William Morton first demonstrated the use of ether anesthesia for a tooth extraction in Boston, Massachusetts on September 30, 1846. At the time, however, neither the patient nor the surgeons of Massachusetts General Hospital knew exactly how the patient was rendered insensible to pain. In an effort to preserve his monopoly on his discovery and to increase his notoriety, Morton colored his new drug red, added perfume to disguise its smell, and called it “Letheon.” Morton subsequently wrote an article about his success with his “ethereal preparation” and published it, along with an advertisement, in the Boston Daily Journal.
Shortly thereafter, he approached John Collins Warren, Harvard’s second professor of surgery, for permission to use his new compound on a patient undergoing a more complicated surgical procedure. The ethical guidelines of Massachusetts General Hospital, however, prohibited the use of secret preparations, called “nostrums.” Collins responded by inviting Morton into the operating theater to observe an amputation of a young woman’s leg. When Morton arrived, Warren questioned him about the composition of his new compound. Morton initially hedged, but under pressure from the surgeons and in front of the patient, he was eventually forced to reveal that his magical substance was nothing more than diethyl ether. Morton’s secret was out, and the medical community was finally able to conduct a real evaluation of this revolutionary discovery (1).
In this issue, Rampersad and Mulroy(2) report a case in which an Aspect Bispectral Index (BIS) monitor that was being used during the procedure indicated loss of consciousness, but in which the patient experienced awareness and recall. The fact that the patient experienced awareness is not in itself unexpected; his poor physical condition and hemodynamic instability under anesthesia placed him at high risk for this complication. What is remarkable, however, is the authors’ reliance on a device to prevent this complication and their inability to explain exactly why it did not work for this patient. The case report illustrates a fundamental problem with the use of monitors that use proprietary algorithms.
The purpose of discussing adverse events at quality assurance conferences is to determine the cause of a particular problem and develop a plan to prevent its reoccurrence and then to disseminate this information to others. Publication of a case report places such an event, and the steps that the authors would have taken to prevent it, into the literature, where it can be read by practitioners around the world and preserved indefinitely. Unfortunately, the proprietary nature of many new medical devices makes it impossible to identify the root cause of the error, let alone take steps to correct it. The authors are left with the rather unsatisfying conclusion that for this patient the device failed for reasons that may be known only to the engineers who developed it.
Digital signal processing was first introduced to reduce the number of artifacts presented to the clinician (for example, removing movement artifact from a pulse oximeter display). As the technology improved, newly developed monitors offered signal interpretation. Monitors capable of monitoring the electrocardiogram could, for example, automatically detect arrhythmias. Sophisticated digital signal processing does not come without a price, however. In some monitors, the relationship between the original signal obtained from the patient and the information presented to the clinician is often unclear. Many new devices require that clinicians interpret artificially generated numbers or waveforms that have undergone complex or undisclosed signal processing. In these devices, there is an unclear relationship between the number or waveform and the original signal.
In the end, clinicians are left to interact with many medical devices as if they were “black boxes” in which raw data from the patient enter one side and a highly processed signal emerges from the other. Because of the highly competitive nature of the medical device industry some manufacturers are very clandestine regarding the details of these processes. The physician is presented with a significant problem if the device fails. Without an understanding of how a number or a waveform is generated, how is the clinician to know if it is valid? Again, the clinician is forced to rely on an “artifact” indicator that may or may not work as described. In Drs. Rampersad and Mulroy’s case report the result was a patient who experienced intraoperative awareness.
“Black box” medical devices may impede medical research. Researchers are forced to treat the device as if it were a natural phenomenon and to design studies that probe it by evaluating its waveforms or numbers under various conditions. It is disheartening to read and review scientific studies designed with the sole objective of discovering these proprietary algorithms. One can only wonder about how much time and how many resources are misdirected in this process. More often than not, the medical academic community is left to writing “Consumer Reports” type studies (i.e., Device A did better with motion artifact while Device B did better with low perfusion states). The end result is that each manufacturer determines its own optimal device configuration and clinicians are forced to trust manufacturers. Although this may be perfectly acceptable for evaluating television sets, it is inadequate for the evaluation of medical equipment. Such studies do not really contribute to the body of knowledge nor do they encourage the development of new technology. Like the physicians at Massachusetts General Hospital in the 19th century, they are asked to accept a newly developed product without knowing how it works, or sometimes even if it works, for a given purpose.
What would happen if the major oncology centers competed in such a manner? Each center would keep chemotherapy protocols a highly guarded secret. One would be left with the choice of using one center’s Protocol B or another center’s Plan 9 when choosing a patient care plan. This is not the way for medical science to work. No physician would tolerate undisclosed formulations from the pharmaceutical industry (“Give this blue pill to your leukemia patients. It will make them better.”). However, health care professionals use heavily filtered signals and artificially generated numbers and waveforms without question when caring for patients in the operating room.
Medical science has, in some ways, returned to its origins, but instead of patent medicines, manufacturers now sell products with secret algorithms. Ultimately, regulatory authorities should mandate disclosure of the algorithms employed by medical devices used in patient care as a condition of certification. Until regulatory agencies can be convinced of the potential problems caused by the current practice, individual clinicians can help to improve the situation. Physicians should refuse to purchase a new “wonder” monitor without understanding exactly how the numbers and waveforms are generated. If medical device manufacturers find that health care providers demand a detailed explanation of how their devices work as a condition of purchase, they will ultimately have to release their algorithms. It worked for Dr. Warren in the Etherdome.