A Comprehensive Systematic Review and Meta-Analysis: Evaluating the Effectiveness and Integration Obstacles of Artificial Intelligence (AI) within Anesthesia Departments.

A Comprehensive Systematic Review and Meta-Analysis: Evaluating the Effectiveness and Integration Obstacles of Artificial Intelligence (AI) within Anesthesia Departments. | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Systematic Review A Comprehensive Systematic Review and Meta-Analysis: Evaluating the Effectiveness and Integration Obstacles of Artificial Intelligence (AI) within Anesthesia Departments. Hany A. Zaki, Eman E. Shaban, Nabil Shallik, Ahmed Shaban, Amira Shaban, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4599435/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Artificial intelligence (AI) is a multidisciplinary field focusing on expanding and generating intelligent computer algorithms to carry out simple to more complex tasks traditionally performed using human intelligence. In anesthesia, AI is rapidly becoming a transformative technology. However, its efficacy in anesthesia is still unknown. Therefore, the current study analyzed the efficacy of AI in anesthesia by studying two main applications of AI, i.e., predicting events related to anesthesia and assisting anesthesia-related procedures. Furthermore, this study explored some of the challenges of integrating AI in the anesthesia field. Methods PubMed, Google Scholar, IEEE Xplore, and Web of Science databases were thoroughly searched for articles relevant to the objective of the current study. The Comprehensive Meta-analysis software and STATA 16.0 were used for statistical analyses, while the Newcastle Ottawa Scale was used for quality evaluation. Results Twenty studies satisfying the eligibility criteria were used for review and analysis. A subgroup analysis showed that models incorporating machine learning algorithms were superior in predicting postinduction hypotension (AUROC: 0.93). ANN and SANN models also showed a good discriminatory capacity in predicting postinduction hypotension (AUROC: 0.82 and 0.80, respectively). Similarly, the subgroup analysis showed that ANN and GBM models had a good discriminatory capacity when predicting hypoxemia (AUROC: 0.8 and 0.81, respectively). Furthermore, SVM, ANN, and fuzzy logic models had a relatively good differentiation ability in predicting postoperative nausea and vomiting (AUROC: 0.93, 0.77, and 0.72, respectively). On the other hand, the subgroup analysis showed that robotically-assisted tracheal intubations were highly successful in both mannikins and humans (success rate: 98% and 92%, respectively). Similarly, robotically-assisted ultrasound-guided nerve blocks were highly successful in mannikins and humans (Success rate: 96% for humans and mannikins, respectively). Conclusion The current study suggests that AI is useful in predicting anesthesia-related events and automating procedures such as tracheal intubation and ultrasound-guided nerve block. However, there are multiple barriers hindering the integration of AI in anesthesia that need to be addressed. Anesthesiology & Pain Medicine Artificial intelligence machine learning neural networks Bayesian methods anesthesia and anesthesiology Figures Figure 1 INTRODUCTION Artificial intelligence (AI) is a multidisciplinary field focusing on expanding and generating intelligent computer algorithms to carry out simple to more complex tasks traditionally performed using human intelligence. These tasks range from the intellectual ability to learn and critical thinking to problem-solving, object and word recognition, inference of world states, and development of philological aspects of language [1,2]. While AI is thought to be related exclusively to computers or robots, its roots can be traced to multiple fields, including philosophy, psychology, linguistics, and statistics [3]. Moreover, AI comprehends several subfields, such as robotics, natural language processing (NLP), and machine learning, which enable computers to evolve by understanding and comprehending human languages as well as learning new tasks that were not initially programmed [4]. In the anesthesia field, AI has been a transformational force for many decades. Earlier researchers developed algorithms that enabled computer programs to resemble the judgment of human experts. These algorithms were primarily designed to help anesthesiologists with drug administration and patient monitoring. For example, in the late 1990s, the “Diprifusor,” a propofol target-controlled system, was developed [5]. This system was designed to use the pharmacokinetic properties of propofol for drug administration, thus sustaining the desired plasma concentration of the drug. Similarly, the “closed loop anesthesia drug administration” (CLAD) system, which uses a fuzzy logic controller, was developed to administer propofol and remifentanil while sustaining the needed depth of anesthesia [6,7]. Moreover, as the technology advanced, the attention shifted to developing more complex AI programs that integrated machine learning and deep learning. As a result, AI systems capable of carrying out difficult tasks, such as predicting events related to anesthesia, recognizing risk factors associated with complications, and aiding in anesthesia during surgical procedures [8–10], were developed. A good example of these systems is the AIMS, which uses AI to gather, store, and evaluate patient data, providing anesthesiologists with real-time information on how to manage patients better [11]. Another example is the SAM system, which works in conjunction with AIMS to evaluate patient data and recommend anesthetic management using machine learning technology [12]. Although research has shown that AI is well positioned in the anesthesiology field, the efficacy of integrating AI in anesthesia still needs to be understood. Therefore, we carried out the present systematic review and meta-analysis to analyze the effectiveness of various AI applications in the anesthesia department. Furthermore, we reviewed some of the challenges that hinder the integration of AI in the anesthesia field. METHODS Information Sources and Literature Search The PubMed, Google Scholar, IEEE Xplore, and Web of Science databases were systematically searched for articles published between January 1, 2000, and March 2024. The search involved the combination of Keywords such as Artificial intelligence, machine learning, neural networks, Bayesian methods, anesthesia, and anesthesiology. Furthermore, reference lists of all articles related to our research objective were scrutinized for additional studies. Duplicate papers and grey literature containing unpublished data were excluded since they would have undercut the current study's scientific purpose and jeopardized the statistical analysis. The search strategy employed in each electronic database is detailed in the appendix section (Appendix A). Eligibility Criteria Two independent reviewers screened the full-text records of publications retrieved from the electronic databases mentioned earlier and included those that focused on the application of AI-based algorithms or systems in the practice of anesthesia. Furthermore, all the articles had to have been authored in English to be included in the present systematic review. This criterion was vital as we wanted to avoid direct translation of scientific terms. On the other hand, articles designed as narrative reviews, case reports, animal studies, editorials, letters to the editors, conference proceeding papers, or abstracts were excluded. Moreover, papers that reported the application of AI in medical fields but were not specific to the practice of anesthesia and those that contained simulated data were excluded. All discrepancies during the selection process were resolved by consulting a third reviewer. Data Extraction and Data Items Two independent researchers assessed the included studies and collected the data required for review and analysis into separate standardized Excel spreadsheets. Afterward, the extracted data was harmonized into a single characteristic table. The data extracted by the reviewers is as follows: Study ID (name of the primary author and the year of publication), study design, characteristics of the enrolled participants (the population enrolled and sample size of the test data), AI techniques employed, specific applications of the AI in anesthesia, and analyzed outcomes. All inconsistencies during the data extraction were resolved either by constructive discussion between the two reviewers or by consulting a third reviewer. The outcomes in our study were grouped depending on the application of AI in anesthesia. The first outcome was model performance in predicting anesthesia-related events, estimated by analyzing the area under the receiver operating characteristic curves (AUROC). An AUROC value of 1 represented flawless discrimination, and a value greater than or equal to 0.8 meant that the model had a high degree of discrimination. On the other hand, a model was interpreted to have fair or poor discrimination for 0.7 ≤ AUROC < 0.8 and 0.6 ≤ AUROC < 0.7. Moreover, a model with an AUROC value between 0.5 and 0.6 was interpreted to have failed since the prediction was purely random [13]. The other outcome was success rate, which was used to analyze the efficacy of AI in guiding anesthesia procedures such as tracheal intubation and nerve blocks. Quality Appraisal All studies included in the present review were observational, therefore methodological quality evaluation was conducted using the Newcastle Ottawa Scale (NOS). The NOS consists of three categories: an outcome domain (maximum of three NOS stars for assessment of outcome, follow-up long enough for outcomes to occur, and adequacy of the follow-up for cohorts), a comparability domain (maximum of two NOS stars for comparability of the cohorts on the basis of the design or analysis controlled for confounders), and a selection domain (maximum of four NOS stars for representativeness of the exposed cohort, selection of the nonexposed cohort, ascertainment of exposure, and demonstration that the outcome of interest was not present at the beginning of the study). Moreover, the overall methodological quality of each study was categorized as poor (NOS score 0–3), fair (NOS score 4–6) or good (NOS score 7–9). Data Synthesis Statistical analyses in the current study were performed using two different statistical tools. STATA 16.0 software was used to pool the AUROC data, while the Comprehensive Meta-analysis software was used to pool the data related to success rate. All the analyses were pooled using the random effects model, which can counter the anticipated heterogeneity and provide more conservative estimates. Moreover, the effect sizes were presented in forest plots together with their 95% Confidence interval. The interstudy heterogeneity was calculated using I 2 statistics, of which values greater than 50% were deemed substantial according to previous guidelines and recommendations [14,15]. Whenever possible, subgroup analyses were performed based on the AI technique employed. RESULTS Study Selection The initial database search resulted in 1420 potential articles. Additionally, 67 articles were identified after hand-searching reference lists of potential studies retrieved from the electronic databases. After 603 duplicate articles were removed, titles and abstracts of the remaining 884 records were screened, of which 729 were excluded. Out of the remaining 155 records, 99 were disqualified before screening for eligibility because they were case reports, editorials, or reviews. Finally, after a thorough screening process and an evaluation of eligibility, 20 studies were included in the review. The other 36 records were excluded due to the following reasons: 13 contained simulated data only, 3 were authored in different languages, and 21 reported the application of AI in other medical fields. The full selection criteria is outlined in Fig. 1 . Study Characteristics. Our study only focused on two applications of AI in the anesthesia department. The first application was the use of AI in predicting events related to anesthesia, of which 15 studies were included (Table 1 ). 8 of these studies reported the use of AI in predicting hypotension, 2 reported the prediction of hypoxemia, and the remaining 5 reported prediction of postoperative nausea and vomiting (PONV). Several AI techniques, including artificial neural network (ANN), grade boosting machine learning (GBM), machine learning algorithm, simplified artificial neural network (SANN), random forest (RF), support vector machine (SVM), fuzzy logic, K-nearest neighbors (KNN), and decision tree (DT) were used to predict these events. The second application was the use of AI in guiding anesthesia procedures, of which 5 studies were included (Table 3 ). 3 of these studies reported the use of AI in tracheal intubation and 2 reported the use of AI in ultrasound-guided nerve blocks. Table 1 Characteristics of studies evaluating the application of AI in predicting events related to anesthesia. Study ID Study Design Patient characteristics AI method Outcomes Population enrolled Sample size (n) Hypotension Prediction Palla et al.,2022[16] Mixed methods study Patients who had undergone surgery and were admitted to PACU. 17029 GBM AUROC: 0.82 Hatib et al.,2018[17] Retro/prospective cohort study. Patients in the OR and ICU receiving anesthesia. 350 Machine-learning algorithm AUROC: 0.97, 0.95, and 0.95 at 5 minutes, 10 minutes, and 15 minutes before the hypotensive event, respectively. Wijnberge et al.,2021[18] Prospective cohort study Patients undergoing a variety of surgical procedures under anesthesia. 568 Machine-learning algorithm AUROC: 0.93, 0.91, and 0.90 at 5, 10, and 15 min before the hypotensive event, respectively. Lin et al.,2008[19] Cohort study Patients undergoing surgery under spinal anesthesia. 375 ANN and SANN AUROC: 0.796 and 0.798 for ANN and SANN, respectively. Lin et al.,2011[20] Cohort study Patients undergoing surgery under general anesthesia. 294 ANN AUROC: 0.893 Gratz et al.,2020[21] Cohort study Patients undergoing cesarean section under general anesthesia. 45 ANN AUROC: 0.89 Kendale et al.,2018[22] Cohort study Patients undergoing surgery under general anesthesia 13323 GBM, ANN, RF, and SVM AUROC: 0.74, 0.70, 0.73, and 0.64, respectively. Li et al.,2021[23] Retrospective cohort study Patients undergoing cardiac surgery under anesthesia. 909 RF AUROC: 0.843 Hypoxemia prediction Geng et al.,2019[24] Prospective observational study Patients undergoing gastroscopy and/or colonoscopy after sedation 220 ANN AUROC: 0.80 Lundberg et al.,2018[25] Cohort study Patients undergoing surgery under anesthesia 53126 GBM AUROC: 0.83 and 0.81 for initial and real-time predictions, respectively. Prediction of PONV Kim et al.,2023[26] Retrospective observational study Patients undergoing surgery under general anesthesia 21372 RF and GBM AUROC: 0.63 and 0.60, respectively. Gong et al.,2014[27] Retrospective study. Orthopedic patients subjected to PCEA 195 ANN AUROC: 0.663, 0.900, and 0.847 for 8, 10, 12 neuron ANN. Shim et al.,2022[28] Retrospective study. Patients undergoing non-cardiac surgery subjected to IV-PCA. 2149 KNN, DT, RF, GBM, SVM and ANN AUROC: 0.597, 0.561, 0.610, 0.580, 0.649, and 0.686, respectively. Bassanezi et al.,2013[29] Prospective study. Pediatric oncology patients undergoing surgery under anesthesia. 198 Fuzzy logic model AUROC: 0.72 Wu et al.,2016[30] Retrospective study Orthopedic patients subjected to PCEA 195 SVM AUROC: 0.929 Note : AI: Artificial intelligence; ANN: artificial neural network; GBM: grade boosting machine learning; SANN: simplified artificial neural network; RF: random forest; SVM: support vector machine; KNN: K-nearest neighbors; DT: decision tree (DT); AUROC: area under the receiver operating characteristic curve; PCEA: Patient-Controlled Epidural Analgesia; PCA: Patient-controlled analgesia; PONV: postoperative nausea and vomiting; ICU: intensive care unit; PACU: Post anesthesia care unit; OR: Operating room. The Application of AI in Event Prediction. A total of 8 studies reported the use of AI in predicting postinduction hypotension. A subgroup analysis of data from these studies showed that models incorporating machine learning algorithms were superior in predicting postinduction hypotension (AUROC: 0.93). Additionally, ANN and SANN models showed a good discriminatory capacity in predicting postinduction hypotension (AUROC: 0.82 and 0.80, respectively). On the other hand, RF and GBM models showed a fair differentiation ability (AUROC: 0.79 and 0.78, respectively), while SVM demonstrated a poor discriminatory ability in predicting postinduction hypotension (AUROC: 0.64) (Table 2 ). Regarding the use of AI in predicting hypoxemia, we found only two relevant studies. One of these studies reported an ANN model, while the other study reported a GBM model. A subgroup analysis of data from these studies showed that both models had a good discriminatory capacity when predicting hypoxemia (AUROC: 0.8 and 0.81, respectively) (Table 2 ). In addition, 5 observational studies reported the use of AI in predicting PONV. A subgroup analysis of data from these studies revealed that an SVM model had a high discriminatory ability for hypoxemia prediction (AUROC: 0.93). On the other hand, ANN and fuzzy logic models had a fair discriminatory ability, while RF, GBM, and KNN models had poor differentiation capacity in terms of predicting PONV. The DT model failed to predict PONV (AUROC: 0.56) (Table 2 ). Table 2 Meta-analytic results on the efficacy of AI in predicting events related to anesthesia Event predicted AI technique No. of studies AUROC (95% CI) I 2 (%) Hypotension ANN 4 0.82 (0.72–0.91) 96.80 GBM 2 0.78 (0.70–0.86) 97.05 Machine learning algorithm 2 0.93 (0.88–0.97) 98.58 RF 2 0.79 (0.68–0.90) 95.74 SANN 1 0.80 (0.77–0.83) N/A SVM 1 0.64 (0.76–0.86) N/A Hypoxemia ANN 1 0.8 (0.76–0.84) N/A GBM 1 0.81 (0.79–0.84) N/A PONV ANN 2 0.77 (0.61–0.93) 95.32 GBM 2 0.60 (0.59–0.60) 0 RF 2 0.63 (0.62–0.63) 0 DT 1 0.56 (0.50–0.62) N/A Fuzzy logic 1 0.72 (0.66–0.78) N/A KNN 1 0.60 (0.62–0.63) N/A SVM 1 0.93 (0.90–0.96) N/A Table 3 Characteristics of studies evaluating the application of AI in guiding anesthesia procedures Study ID Study Design Patient characteristics AI model Outcomes Population Sample size Tracheal Intubation Hermmerling et al.,2012[31] Feasibility pilot study Mannikin 90 Robotic system: KIS First attempt success rate: 100%. Intubation time: 46, 51, and 41 seconds for direct view, indirect view, and semiautomated groups, respectively. Biro et al.,2020[32] Proof-of-concept study Mannikin 42 Robotic system: REALITI First attempt success rate: 95%. Intubation time: 15 and 17 seconds for automated and manual insertions. Hemmerling et al.,2012[33] Pilot clinical study Humans 12 Robotic system: KIS First attempt success rate: 92%. Intubation time: 57 seconds. Ultrasound-guided nerve block Morse et al.,2014[34] Single-center observational comparative study Mannikin 10 Robotic system: Magellan First attempt success rate: 100%. Hemmerling et al.,2013[35] Pilot clinical study Humans 13 Robotic system: Magellan First attempt success rate: 100%. Performance time: 189 seconds. Note : KIS: Kepler intubation system; REALITI Robotic endoscope-automated via laryngeal imaging for tracheal intubation AI application as an assistance tool . Only 3 studies have reported the use of AI in assisting tracheal intubation. Two of these studies were performed in mannikins and one in Humans. The subgroup analysis has shown that robotically-assisted tracheal intubation was highly successful in both mannikins and humans (success rate: 98% and 92%, respectively) (Table 4 ). On the other hand, two studies reported the use of AI in assisting ultrasound-guided nerve blocks. A subgroup analysis of data from these studies showed that robotically-assisted ultrasound-guided nerve block was highly successful in mannikins and humans (Success rate: 96% for humans and mannikins, respectively) (Table 4 ). Table 4 Meta-analytic results on the success rate of robotically-assisted tracheal intubations and ultrasound guided nerve blocks . Procedure Subgroup No. of studies ER (95% CI) I 2 (%) Tracheal intubation Mannikin 2 0.98 (0.85–1.00) 47.73 Humans 1 0.92 (0.59–0.99) N/A Ultrasound-guided nerve block Mannikin 1 0.96 (0.55 -1.00) N/A Humans 1 0.96 (0.62–1.00) N/A Quality Appraisal outcomes Table 5 shows results of the quality appraisal using the Newcastle Ottawa Scale (Appendix B). From the assessment only 4 studies had good methodological quality and 16 had fair methodological quality, therefore, the risk of bias from these studies was minimal. Furthermore, we have seen most of the studies were unable to attain maximum score under the selection domain because they had small sample sizes (i.e., less than 2000) or were carried out on mannikins. Additionally, none of the studies attained maximum scores under the outcome domain because there was no information given about the follow-up duration and how the outcomes were assessed. DISCUSSION AI is increasingly being integrated into the medical field, and its application in anesthesia has been gaining interest in the past decade [3]. Therefore, the current meta-analysis has attempted to analyze the efficacy of AI in the anesthesiology department through two main applications, i.e., prediction of events associated with anesthesia and assisting anesthesia-related procedures. Moreover, our study also reviews some of the challenges faced during the integration of AI in the anesthesia field. The efficacy of AI in predicting anesthesia-related events was studied by analyzing the prediction of postinduction hypotension, hypoxemia, and PONV. Postinduction hypotension is a common complication, with a reported incidence of around 20% [36]. This complication is known to be related to significant post-operative adverse events. For instance, Maheshwari and colleagues reported that the occurrence of postinduction hypotension resulted in prolonged postoperative hospital stay and death [37]. Therefore, such events suggest the need to predict postinduction hypotension, which may eventually help anesthesiologists tailor their induction agents according to the population and monitor the risk of hypotension. Our meta-analysis found that algorithms using multiple machine-learning techniques had the highest discriminatory ability when predicting postinduction hypotension. This result was not surprising given that machine learning is known to have a superior data processing ability. Furthermore, we investigated the efficacy of various machine-learning techniques and found that models based on ANN and SANN had a higher discriminatory ability compared to the other techniques. However, this does not mean that other machine-learning techniques are inferior, as evidence suggests that the other techniques can also predict postinduction hypotension with a high differentiation capacity. For example, Li and colleagues found that an RF model had a high differentiation ability in identifying patients at a high risk of hypotensive events during cardiac surgery (AUROC: 0.843) [23]. Similarly, Palla and colleagues found that a GBM model improved the ability of anesthesiologists to predict postinduction hypotension (AUROC: 0.82) [16]. Therefore, these results suggest that machine-learning models are powerful data analysis tools that can assist anesthesiologists and other clinicians make informed decisions and thus help improve the outcomes of patients after anesthesia. On the other hand, hypoxemia is a physiological condition that can cause serious harm to patients during general anesthesia and surgery [38]. Research shows that hypoxemia is linked to cardiac arrest, cardiac arrhythmias, post-operative infections, impaired cognitive function, delirium, and cerebral ischemia via several metabolic pathways [39]. Therefore, it is essential to predict hypoxemia before it occurs as it may aid anesthesiologists to prevent it proactively, thus minimizing patient harm. In our study, only two articles have reported the use of AI in predicting hypoxemia. One of these studies showed that an ANN model based on three variables (i.e., Body mass index, habitual snoring, and neck circumference) was useful in predicting hypoxemia during sedation for gastrointestinal endoscopy [24]. Similarly, the other study reported that a GBM model trained on real-time data from electronic medical records of over 50000 surgeries was sufficient to predict hypoxemia [25]. Therefore, the overall results indicate that machine-learning models efficiently predict hypoxemia during anesthesia. In addition, the current study analyzed the ability of various machine learning techniques to predict PONV. The subgroup analysis showed that an SVM model had the highest discrimination ability in predicting PONV (AUROC: 0.93). Other machine-learning models, such as the ANN model, also performed relatively well in identifying patients at a high risk of PONV (AUROC: 0.77). Similarly, one of the studies reported that a fuzzy logic model was adequate to estimate the risk of postoperative vomiting in children with cancer [29]. Therefore, we believe that AI models, especially those based on SVM, ANN, and fuzzy logic, can aid anesthesiologists decide whether they should undertake preemptive measures such as preparing an antiemetic in advance or continuing with follow-up and observation. The present study also analyzed the application of AI in assisting anesthesia-related procedures. First, we evaluated the efficacy of robotics in carrying out tracheal intubations and found that robotically-assisted tracheal intubations were highly successful in both mannikins and humans. However, it is important to note that two different robotic intubation systems (Kepler intubation system (KIS) and Robotic endoscope-automated via laryngeal imaging for tracheal intubation (REALITI)) have been reported. The KIS was successful in 100% of mannikins and 91% (11/12) successful in humans. The KIS was unsuccessful in one patient because of fogging of the Pentax video laryngoscope, which is more common with this type of video laryngoscope [40]. On the other hand, evidence showed that REALITI can be used to automate tracheal intubations and can be used even by individuals without any medical training. Additionally, our results suggest that ultrasound-guided nerve blocks performed using the Magellan system are highly successful in both mannikins and humans. Challenges of integrating AI in Anesthesia. Although our study has shown that AI has enormous potential in predicting events related to anesthesia and automating procedures such as tracheal intubation and nerve blocks, there are several challenges faced when integrating AI into the anesthesia field. First, AI models, especially those based on machine learning and deep learning concepts, usually need high-quality data to work efficiently. However, obtaining this kind of data can be highly challenging. Therefore, it is difficult to guarantee data accuracy, thoroughness, and consistency. Thus, this jeopardizes patient safety due to incomplete or faulty data [41,42]. Moreover, since safeguarding patient information and privacy is paramount, the need for large amounts of data raises privacy and security concerns [43,44]. Second, AI integration in anesthesia may be hampered by several technical constraints. For instance, AI systems often rely on data used in training them and programming; thus, they can produce inaccurate predictions and recommendations if they are trained on biased or unrepresentative data [45]. Additionally, AI systems are devoid of empathy and human judgment; therefore, they may not consider patient’s concerns or anxiety regarding the recommended anesthetic strategy. Third, AI systems are subject to “Black box.” This is a phenomenon where AI systems can identify patterns and make predictions but cannot explain the clinical relationship between variables [46]. Therefore, in medical fields such as anesthesiology, where it is vital to understand the physiological concepts informing a particular intervention, this constraint can create trust and transparency issues between clinicians and AI. Finally, AI integration in anesthesia might be hampered by legal and ethical issues. Several ethical concerns have been raised regarding AI. For example, if an AI makes a mistake, who is liable? Is it the anesthesiologist who used it, the programmer, or the hospital that authorized its use? The law regarding this liability matter is still ambiguous [47]. Additionally, patient consent is an ethical issue that may hinder the use of AI. Although AI can automate some tasks, the human touch is still essential in healthcare; therefore, convincing patients to consent to purely automated systems can be difficult. Limitations The present study is not without its limitations. First, a high interstudy heterogeneity persisted in most of the subgroup analyses. This heterogeneity might have resulted from the variation in sample sizes and the population analyzed. Second, all studies in the present review were observational, therefore increasing the risk of selection bias in the statistical analyses. Third, we considered publications available in English only; hence, the data that might have raised the statistical power of our analysis but presented in studies published in other languages had been eliminated. Finally, although AI has several applications in the anesthesiology department, we only considered two applications for analysis; therefore, we cannot make any judgment on the efficacy of AI out of these two applications. Conclusion AI has great potential to transform the field of anesthesia. Our study has shown that AI systems using machine-learning models have the ability to predict postinduction hypotension and hypoxemia. Moreover, machine learning models, especially those based on SVM, ANN, and fuzzy logic, are adequate in predicting PONV. Therefore, AI can help anesthesiologists predict anesthesia-related events and take preemptive measures. Additionally, AI can improve tracheal intubations and ultrasound-guided nerve blocks by automating the systems. However, there are multiple barriers hindering the integration of AI in anesthesia that need to be addressed. References Bellman R: An Introduction to Artificial Intelligence: Can Computers Think? Boyd & Fraser Publishing Company; 1978. Bellini V, Valente M, Gaddi AV, Pelosi P, Bignami E: Artificial intelligence and telemedicine in anesthesia: potential and problems. Minerva Anestesiol. 2022, 88:. 10.23736/S0375-9393.21.16241-8 Hashimoto DA, Witkowski E, Gao L, Meireles O, Rosman G: Artificial Intelligence in Anesthesiology: Current Techniques, Clinical Applications, and Limitations. Anesthesiology. 2020, 132:379–94. 10.1097/ALN.0000000000002960 Russell SJ, Norvig P: Artificial intelligence: a modern approach. Third edition, Global edition. Pearson: Boston Columbus Indianapolis New York San Francisco Upper Saddle River Amsterdam Cape Town Dubai London Madrid Milan Munich Paris Montreal Toronto Delhi Mexico City Sao Paulo Sydney Hong Kong Seoul Singapore Taipei Tokyo; 2016. Glen JB: The development of ‘Diprifusor’: a TCI system for propofol. Anaesthesia. 1998, 53:13–21. 10.1111/j.1365-2044.1998.53s115.x Liberman MY, Ching S, Chemali J, Brown EN: A closed-loop anesthetic delivery system for real-time control of burst suppression. J Neural Eng. 2013, 10:046004. 10.1088/1741-2560/10/4/046004 Liu N, Chazot T, Hamada S, et al.: Closed-Loop Coadministration of Propofol and Remifentanil Guided by Bispectral Index: A Randomized Multicenter Study. Anesthesia & Analgesia. 2011, 112:546. 10.1213/ANE.0b013e318205680b Solanki SL, Pandrowala S, Nayak A, Bhandare M, Ambulkar RP, Shrikhande SV: Artificial intelligence in perioperative management of major gastrointestinal surgeries. World Journal of Gastroenterology. 2021, 27:2758–70. 10.3748/wjg.v27.i21.2758 Chiew CJ, Liu N, Wong TH, Sim YE, Abdullah HR: Utilizing Machine Learning Methods for Preoperative Prediction of Postsurgical Mortality and Intensive Care Unit Admission. Annals of Surgery. 2020, 272:1133. 10.1097/SLA.0000000000003297 Corey KM, Kashyap S, Lorenzi E, et al.: Development and validation of machine learning models to identify high-risk surgical patients using automatically curated electronic health record data (Pythia): A retrospective, single-site study. PLOS Medicine. 2018, 15:e1002701. 10.1371/journal.pmed.1002701 Ehrenfeld JM, Rehman MA: Anesthesia information management systems: a review of functionality and installation considerations. J Clin Monit Comput. 2011, 25:71–9. 10.1007/s10877-010-9256-y Nair BG, Newman S-F, Peterson GN, Schwid HA: Smart Anesthesia Manager \rm TM (SAM)—A Real-time Decision Support System for Anesthesia Care during Surgery. IEEE Transactions on Biomedical Engineering. 2013, 60:207–10. 10.1109/TBME.2012.2205384 Çorbacıoğlu ŞK, Aksel G: Receiver operating characteristic curve analysis in diagnostic accuracy studies: A guide to interpreting the area under the curve value. Turk J Emerg Med. 2023, 23:195–8. 10.4103/tjem.tjem_182_23 Higgins J, Thompson S, Deeks J, Altman D: Statistical heterogeneity in systematic reviews of clinical trials: a critical appraisal of guidelines and practice. J Health Serv Res Policy. 2002, 7:51–61. 10.1258/1355819021927674 Higgins JPT, Thompson SG, Deeks JJ, Altman DG: Measuring inconsistency in meta-analyses. BMJ. 2003, 327:557–60. 10.1136/bmj.327.7414.557 Palla K, Hyland SL, Posner K, et al.: Intraoperative prediction of postanaesthesia care unit hypotension. Br J Anaesth. 2022, 128:623–35. 10.1016/j.bja.2021.10.052 Hatib F, Jian Z, Buddi S, et al.: Machine-learning Algorithm to Predict Hypotension Based on High-fidelity Arterial Pressure Waveform Analysis. Anesthesiology. 2018, 129:663–74. 10.1097/ALN.0000000000002300 Wijnberge M, van der Ster BJP, Geerts BF, et al.: Clinical performance of a machine-learning algorithm to predict intra-operative hypotension with noninvasive arterial pressure waveforms: A cohort study. Eur J Anaesthesiol. 2021, 38:609–15. 10.1097/EJA.0000000000001521 Lin C-S, Chiu J-S, Hsieh M-H, Mok MS, Li Y-C, Chiu H-W: Predicting hypotensive episodes during spinal anesthesia with the application of artificial neural networks. Comput Methods Programs Biomed. 2008, 92:193–7. 10.1016/j.cmpb.2008.06.013 Lin C-S, Chang C-C, Chiu J-S, et al.: Application of an artificial neural network to predict postinduction hypotension during general anesthesia. Med Decis Making. 2011, 31:308–14. 10.1177/0272989X10379648 Gratz I, Baruch M, Takla M, Seaman J, Allen I, McEniry B, Deal E: The application of a neural network to predict hypotension and vasopressor requirements non-invasively in obstetric patients having spinal anesthesia for elective cesarean section (C/S). BMC Anesthesiol. 2020, 20:98. 10.1186/s12871-020-01015-9 Kendale S, Kulkarni P, Rosenberg AD, Wang J: Supervised Machine-learning Predictive Analytics for Prediction of Postinduction Hypotension. Anesthesiology. 2018, 129:675–88. 10.1097/ALN.0000000000002374 Li X-F, Huang Y-Z, Tang J-Y, Li R-C, Wang X-Q: Development of a random forest model for hypotension prediction after anesthesia induction for cardiac surgery. World J Clin Cases. 2021, 9:8729–39. 10.12998/wjcc.v9.i29.8729 Geng W, Tang H, Sharma A, Zhao Y, Yan Y, Hong W: An artificial neural network model for prediction of hypoxemia during sedation for gastrointestinal endoscopy. J Int Med Res. 2019, 47:2097–103. 10.1177/0300060519834459 Lundberg SM, Nair B, Vavilala MS, et al.: Explainable machine-learning predictions for the prevention of hypoxaemia during surgery. Nat Biomed Eng. 2018, 2:749–60. 10.1038/s41551-018-0304-0 Kim J-H, Cheon B-R, Kim M-G, Hwang S-M, Lim S-Y, Lee J-J, Kwon Y-S: Postoperative Nausea and Vomiting Prediction: Machine Learning Insights from a Comprehensive Analysis of Perioperative Data. Bioengineering. 2023, 10:1152. 10.3390/bioengineering10101152 Gong C-SA, Yu L, Ting C-K, Tsou M-Y, Chang K-Y, Shen C-L, Lin S-P: Predicting postoperative vomiting for orthopedic patients receiving patient-controlled epidural analgesia with the application of an artificial neural network. Biomed Res Int. 2014, 2014:786418. 10.1155/2014/786418 Shim J-G, Ryu K-H, Cho E-A, Ahn JH, Cha YB, Lim G, Lee SH: Machine learning for prediction of postoperative nausea and vomiting in patients with intravenous patient-controlled analgesia. PLoS One. 2022, 17:e0277957. 10.1371/journal.pone.0277957 Bassanezi BSB, de Oliveira-Filho AG, Jafelice RSM, Bustorff-Silva JM, Udelsmann A: Postoperative vomiting in pediatric oncologic patients: prediction by a fuzzy logic model. Paediatr Anaesth. 2013, 23:68–73. 10.1111/pan.12000 Wu H-Y, Gong C-SA, Lin S-P, Chang K-Y, Tsou M-Y, Ting C-K: Predicting postoperative vomiting among orthopedic patients receiving patient-controlled epidural analgesia using SVM and LR. Sci Rep. 2016, 6:27041. 10.1038/srep27041 Hemmerling TM, Wehbe M, Zaouter C, Taddei R, Morse J: The Kepler intubation system. Anesth Analg. 2012, 114:590–4. 10.1213/ANE.0b013e3182410cbf Biro P, Hofmann P, Gage D, et al.: Automated tracheal intubation in an airway manikin using a robotic endoscope: a proof of concept study. Anaesthesia. 2020, 75:881–6. 10.1111/anae.14945 Hemmerling TM, Taddei R, Wehbe M, Zaouter C, Cyr S, Morse J: First robotic tracheal intubations in humans using the Kepler intubation system. Br J Anaesth. 2012, 108:1011–6. 10.1093/bja/aes034 Morse J, Terrasini N, Wehbe M, Philippona C, Zaouter C, Cyr S, Hemmerling TM: Comparison of success rates, learning curves, and inter-subject performance variability of robot-assisted and manual ultrasound-guided nerve block needle guidance in simulation. Br J Anaesth. 2014, 112:1092–7. 10.1093/bja/aet440 Hemmerling TM, Taddei R, Wehbe M, Cyr S, Zaouter C, Morse J: Technical communication: First robotic ultrasound-guided nerve blocks in humans using the Magellan system. Anesth Analg. 2013, 116:491–4. 10.1213/ANE.0b013e3182713b49 Walsh M, Devereaux PJ, Garg AX, et al.: Relationship between Intraoperative Mean Arterial Pressure and Clinical Outcomes after Noncardiac Surgery: Toward an Empirical Definition of Hypotension. Anesthesiology. 2013, 119:507–15. 10.1097/ALN.0b013e3182a10e26 Maheshwari K, Turan A, Mao G, et al.: The association of hypotension during non-cardiac surgery, before and after skin incision, with postoperative acute kidney injury: a retrospective cohort analysis. Anaesthesia. 2018, 73:1223–8. 10.1111/anae.14416 Dunham CM, Hileman BM, Hutchinson AE, Chance EA, Huang GS: Perioperative hypoxemia is common with horizontal positioning during general anesthesia and is associated with major adverse outcomes: a retrospective study of consecutive patients. BMC Anesthesiol. 2014, 14:43. 10.1186/1471-2253-14-43 Strachan L, Noble DW: Hypoxia and surgical patients--prevention and treatment of an unnecessary cause of morbidity and mortality. J R Coll Surg Edinb. 2001, 46:297–302. Teoh WHL, Shah MK, Sia ATH: Randomised comparison of Pentax AirwayScope and Glidescope for tracheal intubation in patients with normal airway anatomy. Anaesthesia. 2009, 64:1125–9. 10.1111/j.1365-2044.2009.06032.x Raimundo R, Rosário A: The Impact of Artificial Intelligence on Data System Security: A Literature Review. Sensors. 2021, 21:7029. 10.3390/s21217029 Harvey HB, Gowda V: Regulatory Issues and Challenges to Artificial Intelligence Adoption. Radiol Clin North Am. 2021, 59:1075–83. 10.1016/j.rcl.2021.07.007 Coppola L, Cianflone A, Grimaldi AM, et al.: Biobanking in health care: evolution and future directions. Journal of Translational Medicine. 2019, 17:172. 10.1186/s12967-019-1922-3 Keskinbora KH: Medical ethics considerations on artificial intelligence. Journal of Clinical Neuroscience. 2019, 64:277–82. 10.1016/j.jocn.2019.03.001 Belenguer L: AI bias: exploring discriminatory algorithmic decision-making models and the application of possible machine-centric solutions adapted from the pharmaceutical industry. AI Ethics. 2022, 2:771–87. 10.1007/s43681-022-00138-8 Lopes S, Rocha G, Guimarães-Pereira L: Artificial intelligence and its clinical application in Anesthesiology: a systematic review. J Clin Monit Comput. 2024, 38:247–59. 10.1007/s10877-023-01088-0 Farhud DD, Zokaei S: Ethical Issues of Artificial Intelligence in Medicine and Healthcare. Iranian Journal of Public Health. 2021, 50:i. 10.18502/ijph.v50i11.7600 Additional Declarations The authors declare no competing interests. Supplementary Files APPENDIX.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4599435","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Systematic Review","associatedPublications":[],"authors":[{"id":315871187,"identity":"445fa362-87f7-4dcb-b57a-ca69763f6599","order_by":0,"name":"Hany A. Zaki","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Hany","middleName":"A.","lastName":"Zaki","suffix":""},{"id":315873051,"identity":"d2c711ee-465a-4275-a0d0-e76e46a65bf1","order_by":1,"name":"Eman E. Shaban","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Eman","middleName":"E.","lastName":"Shaban","suffix":""},{"id":315873052,"identity":"fbfe9fac-e062-42fd-b9d1-88b3c363ff6d","order_by":2,"name":"Nabil Shallik","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyUlEQVRIiWNgGAWjYFACHgaGBwUSBvwSIA4bsVoSDCQMJGeQqIXBwOAGsVrM+c8e/JBgYGFsfLv56YYfZQyy/Q0EtFjOyEuWADrMzOzOMbObPecYjGccIKDF4AaPAUiLjdmNBLMbvG0MiQ0EtZw/Y/wDpMV4Rvq3m3+BWuYT1HIgxwzsMAOJHLPbIFs2EHZYjpkFUIuxxI2cstsy5ySMNxLjsBsfKuoM+2ekb7v5psxGdh4hLehAgrGBRB0MDGRoGQWjYBSMguEOAIGKQzYynLOWAAAAAElFTkSuQmCC","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Nabil","middleName":"","lastName":"Shallik","suffix":""},{"id":315873053,"identity":"e14589ba-7bd4-46e8-adc8-b0a479aa14b9","order_by":3,"name":"Ahmed Shaban","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Ahmed","middleName":"","lastName":"Shaban","suffix":""},{"id":315873054,"identity":"eb847da9-f9ca-4c27-8eec-22d1ce35f728","order_by":4,"name":"Amira Shaban","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Amira","middleName":"","lastName":"Shaban","suffix":""},{"id":315873055,"identity":"cc46e30a-94b8-44f2-89fa-0b3e22e4615d","order_by":5,"name":"Mohamed Elgassim","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Mohamed","middleName":"","lastName":"Elgassim","suffix":""}],"badges":[],"createdAt":"2024-06-18 10:56:27","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-4599435/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4599435/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":58665490,"identity":"29a0fa33-99e2-4444-8398-13ea0abf81cc","added_by":"auto","created_at":"2024-06-19 13:29:35","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":31782,"visible":true,"origin":"","legend":"\u003cp\u003ePRISMA flow chart for study selection\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-4599435/v1/cd39fcbe08a760d1159687df.png"},{"id":59025252,"identity":"634732ae-c053-492e-b782-7a0b5d0667f5","added_by":"auto","created_at":"2024-06-25 12:56:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":728606,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4599435/v1/11335d10-5fe5-4ecb-9453-9bff322ad48d.pdf"},{"id":58665482,"identity":"18722806-ee1d-42a4-b611-3ae3b156f893","added_by":"auto","created_at":"2024-06-19 13:29:34","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":39333,"visible":true,"origin":"","legend":"","description":"","filename":"APPENDIX.docx","url":"https://assets-eu.researchsquare.com/files/rs-4599435/v1/55f2e020e462d596db7d3ac8.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eA Comprehensive Systematic Review and Meta-Analysis: Evaluating the Effectiveness and Integration Obstacles of Artificial Intelligence (AI) within Anesthesia Departments.\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eArtificial intelligence (AI) is a multidisciplinary field focusing on expanding and generating intelligent computer algorithms to carry out simple to more complex tasks traditionally performed using human intelligence. These tasks range from the intellectual ability to learn and critical thinking to problem-solving, object and word recognition, inference of world states, and development of philological aspects of language [1,2]. While AI is thought to be related exclusively to computers or robots, its roots can be traced to multiple fields, including philosophy, psychology, linguistics, and statistics [3]. Moreover, AI comprehends several subfields, such as robotics, natural language processing (NLP), and machine learning, which enable computers to evolve by understanding and comprehending human languages as well as learning new tasks that were not initially programmed [4].\u003c/p\u003e \u003cp\u003eIn the anesthesia field, AI has been a transformational force for many decades. Earlier researchers developed algorithms that enabled computer programs to resemble the judgment of human experts. These algorithms were primarily designed to help anesthesiologists with drug administration and patient monitoring. For example, in the late 1990s, the \u0026ldquo;Diprifusor,\u0026rdquo; a propofol target-controlled system, was developed [5]. This system was designed to use the pharmacokinetic properties of propofol for drug administration, thus sustaining the desired plasma concentration of the drug. Similarly, the \u0026ldquo;closed loop anesthesia drug administration\u0026rdquo; (CLAD) system, which uses a fuzzy logic controller, was developed to administer propofol and remifentanil while sustaining the needed depth of anesthesia [6,7]. Moreover, as the technology advanced, the attention shifted to developing more complex AI programs that integrated machine learning and deep learning. As a result, AI systems capable of carrying out difficult tasks, such as predicting events related to anesthesia, recognizing risk factors associated with complications, and aiding in anesthesia during surgical procedures [8\u0026ndash;10], were developed. A good example of these systems is the AIMS, which uses AI to gather, store, and evaluate patient data, providing anesthesiologists with real-time information on how to manage patients better [11]. Another example is the SAM system, which works in conjunction with AIMS to evaluate patient data and recommend anesthetic management using machine learning technology [12].\u003c/p\u003e \u003cp\u003eAlthough research has shown that AI is well positioned in the anesthesiology field, the efficacy of integrating AI in anesthesia still needs to be understood. Therefore, we carried out the present systematic review and meta-analysis to analyze the effectiveness of various AI applications in the anesthesia department. Furthermore, we reviewed some of the challenges that hinder the integration of AI in the anesthesia field.\u003c/p\u003e"},{"header":"METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eInformation Sources and Literature Search\u003c/h2\u003e \u003cp\u003eThe PubMed, Google Scholar, IEEE Xplore, and Web of Science databases were systematically searched for articles published between January 1, 2000, and March 2024. The search involved the combination of Keywords such as Artificial intelligence, machine learning, neural networks, Bayesian methods, anesthesia, and anesthesiology. Furthermore, reference lists of all articles related to our research objective were scrutinized for additional studies. Duplicate papers and grey literature containing unpublished data were excluded since they would have undercut the current study's scientific purpose and jeopardized the statistical analysis. The search strategy employed in each electronic database is detailed in the appendix section (Appendix A).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eEligibility Criteria\u003c/h2\u003e \u003cp\u003eTwo independent reviewers screened the full-text records of publications retrieved from the electronic databases mentioned earlier and included those that focused on the application of AI-based algorithms or systems in the practice of anesthesia. Furthermore, all the articles had to have been authored in English to be included in the present systematic review. This criterion was vital as we wanted to avoid direct translation of scientific terms. On the other hand, articles designed as narrative reviews, case reports, animal studies, editorials, letters to the editors, conference proceeding papers, or abstracts were excluded. Moreover, papers that reported the application of AI in medical fields but were not specific to the practice of anesthesia and those that contained simulated data were excluded. All discrepancies during the selection process were resolved by consulting a third reviewer.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eData Extraction and Data Items\u003c/h2\u003e \u003cp\u003eTwo independent researchers assessed the included studies and collected the data required for review and analysis into separate standardized Excel spreadsheets. Afterward, the extracted data was harmonized into a single characteristic table. The data extracted by the reviewers is as follows: Study ID (name of the primary author and the year of publication), study design, characteristics of the enrolled participants (the population enrolled and sample size of the test data), AI techniques employed, specific applications of the AI in anesthesia, and analyzed outcomes. All inconsistencies during the data extraction were resolved either by constructive discussion between the two reviewers or by consulting a third reviewer.\u003c/p\u003e \u003cp\u003eThe outcomes in our study were grouped depending on the application of AI in anesthesia. The first outcome was model performance in predicting anesthesia-related events, estimated by analyzing the area under the receiver operating characteristic curves (AUROC). An AUROC value of 1 represented flawless discrimination, and a value greater than or equal to 0.8 meant that the model had a high degree of discrimination. On the other hand, a model was interpreted to have fair or poor discrimination for 0.7\u0026thinsp;\u0026le;\u0026thinsp;AUROC\u0026thinsp;\u0026lt;\u0026thinsp;0.8 and 0.6\u0026thinsp;\u0026le;\u0026thinsp;AUROC\u0026thinsp;\u0026lt;\u0026thinsp;0.7. Moreover, a model with an AUROC value between 0.5 and 0.6 was interpreted to have failed since the prediction was purely random [13]. The other outcome was success rate, which was used to analyze the efficacy of AI in guiding anesthesia procedures such as tracheal intubation and nerve blocks.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eQuality Appraisal\u003c/h2\u003e \u003cp\u003e All studies included in the present review were observational, therefore methodological quality evaluation was conducted using the Newcastle Ottawa Scale (NOS). The NOS consists of three categories: an outcome domain (maximum of three NOS stars for assessment of outcome, follow-up long enough for outcomes to occur, and adequacy of the follow-up for cohorts), a comparability domain (maximum of two NOS stars for comparability of the cohorts on the basis of the design or analysis controlled for confounders), and a selection domain (maximum of four NOS stars for representativeness of the exposed cohort, selection of the nonexposed cohort, ascertainment of exposure, and demonstration that the outcome of interest was not present at the beginning of the study). Moreover, the overall methodological quality of each study was categorized as poor (NOS score 0\u0026ndash;3), fair (NOS score 4\u0026ndash;6) or good (NOS score 7\u0026ndash;9).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eData Synthesis\u003c/h2\u003e \u003cp\u003eStatistical analyses in the current study were performed using two different statistical tools. STATA 16.0 software was used to pool the AUROC data, while the Comprehensive Meta-analysis software was used to pool the data related to success rate. All the analyses were pooled using the random effects model, which can counter the anticipated heterogeneity and provide more conservative estimates. Moreover, the effect sizes were presented in forest plots together with their 95% Confidence interval. The interstudy heterogeneity was calculated using I\u003csup\u003e2\u003c/sup\u003e statistics, of which values greater than 50% were deemed substantial according to previous guidelines and recommendations [14,15]. Whenever possible, subgroup analyses were performed based on the AI technique employed.\u003c/p\u003e \u003c/div\u003e"},{"header":"RESULTS","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eStudy Selection\u003c/h2\u003e \u003cp\u003eThe initial database search resulted in 1420 potential articles. Additionally, 67 articles were identified after hand-searching reference lists of potential studies retrieved from the electronic databases. After 603 duplicate articles were removed, titles and abstracts of the remaining 884 records were screened, of which 729 were excluded. Out of the remaining 155 records, 99 were disqualified before screening for eligibility because they were case reports, editorials, or reviews. Finally, after a thorough screening process and an evaluation of eligibility, 20 studies were included in the review. The other 36 records were excluded due to the following reasons: 13 contained simulated data only, 3 were authored in different languages, and 21 reported the application of AI in other medical fields. The full selection criteria is outlined in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eStudy Characteristics.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eOur study only focused on two applications of AI in the anesthesia department. The first application was the use of AI in predicting events related to anesthesia, of which 15 studies were included (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). 8 of these studies reported the use of AI in predicting hypotension, 2 reported the prediction of hypoxemia, and the remaining 5 reported prediction of postoperative nausea and vomiting (PONV). Several AI techniques, including artificial neural network (ANN), grade boosting machine learning (GBM), machine learning algorithm, simplified artificial neural network (SANN), random forest (RF), support vector machine (SVM), fuzzy logic, K-nearest neighbors (KNN), and decision tree (DT) were used to predict these events.\u003c/p\u003e \u003cp\u003eThe second application was the use of AI in guiding anesthesia procedures, of which 5 studies were included (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). 3 of these studies reported the use of AI in tracheal intubation and 2 reported the use of AI in ultrasound-guided nerve blocks.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCharacteristics of studies evaluating the application of AI in predicting events related to anesthesia.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eStudy ID\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eStudy Design\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003ePatient characteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAI method\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eOutcomes\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePopulation enrolled\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSample size (n)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eHypotension Prediction\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePalla et al.,2022[16]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMixed methods study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients who had undergone surgery and were admitted to PACU.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e17029\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.82\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHatib et al.,2018[17]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRetro/prospective cohort study.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients in the OR and ICU receiving anesthesia.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e350\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMachine-learning algorithm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.97, 0.95, and 0.95 at 5 minutes, 10 minutes, and 15 minutes before the hypotensive event, respectively.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWijnberge et al.,2021[18]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProspective cohort study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing a variety of surgical procedures under anesthesia.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e568\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMachine-learning algorithm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.93, 0.91, and 0.90 at 5, 10, and 15 min before the hypotensive event, respectively.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLin et al.,2008[19]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCohort study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing surgery under spinal anesthesia.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e375\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eANN and SANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.796 and 0.798 for ANN and SANN, respectively.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLin et al.,2011[20]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCohort study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing surgery under general anesthesia.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e294\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.893\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGratz et al.,2020[21]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCohort study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing cesarean section under general anesthesia.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.89\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKendale et al.,2018[22]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCohort study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing surgery under general anesthesia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13323\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGBM, ANN, RF, and SVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.74, 0.70, 0.73, and 0.64, respectively.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLi et al.,2021[23]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRetrospective cohort study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing cardiac surgery under anesthesia.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e909\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.843\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eHypoxemia prediction\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGeng et al.,2019[24]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProspective observational study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing gastroscopy and/or colonoscopy after sedation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e220\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLundberg et al.,2018[25]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCohort study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing surgery under anesthesia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e53126\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.83 and 0.81 for initial and real-time predictions, respectively.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePrediction of PONV\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKim et al.,2023[26]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRetrospective observational study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing surgery under general anesthesia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e21372\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eRF and GBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.63 and 0.60, respectively.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGong et al.,2014[27]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRetrospective study.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOrthopedic patients subjected to PCEA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e195\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.663, 0.900, and 0.847 for 8, 10, 12 neuron ANN.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eShim et al.,2022[28]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRetrospective study.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePatients undergoing non-cardiac surgery subjected to IV-PCA.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2149\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eKNN, DT, RF, GBM, SVM and ANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.597, 0.561, 0.610, 0.580, 0.649, and 0.686, respectively.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBassanezi et al.,2013[29]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProspective study.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePediatric oncology patients undergoing surgery under anesthesia.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e198\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFuzzy logic model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.72\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWu et al.,2016[30]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRetrospective study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOrthopedic patients subjected to PCEA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e195\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAUROC: 0.929\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"6\"\u003e\u003cb\u003eNote\u003c/b\u003e: AI: Artificial intelligence; ANN: artificial neural network; GBM: grade boosting machine learning; SANN: simplified artificial neural network; RF: random forest; SVM: support vector machine; KNN: K-nearest neighbors; DT: decision tree (DT); AUROC: area under the receiver operating characteristic curve; PCEA: Patient-Controlled Epidural Analgesia; PCA: Patient-controlled analgesia; PONV: postoperative nausea and vomiting; ICU: intensive care unit; PACU: Post anesthesia care unit; OR: Operating room.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eThe Application of AI in Event Prediction.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eA total of 8 studies reported the use of AI in predicting postinduction hypotension. A subgroup analysis of data from these studies showed that models incorporating machine learning algorithms were superior in predicting postinduction hypotension (AUROC: 0.93). Additionally, ANN and SANN models showed a good discriminatory capacity in predicting postinduction hypotension (AUROC: 0.82 and 0.80, respectively). On the other hand, RF and GBM models showed a fair differentiation ability (AUROC: 0.79 and 0.78, respectively), while SVM demonstrated a poor discriminatory ability in predicting postinduction hypotension (AUROC: 0.64) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eRegarding the use of AI in predicting hypoxemia, we found only two relevant studies. One of these studies reported an ANN model, while the other study reported a GBM model. A subgroup analysis of data from these studies showed that both models had a good discriminatory capacity when predicting hypoxemia (AUROC: 0.8 and 0.81, respectively) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn addition, 5 observational studies reported the use of AI in predicting PONV. A subgroup analysis of data from these studies revealed that an SVM model had a high discriminatory ability for hypoxemia prediction (AUROC: 0.93). On the other hand, ANN and fuzzy logic models had a fair discriminatory ability, while RF, GBM, and KNN models had poor differentiation capacity in terms of predicting PONV. The DT model failed to predict PONV (AUROC: 0.56) (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMeta-analytic results on the efficacy of AI in predicting events related to anesthesia\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEvent predicted\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAI technique\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo. of studies\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAUROC (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e(%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHypotension\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.82 (0.72\u0026ndash;0.91)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e96.80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.78 (0.70\u0026ndash;0.86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e97.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMachine learning algorithm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.93 (0.88\u0026ndash;0.97)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e98.58\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.79 (0.68\u0026ndash;0.90)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e95.74\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.80 (0.77\u0026ndash;0.83)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.64 (0.76\u0026ndash;0.86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHypoxemia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.8 (0.76\u0026ndash;0.84)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.81 (0.79\u0026ndash;0.84)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePONV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eANN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.77 (0.61\u0026ndash;0.93)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e95.32\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.60 (0.59\u0026ndash;0.60)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.63 (0.62\u0026ndash;0.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.56 (0.50\u0026ndash;0.62)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFuzzy logic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.72 (0.66\u0026ndash;0.78)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.60 (0.62\u0026ndash;0.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.93 (0.90\u0026ndash;0.96)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCharacteristics of studies evaluating the application of AI in guiding anesthesia procedures\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eStudy ID\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eStudy Design\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e \u003cp\u003ePatient characteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eAI model\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eOutcomes\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePopulation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSample size\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTracheal Intubation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHermmerling et al.,2012[31]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFeasibility pilot study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMannikin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eRobotic system: KIS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eFirst attempt success rate: 100%.\u003c/p\u003e \u003cp\u003eIntubation time: 46, 51, and 41 seconds for direct view, indirect view, and semiautomated groups, respectively.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBiro et al.,2020[32]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eProof-of-concept study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMannikin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eRobotic system: REALITI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eFirst attempt success rate: 95%.\u003c/p\u003e \u003cp\u003eIntubation time: 15 and 17 seconds for automated and manual insertions.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHemmerling et al.,2012[33]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePilot clinical study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHumans\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eRobotic system: KIS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eFirst attempt success rate: 92%.\u003c/p\u003e \u003cp\u003eIntubation time: 57 seconds.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c6\" namest=\"c1\"\u003e \u003cp\u003eUltrasound-guided nerve block\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMorse et al.,2014[34]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSingle-center observational comparative study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMannikin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eRobotic system: Magellan\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eFirst attempt success rate: 100%.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHemmerling et al.,2013[35]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePilot clinical study\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHumans\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eRobotic system: Magellan\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eFirst attempt success rate: 100%.\u003c/p\u003e \u003cp\u003ePerformance time: 189 seconds.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"6\"\u003e\u003cb\u003eNote\u003c/b\u003e: KIS: Kepler intubation system; REALITI Robotic endoscope-automated via laryngeal imaging for tracheal intubation\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eAI application as an assistance tool\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eOnly 3 studies have reported the use of AI in assisting tracheal intubation. Two of these studies were performed in mannikins and one in Humans. The subgroup analysis has shown that robotically-assisted tracheal intubation was highly successful in both mannikins and humans (success rate: 98% and 92%, respectively) (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). On the other hand, two studies reported the use of AI in assisting ultrasound-guided nerve blocks. A subgroup analysis of data from these studies showed that robotically-assisted ultrasound-guided nerve block was highly successful in mannikins and humans (Success rate: 96% for humans and mannikins, respectively) (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003eMeta-analytic results on the success rate of robotically-assisted tracheal intubations and ultrasound guided nerve blocks\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eProcedure\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSubgroup\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNo. of studies\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eER (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eI\u003csup\u003e2\u003c/sup\u003e (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eTracheal intubation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMannikin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.98 (0.85\u0026ndash;1.00)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e47.73\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHumans\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.92 (0.59\u0026ndash;0.99)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eUltrasound-guided nerve block\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMannikin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.96 (0.55 -1.00)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHumans\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.96 (0.62\u0026ndash;1.00)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eQuality Appraisal outcomes\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows results of the quality appraisal using the Newcastle Ottawa Scale (Appendix B). From the assessment only 4 studies had good methodological quality and 16 had fair methodological quality, therefore, the risk of bias from these studies was minimal. Furthermore, we have seen most of the studies were unable to attain maximum score under the selection domain because they had small sample sizes (i.e., less than 2000) or were carried out on mannikins. Additionally, none of the studies attained maximum scores under the outcome domain because there was no information given about the follow-up duration and how the outcomes were assessed.\u003c/p\u003e "},{"header":"DISCUSSION","content":"\u003cp\u003eAI is increasingly being integrated into the medical field, and its application in anesthesia has been gaining interest in the past decade [3]. Therefore, the current meta-analysis has attempted to analyze the efficacy of AI in the anesthesiology department through two main applications, i.e., prediction of events associated with anesthesia and assisting anesthesia-related procedures. Moreover, our study also reviews some of the challenges faced during the integration of AI in the anesthesia field.\u003c/p\u003e \u003cp\u003eThe efficacy of AI in predicting anesthesia-related events was studied by analyzing the prediction of postinduction hypotension, hypoxemia, and PONV. Postinduction hypotension is a common complication, with a reported incidence of around 20% [36]. This complication is known to be related to significant post-operative adverse events. For instance, Maheshwari and colleagues reported that the occurrence of postinduction hypotension resulted in prolonged postoperative hospital stay and death [37]. Therefore, such events suggest the need to predict postinduction hypotension, which may eventually help anesthesiologists tailor their induction agents according to the population and monitor the risk of hypotension. Our meta-analysis found that algorithms using multiple machine-learning techniques had the highest discriminatory ability when predicting postinduction hypotension. This result was not surprising given that machine learning is known to have a superior data processing ability. Furthermore, we investigated the efficacy of various machine-learning techniques and found that models based on ANN and SANN had a higher discriminatory ability compared to the other techniques. However, this does not mean that other machine-learning techniques are inferior, as evidence suggests that the other techniques can also predict postinduction hypotension with a high differentiation capacity. For example, Li and colleagues found that an RF model had a high differentiation ability in identifying patients at a high risk of hypotensive events during cardiac surgery (AUROC: 0.843) [23]. Similarly, Palla and colleagues found that a GBM model improved the ability of anesthesiologists to predict postinduction hypotension (AUROC: 0.82) [16]. Therefore, these results suggest that machine-learning models are powerful data analysis tools that can assist anesthesiologists and other clinicians make informed decisions and thus help improve the outcomes of patients after anesthesia.\u003c/p\u003e \u003cp\u003eOn the other hand, hypoxemia is a physiological condition that can cause serious harm to patients during general anesthesia and surgery [38]. Research shows that hypoxemia is linked to cardiac arrest, cardiac arrhythmias, post-operative infections, impaired cognitive function, delirium, and cerebral ischemia via several metabolic pathways [39]. Therefore, it is essential to predict hypoxemia before it occurs as it may aid anesthesiologists to prevent it proactively, thus minimizing patient harm. In our study, only two articles have reported the use of AI in predicting hypoxemia. One of these studies showed that an ANN model based on three variables (i.e., Body mass index, habitual snoring, and neck circumference) was useful in predicting hypoxemia during sedation for gastrointestinal endoscopy [24]. Similarly, the other study reported that a GBM model trained on real-time data from electronic medical records of over 50000 surgeries was sufficient to predict hypoxemia [25]. Therefore, the overall results indicate that machine-learning models efficiently predict hypoxemia during anesthesia.\u003c/p\u003e \u003cp\u003eIn addition, the current study analyzed the ability of various machine learning techniques to predict PONV. The subgroup analysis showed that an SVM model had the highest discrimination ability in predicting PONV (AUROC: 0.93). Other machine-learning models, such as the ANN model, also performed relatively well in identifying patients at a high risk of PONV (AUROC: 0.77). Similarly, one of the studies reported that a fuzzy logic model was adequate to estimate the risk of postoperative vomiting in children with cancer [29]. Therefore, we believe that AI models, especially those based on SVM, ANN, and fuzzy logic, can aid anesthesiologists decide whether they should undertake preemptive measures such as preparing an antiemetic in advance or continuing with follow-up and observation.\u003c/p\u003e \u003cp\u003eThe present study also analyzed the application of AI in assisting anesthesia-related procedures. First, we evaluated the efficacy of robotics in carrying out tracheal intubations and found that robotically-assisted tracheal intubations were highly successful in both mannikins and humans. However, it is important to note that two different robotic intubation systems (Kepler intubation system (KIS) and Robotic endoscope-automated via laryngeal imaging for tracheal intubation (REALITI)) have been reported. The KIS was successful in 100% of mannikins and 91% (11/12) successful in humans. The KIS was unsuccessful in one patient because of fogging of the Pentax video laryngoscope, which is more common with this type of video laryngoscope [40]. On the other hand, evidence showed that REALITI can be used to automate tracheal intubations and can be used even by individuals without any medical training. Additionally, our results suggest that ultrasound-guided nerve blocks performed using the Magellan system are highly successful in both mannikins and humans.\u003c/p\u003e \u003cp\u003e \u003cb\u003eChallenges of integrating AI in Anesthesia.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eAlthough our study has shown that AI has enormous potential in predicting events related to anesthesia and automating procedures such as tracheal intubation and nerve blocks, there are several challenges faced when integrating AI into the anesthesia field. First, AI models, especially those based on machine learning and deep learning concepts, usually need high-quality data to work efficiently. However, obtaining this kind of data can be highly challenging. Therefore, it is difficult to guarantee data accuracy, thoroughness, and consistency. Thus, this jeopardizes patient safety due to incomplete or faulty data [41,42]. Moreover, since safeguarding patient information and privacy is paramount, the need for large amounts of data raises privacy and security concerns [43,44]. Second, AI integration in anesthesia may be hampered by several technical constraints. For instance, AI systems often rely on data used in training them and programming; thus, they can produce inaccurate predictions and recommendations if they are trained on biased or unrepresentative data [45]. Additionally, AI systems are devoid of empathy and human judgment; therefore, they may not consider patient\u0026rsquo;s concerns or anxiety regarding the recommended anesthetic strategy.\u003c/p\u003e \u003cp\u003eThird, AI systems are subject to \u0026ldquo;Black box.\u0026rdquo; This is a phenomenon where AI systems can identify patterns and make predictions but cannot explain the clinical relationship between variables [46]. Therefore, in medical fields such as anesthesiology, where it is vital to understand the physiological concepts informing a particular intervention, this constraint can create trust and transparency issues between clinicians and AI. Finally, AI integration in anesthesia might be hampered by legal and ethical issues. Several ethical concerns have been raised regarding AI. For example, if an AI makes a mistake, who is liable? Is it the anesthesiologist who used it, the programmer, or the hospital that authorized its use? The law regarding this liability matter is still ambiguous [47]. Additionally, patient consent is an ethical issue that may hinder the use of AI. Although AI can automate some tasks, the human touch is still essential in healthcare; therefore, convincing patients to consent to purely automated systems can be difficult.\u003c/p\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eThe present study is not without its limitations. First, a high interstudy heterogeneity persisted in most of the subgroup analyses. This heterogeneity might have resulted from the variation in sample sizes and the population analyzed. Second, all studies in the present review were observational, therefore increasing the risk of selection bias in the statistical analyses. Third, we considered publications available in English only; hence, the data that might have raised the statistical power of our analysis but presented in studies published in other languages had been eliminated. Finally, although AI has several applications in the anesthesiology department, we only considered two applications for analysis; therefore, we cannot make any judgment on the efficacy of AI out of these two applications.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eAI has great potential to transform the field of anesthesia. Our study has shown that AI systems using machine-learning models have the ability to predict postinduction hypotension and hypoxemia. Moreover, machine learning models, especially those based on SVM, ANN, and fuzzy logic, are adequate in predicting PONV. Therefore, AI can help anesthesiologists predict anesthesia-related events and take preemptive measures. Additionally, AI can improve tracheal intubations and ultrasound-guided nerve blocks by automating the systems. However, there are multiple barriers hindering the integration of AI in anesthesia that need to be addressed.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBellman R: An Introduction to Artificial Intelligence: Can Computers Think? Boyd \u0026amp; Fraser Publishing Company; 1978. \u003c/li\u003e\n\u003cli\u003eBellini V, Valente M, Gaddi AV, Pelosi P, Bignami E: Artificial intelligence and telemedicine in anesthesia: potential and problems. Minerva Anestesiol. 2022, 88:. 10.23736/S0375-9393.21.16241-8\u003c/li\u003e\n\u003cli\u003eHashimoto DA, Witkowski E, Gao L, Meireles O, Rosman G: Artificial Intelligence in Anesthesiology: Current Techniques, Clinical Applications, and Limitations. Anesthesiology. 2020, 132:379\u0026ndash;94. 10.1097/ALN.0000000000002960\u003c/li\u003e\n\u003cli\u003eRussell SJ, Norvig P: Artificial intelligence: a modern approach. Third edition, Global edition. Pearson: Boston Columbus Indianapolis New York San Francisco Upper Saddle River Amsterdam Cape Town Dubai London Madrid Milan Munich Paris Montreal Toronto Delhi Mexico City Sao Paulo Sydney Hong Kong Seoul Singapore Taipei Tokyo; 2016. \u003c/li\u003e\n\u003cli\u003eGlen JB: The development of \u0026lsquo;Diprifusor\u0026rsquo;: a TCI system for propofol. Anaesthesia. 1998, 53:13\u0026ndash;21. 10.1111/j.1365-2044.1998.53s115.x\u003c/li\u003e\n\u003cli\u003eLiberman MY, Ching S, Chemali J, Brown EN: A closed-loop anesthetic delivery system for real-time control of burst suppression. J Neural Eng. 2013, 10:046004. 10.1088/1741-2560/10/4/046004\u003c/li\u003e\n\u003cli\u003eLiu N, Chazot T, Hamada S, et al.: Closed-Loop Coadministration of Propofol and Remifentanil Guided by Bispectral Index: A Randomized Multicenter Study. Anesthesia \u0026amp; Analgesia. 2011, 112:546. 10.1213/ANE.0b013e318205680b\u003c/li\u003e\n\u003cli\u003eSolanki SL, Pandrowala S, Nayak A, Bhandare M, Ambulkar RP, Shrikhande SV: Artificial intelligence in perioperative management of major gastrointestinal surgeries. World Journal of Gastroenterology. 2021, 27:2758\u0026ndash;70. 10.3748/wjg.v27.i21.2758\u003c/li\u003e\n\u003cli\u003eChiew CJ, Liu N, Wong TH, Sim YE, Abdullah HR: Utilizing Machine Learning Methods for Preoperative Prediction of Postsurgical Mortality and Intensive Care Unit Admission. Annals of Surgery. 2020, 272:1133. 10.1097/SLA.0000000000003297\u003c/li\u003e\n\u003cli\u003eCorey KM, Kashyap S, Lorenzi E, et al.: Development and validation of machine learning models to identify high-risk surgical patients using automatically curated electronic health record data (Pythia): A retrospective, single-site study. PLOS Medicine. 2018, 15:e1002701. 10.1371/journal.pmed.1002701\u003c/li\u003e\n\u003cli\u003eEhrenfeld JM, Rehman MA: Anesthesia information management systems: a review of functionality and installation considerations. J Clin Monit Comput. 2011, 25:71\u0026ndash;9. 10.1007/s10877-010-9256-y\u003c/li\u003e\n\u003cli\u003eNair BG, Newman S-F, Peterson GN, Schwid HA: Smart Anesthesia Manager\u003csup\u003e\\rm TM\u003c/sup\u003e (SAM)\u0026mdash;A Real-time Decision Support System for Anesthesia Care during Surgery. IEEE Transactions on Biomedical Engineering. 2013, 60:207\u0026ndash;10. 10.1109/TBME.2012.2205384\u003c/li\u003e\n\u003cli\u003e\u0026Ccedil;orbacıoğlu ŞK, Aksel G: Receiver operating characteristic curve analysis in diagnostic accuracy studies: A guide to interpreting the area under the curve value. Turk J Emerg Med. 2023, 23:195\u0026ndash;8. 10.4103/tjem.tjem_182_23\u003c/li\u003e\n\u003cli\u003eHiggins J, Thompson S, Deeks J, Altman D: Statistical heterogeneity in systematic reviews of clinical trials: a critical appraisal of guidelines and practice. J Health Serv Res Policy. 2002, 7:51\u0026ndash;61. 10.1258/1355819021927674\u003c/li\u003e\n\u003cli\u003eHiggins JPT, Thompson SG, Deeks JJ, Altman DG: Measuring inconsistency in meta-analyses. BMJ. 2003, 327:557\u0026ndash;60. 10.1136/bmj.327.7414.557\u003c/li\u003e\n\u003cli\u003ePalla K, Hyland SL, Posner K, et al.: Intraoperative prediction of postanaesthesia care unit hypotension. Br J Anaesth. 2022, 128:623\u0026ndash;35. 10.1016/j.bja.2021.10.052\u003c/li\u003e\n\u003cli\u003eHatib F, Jian Z, Buddi S, et al.: Machine-learning Algorithm to Predict Hypotension Based on High-fidelity Arterial Pressure Waveform Analysis. Anesthesiology. 2018, 129:663\u0026ndash;74. 10.1097/ALN.0000000000002300\u003c/li\u003e\n\u003cli\u003eWijnberge M, van der Ster BJP, Geerts BF, et al.: Clinical performance of a machine-learning algorithm to predict intra-operative hypotension with noninvasive arterial pressure waveforms: A cohort study. Eur J Anaesthesiol. 2021, 38:609\u0026ndash;15. 10.1097/EJA.0000000000001521\u003c/li\u003e\n\u003cli\u003eLin C-S, Chiu J-S, Hsieh M-H, Mok MS, Li Y-C, Chiu H-W: Predicting hypotensive episodes during spinal anesthesia with the application of artificial neural networks. Comput Methods Programs Biomed. 2008, 92:193\u0026ndash;7. 10.1016/j.cmpb.2008.06.013\u003c/li\u003e\n\u003cli\u003eLin C-S, Chang C-C, Chiu J-S, et al.: Application of an artificial neural network to predict postinduction hypotension during general anesthesia. Med Decis Making. 2011, 31:308\u0026ndash;14. 10.1177/0272989X10379648\u003c/li\u003e\n\u003cli\u003eGratz I, Baruch M, Takla M, Seaman J, Allen I, McEniry B, Deal E: The application of a neural network to predict hypotension and vasopressor requirements non-invasively in obstetric patients having spinal anesthesia for elective cesarean section (C/S). BMC Anesthesiol. 2020, 20:98. 10.1186/s12871-020-01015-9\u003c/li\u003e\n\u003cli\u003eKendale S, Kulkarni P, Rosenberg AD, Wang J: Supervised Machine-learning Predictive Analytics for Prediction of Postinduction Hypotension. Anesthesiology. 2018, 129:675\u0026ndash;88. 10.1097/ALN.0000000000002374\u003c/li\u003e\n\u003cli\u003eLi X-F, Huang Y-Z, Tang J-Y, Li R-C, Wang X-Q: Development of a random forest model for hypotension prediction after anesthesia induction for cardiac surgery. World J Clin Cases. 2021, 9:8729\u0026ndash;39. 10.12998/wjcc.v9.i29.8729\u003c/li\u003e\n\u003cli\u003eGeng W, Tang H, Sharma A, Zhao Y, Yan Y, Hong W: An artificial neural network model for prediction of hypoxemia during sedation for gastrointestinal endoscopy. J Int Med Res. 2019, 47:2097\u0026ndash;103. 10.1177/0300060519834459\u003c/li\u003e\n\u003cli\u003eLundberg SM, Nair B, Vavilala MS, et al.: Explainable machine-learning predictions for the prevention of hypoxaemia during surgery. Nat Biomed Eng. 2018, 2:749\u0026ndash;60. 10.1038/s41551-018-0304-0\u003c/li\u003e\n\u003cli\u003eKim J-H, Cheon B-R, Kim M-G, Hwang S-M, Lim S-Y, Lee J-J, Kwon Y-S: Postoperative Nausea and Vomiting Prediction: Machine Learning Insights from a Comprehensive Analysis of Perioperative Data. Bioengineering. 2023, 10:1152. 10.3390/bioengineering10101152\u003c/li\u003e\n\u003cli\u003eGong C-SA, Yu L, Ting C-K, Tsou M-Y, Chang K-Y, Shen C-L, Lin S-P: Predicting postoperative vomiting for orthopedic patients receiving patient-controlled epidural analgesia with the application of an artificial neural network. Biomed Res Int. 2014, 2014:786418. 10.1155/2014/786418\u003c/li\u003e\n\u003cli\u003eShim J-G, Ryu K-H, Cho E-A, Ahn JH, Cha YB, Lim G, Lee SH: Machine learning for prediction of postoperative nausea and vomiting in patients with intravenous patient-controlled analgesia. PLoS One. 2022, 17:e0277957. 10.1371/journal.pone.0277957\u003c/li\u003e\n\u003cli\u003eBassanezi BSB, de Oliveira-Filho AG, Jafelice RSM, Bustorff-Silva JM, Udelsmann A: Postoperative vomiting in pediatric oncologic patients: prediction by a fuzzy logic model. Paediatr Anaesth. 2013, 23:68\u0026ndash;73. 10.1111/pan.12000\u003c/li\u003e\n\u003cli\u003eWu H-Y, Gong C-SA, Lin S-P, Chang K-Y, Tsou M-Y, Ting C-K: Predicting postoperative vomiting among orthopedic patients receiving patient-controlled epidural analgesia using SVM and LR. Sci Rep. 2016, 6:27041. 10.1038/srep27041\u003c/li\u003e\n\u003cli\u003eHemmerling TM, Wehbe M, Zaouter C, Taddei R, Morse J: The Kepler intubation system. Anesth Analg. 2012, 114:590\u0026ndash;4. 10.1213/ANE.0b013e3182410cbf\u003c/li\u003e\n\u003cli\u003eBiro P, Hofmann P, Gage D, et al.: Automated tracheal intubation in an airway manikin using a robotic endoscope: a proof of concept study. Anaesthesia. 2020, 75:881\u0026ndash;6. 10.1111/anae.14945\u003c/li\u003e\n\u003cli\u003eHemmerling TM, Taddei R, Wehbe M, Zaouter C, Cyr S, Morse J: First robotic tracheal intubations in humans using the Kepler intubation system. Br J Anaesth. 2012, 108:1011\u0026ndash;6. 10.1093/bja/aes034\u003c/li\u003e\n\u003cli\u003eMorse J, Terrasini N, Wehbe M, Philippona C, Zaouter C, Cyr S, Hemmerling TM: Comparison of success rates, learning curves, and inter-subject performance variability of robot-assisted and manual ultrasound-guided nerve block needle guidance in simulation. Br J Anaesth. 2014, 112:1092\u0026ndash;7. 10.1093/bja/aet440\u003c/li\u003e\n\u003cli\u003eHemmerling TM, Taddei R, Wehbe M, Cyr S, Zaouter C, Morse J: Technical communication: First robotic ultrasound-guided nerve blocks in humans using the Magellan system. Anesth Analg. 2013, 116:491\u0026ndash;4. 10.1213/ANE.0b013e3182713b49\u003c/li\u003e\n\u003cli\u003eWalsh M, Devereaux PJ, Garg AX, et al.: Relationship between Intraoperative Mean Arterial Pressure and Clinical Outcomes after Noncardiac Surgery: Toward an Empirical Definition of Hypotension. Anesthesiology. 2013, 119:507\u0026ndash;15. 10.1097/ALN.0b013e3182a10e26\u003c/li\u003e\n\u003cli\u003eMaheshwari K, Turan A, Mao G, et al.: The association of hypotension during non-cardiac surgery, before and after skin incision, with postoperative acute kidney injury: a retrospective cohort analysis. Anaesthesia. 2018, 73:1223\u0026ndash;8. 10.1111/anae.14416\u003c/li\u003e\n\u003cli\u003eDunham CM, Hileman BM, Hutchinson AE, Chance EA, Huang GS: Perioperative hypoxemia is common with horizontal positioning during general anesthesia and is associated with major adverse outcomes: a retrospective study of consecutive patients. BMC Anesthesiol. 2014, 14:43. 10.1186/1471-2253-14-43\u003c/li\u003e\n\u003cli\u003eStrachan L, Noble DW: Hypoxia and surgical patients--prevention and treatment of an unnecessary cause of morbidity and mortality. J R Coll Surg Edinb. 2001, 46:297\u0026ndash;302. \u003c/li\u003e\n\u003cli\u003eTeoh WHL, Shah MK, Sia ATH: Randomised comparison of Pentax AirwayScope and Glidescope for tracheal intubation in patients with normal airway anatomy. Anaesthesia. 2009, 64:1125\u0026ndash;9. 10.1111/j.1365-2044.2009.06032.x\u003c/li\u003e\n\u003cli\u003eRaimundo R, Ros\u0026aacute;rio A: The Impact of Artificial Intelligence on Data System Security: A Literature Review. Sensors. 2021, 21:7029. 10.3390/s21217029\u003c/li\u003e\n\u003cli\u003eHarvey HB, Gowda V: Regulatory Issues and Challenges to Artificial Intelligence Adoption. Radiol Clin North Am. 2021, 59:1075\u0026ndash;83. 10.1016/j.rcl.2021.07.007\u003c/li\u003e\n\u003cli\u003eCoppola L, Cianflone A, Grimaldi AM, et al.: Biobanking in health care: evolution and future directions. Journal of Translational Medicine. 2019, 17:172. 10.1186/s12967-019-1922-3\u003c/li\u003e\n\u003cli\u003eKeskinbora KH: Medical ethics considerations on artificial intelligence. Journal of Clinical Neuroscience. 2019, 64:277\u0026ndash;82. 10.1016/j.jocn.2019.03.001\u003c/li\u003e\n\u003cli\u003eBelenguer L: AI bias: exploring discriminatory algorithmic decision-making models and the application of possible machine-centric solutions adapted from the pharmaceutical industry. AI Ethics. 2022, 2:771\u0026ndash;87. 10.1007/s43681-022-00138-8\u003c/li\u003e\n\u003cli\u003eLopes S, Rocha G, Guimar\u0026atilde;es-Pereira L: Artificial intelligence and its clinical application in Anesthesiology: a systematic review. J Clin Monit Comput. 2024, 38:247\u0026ndash;59. 10.1007/s10877-023-01088-0\u003c/li\u003e\n\u003cli\u003eFarhud DD, Zokaei S: Ethical Issues of Artificial Intelligence in Medicine and Healthcare. Iranian Journal of Public Health. 2021, 50:i. 10.18502/ijph.v50i11.7600\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Hamad Medical Corporation","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Artificial intelligence, machine learning, neural networks, Bayesian methods, anesthesia, and anesthesiology","lastPublishedDoi":"10.21203/rs.3.rs-4599435/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4599435/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eArtificial intelligence (AI) is a multidisciplinary field focusing on expanding and generating intelligent computer algorithms to carry out simple to more complex tasks traditionally performed using human intelligence. In anesthesia, AI is rapidly becoming a transformative technology. However, its efficacy in anesthesia is still unknown. Therefore, the current study analyzed the efficacy of AI in anesthesia by studying two main applications of AI, i.e., predicting events related to anesthesia and assisting anesthesia-related procedures. Furthermore, this study explored some of the challenges of integrating AI in the anesthesia field.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003ePubMed, Google Scholar, IEEE Xplore, and Web of Science databases were thoroughly searched for articles relevant to the objective of the current study. The Comprehensive Meta-analysis software and STATA 16.0 were used for statistical analyses, while the Newcastle Ottawa Scale was used for quality evaluation.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eTwenty studies satisfying the eligibility criteria were used for review and analysis. A subgroup analysis showed that models incorporating machine learning algorithms were superior in predicting postinduction hypotension (AUROC: 0.93). ANN and SANN models also showed a good discriminatory capacity in predicting postinduction hypotension (AUROC: 0.82 and 0.80, respectively). Similarly, the subgroup analysis showed that ANN and GBM models had a good discriminatory capacity when predicting hypoxemia (AUROC: 0.8 and 0.81, respectively). Furthermore, SVM, ANN, and fuzzy logic models had a relatively good differentiation ability in predicting postoperative nausea and vomiting (AUROC: 0.93, 0.77, and 0.72, respectively). On the other hand, the subgroup analysis showed that robotically-assisted tracheal intubations were highly successful in both mannikins and humans (success rate: 98% and 92%, respectively). Similarly, robotically-assisted ultrasound-guided nerve blocks were highly successful in mannikins and humans (Success rate: 96% for humans and mannikins, respectively).\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe current study suggests that AI is useful in predicting anesthesia-related events and automating procedures such as tracheal intubation and ultrasound-guided nerve block. However, there are multiple barriers hindering the integration of AI in anesthesia that need to be addressed.\u003c/p\u003e","manuscriptTitle":"A Comprehensive Systematic Review and Meta-Analysis: Evaluating the Effectiveness and Integration Obstacles of Artificial Intelligence (AI) within Anesthesia Departments.","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-19 13:29:22","doi":"10.21203/rs.3.rs-4599435/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"94ab4a51-c0e8-4239-bdbd-a072bcb06ef3","owner":[],"postedDate":"June 19th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":33403581,"name":"Anesthesiology \u0026 Pain Medicine"}],"tags":[],"updatedAt":"2024-06-19T13:29:22+00:00","versionOfRecord":[],"versionCreatedAt":"2024-06-19 13:29:22","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4599435","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4599435","identity":"rs-4599435","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}