Error Prevention Strategies: It's not "Sophie's Choice" folks

Last week on my Medscape medication safety blog On Your Meds, I wrote a piece about how nurses in greater San Francisco area hospitals improved medication safety. The collaborative is reporting an 88% reduction in the incidence of errors in the administration node of the medication use process over a three year period.

At the outset, it's worth noting that these results are astonishing, placing them in the "almost too good to be true" category. The study employed "observed error" methodology, a more robust method of error detection than "reported errors," (the methodology most programs and data sources rely on). The rigor of the detection methodology used in this study adds credence to the results.

But it's worth looking a little more closely at the study design to find the most important take-away lessons.

The nurses tested how adherence to six distinct performance elements in their medication administration process impacted accuracy: [link]

1. Compare medication to medical record
2. Keep medication labeled until administration
3. Check two forms of patient identification
4. Immediately record medication administration in chart
5. Explain the medication to the patient
6. Minimize distractions and disruptions during the administration process

From an engineering standpoint, these elements can be predicted to produce a robust medication administration system. Comparing medications to the medical record and checking two forms of patient identification, for example, add redundancy at high stakes junctures of the process. And "explaining the medication to the patient" creates a recovery opportunity, an engineering control that allows an error that's been set in motion to be detected and remediated before harm occurs. (The practice is also desirable from a participatory care standpoint and also is "the right thing to do" based on variety of ethical principles.)

"Minimizing distractions and disruptions during the medication use process" is the performance element that drew the most attention in the lay press, and it's what I focused on the first time I took on the issue at Medscape. Minimizing distractions at high stakes junctures of performance is a technique that high reliability industries employ. (It's why aviation personnel in the flight deck close the door and why they're subject to tighter performance expectations at altitudes less than 10,000 feet.)

What the San Francisco nurses really studied is whether adherence to a system designed to elicit a specific outcome yields the desired outcome more often than using a loosely defined, variably employed set of expectations does. Minimizing distractions was an important part of the interventions, but it wasn't the only one. The nurses did not find one "magic bullet," but rather moved from an "intention-based" process to a process that was both engineered and adhered to, something that helps explain the very favorable, highly desirable results obtained.

Understanding how these results were obtained is also important before leaping into the comparative arena, especially when the discussion is built around a "forced choice" construct that does not and should not exist. This is what I think is happening in a blog post entitled, Low Tech solution to Med Admin errors better than BCMA?

Designing the most robust system feasible to accomplish a high stakes task is how system engineers approach their work. (Risks surrounding medication administration are well documented and errors at this point remain common.)

Seminal medication safety data show that a substantial portion of errors originate in the administration phase of the medication use process.



Equally important these data reveal that patient harm is highly likely to occur as a result of errors that originate in the administration node.


It's important to recognize that errors in the administration node are problematic, not because nurses are problematic but because the systems nurses rely on and the downstream position of their work confer risk. Managing that risk has been the focus of medication and patient safety specialists over the past decade. IT solutions, specifically the ability to bar code patients and their medications, and to have key patient, drug, and order information integrated and available at the point of care, represent strategies engineers see as reliable, reproducible, and capable of sustaining change over time.

The San Francisco nurses' study did not rely upon bar code medication administration (BCMA) although it appears BCMA was used in at least some of the study sites. But what must be noted is that key performance measures in the study (namely, "compare medications to the medical record" and "check two forms of patient identification") represent standard medication safety practices that are now part of The Joint Commission's healthcare accreditation standards. While they are important elements in the system design the nurses tested, these elements are not "stand alones." They would have occurred, on some level and likely with unwelcome variability, in these hospitals during the study period irrespective of whether they were part of an intervention study.

More important to debunking ill-conceived notions that medication administration accuracy is an "either/or" proposition (pitting low tech performance measures against tech-mediated ones) is the knowledge that BCMA automates key elements of the performance measures the San Francisco nurses built into the system they tested. These include comparing medication to data in the medical record; immediately recording medication administration in the chart; and checking two forms of patient identification. Additionally, BCMA work flows necessarily foster work processes in which medications remain labeled (often in their original packaging) until the point of medication administration.

If BCMA has failed to reach its full potential in the medication administration arena, as John Poikonen questions in his RxInformatics post, the reason has less to do with the inherent fitness of the technology than how user-friendly it is designed to be; how it is incorporated into nurses' work flow; and how it is supported in the aftermath of the initial investment. Most importantly, disappointing results with BCMA likely reflect system design failures that do not take into consideration the limits of human performance when carrying out high stakes tasks. Nurses should rely on automated solutions to accomplish high stakes work and they should not be expected to multitask while using them.

Your pilots get to close the cockpit door when they perform tasks that, if carried out incompletely or incorrectly, could kill the people who depend upon them. Pilots also rely on high tech instrumentation that automates many key performance elements.

Why would you want your nurses to "pick one"?


Note: Representation of the seminal medication error data discussed here was borrowed from similar formats used by the medication safety professionals at the Institute for Safe Medication Practices. I am indebted to them, both for this depiction and the modeling upon which my knowledge of medication safety is based.