The growing role of clinical informaticists in making EHRs more satisfying.
Last month I reflected on the 10 years that have passed since the Meaningful Use incentive program spurred widespread adoption of electronic health records (EHRs). In EDs, we mostly went from an eclectic mix of paper systems and customized (‘best of breed’) ED information systems, to a state where most EDs are functioning on one of several ‘enterprise’ vendors like Epic, Cerner or Meditech.
Using an enterprise EHR system, particularly in the ED, wasn’t very fast or efficient, but there was a hope that the mass influx of customers would drive improvements in usability. Instead, studies show EHR usability remains poor, and in fact contributes to physician burnout.
In Part 1, I covered some modest improvements EHRs have introduced to patient care and safety, but at significant cost. In this final part, I’ll talk about some specific shortcomings of EHRs that, 10 years ago, seemed poised for improvement, and yet continue to persist. I’ll highlight the growing role of clinical informaticists in making EHRs more satisfying and efficient for doctors, and safer for patients.
The Unrealized Promise of Clinical Decision Support
Clinical Decision Support (CDS) takes many forms – from subtle nudges to hard-stops. Ten years ago, CDS was basic, inelegant and heavy-handed – and it hasn’t gotten nearly as sophisticated as I had expected it would.
I can’t say it’s all bad. Protocols implemented through the EHR have led to improved outcomes and efficient throughput, for well-defined entities like stroke and STEMI. We’ve also seen and reduced error and promoted standardization for things like procedural sedation and post-intubation care. But for less clearly-defined ED presentations, like sepsis or CHF, we find ourselves often battling against the EHR’s suggestions.
Trying to manage a young, otherwise healthy ED patient with the flu in 2020 is an exercise in frustration – constantly dismissing pop-ups and alerts about ordering serial lactates, and 30mL/kg of fluids, and prompts for special SEP-1 documentation. This effect is insidious, and ultimately leads to more workups for minor complaints – it’s just the path of least resistance.
We have mountains of evidence that the majority of drug-drug interaction warnings are worse than useless: they’re routinely ignored, which leads to alert fatigue, which means the occasional important warning is also missed.
Countless mitigation strategies have been proposed and evaluated, including pharmacy-driven monitoring of alert responses, and clinician-driven ratings of alerts. Researchers estimate that it takes hundreds of alerts to prevent one true adverse drug event, and that the volume of alerts could be cut by more than half without appreciably increasing risk to patients.
A few institutions have developed progressive alert monitoring strategies: they have infrastructure to collect and act on feedback, and have cut back on the volume of alerts, or tailor future alerts to the appropriate audience. Yet for the most part, US hospital administrations are fearful of rolling back anything that may lead to a bad outcome. Most of us continue to click away, despite the alert fatigue and evidence for safe alternatives.
At the start of the last decade, I naively assumed that a lot of the mindless cruft that got entered into charts would get cleaned up. But it’s 2020 and we still see too many allergy alerts that are meaningless, or just wrong. Wouldn’t someone, or some bot, go into each chart and decide that vomiting with codeine wasn’t actually allergy, and shouldn’t pop up when docs are ordering a fentanyl patch?
Or take the patient who has reported a penicillin allergy — if they’ve received Keflex on other occasions without incident, can’t the EHR suppress a warning when the doc is ordering a cephalosporin? Yet, any kind of rollback of alerts, or bulk-editing of charts, is viewed as too risky.
Further, I had expected for context-aware alerting to be more widespread by now. A good example is mixing heparin and aspirin — EHRs typically discourage mixing anticoagulants with big, bold, red warnings — even in cases where it’s indicated, like an NSTEMI. You’d think by now, an EHR could “detect” an elevated troponin and guide physicians through an NSTEMI protocol, but instead it continues to caution EPs trying to do the right thing.
Similarly, the new radiology decision support mandate (Appropriate Use Criteria for CTs and MRIs) is another a missed opportunity. If we had systems that used natural-language processing of our notes, and took chief complaint and other factors (pregnancy, creatinine) into account, that might be worth something.
But instead, this primitive pop-up usually forces the ED doc to pick a diagnosis from an unintuitively categorized drop-down list that includes mostly non-emergent conditions, and routinely recommends an MRI instead (the system never understands that we’re in the ED and it’s 3am).
After years of hearing about the benefits of big data, we’re only now starting to see alerts based on predictive analytics. Our shop has been looking to implement predictions for inpatient admission — the promise is that, behind the scenes, the EHR can look at a patient’s demographics, vitals and lab results, and predict admission about an hour before the ED doc activates the admission.
Our case managers and bed planners say they’d find this extra lead time helpful, even if it’s occasionally too sensitive. I worry that the computer’s prediction will cause us to second-guess ourselves — this occasionally happens with EKGs, for instance, but at least EKG criteria are finite and readily verifiable — it’ll be harder to debate a Greek Oracle. And, as with sepsis, how many times will we take the path of least resistance, admitting a patient who seems well, rather than taking a risk and documenting our disagreement with an algorithm?
Things EHR could’ve fixed — but didn’t
Some frustrating aspects of practicing medicine in the US today aren’t necessarily the fault of EHRs themselves. They could be the effect of local implementation decisions, or responses to regulations. Yet EHRs seem to have highlighted these problems, or encouraged them.
Take e-prescribing. Please. This was billed as a solution to the problem of Rx pad legibility, and gave agencies access better access to drug prescription data, making prescription drug monitoring programs (PDMPs) possible. Some states like mine mandate it. But, when you have EHRs that can print Rx, you’ve already solved the problem of legibility (and you’ve also got order sets and error-checking to minimize inappropriate choices and doses). Plus, pharmacies already contributed data to prescription drug databases to cut down on drug diversion — they never relied on e-prescribing.
And e-prescribing doesn’t work like modern cloud-based messaging – you can’t write an e-prescription and have your patient pick it up anywhere. Instead, you’ve got to send it to a specific pharmacy. It’s like faxing, only without confirmation the message went through. So what do we gain by e-prescribing?
From my ED perspective, we gain nothing — and we lose a ton of time helping stuck frustrated patients, by re-routing prescriptions to pharmacies that are open later, or aren’t running short of the drug the patient needed. We lose a lot of time entering and re-entering pharmacy addresses into the chart. The patient loses the ability to comparison shop. E-prescribing has been an expensive, frustrating, time-wasting replacement of a serviceable paper system — sort of a microcosm of many complaints of EHR.
The informatics textbooks say, to properly implement an electronic system, you should redesign the entire workflow, and not just digitize a byzantine, paper-based process. But too often, that’s exactly what happens. For instance, there’s a good chance when you order a CT scan with IV contrast, you’ve got to print out a form about contrast reactions, for the patient to fill out and bring to Radiology.
There’s no option for the provider to fill out the questions in the EHR, or to give the patient a web-based form on a tablet — no option to even scan the paper into the chart. Why not? Radiology was probably reluctant to train their techs to use the EHR, to log in to view responses. Fewer clicks for their staff forced a more convoluted process upon the rest of us.
Similarly, anytime I watch a nurse waiting on hold to “give report” before a patient can go upstairs, I see a failure to adapt workflows to the electronic era. Every time I have to recite a patient’s history and labs to a consultant, or call a clinic to arrange a follow-up appointment, it’s a lost opportunity to take advantage of the information in the EHR. I’m not saying “warm handoffs” and human conversation are inherently wasteful, but too many workflows don’t even incorporate the EHR, let alone leverage it to make calls brief and timely.
Enter the Informaticist
Emergency Medicine is still considered a young field. Clinical Informatics (CI) — the study of information technology and how it can be applied in healthcare — is even younger. The first subspecialty CI board exams were offered in 2013, at the height of Meaningful Use adoption. There are now dozens of ACGME-accredited Clinical Informatics fellowships, and ACEP’s Informatics Section boasts 400 members.
It’s a start. There’s emerging evidence that departments with dedicated clinician informaticists have more satisfied docs. This makes sense: informaticists can’t rewrite the code behind an EHR like Meditech or Cerner, but they can smooth out some rough edges, and streamline common workflows. Informaticists can identify docs who aren’t making the most use of EHR automation and personalization tools, and work with them to bring up their efficiency. When a well-meaning administrator wants to ensure compliance with some new mandate, the local informaticist can work to insert it into clinical workflows so it’s not such a disruption.
Informatics is a natural fit for the ED — nowhere else in the hospital is it as crucial to bring together patient data, records and evidence-based guidelines to help bedside decision-making. It’s what initially drew me to this new subspecialty. I’ve since been immersed by efforts to make our system more capable, intuitive and efficient, for me and my colleagues. And there’s plenty of work to be done (I doubt any enterprise EHR vendors made a sale to the hospital C-suite on the strength of their ED module).
Medicine and EM in particular need more informaticists, but I think we can also do more with the informaticists we already have. This fall, I had the honor of being named as the first chair of ACEP’s new HIT committee. Our purpose is twofold — to gather and disseminate best practices for making the most of your EHR — be it documentation, or achieving interoperability or submitting quality metrics. Our second goal is to work directly with vendors, other medical societies and policymakers on improving the experience of EHRs.
Our committee has lofty goals, and the support of ACEP leadership. Most importantly, we are motivated by tens of thousands of practicing EPs who want a better, more efficient experience using their EHR to care for patients. It’s these voices that will guide us, as we work to make sure the next 10 years of EHRs in the ED are more satisfying than the last 10.