The Department of Anesthesiology, Perioperative and Pain Medicine utilizes applications of artificial intelligence, machine learning, and visual analytics techniques within the health care domain to optimize risk assessment and patient outcomes. Much of our work is done in partnership with The Charles Bronfman Institute for Personalized Medicine and operationalized in cooperation with the Mount Sinai Health System Clinical Data Science team. The Clinical Data Science team has implemented a real-time multi-modal streaming data science pipeline that is used to deploy operational machine learning models at the point of care, including in the operating room, at the patient bedside, and in the outpatient arena.  

The Clinical Data Science team recently completed a clinical trial investigating the potential for real-time machine-learning-generated alerts to predict clinical deterioration and reduce escalations in care. While the results showed that the real-time alerts did not reduce escalations, there was a significant correlation between alerts and increased escalations associated with decreased mortality and shorter hospital stays. The study, published in Critical Care Medicine received considerable attention and was accompanied by an editorial. While further research is needed, these findings are promising in terms of using machine learning to provide opportune care in a timely fashion.

Pulse oximetry is a standard for basic anesthetic monitoring. It is used to monitor a patient’s blood oxygen saturation levels (designated SpO2) and pulse rate. This information can be vital in unstable patients and those with chronic heart and lung disease, including congenital heart disease. Recent retrospective studies have demonstrated discrepancies between the SpO2 reading and the hemoglobin arterial oxygen saturation measured using a laboratory co-oximeter (designated SaO2 and considered the “gold standard”) in patients who self-identified as Black or Hispanic. These non-white patients were noted to be at an increased risk of having a large difference between their SpO2 and SaO2 readings, with the SpO2 overestimating the SaO2. This “occult hypoxemia” resulted in patients with reassuring SpO2 readings who actually had low SaO2 values being undertreated with oxygen supplementation, delayed institution of respiratory support, and worse outcomes. Our researchers are conducting a study to further address this issue by investigating the relationship between pulse oximeter accuracy and color spectrophotometry-measured skin pigmentation in patients with congenital heart disease. Color spectrophotometry is a specific method of measuring skin pigmentation using light and energy values. The goal of the study is to identify how skin pigment contributes to discrepancies between simultaneous SaO2 and SpO2 readings.

Cirrhosis and liver transplantation influence every organ system and present complex clinical challenges. While clinicians and researchers are seeking new ways to improve outcomes in liver transplant recipients, research in this area is difficult due to the rarity of the procedure, and the large variation in patient complexity. To address these limitations, our team contributes to a Multicenter Liver Transplant Database sponsored by the Society for the Advancement of Transplant Anesthesia. The database pools anesthesia-related data from 18 transplant centers. The team is currently evaluating a potential prediction index for successful tracheal extubation in the operating room, intraoperative risk factors for postoperative acute kidney injury, and evaluating the impact of machine perfusion of donated livers on intraoperative hemodynamics.

Additionally, our researchers are also taking part in a multi-center clinical trial investigating the use of Angiotensin-II in liver transplantation. Liver transplant recipients often lack a robust response to the usual vasoactive medications and remain hypotensive. The goal of this research is to determine if Angiotensin II, a medication most often used to treat septic shock, may be a useful adjunct in treating hypotension during liver transplantation and potentially decrease perioperative morbidity and mortality, including acute kidney injury.

Obstetrical anesthesia is a major focus of our research. Maternal morbidity and mortality rates in the United States are higher compared to other industrialized countries. Further, morbidity and mortality in Black women remain stubbornly higher than any other group. These outcomes need to be improved and the reasons for disparities elucidated.

Life-threatening post-partum hemorrhage is of particular interest. Improved detection and treatment are necessary. We have several on-going hemostasis and coagulation studies designed to improve diagnosis and guide management. We were an early adopter of viscoelastic testing and have published several investigations using these methods. We use blood conservation strategies and are examining the use of blood shed during vaginal delivery as a source for autologous transfusion, which may significantly change the way blood loss is managed in this population.

Within the field of clinical obstetrics, we are finding innovative ways to address postpartum pain and minimize side effects in our patients with epidurals. Our team recently used epidural morphine in vaginal delivery patients to reduce postpartum pain and use of oral opioids. The study found that a single 2-mg dose of epidural preservative-free morphine administered within one hour of vaginal delivery decreased pain and opioid use in the postpartum period. While these findings also showed there was an increased incidence of pruritus, this research has led to an additional analgesia option for post-partum patients.

Chronic pain is a significant public health problem, with a prevalence varying widely from 2 percent to as high as 64 percent, depending on the clinical populations studied. The presence of chronic pain can lead to a diminished quality of life and serious mental health issues for patients. Therapeutic options to treat chronic pain can include the use of chronic opioids with its associated risk for dependence, tolerance, addiction, and overdose. Our research seeks to reveal what specific characteristics may portend vulnerability to opioid abuse, and disability from back pain, which is a common cause of chronic pain. Specifically, we are investigating behavioral, cognitive, and genomic features using comprehensive phenotyping, informatics, digital health records, and whole genome sequencing. In the future, we can identify patients at high risk, special precautions or more attention to alternative treatment may be recommended.

The Department has a long and rich history of clinical research focused on perioperative outcomes and clinical informatics. We maintain a continuously updated data warehouse that currently contains comprehensive perioperative data for nearly 1.7 million anesthetics. This helps drive a comprehensive program of data-driven retrospective research, all conducted under a global Institutional Review Board protocol that streamlines and facilitates study initiation and approval.

We also have a strong history and active program in pragmatic clinical trials that aim to better predict and improve patient outcomes and understand the trajectory of recovery. Past trials include a prospective randomized study evaluating the effectiveness of alerts indicating a “double low” of low blood pressure and low bispectral index in reducing 90-day post-operative mortality. While mortality was not significantly lower in patients who triggered double-low alerts, prolonged cumulative double-low conditions were strongly associated with increased mortality. The Department is a participating site in the General Anesthetics in CAncer REsection Surgery (GA-CARES), a multicenter trial looking at the influence of anesthetic technique on long-term outcomes after cancer surgery. Results should be available for publication by 2027.

Another area of research is perioperative genomics and disparities among different races and ethnic groups. Our researchers have investigated the relationship between self-reported ancestry and intraoperative blood pressure response to phenylephrine. They found that white Americans have a significantly higher blood pressure response to phenylephrine than Black patients and identified two low frequency variants in these populations where patients show no significant blood pressure response. Illustrating disparities of care, our researchers demonstrated that cerebral oximetry monitoring devices appear to have no racial/ethnic bias, whereas pulse oximeters may. In a retrospective cohort study of more than 46,000 patients, we found that the rate of intraoperative occult hypoxemia as measured by pulse oximetry was significantly higher in Black and Hispanic patients than white patients. The Department continues to seek innovative ways to improve perioperative outcomes for all patients while understanding and minimizing the impact of disparate measurements and responses.