MBA/marketing bozo -- came along and decided to rebrand it as an EHR (Electronic Health Record). Guess what. It's an ELECTRONIC version of the MEDICAL RECORD. I really don't give a shit if some magical-thinking homeopathy-prescribing naturopath, or some nurse practitioner who wants to play doctor, is offended because we are using the term MEDICAL rather than HEALTH. It's an electronic version of the fucking medical record. Get fucking used to it.
6. You ever heard of MUMPS, as in the Multi-User Medical Information System? It's what Epic is based on.
MUMPS was designed in 1966 at the MGH Laboratory for Computer Science by a team working under the direction of Dr. Octo Barnett, a physician/computer scientist. Dr. Barnett is (was) a fucking genius. He figured out how to build a database-enabled multi-user OS and run it on a DEC PDP 6 with only 128K of RAM and 5MB of disk storage. Think about it. That's K, as 128x1024 bytes of RAM. And 5MB, not 5GB, not 5TB, but 5MB of disk storage. Way the hell less RAM and non-volatile storage than in that computer in your pocket that wants you to think it's nothing more than a phone on steroids. Could you write a multi-user OS with persistent storage that run in 128K RAM? I didn't think so.
A few fun facts about MUMPS, the engine that powers Epic:
a) MUMPS uses a hierarchical database design, and was developed four years before Dr. Edgar Codd published his seminal paper on the set-theoretical relational model for data. No self-respecting professional software engineer uses a hierarchical database.
b) MUMPS variable names are/were limited to three characters in length. Imagine writing a complex piece of software where your variable naming scheme is limited to three-character variable names. But remember, this had to run within 128K of RAM.
c) All variables are global in scope. There is no encapsulation. Um. Ever tried to implement recursion where all variable are global in scope? Or how about RPN notation, as in the shunting yard algorithm? Not happening.
d) All character data values are limited in length to 1024 Bytes. So a text note that is 1500 characters has to be stored in a linked list. Or not. Epic just truncates long notes to 1K.
e) All variable are persistent. As in -- they represent data storage elements on the disk. Huh? Again, this was a great design for a PDP 6 with 128K RAM and 5MB disk. But not for today's architectures.
f) No such thing as a transaction. So no commit or rollback or ACID compliance.
g) None of the first tier computer science programs teach their students how to program in MUMPs. Why would they? But I'll bet you can find an open-enrollment "university" -- a diploma mill where your ability to pay tuition is the only requirement for matriculation -- that is willing to teach MUMPS programming skills.
Bottom line: Epic is the technological equivalent of the 1966 Corvair. Advanced for its time. But that was more than 50 years ago. Wouldn’t you rather own and drive a 2019 Tesla Model S?
The only way this may change is if we change how we pay for our care. This is not on the horizon at this time. Even the value based programs that CMS has implemented are really fee for service with a look back and adjustments to payments made based on savings that were encountered.