Let's not conflate the theoretical (and very achievable) advantages, with the shit being peddled to date. The current state of available systems is really abysmal, and I'm sure it's true that using any of most of the systems on the market won't save you money.
"Fundamentally there is not as much value to be obtained by having the ability to do a structured query into multiple patients’ data. Nor is there much value in being able to do a structured query into a single patient’s data."
I supremely disagree with this point, especially the first bit. Specifically, the author is comparing the labor of him looking for information he's thought to look for in one or more patient's records via computer or paper, and finding them similar. That's the sign of a completely failed system. Most EMR systems are built to act like the file cabinets they replace, which is a shame. Meanwhile, my mint.com account sends me emails when I'm charged an ATM fee and facebook knows way too much about my current interest in purchasing a juicer.
I believe the unrealized benefits of automated predictive, proactive, and historical analysis of patient health data is enormous. The author is excused for disagreeing with me though, because presumably billions of dollars at this point have been poured into building software that rivals late nineties Visual Basic POS systems in quality and effectiveness. There's so much untapped potential here it's unreal. It's just such a shame that the Health IT industry has eroded the patience of caregivers over the last decade, especially ones who were eager to explore the possibilities early on.
I have also built EMR systems. They do not achieve much over paper to be honest. Where they do make a difference is at the point of care. Scanning patient, nurse, and medications prior to delivery for example removes errors. But in terms of cost savings it really isn't that significant. Patient satisfaction is not really the issue being addressed here. Cost of delivery and efficient use of limited resources is.
I worked on a browser based full fledged EMR (i.e. Charting, Billing, Scheduling, etc). Besides being an enormously complicated software system in general, there were many factors related to consumers' usability of the product in a real time interaction with patients.
For instance, we built IMHO beautiful note taking app where doctors could free text in notes taken from patient consultation. The beauty of the system is it had a large natural language parsing backend that would help highlight parts of the notes that could be codified in a system such as ICD9 or SNOMED. There was even a rich templatimg system allowing doctors to pull up a note template for something like chest pain and then the system would prompt the doctor to fill in the relevant codifiable sections.
It demoed great to physicians interested in an EMR, but in practice, doctors did not like the extra effort required to record a real time interaction in a codifiable way. I'm not sure what's to blame: EMR usability or hospital staff's lack of time and training to document patient care in a standardized format.
I know a few doctors in the UK and they don't understand two things:
- they are a reasonably efficient expert system
- large scale data mining can find patterns of behaviour and disease that are impossible to find any other way.
Point one can partly be explained by the fact that as a society we rightly respect doctors and as humans over the years they take it to heart that they really are quite special. They also rightly diagnose hundreds of common illnesses a week which boosts their ego.
Point two is a common problem for people outside of the tech world. Doctors are used to small studies growing to large studies about a specific illness. They do use meta studies and use large scale statistics to medicate (everyone over 60 gets cholesterol medication etc) but that is the exception. They don't understand that an abstract statistical analysis of patient records could deliver gold.
So when asked to codify patients notes (something they have to do in the NHS using a ridiculous DOS based app, with over the top word templates for paper forms) they resent it. They don't feel the advantage of this. When the system controls their flow they feel like they are robots and that this gets in the way of their treatment of the patient.
It pains me to have to explain to them that they are, for the vast majority of patient interactions, meat robots. For 99% of their interactions they are slower and less accurate than an expert system (although of course the doctors I know have a more pleasing interface than a robot).
They can't see that if the burden of the 99% was lifted they could genuinely help their patients by having more time. Instead they insist awareness campaigns are the way to go: "If less people came in to the surgery with a cold we could DOUBLE the time they spent with patients"
This person has clearly never worked in a stressful trauma unit. I have actually tried the google docs and drop box think in an outpatient clinic, and it is terrible. People accidentally delete things (sure you can get them back but the non-it person at the coal face wasn't able to). Records get written over and multiple people try to edit the document at the same time and conflict with each other. People forget their logins and then ask you for help working out their password (quote: shall we just ring them?). People accident sign the wrong email address up and distribute access to the wrong person. Privacy? Forget it. Data is stored offsite. Dunno where. Who is accessing the data. Unsure. Does it auto logout? Depends.
Piecemeal electronic record systems are bad, but consumer focused solutions are far worse.
The trauma unit isn't interested in population cholesterol or stuff like that. They want platelet levels now. X-rays now, CT Report, before now. Google and dropbox weren't made for this, and don't work on a day to day basis. I've tried.
Edit: typos. This subject makes me frothy mouthed as the commentator has valid concerns, terrible suggestions. Try it and see.
I appreciate your experience (and it's good to learn that using these tools in the outpatient clinic didn't violate any regulations) but I don't think Phil is really suggesting Google Drive specifically here. He's suggesting instead that medical records should be more like a blank page (document) that physicians can fill in than e.g. a FoxPro app (dbms). Personally I think the "ideal" would fall somewhere in between. Physicians should probably be allowed to keep records however they like, but lives would be saved by the structure of a well-implemented [what a caveat!] digital checklist system.
One of the killer problems we have is that a huge local hospital (servicing probably several hundred thousand in population) won't give us any real access to their RIS and PACS, despite us doing lots of scanning for them. We have to get previous imaging and notes delivered on CD and DVD which is probably the worst type of digital info transfer possible. Even if we did have access it would likely take the IT team a few weeks to cobble together something that worked for us end users, but that would be an amazing improvement.
We just finished a spell using google drive - oh wow it was terrible.
I live in New Zealand btw.
"Some exceptions prove this rule. One of most successful health IT systems in use today is the Department of Veterans Affairs’ Veterans Health Information Systems and Technology Architecture, which actively engaged clinicians in its development process.25 Health IT is widely credited with helping transform the Department of Veterans Affairs into one of our nation’s highest-performing health care systems.26–28"
Something like 8 million people are managed by the VA records system [2]. Perhaps a solution would be to expand this existing system, with a demonstrated track-record, to the broader population.
> But where is the value to the patient? Does it help to have Patient A’s cholesterol information in the same database table when treating Patient B? No.
This is incorrect. Phil might have experience from the DB side, but apparently not sufficent experience from the patient side.
One example where it has major value for the patient: While contemplating a heart valve replacement at age 71, a recent patient had many conflating factors (diabetes etc). What is the previous history of people with the same conflating factors surviving? How about surviving but only with a ventilator?
Only after considering the DB of about 100K entries did the patient decide on the valve replacement. (I'm happy to report she came through with flying colors.)
If you care to disclose, are you at Vanderbilt? I ask because this is one of the few places (known to me) to have such a large, query-able database and I am curious to learn about others.
Thanks! What's hilarious is that I was literally there today and they didn't mention this extraordinary capability during the entire 8 hours I was there. Talk about missed opportunities!
The tl;dr = Society of Thoracic Surgeons reduced a dataset and built a website. For people dealing with some pretty heavy heart issues, it's a damn useful item.
We definitely used that site clinically (STS risk score) when I was rotating through the CVSICU at JHH. I don't usually barf out whatever I'm thinking, but I figured some people on HN would also be interested to know that this is actively used.
Phil has a good point in the comments, though: why waste expensive doctors' and nurses' time reducing the data to structured form when this work could easily be outsourced?
I missed the part where they said how the savings were expected to be realized and weren't. I got that a complex system might cost $1000 per patient, but that isn't even one month's "health care" fee for a family of four, much less the cost of healthcare for the patient over the 5, 10, maybe 30 years they are visiting the doctor. I also missed the point where creating new structured documents out of existing structured documents was harder to do than constructing structured documents out of unstructured documents. And I missed the part where the quality of care in terms of unnecessary tests or procedures was compared at hospitals with such systems and those without. Other than that it seemed pretty reasonable, and that Bill Gates was commenting was interesting.
Most families spend less than $1000/month on medical care, on average. Even for those that spend that much, presumably much of that pays physicians and nurses and buys drugs, rather than paying for IT. It would be difficult to sell e.g. Ford Motor Company on the idea that since people pay $25k for the average car, they should be happy to spend about $25k/customer on IT. The other things you "missed" seem oddly beside the point.
Your right of course, it was more rant than argument.
The paper claims that $1000 per patient is extortionate as compared against the 'free' cost of existing web based tools. However it never supports that claim except through that single comparison. Thus, I am not persuaded that $1000 per patient is extortionate.
Lets take that cost as legitimate. The structured system exists for the life time of the patients who use the hospital, so understanding that cost requires understanding how the cost of that system is recouped over the lifetime of the patient. Picking three cohorts, a 5 year patient, a 10 year patient, and a 25 year patient, they would have lifetime costs of $16/month, $8/month, and $3.33/month to recoup a $1000 cost to both enter them into the system and pay for the system (these costs weren't split out in the article).
One of the arguments for electronic records is a reduction in unneeded tests. So for example when Bob comes in for a physical do you order a cholesterol test or not? If the previous test is still valid, you don't, but if you miss the previous test in the record? The doctor re-orders the test "to be sure." which occurs because insurance companies can find you partially negligent if a test was available and you didn't order it but should have.
In order to quantify the 'value' of electronic records, either structured or un-structured, you really have to compare the patient costs for hospitals with such systems and those without. And any thorough treatment of that question would also cover the ingress process of a new patient coming from a hospital with a function structured electronic record capability and one without.
I can see why its hard to see that this was what I was saying in my original comment.
Thanks for the explanation! I accept that healthcare is different than other industries, but even the lowest IT cost/month you cite seems exorbitant for a customer you're not billing on a regular basis (and many people can go decades without entering hospital). IT is a cost center, and this is just too much cost.
True things are different in different places. But in the US most healthcare is part of an employment package and as an employer I'm paying every month to our healthcare provider whether or not my folks actually get any services.
> Most families spend less than $1000/month on medical care, on average.
Off the top of my head, the healthcare industry is roughly 2.5T, and there are roughly 115M households in the US, which implies health care spending is ~22k per household per year.
I agree that most spend less than $1k/month but the average spend is $1800/month.
"Fundamentally there is not as much value to be obtained by having the ability to do a structured query into multiple patients’ data. Nor is there much value in being able to do a structured query into a single patient’s data."
ie: I can't imagine anyway to do these queries that derives value, so fundamentally there is no value.
I'm far removed from the healthcare industry, but I'm absolutely positive there are correlations hidden in those files. How would you ever spot commonalities between similar patients' histories without structured records?
It's not an argument from ignorance, it's a simple assertion. An argument from ignorance would be saying that because we don't know what value there is to be obtained, there is no value(or the value is limitless.)
Condemning polemic for making assertions is kind of missing the point of polemic. It isn't like he isn't trying multiple arguments to suggest these assertions may be true.
I agree the systems in use seem horrible and expense from a cost, engineering, and IT perspective, but I'm also a big fan of my doc being able to see my records in digestible format, IE graphs showing a series of test results over time and such, rather than just 20 pages of paper to eyeball in a hurry and hopefully spot a trend. Or the old situation where you'd inevitably be asked about a test or other visit and they'd have to get results and such faxed over to look at later basically wasting the whole visit. "Oh, sorry, we haven't received the x-rays/blood work/random other diagnostic, referral, or test yet..."
Now most times they just bring it up on their computer and are like "oh, I saw you had X done 2 weeks ago, no sense in doing it again..."
But that's possible because all my providers are using the same system.
As somebody who's trying to build a startup in this space, I'm not surprised.
The root problem I see is that emphasizing "records" is completely backward. Making health records electronic is about 5% of the real work.
The real value is all in building systems for effective collaboration and operational excellence. Records are almost a byproduct.
My pilot customers have measurable money savings. And my product isn't even a "full-fledged electronic health record", because that's rather beside the point for my customers. Their problem isn't fundamentally about records, it's about coordination between pharmacists, nurses, health aides, and administrators.
You know what would really be great, and would almost certainly save money? If prescriptions could be done electronically. I am amazed by what it is like to get a prescription medicine, especially Schedule II -- the doctor types a lot into a computer, then puts special paper into a printer, prints it, hands it to me (and now we will just assume I cannot figure out how to copy it), then I walk to the pharmacy and what happens? They look at the paper, type more things, and then I get to wait 15 minutes for the medicine to be dispensed.
Now, what if instead, my doctor would type in the information, and then either (a) prints a QR code that I can just bring to the pharmacy or (b) asks if I will be using the same pharmacy as last time, and has the prescription sent their automatically, so it is waiting for me when I arrive? Let each doctor's office create a prescription record in some database, and let each pharmacy update that database when the prescription is filled -- why rely on copy-evident paper, when we can rely on computers instead? We could even connect this to an EMR system, so that doctors not only know what a patient was prescribed, but whether or not the prescription was actually filled (when I was last prescribed an opiate painkiller, I never filled the prescription; yet my followup appointment involved the receptionist printing a sheet that showed the prescription, with no indication one way or the other if I had bothered with it).
I understand that regulations and so forth and probably the doom of these ideas. Still, it would be nice to get something a little better than the ludicrous anachronism we have today.
What you describe here has been happening in Singapore public sector for quite some time. Most patients in a hospital will fill their prescriptions in the hospital pharmacy anyway to enjoy subsidies.
Some patients still prefer to have some kind of paper with them, so the doctor will actually print out the prescription but the pharmacy will still process it from the system anyway. When the pharmacy is not busy, we will also pre-process pending (and short) prescriptions from the system and the patient would not need to wait that long for the medicines. It's also a boon for the pharmacy operations because they can smooth out workload throughout the day by pre-processing prescriptions before the peak hours.
The vast majority of US pharmacies are already able to receive prescriptions electronically from physicians. SureScripts [http://www.surescripts.com/] is the leader in that space.
If you don't already have e-prescribing, it's probably because your physician is behind the times.
The above applies to doctor-to-pharmacy communication. For any other combination, paper and faxes still reign supreme. Examples that are still almost always faxes and paper: pharmacy to visiting nurse, specialist physician to primary care physician, physician to assisted living facility, even pharmacy to original doctor when asking for refill extension.
"You know what would really be great, and would almost certainly save money? If prescriptions could be done electronically."
This can be done today, and frequently is (depending on the organization). Schedule II is a bit nastier since there are regulations about electronic delivery, but the vast majority of medications can be sent electronically in exactly the manner you describe. Just two weeks ago I was in for a follow-up and they ordered some meds that were automatically sent to the pharmacy, where I simply picked them up. Easy.
What health care records really need is a set of RFC-like standards documents. These documents would define the information that a health record must have; or how interrogations of the database should behave. You make sure you include a requirement for the data to be transportable. Once those documents have been done you allow the market to create the software and hospitals (etc) to buy the software.
The UK had a ridiculous mostly failed system that cost £12 billion for a population of about 60 million people. A lot of that is from the UK traditional terrible huge it project implementation, but also from a top-down approach by people who didn't really know what they were doing.
> Let’s consider what an unstructured electronic health record would look like and cost. You’d create a directory in Google Drive or Dropbox and give doctors and hospitals access to this directory. Physicians could type into a shared Google Doc with other physicians or upload scanned output from tests, etc. It would be paperless, organized, and electronic, but not very structured.
With medical records you need to know who has accessed them (when and why); who has changed them (and if they had the permissions level to make changes (eg, receptionist can change appointments, doctors can change prescribed meds, nurse can add test results); who has the notes now (and lock the notes for single person only); all of this needs cryptographic signing. (Because people die when doctors make mistakes, and doctors do make mistakes.)
Much of the EMR system I deal with was built primarily to address billing rather than patient care. The structured data (ICD-9 codes, service dates, etc) exists for that. The crufty part of the system is related to patient test results and other information (pathology reports related to tissue evaluations, etc).
So, the data that is tangential to patient treatment is highly queryable, the data related to patient status is mostly free text. Worse yet, the free text does not follow much of a "standard" entry form - whoever entered the text followed, at best, a departmental standard, at worst, just some quick thoughts randomly sprayed into text.
But even that terribly inefficient and crufty free text is highly useful to staff (MDs, nurses, etc) when reviewing patient treatment during, say, tumor board meetings discussing specific cancer treatments and patient status.
What might be really nice would be to divorce the two functions - separate billing/scheduling from other information. For the patient treatment/status data, there is some fine line in the data model with having some structured data, plus (probably) some flexible key/value annotation system.
One killer issue is that given any software tool, medical staff is smart enough to bend it to do what they want, rather than what it was designed for, if that makes sense.
I fear that as the practice of medicine is increasingly centralized and depersonalized, such quaint questions as "what would be cost-effective?" and "what would be good for the patient?" will be made increasingly irrelevant. The rest of the software industry should carefully consider this case before emulating the "lobbyists as salespeople" model.
The author wrote this in the comments in response to a reader's comment about the value of structured data:
But reflect that a U.S. doctor might earn $600,000 per year, the nurse’s salary might be $120,000 per year, and the same task could be accomplished by a worker in India or Cambodia, starting from the unstructured data, at a cost of $2 per hour.
Let's assume I didn't already think this guy was way off the mark... that comment right there makes me think he's just an arrogant twat. Never mind the huge privacy nightmare of having some "$2/hr worker in Cambodia" sifting through our medical records on Google Drive.
I don't care what it costs or whether it saves money or not. I just want to be able to walk into a medical office and not fill out all the damn forms over and over again.
Sure, but we can fix that in a number of ways. We could, for example, require the use of smartcards to encrypt the records, so that only doctors or the specific patient to who the record pertains can read things. It would be a bit pricey, of course, since each patient would need a smartcard, but it is not as though the problem is insurmountable. It would also be reasonable to just issue smartcards to medical workers, and have patients obtain their records in-person (though I prefer a solution that allows records to be obtained at home).
The problems with security are mainly an artifact of the failure to utilize the available technologies, not that such technologies do not exist. It does not really help that a lot of relevant technologies are horribly patent encumbered (e.g. attritubte-based encryption, which would probably be covered by a number of IBE and ECC patents), but this would not be the first time patents have been the bottleneck in deploying important software.
It only saves trees, time, data loss, and death by decimal. But hey, we haven't found a solution to the problem in 10 years, so we should probably just quit trying.
I think fear of change in dinosaurs is stopping this. Respected doctor = lots of school + lots of practice = old age.
Well, the particular systems that have been proposed might not be that cost effective, but it doesn't mean, that to pursue for a cheaper more effective electronic record is wrong. Putting things in google drive might be quite safe in most ways, but it would still effect patient privacy in long run, mainly because each year millions of accounts are unlawfully accessed due to guessed or phished passwords. Doctors included, as at the end of the day they are still humans.
Just to add to that, I'd estimate that patient records are accessed far less by doctors than by allies health staff, booking clerks, receptionists and such like. Te actual time in front of a doctor is small compared to the amount of time everyone else accesses records. The system for booking in one CT where I work would have probably 5 people see the record, one being a doctor.
indeed, but most of them are from medical field or allied fields, people who have gone through tons of records, and don't really care to snoop anymore even if they unprofessionally did in the past. But putting it on the internet, even if its in the corner, strict no-no. But I wouldn't mind to say this that the person who wrote that article didn't actually do that bad, because its of course natural to perfectly identify a mistake yet not being able to perfectly suggest a alternative. So yeah, he got a point somewhere.
Sorry, rereading what I wrote (argh, typos), perhaps I've been a bit unclear. I don't mean to imply that any of the record viewing is unprofessional (although I know this does occur). The 5 people viewing with only one being the doctor is professional and appropriate. One person books the scan, another greets them and enters their details into the RIS, another scans them, another sees them off, takes payment (or confirms who is paying) another reports the scan, another types the findings, and then someone distributes the findings once authorized by the reporting doctor. Sometimes these roles are filled by the same person (eg typist greets patient, takes payment and types the report), but the point is, there are many steps to the service and appropriate use of the patient record requires many people to have access to it.
If you implement it retardedly, of course they won't. It's just what they were doing before (different formats and methods making it impossible for the hand to know the leg is doing, in terms of communication between multiple or even within a single organization) except with the added layer of electronic technology that means even the people within a single institution won't know how the hell to use it.
10000x this for a profession centering around doctors, since so many of them believe the world of knowledge hasn't changed since when they graduated medical school back in 19xx. :/ They are unaccustomed to accumulating new methodologies as a result of recent advancements.
"Fundamentally there is not as much value to be obtained by having the ability to do a structured query into multiple patients’ data. Nor is there much value in being able to do a structured query into a single patient’s data."
I supremely disagree with this point, especially the first bit. Specifically, the author is comparing the labor of him looking for information he's thought to look for in one or more patient's records via computer or paper, and finding them similar. That's the sign of a completely failed system. Most EMR systems are built to act like the file cabinets they replace, which is a shame. Meanwhile, my mint.com account sends me emails when I'm charged an ATM fee and facebook knows way too much about my current interest in purchasing a juicer.
I believe the unrealized benefits of automated predictive, proactive, and historical analysis of patient health data is enormous. The author is excused for disagreeing with me though, because presumably billions of dollars at this point have been poured into building software that rivals late nineties Visual Basic POS systems in quality and effectiveness. There's so much untapped potential here it's unreal. It's just such a shame that the Health IT industry has eroded the patience of caregivers over the last decade, especially ones who were eager to explore the possibilities early on.
[Full disclosure: I build electronic medical record systems for a living - http://www.quora.com/What-product-is-ElationEMR-building]