In my previous blog post, I wrote about the Nashville, TN nurse RaDonda Vaught who was convicted last month of criminally negligent homicide and abuse of an impaired adult after a medication error contributed to the death of a patient in 2017. In it, I described some of the circumstances leading up to the death of the patient, Charlene Murphey. In this post, I’ll discuss some of the systemic issues in healthcare that contributed to this tragedy. These are issues that the industry should be aware of, as they threaten the quality of care delivered at every institution.
Here’s a summary of the events based on publicly available information: Nurse Vaught was taking care of a patient in a critical care unit at a major hospital. The patient was scheduled for a PET scan, and her physicians ordered an IV sedative called midazolam (that’s the generic name; the brand name is Versed) to alleviate anxiety with the test. Because communication between the hospital’s electronic health record (EHR) and medication dispensing cabinets wasn’t working, the nurse was directed to obtain the medicine by overriding the cabinet controls. She typed in V and E searching for “Versed” but since the cabinet user interface only displayed generic names, the med that came up was vecuronium, which is a paralyzing agent used for intubation. Nurse Vaught didn’t notice that she grabbed vecuronium and not Versed, despite warnings on the vial of vecuronium that it can cause paralysis and hence patients would likely need respiratory support.
Let’s begin the discussion regarding systemic issues by noting that the nurse was orienting a new employee to the neurological intensive care unit (ICU). This is appropriate for an experienced clinician to train new staff in the particularities of an ICU. Even doctors and nurses who have been doing critical care for decades need to understand the specific workflows and “customs” of a new unit. However, in this instance, the clinic operations were not ideal as typically a trainer would not have direct clinical responsibilities in addition to orientation duties. Yet such an ideal operating scenario is now a luxury given the staffing shortage during a global pandemic. On this day, the unit was extremely busy with very sick patients. Additionally, some of the safeguards that were put in place to mitigate medication errors weren’t functional.
As we consider patient safety in a hospital setting, it’s helpful to understand one of the ways that hospitals use structure to minimize the chance of errors. The Swiss Cheese Model helps to explain how errors can occur in complicated systems. Essentially, every slice of cheese is a safety layer that’s in place to prevent errors. Each layer has a few holes where mistakes can occur. As long as the holes in all of the layers don’t align perfectly, errors can’t occur. But infrequently, the holes all line up. This is what happened in this tragic incident.
As I wrote in my last post, modern EHRs are typically interfaced to medication cabinets in busy hospital units. The physician orders the medication, and depending on the specific workflow, when the nurse is ready to administer the medication, they go to the med cabinet, identify themselves, identify the patient, and see a list of meds that have been ordered and “cleared” for the patient. Only in emergencies should the nurse need to override these safety features. When the connection between the medical record and the cabinet isn’t working, the only option is to override.
For high-risk medications (for example, think chemotherapy and powerful opiates), nurses often must ask a colleague to double check that they’ve achieved the Five Rights of Medication Administration: right patient, right drug, right dose, right route, and right time. The Five Rights should always be met whenever administering any med, but especially with high-risk medications. A human double check distinctly decreases the chance of a misadventure, but it can’t occur with every dose for a few reasons. As I alluded to earlier, staffing is a critical issue during the pandemic, so nurses just aren’t easily available to do a routine double check. Further, we have to consider that asking for a double check is disruptive to workflow and concentration for a nurse who is likely focused on other care duties. While a double check might have made sense if the nurse realized she was administering a paralyzing agent, even with adequate staffing, it typically wouldn't occur with a sedative like midazolam.
From a software usability standpoint, auto-completion while searching for a word or phrase is a no-brainer. As you type, the system anticipates what you’re looking for and offers up suggestions. This is how Google search works, and users typically love it. In this particular case, though, a few problems crept up. First, the medication dispensing cabinet was programmed to only accept generic names. Alternatively, the system could have used a combination of generic and brand names to allow clinicians to find a med searching for either. If the medication had been called “midazolam (VERSED)” then theoretically, typing in M and I (for midazolam) or V and E (for Versed) would have identified the same drug. Second, the system might have been programmed to not return search results until a minimum number of letters were entered. Hence, if the system required even just three letters, vecuronium would not have been offered when V, E, and R had been entered.
Most U.S. hospitals now utilize bar code medication administration (BCMA) in at least some of their high-risk units. BCMA ensures the Five Rights are checked every time. The nurse who is to administer the medication first scans the patient’s identification band and then scans the medication (either the manufacturer’s bar code or one added by the hospital’s pharmacy). The ID band tells the nurse they have the right patient; the medication bar code confirms the right drug, dose, route, and time. For this to work, we need the physician’s order already entered into the EHR. Additionally, we need the bar code reader and its interfaces to all be functioning.
If a systematic look at the procedures and technologies in place had resulted in a redesign that made it more difficult for the people in the system to do the wrong thing, the medication error might have been averted. Notice that I wrote that it might have been averted. Even under ideal circumstances, errors can occur in ways that we just can’t predict and/or mitigate. Yet it’s clear that we need as many of these systems to be up and running as possible to decrease the risk of major errors. We need to ensure that user-centered design principles are incorporated into all of our thinking about care delivery to make it easy to do the right thing and quite difficult to do the wrong thing.