“The use of artificial intelligence–derived controllers clearly signals a new era in intraoperative hemo dynamic management.”
The classic 1985 science fiction film Back to the Future transports the erstwhile protagonist (Marty McFly, played by a young Michael J. Fox) 30 yr backwards into the past in the eccentric “Doc” Brown’s custom DeLorean time machine to deal with a series of comedic yet moral quandaries regarding his future existence. A notable quote by Doc Brown is, “The future is whatever you make it, so make it a good one.”
The future of intraoperative and perioperative anesthetic management is of course unfolding every moment, but as in our personal and social lives, seminal events tend to provide “inflection points” when changes in pace allow us to step back and take stock of what “is already there.” For the different generations, those entering anesthesia practice, in the prime of their professional careers, or leaving practice (roughly equating with Millennial, Generation X, and Baby Boomers), such inflection points are either nearly invisible or strikingly obvious. For an aging Boomer whose day would begin nearly four decades ago with my earpiece pinned to my scrub shirt and pen in hand to fill out my paper record, overjoyed that I had arrived early enough to hunt down one of the few automated blood pressure monitors available so I wouldn’t have to manually pump blood pressures myself every 5 min, such points are striking despite progressive age-related memory loss! Obvious advancements in the operating room such as routine pulse oximetry, automated blood pressure monitoring, widespread availability of ultrasound technologies, and videolaryngoscopy make providing anesthesia so much less stressful for the practitioner and safer for the patient while watching surgeons perform near miraculous forms of minimally invasive surgery guided by advanced pre- and intraoperative imaging technologies. But when it comes to actually figuring out what is going on with my patient physiologically during a case, I still often feel the same sense of “unease” I experienced as a novice trainee.
Progress in advancing “interconnectedness” and application of artificial intelligence technologies in the operating room has been steadily progressing given concerted efforts by groups of “visionaries” (along with the requisite industrial collaborators) albeit perhaps at a somewhat slower pace than many of us expected. However, given the piecemeal nature of scientific progress, we are presented with an uncomfortably large number of technological paths to “the future.” For better or worse, many of us have increasingly put a fair amount of faith into clinical “black boxes,” with the processed electroencephalogram for predicting the depth of anesthesia perhaps the most pervasive example overall. In the cardiovascular arena, a plethora of noninvasive monitors of cardiac output displaying a panoply of cardiovascular parameters have been introduced and are often used for “goal directed therapy” for fluid and pressor/inotropic support despite underwhelming data for clinical effectiveness.
Most recently, the US Food and Drug Administration has approved clinical software capable of predicting upcoming episodes of hypotension based on alterations in the invasive arterial waveform configuration highlighted previously in Anesthesiology.1 In this issue, the process by which another type of decision support software, Acumen Assisted Fluid Management (Edwards Lifesciences, USA), was approved is presented.2
The use of artificial intelligence–derived controllers clearly signals a new era in intraoperative hemodynamic management. However, it is troubling to me that neither of these approaches has direct inputs from preoperative patient-specific or other potential anesthetic-specific parameters. Although a recent editorial professed that closed loop anesthesia is likely to be as good as the best human anesthesia,3 I have reservations that simply servo-controlling cardiovascular parameters is adequate to improve outcome and also poses challenges to maintain our specialty’s intellectual capabilities as perioperative physicians.
How we are trained as perioperative physicians is grounded firmly in as best an assessment as possible of our patients’ key preoperative baseline factors, forming the basis for many for our decisions, coupled with expected stressors likely to be encountered in the operating room and pharmacologic effects of drugs. The patient with a very low ejection fraction will respond differently to all aspects of anesthetic, fluid, and pressor/inotropic management than a hyperdynamic, hypertrophic patient. A truly integrated closed loop system should have appropriate baseline preoperative factors integrated into it. For the average “normal” patient, such inputs can easily be modeled, but we must be appropriately leery of inappropriate application of “blanket” software approaches to abnormal patients. These systems depend on input of the patients’ key baseline hemodynamic values, but what is an appropriate baseline value is often controversial or, perhaps better stated, highly dynamic. The importance of surgical manipulation, with alterations of gravity (head up/down influencing venous return, side-to-side alterations that may alter a transducer’s reference point, and others), is also a necessary input. Nearly complete reliance on one particular noninvasive method for determining cardiac index (and/or stroke volume) as the major independent variable for such systems can be problematic.
Another dimension to consider, one that is not unique to anesthesiology or medical practice in general, is that of “deskilling.” However, as pointed out by numerous opinion leaders over the past several decades, of greater immediacy to our daily practices is our intellectual linkage to safety issues in the aviation industry. Widespread application of that industry’s safety practices in the form of “preflight checklists” and development of sophisticated simulators has improved and continues to improve what is already an astoundingly safe industry given how many people are transported annually. However, despite these advancements, there is fear among aviation experts that pilots are losing skills at a faster rate than they are gaining them, a process facilitated by increasing automation on the flight deck. A number of air disasters or near misses over the past several decades related to malfunction or misinterpretation of flight path control systems or internal aircraft software control systems (most notably the infamous Boeing 737 MAX disasters) have been attributed to this phenomenon.
It does seem inevitable that software control of hemodynamics and anesthetic depth will become routine. Thus, we might ask, “What happens to the operator/clinician involved?” Will it be more appropriate for a busy anesthesiologist covering multiple operating rooms to be supervising the admittedly extreme scenario of “information technology experts” ensuring the machines are functioning properly or actual healthcare providers monitoring the patient and not the machine? And what happens when the “computers go down”? Who will rush in to fill the gap? Will the process be “good” or will it be “dystopic”?
Despite predictions of possible doom and gloom by an aging Boomer, there is abundant room for hope and daily evidence that surgical and anesthetic outcomes continue to improve. The maturation of simulation technologies holds great promise. “High-fidelity” simulation strategies are already a routine part of anesthesia training and maintenance of certification and have recently moved into the purely digital realm, allowing even wider exposure and repetition required for retention. One intriguing development is the introduction of “multidimensional decision support displays” and “avatar”-type monitoring display technology that visually integrate hemodynamic parameters more clearly display “wellness” or “danger” on the monitor screen.4,5 Although still early in development and commercial introduction, this approach promises to mirror in real time what is happening to the patient in a visually straightforward manner that should be hard to misinterpret. Of course, these depictions will only be as good as the inputs from the monitoring technologies used to estimate them, but they are a step in the right direction to facilitating close attention to the patient at hand. However, a recent report documenting improvement in a variety of “surrogate” process factors (e.g., blood pressure, tidal volume, and fluid administration) failed to document any improvement in the requisite “hard” clinical outcomes, illustrating the complexity of the challenges ahead.6 Is it possible that many physiologic parameters will be passé in the distant future when “precision medicine” is highly evolved and routine?
The way forward to the operating room of the future is fraught with thousands of incremental advances. Generations of clinicians will come and go, either happy with their practice or bemoaning the excessive troubleshooting involved. Hopefully, medical educators will remain faithful to ensuring that the basic, yet admittedly imperfect, tenets of cardiovascular physiology (cardiac output, venous return, heart rate, preload, afterload, contractility) are thoroughly engrained into all care providers’ brains along with the necessary manual clinical skills honed by many hours of simulation training. With this mantra in mind, I certainly think we can fulfill Doc Brown’s advice, so let’s make it good!
The author is not supported by, nor maintains any financial interest in, any commercial activity that may be associated with the topic of this article.