Machine learning in perioperative medicine: a systematic review

Artificial intelligence in regional anesthesia

PURPOSE OF REVIEW

Artificial intelligence (AI) is having an increasing impact on healthcare. In ultrasound-guided regional anesthesia (UGRA), commercially available devices exist that augment traditional grayscale ultrasound imaging by highlighting key sono-anatomical structures in real-time. We review the latest evidence supporting this emerging technology and consider the opportunities and challenges to its widespread deployment.

RECENT FINDINGS

The existing literature is limited and heterogenous, which impedes full appraisal of systems, comparison between devices, and informed adoption. AI-based devices promise to improve clinical practice and training in UGRA, though their impact on patient outcomes and provision of UGRA techniques is unclear at this early stage. Calls for standardization across both UGRA and AI are increasing, with greater clinical leadership required.

SUMMARY

Emerging AI applications in UGRA warrant further study due to an opaque and fragmented evidence base. Robust and consistent evaluation and reporting of algorithm performance, in a representative clinical context, will expedite discovery and appropriate deployment of AI in UGRA. A clinician-focused approach to the development, evaluation, and implementation of this exciting branch of AI has huge potential to advance the human art of regional anesthesia.

International expert consensus on the current status and future prospects of artificial intelligence in metabolic and bariatric surgery

Artificial intelligence (AI) is transforming the landscape of medicine, including surgical science and practice. The evolution of AI from rule-based systems to advanced machine learning and deep learning algorithms has opened new avenues for its application in metabolic and bariatric surgery (MBS). AI has the potential to enhance various aspects of MBS, including education and training, decision-making, procedure planning, cost and time efficiency, optimization of surgical techniques, outcome and complication prediction, patient education, and access to care. However, concerns persist regarding the reliability of AI-generated decisions and associated ethical considerations. This study aims to establish a consensus on the role of AI in MBS using a modified Delphi method. A panel of 68 leading metabolic and bariatric surgeons from 35 countries participated in this consensus-building process, providing expert insights into the integration of AI in MBS. Of the 28 statements evaluated, a consensus of at least 70% was achieved for all, with 25 statements reaching consensus in the first round and the remaining three in the second round. Experts agreed that AI has the potential to enhance the evaluation of surgical skills in MBS by providing objective, detailed assessments, enabling personalized feedback, and accelerating the learning curve. Most experts also recognized AI's role in identifying qualified candidates for MBS referrals, helping patient and procedure selection, and addressing specific clinical questions. However, concerns were raised about the potential overreliance on AI-generated recommendations. The consensus emphasized the need for ethical guidelines governing AI use and the inclusion of AI's role in decision-making within the patient consent process. Furthermore, the results suggest that AI education should become an essential component of future surgical training. Advancements in AI-driven robotics and AI-integrated genomic applications were also identified as promising developments that could significantly shape the future of MBS.

A novel generative multi-task representation learning approach for predicting postoperative complications in cardiac surgery patients

OBJECTIVE

Early detection of surgical complications allows for timely therapy and proactive risk mitigation. Machine learning (ML) can be leveraged to identify and predict patient risks for postoperative complications. We developed and validated the effectiveness of predicting postoperative complications using a novel surgical Variational Autoencoder (surgVAE) that uncovers intrinsic patterns via cross-task and cross-cohort presentation learning.

MATERIALS AND METHODS

This retrospective cohort study used data from the electronic health records of adult surgical patients over 4 years (2018-2021). Six key postoperative complications for cardiac surgery were assessed: acute kidney injury, atrial fibrillation, cardiac arrest, deep vein thrombosis or pulmonary embolism, blood transfusion, and other intraoperative cardiac events. We compared surgVAE's prediction performance against widely-used ML models and advanced representation learning and generative models under 5-fold cross-validation.

RESULTS

89 246 surgeries (49% male, median [IQR] age: 57 [45-69]) were included, with 6502 in the targeted cardiac surgery cohort (61% male, median [IQR] age: 60 [53-70]). surgVAE demonstrated generally superior performance over existing ML solutions across postoperative complications of cardiac surgery patients, achieving macro-averaged AUPRC of 0.409 and macro-averaged AUROC of 0.831, which were 3.4% and 3.7% higher, respectively, than the best alternative method (by AUPRC scores). Model interpretation using Integrated Gradients highlighted key risk factors based on preoperative variable importance.

DISCUSSION AND CONCLUSION

Our advanced representation learning framework surgVAE showed excellent discriminatory performance for predicting postoperative complications and addressing the challenges of data complexity, small cohort sizes, and low-frequency positive events. surgVAE enables data-driven predictions of patient risks and prognosis while enhancing the interpretability of patient risk profiles.

Artificial intelligence and machine learning in knee arthroplasty

BACKGROUND

Artificial intelligence (AI) and its subset, machine learning (ML), have significantly impacted clinical medicine, particularly in knee arthroplasty (KA). These technologies utilize algorithms for tasks such as predictive analytics and image recognition, improving preoperative planning, intraoperative navigation, and postoperative complication anticipation. This systematic review presents AI-driven tools' clinical implications in total and unicompartmental KA, focusing on enhancing patient outcomes and operational efficiency.

METHODS

A systematic search was conducted across multiple databases including Cochrane Central Register of Controlled Trials, Embase, OVID Medline, PubMed, and Web of Science, following the PRISMA guidelines for studies published in the English language till March 2024. Inclusion criteria targeted adult human models without geographical restrictions, specifically related to total or unicompartmental KA.

RESULTS

A total of 153 relevant studies were identified, covering various aspects of ML application for KA. Topics of studies included imaging modalities (n = 28), postoperative primary KA complications (n = 26), inpatient status (length of stay, readmissions, and cost) (n = 24), implant configuration (n = 14), revision (n = 12), patient-reported outcome measures (PROMs) (n = 11), function (n = 11), procedural communication (n = 8), total knee arthroplasty/unicompartmental knee arthroplasty prediction (n = 6), outpatient status (n = 4), perioperative efficiency (n = 4), patient satisfaction (n = 3), opioid usage (n = 3). A total of 66 ML models were described, with 48.7% of studies using multiple approaches.

CONCLUSION

This review assesses ML applications in knee arthroplasty, highlighting their potential to improve patient outcomes. While current algorithms and AI show promise, our findings suggest areas for enhancement in predictive performance before widespread clinical adoption.

The foundational capabilities of large language models in predicting postoperative risks using clinical notes

Clinical notes recorded during a patient's perioperative journey holds immense informational value. Advances in large language models (LLMs) offer opportunities for bridging this gap. Using 84,875 preoperative notes and its associated surgical cases from 2018 to 2021, we examine the performance of LLMs in predicting six postoperative risks using various fine-tuning strategies. Pretrained LLMs outperformed traditional word embeddings by an absolute AUROC of 38.3% and AUPRC of 33.2%. Self-supervised fine-tuning further improved performance by 3.2% and 1.5%. Incorporating labels into training further increased AUROC by 1.8% and AUPRC by 2%. The highest performance was achieved with a unified foundation model, with improvements of 3.6% for AUROC and 2.6% for AUPRC compared to self-supervision, highlighting the foundational capabilities of LLMs in predicting postoperative risks, which could be potentially beneficial when deployed for perioperative care.

Data Science and Geriatric Anesthesia Research: Opportunity and Challenges

With an increase in geriatric population undergoing surgical procedures, research focused on enhancing their perioperative outcomes is of paramount importance. Currently, most of the evidence-based medicine protocols are driven by studies concentrating on adults encompassing all adult age groups. Given the alterations in physiology with aging, geriatric patients respond differently to anesthetics and, therefore, require specific research initiatives to further expound on the same. Large databases and the development of sophisticated analytic tools can provide meaningful insights into this. Here, we discuss a few research opportunities and challenges that data scientists face when focusing on geriatric perioperative research.

The future outlook for data in orthopedic surgery: A new era of real-time innovation

The orthopedic field is on the brink of a significant transformation-a shift from retrospective analysis to real-time decision-making fueled by data. The dependence on historical trends or long-term studies is yielding to an era where data flows dynamically, allowing medical professionals to adjust protocols instantly. This isn't just an evolution; it's a revolution. Data is no longer a passive observer of outcomes-it's an active participant in shaping them.Imagine a future where wearable devices, artificial intelligence (AI) algorithms, and predictive analytics come together to guide surgeons in real time. For example, wearables monitor vital signs during surgery and oversee rehabilitation while AI analyzes data to predict complications. Postoperative protocols adapt to individual recovery journeys, not averages. Complication risks are flagged preemptively, and treatment plans evolve with patient progress. This shift empowers orthopedic professionals to respond and anticipate, creating a level of care precision that was once unimaginable.What if we viewed data not merely as a tool but as collaborators? With AI and machine learning, the surgical suite of tomorrow transforms into ecosystems where data communicates directly providing insights, suggesting strategies, and enhancing outcomes. This collaborative approach encourages our conventional medical mindset to prioritize adaptability and individualization.The provocative truth is that the game-changer in orthopedics isn't a new implant design or surgical technique-it's the mindset shift to trust real-time data as the foundation of every decision. Orthopedics is no longer about perfecting procedures but refining protocols for every patient consistently.As we race toward the future, equitable access becomes crucial. As William Gibson noted, "The future is already here - it's just not very evenly distributed." We must ensure these breakthroughs reach everyone, bridging the gap between potential and practice. The future of orthopedics isn't just a vision - it's a promise, and it's time to deliver.

Protocol for the impact of machine learning-based clinician decision support algorithims in perioperative care (IMAGINATIVE) in Singapore general hospital : a large prospective randomised controlled trial

INTRODUCTION

As surgical accessibility improves, the incidence of postoperative complications is expected to rise. The implementation of a precise and objective risk stratification tool holds the potential to mitigate these complications by early identification of high-risk patients. Moreover, it could address the escalating costs from resource misallocation. In Singapore General Hospital (SGH), we introduced the Combined Assessment of Risk Encountered in Surgery-Machine Learning (CARES-ML) in June 2023, focusing on predicting 30-day postoperative mortality and the need for post-surgery intensive care unit (ICU) stays. The IMAGINATIVE Trial aims to evaluate the efficacy of such systems in a large academic medical centre.

METHODS AND ANALYSIS

This study adopts type 1 effectiveness-implementation study design within a randomised controlled trial framework. Patients will be randomly assigned in a 1:1 ratio to either the CARES-guided group (unblinded to risk level) or the unguided group (blinded to the risk level). A total of 9200 patients will be enrolled in the study, with the inclusion criteria encompassing individuals aged 21-100 years old undergoing elective surgeries except for neurology and cardiology surgeries at SGH. The primary outcome is to evaluate the effectiveness of the Machine Learning Clinical Decision Support (ML-CDS) algorithm in improving perioperative mortality rates when integrated into the clinical workflow.

ETHICS AND DISSEMINATION

The study has been approved by the SingHealth Centralised Institutional Review Board (CIRB Ref: 2023:2114) and is registered on ClinicalTrials.gov (trial number: NCT05809232). All patients will sign an informed consent form before recruitment and translators will be made available to non-English-speaking participants. This study is funded by the National Medical Research Council, Singapore (HCSAINV22jul-0002) and the findings will be published in peer-reviewed journals and presented at academic conferences.

TRIAL REGISTRATION NUMBER

NCT05809232.

Machine learning-augmented interventions in perioperative care: a systematic review and meta-analysis

BACKGROUND

We lack evidence on the cumulative effectiveness of machine learning (ML)-driven interventions in perioperative settings. Therefore, we conducted a systematic review to appraise the evidence on the impact of ML-driven interventions on perioperative outcomes.

METHODS

Ovid MEDLINE, CINAHL, Embase, Scopus, PubMed, and ClinicalTrials.gov were searched to identify randomised controlled trials (RCTs) evaluating the effectiveness of ML-driven interventions in surgical inpatient populations. The review was registered with PROSPERO (CRD42023433163) and conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Meta-analysis was conducted for outcomes with two or more studies using a random-effects model, and vote counting was conducted for other outcomes.

RESULTS

Among 13 included RCTs, three types of ML-driven interventions were evaluated: Hypotension Prediction Index (HPI) (n=5), Nociception Level Index (NoL) (n=7), and a scheduling system (n=1). Compared with the standard care, HPI led to a significant decrease in absolute hypotension (n=421, P=0.003, I2=75%) and relative hypotension (n=208, P<0.0001, I2=0%); NoL led to significantly lower mean pain scores in the post-anaesthesia care unit (PACU) (n=191, P=0.004, I2=19%). NoL showed no significant impact on intraoperative opioid consumption (n=339, P=0.31, I2=92%) or PACU opioid consumption (n=339, P=0.11, I2=0%). No significant difference in hospital length of stay (n=361, P=0.81, I2=0%) and PACU stay (n=267, P=0.44, I2=0) was found between HPI and NoL.

CONCLUSIONS

HPI decreased the duration of intraoperative hypotension, and NoL decreased postoperative pain scores, but no significant impact on other clinical outcomes was found. We highlight the need to address both methodological and clinical practice gaps to ensure the successful future implementation of ML-driven interventions.

SYSTEMATIC REVIEW PROTOCOL

CRD42023433163 (PROSPERO).

Feature selection integrating Shapley values and mutual information in reinforcement learning: An application in the prediction of post-operative outcomes in patients with end-stage renal disease

BACKGROUND

In predicting post-operative outcomes for patients with end-stage renal disease, our study faced challenges related to class imbalance and a high-dimensional feature space. Therefore, with a focus on overcoming class imbalance and improving interpretability, we propose a novel feature selection approach using multi-agent reinforcement learning.

METHODS

We proposed a multi-agent feature selection model based on a comprehensive reward function that combines classification model performance, Shapley additive explanations values, and the mutual information. The definition of rewards in reinforcement learning is crucial for model convergence and performance improvement. Initially, we set a deterministic reward based on the mutual information between variables and the target class, selecting variables that are highly dependent on the class, thus accelerating convergence. We then prioritized variables that influence the minority class on a sample basis and introduced a dynamic reward distribution strategy using Shapley additive explanations values to improve interpretability and solve the class imbalance problem.

RESULTS

Involving the integration of electronic medical records, anesthesia records, operating room vital signs, and pre-operative anesthesia evaluations, our approach effectively mitigated class imbalance and demonstrated superior performance in ablation analysis. Our model achieved a 16% increase in the minority class F1 score and an 8.2% increase in the overall F1 score compared to the baseline model without feature selection.

CONCLUSION

This study contributes important research findings that show that the multi-agent-based feature selection method can be a promising approach for solving the class imbalance problem.

The use of artificial intelligence in reconstructive surgery for head and neck cancer: a systematic review

OBJECTIVES

The popularity of artificial intelligence (AI) in head and neck cancer (HNC) management is increasing, but postoperative complications remain prevalent and are the main factor that impact prognosis after surgery. Hence, recent studies aim to assess new AI models to evaluate their ability to predict free flap complications more effectively than traditional algorithms. This systematic review aims to summarize current evidence on the utilization of AI models to predict complications following reconstructive surgery for HNC.

METHODS

A combination of MeSH terms and keywords was used to cover the following three subjects: "HNC," "artificial intelligence," and "free flap or reconstructive surgery." The electronic literature search was performed in three relevant databases: Medline (Ovid), Embase (Ovid), and Cochrane. Quality appraisal of the included study was conducted using the TRIPOD Statement.

RESULTS

The review included a total of 5 manuscripts (n = 5) for a total of 7524 patients. Across studies, the highest area under the receiver operating characteristic (AUROC) value achieved was 0.824 by the Auto-WEKA model. However, only 20% of reported AUROCs exceeded 0.70. One study concluded that most AI models were comparable or inferior in performance to conventional logistic regression. The highest predictors of complications were flap type, smoking status, tumour location, and age.

DISCUSSION

Some models showed promising results. Predictors identified across studies were different than those found in existing literature, showing the added value of AI models. However, the algorithms showed inconsistent results, underlying the need for better-powered studies with larger databases before clinical implementation.

Machine learning and preoperative risk prediction: the machines are coming

Preoperative risk prediction is an important component of perioperative medicine. Machine learning is a powerful tool that could lead to increasingly complex risk prediction models with improved predictive performance. Careful consideration is required to guide the machine learning approach to ensure appropriate decisions are made with regard to what we are trying to predict, when we are trying to predict it, and what we seek to do with the results.

Intraoperative Hypotension Prediction: Current Methods, Controversies, and Research Outlook

Intraoperative hypotension prediction has been increasingly emphasized due to its potential clinical value in reducing organ injury and the broad availability of large-scale patient datasets and powerful machine learning tools. Hypotension prediction methods can mitigate low blood pressure exposure time. However, they have yet to be convincingly demonstrated to improve objective outcomes; furthermore, they have recently become controversial. This review presents the current state of intraoperative hypotension prediction and makes recommendations on future research. We begin by overviewing the current hypotension prediction methods, which generally rely on the prevailing mean arterial pressure as one of the important input variables and typically show good sensitivity and specificity but low positive predictive value in forecasting near-term acute hypotensive events. We make specific suggestions on improving the definition of acute hypotensive events and evaluating hypotension prediction methods, along with general proposals on extending the methods to predict reduced blood flow and treatment effects. We present a start of a risk-benefit analysis of hypotension prediction methods in clinical practice. We conclude by coalescing this analysis with the current evidence to offer an outlook on prediction methods for intraoperative hypotension. A shift in research toward tailoring hypotension prediction methods to individual patients and pursuing methods to predict appropriate treatment in response to hypotension appear most promising to improve outcomes.

Patient selection for nonoperating room anesthesia

PURPOSE OF REVIEW

Given the rapid growth of nonoperating room anesthesia (NORA) in recent years, it is essential to review its unique challenges as well as strategies for patient selection and care optimization.

RECENT FINDINGS

Recent investigations have uncovered an increasing prevalence of older and higher ASA physical status patients in NORA settings. Although closed claim data regarding patient injury demonstrate a lower proportion of NORA cases resulting in a claim than traditional operating room cases, NORA cases have an increased risk of claim for death. Challenges within NORA include site-specific differences, limitations in ergonomic design, and increased stress among anesthesia providers. Several authors have thus proposed strategies focusing on standardizing processes, site-specific protocols, and ergonomic improvements to mitigate risks.

SUMMARY

Considering the unique challenges of NORA settings, meticulous patient selection, risk stratification, and preoperative optimization are crucial. Embracing data-driven strategies and leveraging technological innovations (such as artificial intelligence) is imperative to refine quality control methods in targeted areas. Collaborative efforts led by anesthesia providers will ensure personalized, well tolerated, and improved patient outcomes across all phases of NORA care.

Machine learning in the prediction and detection of new-onset atrial fibrillation in ICU: a systematic review

Atrial fibrillation (AF) stands as the predominant arrhythmia observed in ICU patients. Nevertheless, the absence of a swift and precise method for prediction and detection poses a challenge. This study aims to provide a comprehensive literature review on the application of machine learning (ML) algorithms for predicting and detecting new-onset atrial fibrillation (NOAF) in ICU-treated patients. Following the PRISMA recommendations, this systematic review outlines ML models employed in the prediction and detection of NOAF in ICU patients and compares the ML-based approach with clinical-based methods. Inclusion criteria comprised randomized controlled trials (RCTs), observational studies, cohort studies, and case-control studies. A total of five articles published between November 2020 and April 2023 were identified and reviewed to extract the algorithms and performance metrics. Reviewed studies sourced 108,724 ICU admission records form databases, e.g., MIMIC. Eight prediction and detection methods were examined. Notably, CatBoost exhibited superior performance in NOAF prediction, while the support vector machine excelled in NOAF detection. Machine learning algorithms emerge as promising tools for predicting and detecting NOAF in ICU patients. The incorporation of these algorithms in clinical practice has the potential to enhance decision-making and the overall management of NOAF in ICU settings.

Application of Machine Learning in Predicting Perioperative Outcomes in Patients with Cancer: A Narrative Review for Clinicians

This narrative review explores the utilization of machine learning (ML) and artificial intelligence (AI) models to enhance perioperative cancer care. ML and AI models offer significant potential to improve perioperative cancer care by predicting outcomes and supporting clinical decision-making. Tailored for perioperative professionals including anesthesiologists, surgeons, critical care physicians, nurse anesthetists, and perioperative nurses, this review provides a comprehensive framework for the integration of ML and AI models to enhance patient care delivery throughout the perioperative continuum.

Artificial intelligence in surgery

Artificial intelligence (AI) is rapidly emerging in healthcare, yet applications in surgery remain relatively nascent. Here we review the integration of AI in the field of surgery, centering our discussion on multifaceted improvements in surgical care in the preoperative, intraoperative and postoperative space. The emergence of foundation model architectures, wearable technologies and improving surgical data infrastructures is enabling rapid advances in AI interventions and utility. We discuss how maturing AI methods hold the potential to improve patient outcomes, facilitate surgical education and optimize surgical care. We review the current applications of deep learning approaches and outline a vision for future advances through multimodal foundation models.

Prognostic Biomarkers and AKI: Potential to Enhance the Identification of Post-Operative Patients at Risk of Loss of Renal Function

Acute kidney injury (AKI) is a common complication after surgery and the more complex the surgery, the greater the risk. During surgery, patients are exposed to a combination of factors all of which are associated with the development of AKI. These include hypotension and hypovolaemia, sepsis, systemic inflammation, the use of nephrotoxic agents, tissue injury, the infusion of blood or blood products, ischaemia, oxidative stress and reperfusion injury. Given the risks of AKI, it would seem logical to conclude that early identification of patients at risk of AKI would translate into benefit. The conventional markers of AKI, namely serum creatinine and urine output are the mainstay of defining chronic kidney disease but are less suited to the acute phase. Such concerns are compounded in surgical patients given they often have significantly reduced mobility, suboptimal levels of nutrition and reduced muscle bulk. Many patients may also have misleadingly low serum creatinine and high urine output due to aggressive fluid resuscitation, particularly in intensive care units. Over the last two decades, considerable information has accrued with regard to the performance of what was termed "novel" biomarkers of AKI, and here, we discuss the most examined molecules and performance in surgical settings. We also discuss the application of biomarkers to guide patients' postoperative care.

Artificial intelligence in academic writing: a detailed examination

INTRODUCTION

As AI tools have become popular in academia, concerns about their impact on student originality and academic integrity have arisen.

METHODS

This quality improvement project examined first-year nurse anesthesiology students' use of AI for an academic writing assignment. Students generated, edited, and reflected on AI-produced content. Their work was analyzed for commonalities related to the perceived ease of use, accuracy, and overall impressions.

RESULTS

Students found AI tools easy to use with fast results, but reported concerns with inaccuracies, superficiality, and unreliable citations and formatting. Despite these issues, some saw potential in AI for brainstorming and proofreading.

IMPLICATIONS FOR INTERNATIONAL AUDIENCE

Clear guidelines are necessary for AI use in academia. Further research should explore AI's long-term impact on academic writing and learning outcomes.

CONCLUSIONS

While AI tools offer speed and convenience, they currently lack the depth required for rigorous academic work.

How anesthesiology can deal with innovation and new technologies?

Innovation and new technologies have always impacted significantly the anesthesiology practice all along the perioperative course, as it is recognized as one of the most transformative medical specialties specifically regarding patient's safety. Beside a number of major changes in procedures, equipment, training, and organization that aggregated to establish a strong safety culture with effective practices, anesthesiology is also a stakeholder in disruptive innovation. The present review is not exhaustive and aims to provide an overview on how innovation could change and improve anesthesiology practices through some examples as telemedicine (TM), machine learning and artificial intelligence (AI). For example, postoperative complications can be accurately predicted by AI from automated real-time electronic health record data, matching physicians' predictive accuracy. Clinical workflow could be facilitated and accelerated with mobile devices and applications, assuming that these tools should remain at the service of patients and care providers. Care providers and patients connections have improved, thanks to these digital and innovative transformations, without replacing existing relationships between them. It also should give time back to physicians and nurses to better spend it in the perioperative care, and to provide "personalized" medicine keeping a high level of standard of care.

Machine Learning Algorithm to Perform the American Society of Anesthesiologists Physical Status Classification

OBJECTIVE

The American Society of Anesthesiologists (ASA) Physical Status (PS) Classification System defines perioperative patient scores ranging from 1 to 6 (healthy to brain dead, respectively). The scoring is performed and used by physician anesthesiologists and providers to classify surgical patients based on co-morbidities and various clinical characteristics. There is potentially a variability in scoring stemming from individual biases. The biases impact the prediction of operating times, length of stay in the hospital, anesthetic management, and billing. This study's purpose was to develop an automated system to achieve reproducible scoring.

METHODS

A machine learning (ML) model was trained on already assigned ASA PS scores of 12,064 patients. The ML algorithm was automatically selected by Wolfram Mathematica (Wolfram Research, Champaign, IL) and tested with retrospective records not used in training. Manual scoring was performed by the anesthesiologist as part of the standard preoperative evaluation. Intraclass correlation coefficient (ICC) in R (version 4.2.2; R Development Core Team, Vienna, Austria) was calculated to assess the consistency of scoring.

RESULTS

An ML model was trained on the data corresponding to 12,064 patients. Logistic regression was chosen automatically, with an accuracy of 70.3±1.0% against the training dataset. The accuracy against 1,999 patients (the test dataset) was 69.6±1.0%. The ICC for the comparison between ML and the anesthesiologists' ASA PS scores was greater than 0.4 ("fair to good").

CONCLUSIONS

We have shown the feasibility of applying ML to assess the ASA PS score within an oncology patient population. Though our accuracy was not very good, we feel that, as more data are mined, a valid foundation for refinement to ML will emerge.

Data Science and Geriatric Anesthesia Research: Opportunity and Challenges

With an increase in geriatric population undergoing surgical procedures, research focused on enhancing their perioperative outcomes is of paramount importance. Currently, most of the evidence-based medicine protocols are driven by studies concentrating on adults encompassing all adult age groups. Given the alterations in physiology with aging, geriatric patients respond differently to anesthetics and, therefore, require specific research initiatives to further expound on the same. Large databases and the development of sophisticated analytic tools can provide meaningful insights into this. Here, we discuss a few research opportunities and challenges that data scientists face when focusing on geriatric perioperative research.

A Comprehensive Analysis and Review of Artificial Intelligence in Anaesthesia

In the field of anaesthesia, artificial intelligence (AI) has become a game-changing technology. Applications of AI include keeping records, monitoring patients, calculating and administering drugs, and carrying out mechanical procedures. This article explores the current uses, challenges, and prospective applications of AI in anaesthesia practices. This review discusses AI-supported systems like anaesthesia information management systems (AIMS), mechanical robots for carrying out procedures, and pharmacological models for drug delivery. AIMS has helped in automated record-keeping, predicting bad events, and monitoring the vital signs of the patient. Their application has a vital role in improving the efficacy of anaesthesia management and patient safety. The application of AI in anaesthesia comes with its own unique difficulties. Noteworthy obstacles include issues with data quantity and quality, technical limitations, and moral and legal dilemmas. The key to overcoming these barriers is to set guidelines for the ethical use of AI in healthcare, improve the reliability and comprehension of AI systems, and certify the health data precision and security. AI has very bright potential. Exciting future directions include developments in AI and machine learning thus development of new applications, and the possible enhancement in training and education. Potential research areas include the application of AI to chronic disease management, pain management, and the reinforcement of anaesthesiologists' education. AI could be used to design authentic lifelike training simulations and individualized student feedback systems, hence transforming anaesthesia education and training methodology. For this review, we conducted a PubMed, Google Scholar, and Cochrane Database search in 2022-2023 and retrieved articles on AI and its uses in anaesthesia. Recommendations for future research and development include strengthening the safety and reliability of health data, building a better understanding of AI systems, and looking into new areas of use. The power of AI can be used to innovate anaesthesia practices by concentrating on these areas.

Artificial intelligence and "the Art of Kintsugi" in Anesthesiology: ten influential papers for clinical users

Artificial intelligence refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. In the present review we chose ten influential papers from the last five years and through Kintsugi, shed the light on recent evolution of artificial intelligence in anesthesiology. A comprehensive search in in Medline, Embase, Web of Science and Scopus databases was conducted. Each author searched the databases independently and created a list of six articles that influenced their clinical practice during this period, with a focus on their area of competence. During a subsequent step, each researcher presented his own list and most cited papers were selected to create the final collection of ten articles. In recent years purely methodological works with a cryptic technology (black-box) represented by the intact and static vessel, translated to a "modern artificial intelligence" in clinical practice and comprehensibility (glass-box). The purposes of this review are to explore the ten most cited papers about artificial intelligence in anesthesiology and to understand how and when it should be integrated in clinical practice.

Three Logistic Predictive Models for the Prediction of Mortality and Major Pulmonary Complications after Cardiac Surgery

Background and Objectives: Pulmonary complications are a leading cause of morbidity after cardiac surgery. The aim of this study was to develop models to predict postoperative lung dysfunction and mortality. Materials and Methods: This was a single-center, observational, retrospective study. We retrospectively analyzed the data of 11,285 adult patients who underwent all types of cardiac surgery from 2003 to 2015. We developed logistic predictive models for in-hospital mortality, postoperative pulmonary complications occurring in the intensive care unit, and postoperative non-invasive mechanical ventilation when clinically indicated. Results: In the "preoperative model" predictors for mortality were advanced age (p < 0.001), New York Heart Association (NYHA) class (p < 0.001) and emergent surgery (p = 0.036); predictors for non-invasive mechanical ventilation were advanced age (p < 0.001), low ejection fraction (p = 0.023), higher body mass index (p < 0.001) and preoperative renal failure (p = 0.043); predictors for postoperative pulmonary complications were preoperative chronic obstructive pulmonary disease (p = 0.007), preoperative kidney injury (p < 0.001) and NYHA class (p = 0.033). In the "surgery model" predictors for mortality were intraoperative inotropes (p = 0.003) and intraoperative intra-aortic balloon pump (p < 0.001), which also predicted the incidence of postoperative pulmonary complications. There were no specific variables in the surgery model predicting the use of non-invasive mechanical ventilation. In the "intensive care unit model", predictors for mortality were postoperative kidney injury (p < 0.001), tracheostomy (p < 0.001), inotropes (p = 0.029) and PaO2/FiO2 ratio at discharge (p = 0.028); predictors for non-invasive mechanical ventilation were kidney injury (p < 0.001), inotropes (p < 0.001), blood transfusions (p < 0.001) and PaO2/FiO2 ratio at the discharge (p < 0.001). Conclusions: In this retrospective study, we identified the preoperative, intraoperative and postoperative characteristics associated with mortality and complications following cardiac surgery.

Enabling personalized perioperative risk prediction by using a machine-learning model based on preoperative data

Preoperative risk assessment is essential for shared decision-making and adequate perioperative care. Common scores provide limited predictive quality and lack personalized information. The aim of this study was to create an interpretable machine-learning-based model to assess the patient's individual risk of postoperative mortality based on preoperative data to allow analysis of personal risk factors. After ethical approval, a model for prediction of postoperative in-hospital mortality based on preoperative data of 66,846 patients undergoing elective non-cardiac surgery between June 2014 and March 2020 was created with extreme gradient boosting. Model performance and the most relevant parameters were shown using receiver operating characteristic (ROC-) and precision-recall (PR-) curves and importance plots. Individual risks of index patients were presented in waterfall diagrams. The model included 201 features and showed good predictive abilities with an area under receiver operating characteristic (AUROC) curve of 0.95 and an area under precision-recall curve (AUPRC) of 0.109. The feature with the highest information gain was the preoperative order for red packed cell concentrates followed by age and c-reactive protein. Individual risk factors could be identified on patient level. We created a highly accurate and interpretable machine learning model to preoperatively predict the risk of postoperative in-hospital mortality. The algorithm can be used to identify factors susceptible to preoperative optimization measures and to identify risk factors influencing individual patient risk.

Peripheral Intravenous Waveform Analysis Responsiveness to Subclinical Hemorrhage in a Rat Model

BACKGROUND

Early detection and quantification of perioperative hemorrhage remains challenging. Peripheral intravenous waveform analysis (PIVA) is a novel method that uses a standard intravenous catheter to detect interval hemorrhage. We hypothesize that subclinical blood loss of 2% of the estimated blood volume (EBV) in a rat model of hemorrhage is associated with significant changes in PIVA. Secondarily, we will compare PIVA association with volume loss to other static, invasive, and dynamic markers.

METHODS

Eleven male Sprague Dawley rats were anesthetized and mechanically ventilated. A total of 20% of the EBV was removed over ten 5 minute-intervals. The peripheral intravenous pressure waveform was continuously transduced via a 22-G angiocatheter in the saphenous vein and analyzed using MATLAB. Mean arterial pressure (MAP) and central venous pressure (CVP) were continuously monitored. Cardiac output (CO), right ventricular diameter (RVd), and left ventricular end-diastolic area (LVEDA) were evaluated via transthoracic echocardiogram using the short axis left ventricular view. Dynamic markers such as pulse pressure variation (PPV) were calculated from the arterial waveform. The primary outcome was change in the first fundamental frequency (F1) of the venous waveform, which was assessed using analysis of variance (ANOVA). Mean F1 at each blood loss interval was compared to the mean at the subsequent interval. Additionally, the strength of the association between blood loss and F1 and each other marker was quantified using the marginal R2 in a linear mixed-effects model.

RESULTS

PIVA derived mean F1 decreased significantly after hemorrhage of only 2% of the EBV, from 0.17 to 0.11 mm Hg, P = .001, 95% confidence interval (CI) of difference in means 0.02 to 0.10, and decreased significantly from the prior hemorrhage interval at 4%, 6%, 8%, 10%, and 12%. Log F1 demonstrated a marginal R2 value of 0.57 (95% CI 0.40-0.73), followed by PPV 0.41 (0.28-0.56) and CO 0.39 (0.26-0.58). MAP, LVEDA, and systolic pressure variation displayed R2 values of 0.31, and the remaining predictors had R2 values ≤0.2. The difference in log F1 R2 was not significant when compared to PPV 0.16 (95% CI -0.07 to 0.38), CO 0.18 (-0.06 to 0.04), or MAP 0.25 (-0.01 to 0.49) but was significant for the remaining markers.

CONCLUSIONS

The mean F1 amplitude of PIVA was significantly associated with subclinical blood loss and most strongly associated with blood volume among the markers considered. This study demonstrates feasibility of a minimally invasive, low-cost method for monitoring perioperative blood loss.

Impact of the Covid-19 pandemic on the performance of machine learning algorithms for predicting perioperative mortality

BACKGROUND

Machine-learning models are susceptible to external influences which can result in performance deterioration. The aim of our study was to elucidate the impact of a sudden shift in covariates, like the one caused by the Covid-19 pandemic, on model performance.

METHODS

After ethical approval and registration in Clinical Trials (NCT04092933, initial release 17/09/2019), we developed different models for the prediction of perioperative mortality based on preoperative data: one for the pre-pandemic data period until March 2020, one including data before the pandemic and from the first wave until May 2020, and one that covers the complete period before and during the pandemic until October 2021. We applied XGBoost as well as a Deep Learning neural network (DL). Performance metrics of each model during the different pandemic phases were determined, and XGBoost models were analysed for changes in feature importance.

RESULTS

XGBoost and DL provided similar performance on the pre-pandemic data with respect to area under receiver operating characteristic (AUROC, 0.951 vs. 0.942) and area under precision-recall curve (AUPR, 0.144 vs. 0.187). Validation in patient cohorts of the different pandemic waves showed high fluctuations in performance from both AUROC and AUPR for DL, whereas the XGBoost models seemed more stable. Change in variable frequencies with onset of the pandemic were visible in age, ASA score, and the higher proportion of emergency operations, among others. Age consistently showed the highest information gain. Models based on pre-pandemic data performed worse during the first pandemic wave (AUROC 0.914 for XGBoost and DL) whereas models augmented with data from the first wave lacked performance after the first wave (AUROC 0.907 for XGBoost and 0.747 for DL). The deterioration was also visible in AUPR, which worsened by over 50% in both XGBoost and DL in the first phase after re-training.

CONCLUSIONS

A sudden shift in data impacts model performance. Re-training the model with updated data may cause degradation in predictive accuracy if the changes are only transient. Too early re-training should therefore be avoided, and close model surveillance is necessary.

Extended-age Out-of-sample Validation of Risk Stratification Index 3.0 Models Using Commercial All-payer Claims

BACKGROUND

The authors previously reported a broad suite of individualized Risk Stratification Index 3.0 (Health Data Analytics Institute, Inc., USA) models for various meaningful outcomes in patients admitted to a hospital for medical or surgical reasons. The models used International Classification of Diseases, Tenth Revision, trajectories and were restricted to information available at hospital admission, including coding history in the previous year. The models were developed and validated in Medicare patients, mostly age 65 yr or older. The authors sought to determine how well their models predict utilization outcomes and adverse events in younger and healthier populations.

METHODS

The authors' analysis was based on All Payer Claims for surgical and medical hospital admissions from Utah and Oregon. Endpoints included unplanned hospital admissions, in-hospital mortality, acute kidney injury, sepsis, pneumonia, respiratory failure, and a composite of major cardiac complications. They prospectively applied previously developed Risk Stratification Index 3.0 models to the younger and healthier 2017 Utah and Oregon state populations and compared the results to their previous out-of-sample Medicare validation analysis.

RESULTS

In the Utah dataset, there were 55,109 All Payer Claims admissions across 40,710 patients. In the Oregon dataset, there were 21,213 admissions from 16,951 patients. Model performance on the two state datasets was similar or better than in Medicare patients, with an average area under the curve of 0.83 (0.71 to 0.91). Model calibration was reasonable with an R2 of 0.93 (0.84 to 0.97) for Utah and 0.85 (0.71 to 0.91) for Oregon. The mean sensitivity for the highest 5% risk population was 28% (17 to 44) for Utah and 37% (20 to 56) for Oregon.

CONCLUSIONS

Predictive analytical modeling based on administrative claims history provides individualized risk profiles at hospital admission that may help guide patient management. Similar predictive performance in Medicare and in younger and healthier populations indicates that Risk Stratification Index 3.0 models are valid across a broad range of adult hospital admissions.

EDITOR’S PERSPECTIVE

Directed acyclic graphs in perioperative observational research-A systematic review and critique against best practice recommendations

The Directed Acyclic Graph (DAG) is a graph representing causal pathways for informing the conduct of an observational study. The use of DAGs allows transparent communication of a causal model between researchers and can prevent over-adjustment biases when conducting causal inference, permitting greater confidence and transparency in reported causal estimates. In the era of 'big data' and increasing number of observational studies, the role of the DAG is becoming more important. Recent best-practice guidance for constructing a DAG with reference to the literature has been published in the 'Evidence synthesis for constructing DAGs' (ESC-DAG) protocol. We aimed to assess adherence to these principles for DAGs constructed within perioperative literature. Following registration on the International Prospective Register of Systematic Reviews (PROSPERO) and with adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting framework for systematic reviews, we searched the Excerpta Medica dataBASE (Embase), the Medical Literature Analysis and Retrieval System Online (MEDLINE) and Cochrane databases for perioperative observational research incorporating a DAG. Nineteen studies were included in the final synthesis. No studies demonstrated any evidence of following the mapping stage of the protocol. Fifteen (79%) fulfilled over half of the translation and integration one stages of the protocol. Adherence with one stage did not guarantee fulfilment of the other. Two studies (11%) undertook the integration two stage. Unmeasured variables were handled inconsistently between studies. Only three (16%) studies included unmeasured variables within their DAG and acknowledged their implication within the main text. Overall, DAGs that were constructed for use in perioperative observational literature did not consistently adhere to best practice, potentially limiting the benefits of subsequent causal inference. Further work should focus on exploring reasons for this deviation and increasing methodological transparency around DAG construction.

Artificial intelligence-enhanced care pathway planning and scheduling system: content validity assessment of required functionalities

BACKGROUND

Artificial intelligence (AI) and machine learning are transforming the optimization of clinical and patient workflows in healthcare. There is a need for research to specify clinical requirements for AI-enhanced care pathway planning and scheduling systems to improve human-AI interaction in machine learning applications. The aim of this study was to assess content validity and prioritize the most relevant functionalities of an AI-enhanced care pathway planning and scheduling system.

METHODS

A prospective content validity assessment was conducted in five university hospitals in three different countries using an electronic survey. The content of the survey was formed from clinical requirements, which were formulated into generic statements of required AI functionalities. The relevancy of each statement was evaluated using a content validity index. In addition, weighted ranking points were calculated to prioritize the most relevant functionalities of an AI-enhanced care pathway planning and scheduling system.

RESULTS

A total of 50 responses were received from clinical professionals from three European countries. An item-level content validity index ranged from 0.42 to 0.96. 45% of the generic statements were considered good. The highest ranked functionalities for an AI-enhanced care pathway planning and scheduling system were related to risk assessment, patient profiling, and resources. The highest ranked functionalities for the user interface were related to the explainability of machine learning models.

CONCLUSION

This study provided a comprehensive list of functionalities that can be used to design future AI-enhanced solutions and evaluate the designed solutions against requirements. The relevance of statements concerning the AI functionalities were considered somewhat relevant, which might be due to the low level or organizational readiness for AI in healthcare.