Preface by e-Patient Dave:
This is a story of bad data gone wild, wrong info that spreads. It starts with a story from the 1600s, which applies all too aptly to our EMR situation today, in which there are inadequate controls on data quality, and errors that leak can be impossible to contain.
As readers of many healthcare blogs know, David Kibbe M.D. is Senior Advisor for technology to the American Academy of Family Physicians, which is one of the main reasons family doctors have much higher adoption of electronic medical record systems (EMRs) than most physician groups. He’s also chair of ASTM’s Technical Committee on Healthcare Informatics. He has street smarts about the realities of adopting EMRs.
David has a certain joie de vivre that I enjoy, as evidenced by his approach to conveying what “health 2.0” is all about: he started riding his motorcycle around the country, talking to people who are doing it, and made a little movie about it. It’s a pilot for a potential bigger project. And, he’s one of the doctors quoted in the Boston Globe’s earth-shaking story a week ago about the errors we found in my medical records. (Well, it had an earthquake’s effect on my life, anyway.)
The story of how an error in billing records was propagated to another system reminded him of the true story of “the Wicked Bible,” in which an error was propagated with effects that some will find amusing and others will find scandalous. He wrote the following as a guest post.
David’s opinions and proscriptions are his own.
* * * * *
It was a scandal. In 1631 two London printers published an edition of the bible that omitted “not” from the seventh commandment. [It should have said “Thou shalt not commit adultery,” but it didn’t.] The public outrage over what was dubbed the “Wicked Bible” was loud and immediate. King Charles I heard about it, and was incensed. This simple mistake by print compositors landed their employers in the Star Chamber before the infamous Bishop Laud, where they were tried, found guilty, and fined 300 pounds. They also had their print licenses withdrawn; the fine was directed to be used to for a new set of print typefonts and to oversee new quality control practices to prevent such a mistake from ever again occurring in the future.
The episode of the Wicked Bible has historical importance because it demonstrated how the new print technology allowed printers to create “standardized” errors, something impossible in the scribal era when all books were the product of hand copyists. Textual drift – the result of small copyist’s errors in single books, which were then repeated in the next copy, and so on – was no longer possible, replaced by the textual fixity of print type. If printing presses could greatly lower the costs of producing books, and make them available to whole new classes of people to read, they were also capable of mass producing errors!
Enter e-Patient Dave. As we all know by now, Dave asked to have his hospital’s electronic medical record system upload his health data to his Google Health account, only to find that the diagnoses transferred were claims data that were largely unintelligible and meaningless to Dave, and some of the problems listed were downright inaccurate or false.
Wicked EMR! How is it possible that that such mistakes could be made? Not exactly the Word of God, but most people trust that their health information is accurately recorded inside the EHR technology of the hospitals where they are cared for and treated.
Plus, since insurance billing records are transferred to the MIB, an insurance industry database that insurers use to check patients for pre-existing conditions, errors in billing records can have serious effects, as the Consumer Reports blog reported last August. A truly wicked consequence of a propagated error.
Hundreds of blog posts later and two articles in the Boston Globe, here are my takeaways from the Parable of the Wicked EMR:
- Hospitals must recognize that more and more of its customers will want their medical records in electronic format, and help filter and organize these data, rather than just “dump” them to the patient’s chosen PHR, in this case Google Health.
- Dave’s healthcare providers need to help keep the data and information available in terms that patients can understand, along with coded data, and be aware that reconciliation at discharge in CCR or CCD format will be valuable to them. This will help them check for errors (free quality control!) and empower them to be increasingly responsible for their medical information.
- And the PHR companies need to continue to help bridge the gaps that exist between health data in EHRs and IT systems, some of which is largely incomprehensible, and organized sets of information available in patient-understood terminology on the Web.
- Finally, as Dave is proving every day, the patients/consumers have to take some responsibility for feedback and additional commentary until we all get this right.
The good news in all of this is that so many people actually care about e-Patient Dave’s experience getting better. It’s lit up the blogosphere because it’s important. This isn’t about blame – it’s about improvement to the point that patients get accurate and up-to-date summary health information about themselves at every point in the health care system.
A few questions that we might want to answer before this is all over:
- How can it be that a doctor’s list of problems/diagnoses and those that the hospital uses are not the same? Is this an error, or is there upcoding and possibly abuse of the system going on?
- If Dave’s doctors had acted on the data sent from the hospital to Google that was incorrect, and Dave was harmed in some way, would he have a legal cause for action against the hospital? Against Google?
- If these billing data are inaccurate, wildly so in some cases, then why are we using them for analytics and quality research? For disease management?
- If Dave’s billing data in the hospital EHR/EMR system is actually data from someone else, ie. another patient, then is Dave prohibited from seeing his own chart due to HIPAA privacy rules?
- Isn’t it time for there to be a patient right to summary health data that is digital, up-to-date, and accurate?
We don’t have access to the same recourse King Charles had; we’re not likely to arrest and fine those who mismanaged the “sacred” data. But if you ask me, we ought to have the same sense of indignation, and the same commitment to hunt down and eradicate the Wicked EMR.
A couple of things to note here – first of all there are, in general, two types of EHR data:
– Discrete: the type of data that is easily queried, has understood distinct and discrete values, is coded, etc
– Free-text: Generally what doctors use, that is – notes, pathology reports, radiology reports, etc
Being a computer scientist (and budding biomedical informaticist) my feeling is “well just make the EHR collect all information discretely” but there are many problems with this approach. It is insufficiently holistic with regard to actually speaking about a patient’s health, it is not how doctors usually “think” about patient information, it is MUCH harder (prohibitively) to collect, and the field of medicine changes so quickly that it is very difficult to keep your “library” of variables up to date. And thus, unfortunately, shall it always be.
Hence the dichotomy between the discrete (read: insurance billing) and the free-text (read: patient care) types of data. The key here is that I suspect the folks DESIGNING the google systems are some mix of MDs and Strict Computer Scientists, whereas they really should have bioinformatics people designing it.
In response to your specific questions raised:
1) Billing data is discrete, hence it cannot by definition explain fully patient conditions. Unless and until there is a perfect mapping between free-text entry and discretely defined patient data, coded billing data will always be “confusing” if viewed as diagnoses.
2) My feeling is that this is GOOGLE’s fault (see my point above about who is working on the google product). Doctors typically use PATIENT CHARTS and have absolutely no interaction with billing codes.
3) Because without using complicated programming like natural language processing (NLP, which is always only partially successful in representing free-text), billing data is often the only discrete data available. My guess is that nobody is using billing data for disease management. :)
4) This is a VERY interesting question. It would only be posed by auditors. They would look at your scenario and see that you received what you thought was ONLY your own data and looked at it in good faith. My feeling is that the results of this audit would say “correct your systems so that patient data is associated with the correct patient.” In other words an IT compliance problem and not a patient or doctor behavior problem.
5) Good question, harkens back to 4. I work in medical IT and one of the biggest barriers to a real cross-institution EHR are complex and scary privacy laws which cause us to ask questions like the one you pose above. There is of course still the challenge of GETTING patient charts (read: the free-text stuff) into an electronic format (as opposed to billing information, which already is for the most part).
Dear Ben: Good comments all. I’d like to add that I think E-patient Dave’s story of the Wicked EMR has been so popular because it’s an understandable tale about things people find hard to understand. You’re doing us all a favor by diving into the expert domain, and this will help us make things better in the future. DCK