EHR Usability Problems Linked To Potential Patient Harm

If you’re a clinician, you’ve probably always felt that EHR usability problems were a factor in some patient care glitches. Now, there’s some research backing up this hunch. While the numbers of EHR-specific events represented in the study are relatively low, its lead researcher said that it probably underestimated the problem by several orders of magnitude.

The study, which was profiled in the American Journal of Managed Care concluded, that at least some patient safety events were attributable to usability issues. The study, which was just published in JAMA, involved the analysis of nearly 2 million reported safety events taking place from 2013 to 2016 in 571 healthcare facilities in Pennsylvania. The data also included records from a large mid-Atlantic multi-hospital academic medical system.

Of the 1.735 million reports, 1,956 (0.11%) directly mentioned an EHR vendor or product. Also, 557 (0.03%) include language explicitly suggesting that usability concerns played a role in possible patient harm, AJMC reported.

Meanwhile, of the 557 events, 84% involved a situation where patients needed to be monitored to preclude harm, 14% of events potentially caused temporary harm, 1% potentially caused permanent harm and under 1% (2 cases), resulted in death.

The lead researcher on the study, Raj Ratwani, PhD, MA, told the AJMC that these issues are unlikely to resolve unless EHR vendors better understand how providers manage the rollout of their products.

Even if the vendor has done a good job with usability, he suggests, healthcare organizations adopting the platform sometimes make changes to the final configuration during their implementation of the product, something which could be undoing some of the smart usability choices and safety choices made by the vendor. “We really need to focus on the variability that’s occurring during the implementation and ensuring that vendors and providers are working together,” Ratwani said.

Along the way, it’s worth pointing out that the researchers themselves feel that the actual number of usability-related patient safety events could be far higher than the study would suggest.

Ratwani cautioned that he and his team took a “very, very conservative approach” to how they analyzed the patient safety reports. In fact, he suspects that since patient safety events are substantially underreported, the number of events related to poor usability is probably also very understated as well.

He also noted that while the study only included reports that explicitly mentioned the name of the vendor or product, clinicians usually don’t include such names when their writing up a safety report.

About the author

Anne Zieger

Anne Zieger

Anne Zieger is a healthcare journalist who has written about the industry for 30 years. Her work has appeared in all of the leading healthcare industry publications, and she's served as editor in chief of several healthcare B2B sites.


  • Over the past few years we’ve worked with a number of EHR vendors on improving the usability of their solutions. We’ve noticed a number of items that seem to common to many of the systems, and this list contains some of the most common and highest priority usability issues that should be avoided in your EHR designs.

    Here are the top 5 issues we’ve seen across multiple EHRs we’ve tested:

    1) Required fields not clearly indicated

    This, by far, is the most frequent, and most annoying usability issue that has shown up in our studies. Most people expect required fields to be identified with a “*” or some other commonly used interface convention

    Make sure that you identify ALL of the required fields (and don’t use a single label for a group of fields) and provide some type of embedded assistance around the more complex conditionally required fields.

    2) Button nomenclature, and button consistency.

    It is hard to believe that this issue still shows up in software. It is important to make sure that the wording, and the order of the action buttons within a system are consistent.

    For example use: “Cancel/OK” or “OK/Cancel.” Pick one and stick with it (or better yet, sniff the browser and if it is a MAC use “Cancel/OK”). Also, make sure that you use industry standard nomenclature and icons. Make sure that you don’t use the same icon for multiple actions.

    3) Default sort settings are the same for every list.

    Most systems seemed to have an ascending alphabetical sort for all or most of the lists of items provided. Each default sort across the entire application should be carefully examined and set to minimize user action while also maximizing patient safety.

    This is a situation where consistency reduces the usability of an EHR. Each list of items should have it’s own default sort order that reduces the amount of actions and decisions required to correctly choose what the user is searching for.

    An example of this can be taken from a recent evaluation we conducted.The user was presented a list of “Common Medication Allergies” and they were to select an item that their patient was allergic to.The list was sorted, as are most lists, in ascending alphabetic order.Penicillin is one of the most common drug allergies, but because the default sort was wrong, the user had to scroll down the list to find Penicillin.Sorting this list by the frequency of the allergy makes more sense.

    Look at every list that is presented to your users and determine what would be the best way to sort that list to minimize the amount of mental effort, and user action.

    4) Empty lists.

    You might have laughed at some computer manuals with the phrase “This page left intentionally blank.” Those of us that have been using personal computers for as long as we have will understand why we call this the “WordPerfect 5.1Null list” or the “Null list” problem (yeah, this problem as been around since back in the DOS computing era). This is a very common and easy to fix issue.

    When the results of a search, such as a list of patients, or some other query comes up blank, don’t just display a “Null result.”You need to think about all of the possible reasons that the query failed. Was it a misspelled word? Were there not any patients that were born in that year, did your user enter a SSN incorrectly?

    Tell the user that their search came up blank, Don’t just present a null list. Tell them what terms were used for their search. Offer up a way for them to modify their search. Make it easy for them to add a new record to the list you are displaying. Remember, as in item 3 above, each list is different. Don’t use the same formula for each list—unless it makes perfect sense to do so.

    5) Provide embedded assistance on more complex controls. Deep within many of the systems we evaluated lies a complex, and sometimes non-standard UI control feature that is critical to performing an important task.

    In the ONC 2014 Edition Summative tests, this control was usually related to the “Clinical information reconciliation task.” The control was usually some type of “Shuttle” widget where the user selects multiple items from one or more lists and places them into a single reconciled list (of Problems, Allergies, and Medications)

    When using a complex and/or non-standard UI it is very important to provide a small amount of embedded assistance along with the control. A short sentence or phrase telling the user to “Select items from List A to add to the reconciled list” goes a very long way to helping them understand what they need to so in order to solve the task at hand. Notice that we said a “small amount” of text.

  • You don’t say . . . “healthcare organizations adopting the platform sometimes make changes to the final configuration during their implementation of the product, something which could be undoing some of the smart usability choices and safety choices made by th

    Allowing the tail to wag the dog is not the way to go.

    Healthcare organization know or should know what their protocol needs are and configure their platforms accordingly.

    If they find they can’t, they should move from to a platform that allows them to build and extend workflow and workload according to healthcare organization internal “best practices”.

Click here to post a comment