September 29, 2014

An article in the September 21 New York Times Sunday Review asked the question Can a Computer Replace Your Doctor? The author, Elisabeth Rosenthal (who identified herself as a “former physician”) opened with the following:

I shivered a bit when I heard Dr. Vivek Wadhwa say he would rather have an artificial-intelligence doctor than a human one. “I would trust an A.I. over a doctor any day,” he proclaimed at a recent health innovation conference in San Francisco, noting that artificial intelligence provided “perfect knowledge.” When asked to vote, probably a third of those in attendance agreed.

Artificial intelligence is obviously far more than data collection and management. Data collection and reporting is where much health care technology is today, however. Rosenthal noted devices that could turn an iPhone into an otoscope, blood alcohol measurement gadgets, home cholesterol test kits, cardiac trackers worn for more than a few days, devices that record sympathetic nervous system activity as a measure of stress and fitness trackers, all of which have limited potential to improve health, without more. Data on their own can be misleading. Lab tests can show everything within normal limits, and yet the patient dies. Spinal scans can show bulging discs in people who have no back pain. Data need to be understood in the right context to be meaningful. Rosenthal concludes with the hope “that we have the wisdom to ignore [health information technology], as we would a GPS device, when it leads us in the wrong direction—or nowhere at all.”

A number of devices used in anesthesiology practice go a lot farther toward replacing the doctor (or the nurse anesthetist). The SEDASYS®-Computer-Assisted Personalized Sedation System received FDA approval in the delivery of the drug propofol to healthy adults undergoing sedation for upper (esophageal) and lower (colon) cancer screenings.

The SEDASYS System is a computer-assisted personalized sedation system integrating propofol delivery with patient monitoring. It infuses an initial dose of propofol for 3 minutes, and then begins a maintenance infusion of propofol at a pre-programmed rate. If the monitors detect signs of over-sedation, e.g. falling oxygen saturation, depressed respiratory rate, or a failure of the end-tidal CO2 curve, the propofol infusion is stopped automatically. The machine even asks the patient to squeeze a hand-held gripper device at set intervals. If the patient does not respond and squeeze the gripper, the propofol infusion automatically stops.

Although the SEDASYS System goes much farther than the devices mentioned in the New York Times article, it cannot fully replace the anesthesiologist or other clinicians trained to manage loss of consciousness. In his forthcoming article Computer-Assisted Personalized Sedation (CAPS): Will it change the way moderate sedation is administered? (Communique, Fall 2014), Steven Dale Boggs, MD, MBA states that “For anesthesia providers, perhaps the most concerning aspect of the SEDASYS System is the inability of the System to prevent or manage loss of consciousness.” To address this concern, the manufacturers have developed a ten-hour clinical training program that is endorsed by the International Society for Anesthetic Pharmacology and is provided by an independent, qualified third party with expertise in airway management. They also require that an anesthesia provider be “immediately available,” albeit without defining the term. These two requirements, and the recommendations in the American Society of Anesthesiologists’ paper entitled Guidance for Directors of Anesthesia Service for Computer-Assisted Personalized Sedation (CAPS) Devices, make it clear that the SEDASYS System does not replace the anesthesiologist, even for the limited set of patients for whom it is intended to be used, i.e., ASA physical status I and II patients over the age of 18 undergoing minimal to moderate sedation.

In his Theanesthesiaconsultant Blog entry for November 1, 2012, Robot Anesthesia, Richard Novak, MD asks the question, “Will robots replace anesthesiologists?” He reviews the SEDASYS system and four “other forays into robotic anesthesia:”

1. The Kepler Intubation System (KIS) intubating robot, designed to utilized video laryngoscopy and a robotic arm to place an endotracheal tube (Curr Opin Anaesthesiol. 2012 Oct 25. Robotic anesthesia: not the realm of science fiction any more. Hemmerling TM, Terrasini N. Departments of Anesthesia, McGill University),
2. The McSleepy intravenous sedation machine, designed to administer propofol, narcotic, and muscle relaxant to patients to control hypnosis, analgesia, and muscle relaxation. (Curr Opin Anaesthesiol. 2012 Dec;25(6):736-42. Robotic anesthesia: not the realm of science fiction any more. Hemmerling TM, Terrasini N.)
3. The use of the DaVinci surgical robot to perform regional anesthetic blockade. (Anesth Analg. 2010 Sep;111(3):813-6. Epub 2010 Jun 25. Technical communication: robot-assisted regional anesthesia: a simulated demonstration. Tighe PJ, Badiyan SJ, Luria I, Boezaart AP, Parekattil S.).
4. The use of the Magellan robot to place peripheral nerve blocks (Anesthesiology News, 2012, 38:8)

Dr. Novak concludes that, although “Inventors are edging in that direction . . . . It will be a long time before anyone invents a machine or a robot that can perform mask ventilation. SEDASYS is designed for conscious sedation, not deep sedation or general anesthesia. Anyone or anything that administers general anesthesia without expertise in mask ventilation and all facets of airway management is courting disaster.”

Famously, computer scientists Clifford Stoll and Gary Schubert stated that “Data is not information, information is not knowledge, knowledge is not understanding, understanding is not wisdom.” Until a robot or other device employing information technology can function at the highest levels of the “knowledge hierarchy” also known as the “DIKW hierarchy,” there will be a need for doctors in medicine. The model and the relationships between data, information, knowledge, understanding and wisdom appear in the following graph from Gene Bellinger, Data, Information, Knowledge and Wisdom on The Way of Systems website :

The health technology devices mentioned in this Alert all provide data. The word “data” has been defined variously as facts, signals or symbols that must be organized into relationships in order to provide information. Fitness tracking device values, or, in anesthesiology practice, blood pressure values collected intraoperatively or (any of about 40 other patient monitoring values updated regularly and frequently in an anesthesia information management system [AIMS]) are data that do not have any meaning until the relationships between them are understood.

To become “information,” data must be organized or structured in such a way that they can be used—“processed in such a way that the information now has relevance for a specific purpose or context, and is therefore meaningful, valuable, useful and relevant" (Rowely J. "The wisdom hierarchy: representations of the DIKW hierarchy". Journal of Information Science 2007;33 (2): 163–180.) Home cholesterol kits or cardiac trackers are intended to show not just data, but information.

The next phase, “knowledge,” involves understanding the patterns, context and cross-connections of all of the relevant information available from various sources. The AIMS may, or may not, be able to transform information on the patient’s blood volume into a warning that the patient needs fluid. The blood pressure machine in the OR showing a reading of 30/0 or worse does not know whether the arterial catheter is damped or whether an antibiotic was just administered or the machine has been malfunctioning. The anesthesiologist must quickly form a comprehensive understanding of the entire situation and be able to choose the best course of action to treat the patient.

Arguably the alarm mechanisms in the SEDASYS system represent knowledge, but for a number of conditions that set off the alarms, the physician or nurse must decide when and whether to restart the propofol infusion, and at what rate. In other words, the system’s knowledge is incomplete in a very material way. As stated on the SEDASYS website,

For physiological conditions that have a high correlation with over-sedation, hypoxemia and prolonged apnea, the System’s red alarms stop delivery of propofol and instruct the patient to take a deep breath. After this red alarm clears, the clinician must manually restart the delivery of propofol….

The System does not stop propofol delivery for bradycardia, tachycardia, hypotension, hypertension or high respiratory rate alarms. These conditions do not have a high correlation with over-sedation. Clinically, it may not be appropriate to stop or reduce propofol delivery for hypotension induced by a vasovagal reaction. These alarms alert the clinician, enabling him or her to assess the patient and determine the appropriate course of action.

Only the clinician has the capability to evaluate the patient and determine the appropriate actions. Technology like Sedasys can signal a hypoxia problem but they cannot determine or effect a rescue.

Another area in which computers’ “knowledge” only goes so far is quality improvement, a hallmark of professionalism and a prerequisite to the advancement of the science of medicine. Steven R. Tosone, MD stated in his article Why Measure? From Quality Measurement to Quality Improvement published in the ASA Newsletter in February 2014 that “An area that has received too little emphasis is how we take data from measurements of quality and convert that into actual, real-world improvement.” To use such data for improvement requires feedback or communication of the status quo. Often, variability between practitioners, when clearly demonstrated, allows us to learn from one another what works best in the measured parameter and in our local practice.”

Translating knowledge into clinical practice will frequently involve the final component in the DIKW hierarchy: wisdom. Bellinger summarizes the definition of “wisdom” by Russell Ackoff, a leading systems theorist, thus:

Wisdom is an extrapolative and non-deterministic, non-probabilistic process. It calls upon all the previous levels of consciousness, and specifically upon special types of human programming (moral, ethical codes, etc.). … Wisdom is therefore, the process by which we also discern, or judge, between right and wrong, good and bad. I personally believe that computers do not have, and will never have the ability to possess wisdom. Wisdom is a uniquely human state….”

Asking questions that may seem unrelated to the patient’s symptoms and listening closely to the answers, and then honoring the patient’s choices are often uniquely human actions. While technologies that collect and analyze data and even in some cases apply specialized knowledge will continue to appear, they must allow override by a physician. As Rosenthal stated in the Sunday Times article, it is vital “that we have the wisdom to ignore [health information technology], as we would a GPS device, when it leads us in the wrong direction—or nowhere at all.” When it comes to anesthesia, there will always be a need for the experienced and wise anesthesiologist.

With best wishes,

Tony Mira
President and CEO