AI in Anesthesiology: Lowering the Risk of Surgical Complications and Adverse Outcomes

Our Take:

Artificial intelligence (AI) has the potential to transform the field of anesthesiology by improving patient safety and patient outcomes. It has the potential to contribute significantly to the advancement of healthcare practices, offering innovative solutions to address critical issues in the field of anesthesiology. Pilot AI projects have already been successfully utilized to classify high-risk surgical patients, continuously administer medications throughout surgery, and aid in clinical decision making. Further opportunities exist for AI to transform the field of anesthesiology by improving the safety of critical procedures like tracheal intubation and nerve blocks.

With the current shortage of anesthesiologists, a forecasted increase in the demand for surgical procedures, and an increasingly sick patient population, AI may allow anesthesiologists to provide safe high-quality care for these complex patients and decrease the risks associated with invasive procedures. Despite the promising outlook, the incorporation of AI into anesthesiology requires a deliberate and responsible strategy. The implementation of AI technology should be done carefully to navigate legal, ethical, and safety concerns.

Key Takeaways:

In 2020, workplace staffing shortages affected 35.1% of US anesthesiologists, a figure that rose to 78.4% in 2022 (Anesthesiology)
Experts project approximately a 100% increase in Americans aged 50 years and older with at least one chronic disease by 2050 leading to an increase in patient complexity and surgical demand (Frontiers in Public Health)
An automated machine learning model that analyzed a dataset of over 1M patient encounters effectively identified patients at high risk of postoperative adverse outcomes, showcasing the potential for AI to enhance risk prediction in surgical settings (Journal of the American Medical Association)
Anesthesiologists successfully used a closed-loop system to maintain blood pressure within 10% of a target range for >90% of the case time in patients undergoing abdominal surgery (Journal of Personalized Medicine)

The Problem:

Artificial intelligence has emerged as a transformative force in the field of anesthesiology, presenting numerous applications that could greatly improve patient care. From optimizing preoperative patient conditions to calculating the risk of adverse events during the perioperative period, AI may also soon perform routine procedures while augmenting closed-loop systems for automated medication and fluid delivery. Despite the many benefits of utilizing AI in the field of anesthesiology, multiple critical issues must be tackled before anesthesiologists can fully embrace and trust this technology.

First and foremost, the adaptability and efficiency of AI in anesthesiology may be impeded by safety considerations. Notably, Ethicon Inc’s Sedasys employed an AI model to administer propofol – an anesthetic and sedative medication – aiming to achieve mild to moderate sedation during gastrointestinal (GI) endoscopic procedures. This closed loop general anesthesia system maintained a continuous propofol infusion to sustain sedation throughout a GI procedure, automatically adjusting the infusion based on vital signs. Due to safety measures, the Sedasys machine was unable to increase the dose of propofol if the patient were to become “light” during a procedure – a significant limitation that decreased the adaptability of this technology. Additionally, Sedasys’ safety mechanism of administering fentanyl followed by a 3-minute waiting period before propofol administration compromised efficiency, particularly given the swift completion of most diagnostic upper GI procedures within 5 to 10 minutes.

Secondly, the integration of AI in anesthesiology raises medical legal concerns. Medicine, being a highly complex field with nuanced variables, relies on human judgment for clinical decision-making. The intricacies that human anesthesiologists consider, especially in novel clinical situations, may not be fully appreciated by AI systems. In the event of an anesthetic error during a surgery involving AI technology, determining responsibility becomes a complex matter — whether it lies with the anesthesiologist, the AI-developing company, or the hospital. Moreover, issues of patient consent arise, as individuals may not fully comprehend the implications, both positive and negative, of AI technologies and may harbor concerns regarding the security and privacy of their health data.

Lastly, a crucial challenge in the utilization of AI in anesthesia pertains to the quality of data it relies on for training and learning. While systems are generally trained on data from electronic health records, patient monitors, and anesthesia machines, these diverse data sets as well as methods of data acquisition, often subjectively documented by clinicians, pose challenges in ensuring the accuracy, consistency, and completeness of the database. Consequently, faulty training data can lead to AI systems making incorrect judgments, potentially resulting in adverse health outcomes. For instance, an AI system may erroneously flag a patient as high-risk before an operation, prompting invasive monitoring during a surgical procedure which introduces associated risks. All these concerns contribute to the hesitation among anesthesiologists to fully embrace AI technology. These are further compounded when practitioners' concerns about job security and the fear of “black box” AI systems making treatment recommendations are added into the clinician’s calculus.

The Backdrop:

A recent report from the American Society of Anesthesiologists reveals that almost 40% of anesthesiologists plan on early retirement, while about a quarter have already reduced or plan to reduce their working hours. In 2020, workplace staffing shortages affected 35.1% of US anesthesiologists, a figure that rose to 78.4% in 2022. Amid escalating production pressures from private equity firms, financial struggles in hospitals, and alarming rates of burnout among anesthesiologists, AI emerges as a strategic tool that can reduce the stress that anesthesiologists face by helping them provide high-quality care and improve patient safety.

Demographic shifts, such as increased life expectancy and the aging "baby boomer" generation, forecast a surge in demand for surgical procedures in the near future. For example as noted in “Retooling for an Aging America: Building the Health Care Workforce”, the authors note, "older patients use two-to-three times as many medical services as younger patients, and the number of people over age 65 will increase by almost 50%, just in the next 10 to 15 years alone”. A 2022 Frontiers in Public Health paper predicts a 99.5% increase in Americans aged 50 years and older with at least one chronic disease by 2050. As surgical volume and complexity rise, anesthesiologists face increasing obstacles in ensuring patient well-being throughout the entire surgical process. AI, through the use of risk calculators, can assist by identifying high-risk patients based on preoperative variables, allowing for optimized resource allocation and patient preparation. A noteworthy example comes from a 2023 Journal of the American Medical Association paper, where researchers employed an automated machine learning model, analyzing a dataset of over a million patient encounters. This model effectively identified patients at high risk of postoperative adverse outcomes, showcasing the potential for AI to enhance risk prediction and resource optimization in surgical settings. Another machine-learning model published in the journal Anesthesiology was able to successfully predict a significant risk factor for adverse perioperative outcomes: post-induction hypotension, defined as low blood pressure in the first 20 minutes after administering anesthetic medication.

Moreover, researchers from Japan published a 2021 paper in the Journal of Intensive Care describing the use of a deep-learning AI model capable of classifying intubation difficulty by analyzing face images of patients. Tracheal intubation, a critical step at the beginning of surgical cases, demands precision to avoid complications such as airway damage, bleeding, and prolonged deoxygenation. The AI model exhibited an 80.5% predictive value, providing anesthesiologists with valuable information to prepare advanced techniques and equipment ahead of “difficult airway” situations, significantly improving patient safety. This development aligns with the broader trend of integrating AI into robotic intubation systems, exemplified by machines like the Da Vinci surgical system and the Kepler intubation system, which show promise in automating and enhancing the safety of intubation procedures.

With further developments in AI, it is possible that artificial intelligence may one day be integrated into these robotic intubation systems allowing for safe and automated procedures to be performed. As shown by a 2021 paper published in the journal Clinical Anatomy, AI has even augmented ultrasound-guided regional anesthesia procedures by aiding in the identification of anatomical structures on ultrasound. In the study, 97.5% or more of the expert anesthesiologists agreed that the AI-assistant would assist in confirmation of anatomical structures on ultrasound for less experienced practitioners.

Another key issue that anesthesiologists face is a phenomenon called “alarm fatigue.” “Alarm fatigue” refers to an increase in a health provider’s response time or decrease in response rate to an electronic alarm alert from a medical or patient monitoring device due to the excessive frequency of alarms. This is especially concerning for anesthesiologists who hear many alarms for blood pressure, heart rate, heart rhythm, oxygen saturation, temperature, and more. Even during high-risk cardiac surgeries, 80% of alarms were deemed useless. In an article published in the Health Informatics Journal entitled “Machine learning in anesthesiology: Detecting adverse events in clinical practice,” the authors propose the possibility of AI systems that can be used to generate meaningful and reliable alarms which can mitigate “alarm fatigue.”

Lastly, closed-loop systems integrated with AI that can deliver medications to induce anesthesia in patients, show promising outcomes. AI closed-loop systems may eventually be used to control other factors throughout a surgical case as well – blood pressure, neuromuscular blockade, vent management, and pain control. For example, a 2022 paper published in the Journal of Personalized Medicine described the successful use of a closed-loop system to maintain blood pressure within 10% of a target range for >90% of the case time in patients undergoing abdominal surgery. Another 2020 article in the journal Anesthesiology found that closed-loop systems had a positive impact on delayed neurocognitive recovery and outperformed manual control by anesthesiologists in managing anesthetic medication, fluids, and ventilation variables.

Implications:

Artificial intelligence holds substantial promise for enhancing patient safety and outcomes in the field of anesthesiology. AI enables anesthesiologists to adopt a proactive approach by identifying high-risk surgical patients and optimizing patient preparation. During a surgical procedure, customized and targeted alarms can help reduce "alarm fatigue," and the integration of robotic-AI devices may improve the safety of procedures performed by anesthesiologists such as tracheal intubation and ultrasound-guided regional anesthesia. These measures collectively improve patient outcomes while avoiding adverse health events, paving the way for the future of anesthesiology.

As the demand for surgical procedures rises with an aging population and an increase in chronic diseases, anesthesiologists stand to benefit from AI to navigate the complexities of patient care and aid in clinical decision-making. AI may even play a role in helping to alleviate the aforementioned shortage of anesthesiologists anticipated over the next several years. For example, AI may enable anesthesiologists to supervise less credentialed but highly capable clinicians such as nurse anesthetists to broaden care all without compromising patient care quality. The integration of AI into the field of anesthesiology holds the potential to improve care and lower costs all while helping the evolution of healthcare practice.

Related Reading: