Electronic medical records don’t save money

The New York Times reports today on a RAND study that found that government-mandated (and taxpayer-subsidized) electronic medical records don’t save money. I predicted this in my health care reform proposal from 2009, under “Scrap Government-mandated Health Care IT”. Because we as a society were going to invest in IT in the dumbest possible way (hundreds of incompatible systems installed nationwide, each one chosen by busy physicians), there was no way that it could be anything other than a huge waste of money.

The report, the industry, and the country refuse to deal with the deeper issue: why did we ever think that this kind of IT would save money and improve care?

There are two ways to do an electronic health record: structured and unstructured.

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. To figure out what a cholesterol test result was from 2003, for example, one might have to read through a document in English rather than executing an SQL query. What would this cost? Nothing, basically (which is why Google and Dropbox can give you these services more or less for free). How about doctors and nurses? If they can create a Google Doc and type at a computer then they can easily add notes, test results, instructions, etc.

How about structure? A system for a hospital typically costs about $100 million to install. If it serves a community of 100,000 patients that’s $1000 per patient. Physicians and other providers spend a lot of time in training and a lot of time navigating through screens to get the information into exactly the right table, so the ongoing cost in time and effort is enormous. What value in terms of health benefits is delivered for all of that effort? Epidemiologists might be happy because they can now get an answer to the question “What’s the average cholesterol level for patients in this hospital?” by typing a three-line SQL query (assuming they have first hired 10 lawyers to address patient-privacy issues!). 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. That is fundamentally why little value is delivered in exchange for the $100-200 million cost of using a structured system.

Why is IT so popular in other industries? Consider a bank. There is a lot of value to having all of the accounts in one big structured database. It then becomes easy to ask “How much money is in the bank right now?” An airline can similarly benefit from a centralized and comprehensive reservations database in order to handle queries such as “How many empty seats are there on all of our flights and can we sell an unusual two-stop route from NY to San Francisco at a low price to fill up some of those empty seats?”

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. Unless you go to the doctor every day of your life, it really isn’t that hard to scan up and down in a big word processing document or spreadsheet and/or use conventional word processing “find in document” tools and/or use Google Drive’s “search a bunch of documents” tool. I did a quick Google search and learned that the average size of an electronic health record, not counting images (which can easily be stored by date as regular files), is between 1 and 40 MB. The absolute top end of the range is 3-5 GB “for a person with several health issues including images”. In other words, your health record could fit into the memory of a modern toaster oven.  [A doctor could keep a complete synced copy of all of his or her patients' records on a mobile telephone.] Anything this small can be searched pretty easily, either by computer programs or humans.

[I actually asked a cardiologist who had recently invested in a $30,000+ system for her solo practice and who liked the system "How would this compare to using Google Docs to share referral letters and reports with the other doctors and hospitals that you work with? And then simply scanning test results and storing those as PDFs?" She replied that the Google Docs/scanned document system would be better for her and the other doctors, both in terms of ease of access and saving time for data entry but that it wouldn't comply with all of the government regulations for both privacy and "meaningful use".

Separately, lest anyone think that I am simply hostile to the brave new world of computers on doctors' desks, note that I built what is believed to be the first Web-based electronic medical record system. This 1994 effort is described in a 1996 journal paper: http://www.ncbi.nlm.nih.gov/pmc/articles....]


  1. ScottE

    January 11, 2013 @ 1:22 pm


    I think people are suckered into thinking that healthcare IT would be a boon by their experience of answering the same questions multiple times as you move through the system: “Well if it were all computerized they’d already know this stuff about me”, without realizing there’s good reasons those questions get asked a few times.

  2. billg

    January 11, 2013 @ 1:25 pm


    The medical industry wails its siren song and sucks tech startups to their gruesome deaths. Andy Grove, founder of Intel, tried to implement electronic medical records and said the industry needs to start with “embarrassingly simple” products, but there was too little profit and too many cultural barriers for that to succeed.

    There was a great article in the Economist about doctors being reluctant to adopt new technology. The story went on to quote doctors saying they were too busy to learn new tech and generally doubted the possible benefits of using it. The author then revealed the article was originally written in the late 1800′s and the new technology was the stethoscope.

  3. Dan Weber

    January 11, 2013 @ 1:36 pm


    Fundamentally there is not as much value to be obtained by having the ability to do a structured query into multiple patients’ data.

    In terms of figuring out what works, what you call structured data could be very very useful. Mayo Clinic and Cleveland Clinic do a lot of this to figure out which procedures are worth the dollars and which aren’t.

    That (figuring out if this procedure is worth doing) is something largely anathema to the American public at the moment, though.

  4. billg

    January 11, 2013 @ 1:38 pm


    BTW, your idea about scanning records into PDFs then storing them on Google Docs would require a $1000/employee(!) license from a patent troll who is currently terrorizing dozens of small-medium businesses who have networked photocopiers capable of scanning to PDFs.

  5. TK

    January 11, 2013 @ 1:52 pm


    I can’t resist. Sounds like a good use case for a flexible schema data store like some noSQL solutions.

  6. CHenry

    January 11, 2013 @ 3:21 pm


    EHRs cost money, not just for hardware/software/training, but in physician productivity. And now, the government is realizing that EHRs cost them more money now that more consistently and completely documented services are finding themselves being coded to the appropriate level of service. In the past, with paper charting, doctors fearful of penalties for overcoding for their services defaulted many times to undercoding. With charting analysis by the EHR, compliance with government coding requirements is more precise and the government is seeing the charges for what in many cases they should have been paying all along.

    The real purpose is to provide a centralized repository of mine-able records that can be used to look at longitudinal progress of many thousands of patients (enhanced by new ICD-10 diagnostic coding soon to be implemented) and to look for cohorts of patients that consume the least amount of services without developing any significant complications of their disease. Those cohorts will likely become benchmarks for the limits in medicare payments for given diagnoses over specific intervals. So the long-term purpose will be to save money. Whether EHR ever saves doctors money is not the point.

  7. philg

    January 11, 2013 @ 3:27 pm


    TK: Your idea of using a noSQL database behind the thousands of forms that a doctor or nurse might fill in on a patient is a great one… adopted in the 1960s by Massachusetts General Hospital. They even wrote their own noSQL database: http://en.wikipedia.org/wiki/MUMPS. The real beauty of this design is that you can buy it for $100+ million today, e.g., as part of an Epic Systems electronic health record.

    Whether structured data are kept in an RDBMS or MUMPS, though, does not affect the analysis in the original posting.

    Dan: Your idea of making doctors and nurses spend time that they could be spending caring for patients instead using various forms in an information system so that one day a statistical study can be done sounds like a good one. 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.

  8. Dennis

    January 11, 2013 @ 4:26 pm


    They might not save money all by themselves. But…in Germany, there’s a national EMR system. Patient shows up, doc swipes his card, records come up on screen. Doctor does whatever’s needed, puts it in the computer and he’s guaranteed to get paid in a week, because the law says so. In this case the EMR system *does* save money, because a lot of German doctors don’t bother hiring office staff. (Source: The Healing of America, by T.R. Reid.)

    There’s an opensource EMR system that does the job pretty well, because it was built by doctors and interns for their own needs. It just doesn’t do claims and billing, because they were at the V.A. But for everything else, it’s great. No waiting for test results to come back, as soon as they’re done the doctors sees them on screen. That cuts down on retests. Few medication errors, because the nurse scans the patient’s barcode, her own, and the medication’s, and it alerts if there’s a mistake. It also tells the head nurse if she’s late. And when Katrina hit, a lot of people lost their medical records. Not the V.A., it was all backed up and available at other hospitals. Several countries have adopted the software for their national healthcare systems. (Source: Best Care Anywhere, by Phillip Longman)

    I’ll mention too that there’s opportunity for IT to help with the administrative side. I used to work for a startup that saved medical office staff a lot of time doing referrals, eligibility checks, and claim status checks.

  9. philg

    January 11, 2013 @ 4:32 pm


    Dennis: Thanks for the reference on Germany. That’s the kind of system that I wrote about in my health care reform article back in 2009. I think Israel has a similar system that combines health record and payment for doctors. But we don’t have it and we’re not going to have it so we might as well go back either to paper or to a “paperless” system such as Google Drive or Dropbox. To say that “All of a sudden the U.S. is going to become smart/rational like Germany” is unreasonable. (I wrote about this in regards to gun regulations. It doesn’t work to say that we’ll pull in a law that they have in, say, Japan, and then some aspect of our lives will be just as good as what they have in Japan. We’re not Japanese, German, or Israeli and therefore we will be able to do some things better and some things worse than those other countries/cultures.)

  10. Dennis

    January 11, 2013 @ 4:37 pm


    Just read the NYTimes article…sounds like the real problem is that “the available systems seem to be aimed more at increasing billing by providers than at improving care or saving money.”

    The technology seems to be working, it’s just not doing what the rest of us want, because commercial healthcare has a different incentive structure than the V.A.

  11. Senorpablo

    January 11, 2013 @ 4:50 pm


    This seems like an area in which our “free enterprise” system falls down. You’ve got many interests who are free to do whatever they like, and none of them really have any incentive to do what might be best as a whole. Certainly, there’s no incentive to do what’s best for the ultimate customer–the patient–when all these parties are doing quite well with the status quo.

    Just think of all the interested parties: Software companies, software engineers, software sales people, database software companies/suppliers/sales, hospitals, doctors, nurses and their unions, health insurance companies, drug companies, medical supply companies, pharmacies, government regulators, FDA, medicare, medicaid, and on and on… These folks all get paid regardless and pass the burden down the line to the patient. Why should they be motivated to do anything which increases the efficiency beyond one degree of separation which effects their bottom line?

    All those folks are certain they know what’s best for them, and they want choice. When in reality, what’s best for everyone might be one common, good system; perhaps not the absolute best for every case. That’s not how we do it in the US however. Everything here has to be about the best and being able to chose the best and being agile and free market and states knowing what’s best for their residents.

    This is similar to how we do nuclear power, versus France. In France, they saw value in picking one system which was good enough in the long term. They may not have the most up to date, most efficient reactors by today’s standards, but they saw value in having one well known system over the long term. The nuclear power industry and the health care industry, at least here in the US, are both terribly bureaucratic.

    There must be socialized countries which use common IT systems for reference?

  12. philg

    January 11, 2013 @ 4:53 pm


    Dennis: To your second comment… the U.S. isn’t going to change dramatically. We’re not going to make some kind of grand correction in incentives or in how things work. So the way that EHRs have worked in the U.S. for the past 40 years is very likely how they are going to work for the next 40 years. It isn’t reasonable to say “But it could have been done better by Americans”. If it could have been done better then it would have been done better in the 1980s or 1990s. So now that we know that we can’t do EHRs in private hospitals in a way that saves money we should at least stop putting tax dollars into them.

  13. philg

    January 11, 2013 @ 4:59 pm


    Senorpablo: I don’t think that you can use anything about health care in the U.S. as a guide to free enterprise. The government pays roughly half the bills from this system. Insurance companies pay most of the rest. There is no consumer. It is illegal in most cases for an insurance company in one state to offer a policy to a person in another state.

    Same deal with power generation. In the U.S., especially during the heyday of nuclear power plant construction, we had a combination of regulated monopoly and crony capitalism in the electricity market. We don’t have any idea what would have happened with a free market approach to nuclear power (well, maybe we do! Probably the state-of-the-art plants would have blown up, just like Deepwater Horizon! Perhaps the answer is that Americans shouldn’t be allowed near anything dangerous/high-tech).

  14. anon

    January 11, 2013 @ 8:22 pm


    “This seems like an area in which our ‘free enterprise’ system falls down.”

    Hahahaha. Healthcare IT is crony capitalism at its finest. Crony capitalism is not “free enterprise” and most certainly not free markets.

  15. Steve

    January 11, 2013 @ 8:25 pm


    Good comment re: the VA system. It is free, and open-source so providers could cost effectively customize it, and those efforts would benefit the greater good. Evidently (based on recollection) the VA is quite state-of-the-art, and is very efficient and effective in its user. However (again based on recollection) the language of the new health care act does not allow hospitals to consider the VA system. EMR has to be sourced from a commercial venture. If true, then this type of perversity tells you everything you need to know re: whether or not our society is going to benefit from EMR.

  16. philg

    January 11, 2013 @ 8:32 pm


    Steve: I am not aware that there is any legal or regulatory obstacle to a hospital adopting the VA software or a derivative. It doesn’t do billing, though, which is a deal-killer for hospitals, and it is based on a non-relational database (customers would prefer to have a standard RDBMS such as Oracle). There is another free and open-source system from the Department of Defense’s hospital system. That one does use conventional RDBMS as a back-end but, again, does not handle billing. No matter which of these a hospital adopts, of course, they won’t be able to communicate with any hospital that has adopted something else.

    But the larger point is that even the VA may not have benefited from using their fancy EMR instead of something like Google Drive. Why does every item of medical data about a person need to be at the intersection of a table row and column?

  17. @milr0c

    January 11, 2013 @ 8:41 pm



    I understand where you’re coming from, but data adds a huge value to both health professionals and patients. Information availability is great. I think perhaps you should reframe your argument to emphasize that storage solutions that are not in a centralized (and anonymized) manner do not add value to the industry or its patients.

    By decentralizing the costs of storage (ignoring the 100MM price tag), the amount of information is greatly diminished. While I’m not a fan of the term “big data” a centralized solution for health care benefits the industry (less so if insurance would have access to these records to compare to their customers), that professionals can:

    1. Run something like KNN, given a set of criteria, finding anonymized patients that share similarities (albeit a hard problem to work with) and what techniques
    2. Create new standards that will progress in a much faster (which I suppose could be potentially negative) manner.
    3. Learn at a much faster pace or at least with more depth (having every case study would likely add a lot of value to students in a health profession).
    4. Apply research level analysis on (e.g clustering like patients, large scale symptom to disease correlation, realtime breakout analytics).
    and more.
    5. Not require extensive support for IT, considering the system is centralized.

    There are problems (glaring ones) with this method of working with the information:

    1. Lots of regulation (the query may require approval processes, since it impacts the treatment of the patient).
    2. Data cleanliness will tighten (even simple reports (someone with a cough) can help add value to research, thus likely increased ‘paper’ work for doctors.
    3. Even more regulation, security implications aside, the privacy need is one of the strongest needs for this system, ensuring anonymity while offering non-anonymous access a case by case basis, is imperative.
    4. Improving data access for all hospitals will likely increase the opportunity for other hospitals to learn from the value added from top tier hospitals, thus this system could lower competition (or say between two extreme specialists if they both have access to the same information the primary delimiting factor is personal success rates and I suppose intuition, which arguably is the case in extreme niches of any area of medicine).
    (Likely more that I cannot think of)

    This could be handled by government or by a private companies, which would diminish the set of data but a winner will likely prevail.

    Assuming you’re not against it simply from the 1984 Big Brother aspect, this will help push the industry into a direction where the service as a whole they offer the patients increase, *ideally* allowing for more innovation opportunities.

  18. philg

    January 11, 2013 @ 8:49 pm


    @milr0c: Agreed that the system you propose would be a nice thing for a country to have (and in fact Germany and Israel seem already to have it). But there is no legislation or regulation that would move the U.S. in this direction and having each hospital spend $100 million on a proprietary incompatible system does not do anything to push us towards the glorious future that you sketch. It just leaves us $100 million poorer times however many hospitals there are in the U.S.

    The fact that something unrelated (the big central system) involving health care and IT is useful should not prompt us as a society to spend tens or hundreds of billions of dollars on a very different form of health care IT (i.e., one that is not useful).

  19. Giacomo Balli

    January 11, 2013 @ 9:07 pm


    1) $100-200 million cost for a system is widely debatable.
    2) lower errors, higher accuracy and up-to-date records are benefits
    3) once data is digital there are infinite ways it can be (anonymously) used to hopefully find something good.

  20. J. Peterson

    January 11, 2013 @ 10:01 pm


    Random data point: The large clinic our family uses has a custom data system that looks like it works quite well. Doctors/nurses etc. quickly call up records, easily plot charts, bring up data (X-rays, videos of ultrasound scans etc.). Prescriptions are sent off to your local pharmacy with just a few clicks. I’ve never seen a provider struggle with it, lose track of information, or need to get help. Billing and insurance processing is generally accurate and reliable.

    I have no idea how much it cost to install/maintain, but seeing it in action leads me to believe that medical IT does not have to be a disaster.

  21. philg

    January 11, 2013 @ 10:46 pm


    Giacomo: (1) I have been personally involved with at least two hospitals that spent over $100 million; (2) the U.S. is a leader in using health care IT and we don’t have especially good health care results, (3) there are already collections of health records for millions of people yet I’m not aware of anyone “finding something good” from these data. If all of Germany is in one big system, for example, (as noted above in a comment) and there is gold to be mined from those data, then why isn’t Germany the world’s leading source of new medical knowledge?

    J. Peterson: You note that your clinic is successfully using IT instead of a filing cabinet, but you don’t know what it cost and you don’t know if they would have been just as successful or more so with Google Drive (free). The RAND study conclusions cited in the original posting do not depend on medical IT being a “disaster” as you put it. For medical IT as the U.S. currently practices it to fail to save money, all that needs to happen is that the price be high.

  22. CurtisP

    January 11, 2013 @ 11:19 pm


    Having data in structured format eg in an RDBMS, is beneficial even if only for the purpose of printing forms, prescriptions, scheduling appointments etc.

  23. Colin

    January 12, 2013 @ 9:21 am


    I’d love to see an analysis of “real” error rates in EMR. I rarely write “X” when I mean “ABC”, but I do click the wrong item in electronic forms, and don’t notice it.

    Our pediatrician office has EMR, and I’ve seen how easy it is for something to be coded incorrectly when clicking madly on a laptop. And if it weren’t for schools requiring certificates of health, we’d never know it…..

    I’ve been saying for years: just because you CAN automate a process does not mean one SHOULD.

  24. Wayne

    January 12, 2013 @ 7:46 pm


    Hi All,

    Here is an opinion of a different type. I come from a system in which there is a government sponsered EMR medical environment. I have used EMR’s as a family doctor since 2002 and would never go paperless. We have transitioned to a cloud based EMR system and have an extremely secure system that is also efficient. It has definitely improved my care of patients.

    There is also a provincial wide database repository (yes, you guessed it I am from Canada more specifically Alberta) that docs can access lab work, specialist letters etc. It actually works.

    Additionally, the provicince is working on a province wide patient portal that some day patients will be able to access their own results on line and securely message with their health provider.

    But all this actually is not a technical problem but really an issue for government to sit down with health providers and work out an agreement first and foremost. I think that the governance infrastructure needs to be in place first before implementing the technology. the techie stuff is the easy part actually.

    We are also on the cusp of trying to understand primary care activity on a much wider basis. It has larger implications on a public health scale.

    So concentrate less on choosing the EMR (yes, choose only a few and not hundreds) and technical considerations and more on what the system can do for patient care and health.

    Just my two cents worth.

  25. brian gulino

    January 13, 2013 @ 8:41 pm


    As a consumer of far too much medical care over the past 20 years, I have found the records used by Doctors to be strikingly inefficient. The records appear to be designed to insure that no procedure goes unbilled, i.e. 3 sheets of paper with 360 check boxes on which 3 boxes are checked. A traditional justification for a DBMS is that you store the data centrally, and display it in a different format depending on who needs its. The doctor gets the patient record in one form, and the insurance company in another form.

    I am also struck by the amount of time medical staff spends faxing records hither and yon. The doctor orders x-rays, blood tests, and refers you to a specialist. All generate their own paper trail. All of them need to be coordinated. I doubt that shared docs have enough structure to make this work without errors and ambiguous results. Systems such as the VA’s or Kaiser’s where everyone is looking at the same medical records generate less errors and save time coordinating.

    Finally, public health. I live in California, where my children participate in the vaccination monitoring program. We are issued a yellow postcard on which our pediatricians record our children’s vaccination history. Lose the card, you child can’t register for school until you get a new one. Don’t update the card with the latest vaccination, your child can’t register for school. Now, my child’s school has a high speed internet connection, the pediatrician’s office has a high speed internet connection, and I have a high speed internet connection. It seems almost trivial to keep a statewide vaccination compliance database, but the task eludes us. This task, low cost, high numbers of people, is typical of public health initiatives, many of which would benefit greatly by being in some sort of computer database.

    Oh yeah, what about off label drug use? Currently not monitored until somebody notices a problem.

Log in