The promise and perils of electronic health records now extend to metadata.
You’re reading the polished, final version of this article. With any luck, EPMonthly’s editors have scrubbed the typos away and tweaked some of our awkward phrasings. The comments we made to each other, the tracked changes – all gone.
Thank goodness, right? We might be embarrassed if you could see how little time we spent on the first draft. Or the confusion reflected in our comments about our sources. And it’s probably for the best you don’t know about the recently deleted paragraph that, though fascinating, muddled our main point.
You’d see all that, though, if you had access to the MS Word file we used to write this document. In addition to this fine prose, Word files contain metadata – data about the data – that lets people read quite a bit between the lines.
Metadata is in every modern electronic file, and it’s been attracting attention. In a 2004 trial over the role of Vioxx in the cardiac death of a Houston man, plaintiffs subpoenaed the NEJM editor who published the results of a trial asserting its safety. While preparing for the trial, it was discovered that Merck had deleted damaging cardiovascular data from the submission before filing it with the journal. The metadata contained evidence of the deceptive deed and, while the drug had already been withdrawn from the market, it certainly didn’t help Merck’s standing with the jury.
Today, we routinely recommend that our residents check out ClinicalTrials.gov when reviewing a paper for Journal Club. Studies logged in ClinicalTrials.gov must retain a study changelog, which is a kind of metadata about the study methods. Even a trial with convincing data, published in a strong journal, starts to smell fishy when the changelog reveals the authors altered their methodology, mid-study.
It should come as no surprise that metadata exists in electronic health records as well. Metadata standards for health information exchange are being debated in DC; eventually even discrete bits of health data such as medication lists will have metadata tags associated with them, distinguishing between, for instance, admission medication lists and discharge medication lists.
Vendors have metadata baked into their products, too. While software and standards differ, in general, when you log in to your EHR and click a chart, you’re leaving digital fingerprints, and your access to various parts of the chart is timestamped. While these features may be helpful to protect patient privacy and ensure best practices, they’ve also found their way into malpractice cases.
The most famous case –which went to trial in 2002 – involved a bad outcome discovered after surgery. Prosecutors honed in on the anesthesiologists’ spotty logging of vital signs during the procedure and asked for the metadata. That’s when they learned he had attested to being present for the full operation…before it had concluded. Even though he had stuck around the whole time (it turns out these early attestations were common practice at his hospital) the metadata made him look dishonest. He ended up settling the case.
Then, in 2006, the Supreme Court clarified the Federal Rules of Civil Procedure (FRCP). Suddenly, lawyers were learning about “e-discovery” and plaintiffs were entitled to native files – not just a PDF or printout, but the actual data files, and the metadata therein.
After the FRCP changes, some experts expected a wave of metadata malpractice cases. Just how many ways can metadata be damaging? Primarily, it’s the timestamps, which can make it look like you barely glanced at the patient’s results, or didn’t look at it at all until after discharge. It doesn’t matter if your nurse relayed the results. Unfortunately, that kind of patient care doesn’t leave metadata, and thus is harder to prove. With time, conversations are forgotten, and you’re left trying to explain timestamps that make you look hasty or negligent. A more subtle metadata gap can occur if your EHR has a “chart preview” mode enabled with a mouseover or single click from the trackboard. These views don’t typically generate a timestamp. Being aware of this software quirk – and explaining it to the jury – can mean the difference between a physician appearing cavalier or competent.
Decision support – which has great promise to make patient-specific reminders and standardize practice patterns, but in the current state only generates a lot of annoying and mostly obvious drug warnings – also generates metadata. Maybe you start to ignore the warnings, and after the 10th time, or the 100th time, you click “ok” without really paying attention. The metadata will reflect that you acknowledged a potential drug interaction but wrote the prescription anyway.
Earlier this year, the Wired ED series covered Greg Husk’s research on EHR safety, which made use of physician responses to EHR hard stops, selected from a drop-down list. Husk’s group actually could characterize when physicians were selecting choices “honestly” or just clicking the first option available, to get past the pop-up. Decision support metadata, as far as we know, has never come up in court. But with more EHRs and more pop-ups, we suspect it’s only a matter of time.
A related risk of EHRs is based on another feature: auto-population of forms based on data collected during prior visits. Auto-population speeds up your charting and improves billing, letting doctors claim they reviewed the history, medication lists and allergies – because it’s all in the chart. But when old errors are reproduced in new records, and you didn’t intervene to correct them, you appear sloppy, untrustworthy, and vulnerable.
With all this fear mongering about the peril of EHR metadata, is there room for optimism? Well, while it’s true that metadata poses a theoretical risk, there’s no question of the benefits that come with EHR in the legal sphere. For years, it’s been the practice of many malpractice insurance companies to drop premiums upon adoption of EHR. While a single ED visit can now generate, through EHR, a 20-page printout, the more complete documentation of HPI, ROS and exam, nursing actions and vital sign trends seems to protect defendants when it comes to litigation.
We’d also like to think that metadata will help the cause of emergency medicine. Never before has such insight into how medical care is delivered, how physicians perform, been available. The possibilities for research, and process improvement, are manifest. Metadata can demonstrate, in a way that stories cannot, just how many interruptions we tolerate, and how much patient information we’re expected to absorb during a shift. Indeed, metadata may help define what the standard practice in the ED is – and that may protect us more than anything else.
In the end, the creation and monitoring of metadata is one more feature for your implementation team to be mindful of as they work with vendors and other hospital stakeholders, like pharmacy and radiology, to craft and streamline the user experience in your department. Learn what actions trigger, or skip – timestamps. Work to minimize the frequency of “hard stops,” and make the most commonly selected acknowledgments produce verbiage that’s not incriminating.
The wave of malpractice suits decided on metadata, which some experts expected, hasn’t materialized yet. But with the recent rise of electronic records in the US, it’s probably too soon to say whether metadata will do more harm than good to physicians. Just don’t ask us what we think will happen. We came up with a strong conclusion, but wrote it on a piece of paper that’s since gone missing. You’ll just have to take our word for it.
1. Koppel R, Metlay JP, Cohen A, et al. Role of computerized physician order entry systems in facilitating medication errors. JAMA. 2005;293(10):1197–1203. Free Full Text
2. Kuperman GJ, Bobb A, Payne TH, et al. Medication-related clinical decision support in computerized provider order entry systems: a review. J Am Med Inform Assoc. 2007;14(1):29–40. Free Full Text
3. van der Sijs H, Aarts J, Vulto A, Berg M. Overriding of drug safety alerts in computerized physician order entry. J Am Med Inform Assoc. 2006;13(2):138–147. Free Full Text
4. Aarts J, Koppel R. Implementation of computerized physician order entry in seven countries. Health Aff (Millwood). 2009;28(2):404–414. Free Full Text
5. Karsh BT, Weinger MB, Abbott PA, Wears RL. Health information technology: fallacies and sober realities. J Am Med Inform Assoc. 2010;17(6):617–623. Free Full Text
6. Wolfstadt JI, Gurwitz JH, Field TS, et al. The effect of computerized physician order entry with clinical decision support on the rates of adverse drug events: a systematic review. J Gen Intern Med. 2008;23(4):451–458. CrossRefMedline
7. Spina JR, Glassman PA, Belperio P, Cader R, Asch S, Primary Care Investigative Group of the VA Los Angeles Healthcare System. Clinical relevance of automated drug alerts from the perspective of medical providers. Am J Med Qual. 2005;20(1):7–14.