Medicine's human-machine interface: friend or foe?

Automation is designed to simplify activities in technical fields, including healthcare. Yet, as the Boeing 737 Max aircraft accidents in Indonesia and Ethiopia testify to, automation can also be linked to death. In the Boeing 737Max accidents an integrated automatic flight control function that moved the elevator flight control surfaces, intended to prevent excessive angle of attack (really an aircraft stall prevention tool), experienced sensor malfunctioning. This malfunction and the misalignment of the human-machine interface in these aircraft caused the crash deaths of more than 350 individuals shortly after take-off.  

There are examples of other aviation-related human-machine interface accidents that caused deaths. One example is the landing stall of an Asiana Airlines 777 at San Francisco International Airport in 2013. The National Transportation and Safety Board concluded the accident was flight crew’s mismanagement of the aircraft’s final approach to the runway. Also blamed in this accident were Boeing’s documentation for the complex flight control systems and inadequate Asiana’s pilot training. Many with aviation experience believe that the aircraft was “low and slow” and pilots fell behind the power curve for glidepath control during the landing sequence due to the human-machine interface of the autopilot and autothrottle linkages. 

These are just two examples of the lack of integration of training and mechanics of the human-machine interface in aviation, my avocation, and leads me to a more focused question about healthcare, my professional career. Are there similar human-machine interface functions inside healthcare that also impact safety for our patients? Indeed, automation and the human-machine interface impacts physicians and nurses daily within the wards and intensive care units. The human-machine interface in healthcare also impacts the electronic health record processes.  

In April of 2019, a former nurse at Vanderbilt University Medical Center withdrew vecuronium (a muscle paralyzing drug) from an automated medication dispensing machine, rather than her goal of obtaining versed (midazolam – a sedative). Thus, instead of administering the intended anxiolysis (anxiety relief) from the versed, the aging patient became paralyzed from the vecuronium and ultimately died from complications. The first two letters of the drugs are similar, “ve”; when the nurse overrode the safety mechanism in the machine, based on drug name, she pulled the wrong medication for administration. A theoretically avoidable human-machine interface error.

Also impacting patients are human machine interface functionsimbedded more widely in the electronic health records (EHRs). EHRs are now present in almost all acute care hospitals. As Fred Schulte and Erika Fry reported in the Fortune magazine issue of March 18, 2019, patient-harming errors are frequent inside healthcare, and many are related to “buggy” EHRs. As example, one Vermont attorney turned innkeeper, died, probably unnecessarily, due to software in the eClinicalWorks (eCW) electronic health record that failed to send to the radiology team a request for an imaging scan that should have picked up her intracranial aneurysm. Following this error within the automated EHR ordering function, this 47-year-old woman died of a subarachnoid hemorrhage that could have been prevented by the scan that was ordered, yet remained unsent to radiology department within the EHR.  

Automation in the human-machine interface is meant to make our work in both aviation and healthcare safer, yet the human-machine interface demands familiarity, training, and realization that the “human” is still responsible for monitoring the automation. The busyness of modern healthcare, the production pressure of clinical environment, the reliance on automation tools and the confusing human-machine interfaces can contribute to injury and death in both environments.