Enhanced Recovery After Surgery (ERAS) protocols represent a paradigm shift in perioperative care, aiming to reduce surgical stress, optimize physiological function, and accelerate recovery without compromising patient safety (1, 2). First introduced in the late 1990s for colorectal surgery, ERAS guidelines have since been widely adopted across multiple surgical specialties, including orthopaedics, gynaecology, urology, hepatobiliary, and cardiac surgery (3, 4). The success of ERAS lies in its multidisciplinary, evidence-based approach that integrates preoperative optimization, minimally invasive techniques, multimodal analgesia, and early mobilization (5, 6). Among these components, anesthesia and pain management play a pivotal role, as effective control of intraoperative and postoperative nociception directly influences recovery time, complication rates, and overall patient satisfaction (6).

In recent years, the intersection of perioperative medicine and artificial intelligence (AI) has opened new frontiers for optimizing anesthetic delivery and analgesic strategies within ERAS frameworks (7, 8). AI technologies—including machine learning (ML), deep learning (DL), and decision-support algorithms—are increasingly being deployed to enhance precision in drug dosing, predict patient responses, anticipate complications, and tailor analgesia regimens (9-11). These innovations promise not only to improve clinical outcomes but also to enable real-time adaptability, thereby aligning with ERAS objectives of individualized, patient-centred care (12).

AI-driven anesthesia systems, often integrated with automated drug delivery pumps and physiological monitoring devices, are capable of dynamically adjusting sedative and analgesic administration based on continuous data streams such as electroencephalography (EEG), hemodynamic parameters, and nociception indices (13, 14). This level of precision reduces the risk of under- or over-sedation, minimizes opioid consumption, and facilitates faster emergence from anesthesia (15). In the postoperative phase, AI-enabled pain assessment tools—leveraging facial recognition, natural language processing, and predictive modelling—support early detection of breakthrough pain and guide titration of multimodal analgesic regimens, a key ERAS tenet (16-18).

The application of AI in ERAS also extends beyond intraoperative monitoring to preoperative risk stratification and postoperative recovery prediction (19). Preoperatively, AI algorithms can analyze large datasets—including demographic, comorbidity, laboratory, and imaging variables—to predict individual risks for complications such as postoperative nausea and vomiting (PONV), delirium, or chronic post-surgical pain (8, 20). These predictions enable anesthesiologists to proactively modify anesthetic plans, select optimal agents, and personalize analgesia strategies (21). Postoperatively, AI-based mobile health (mHealth) platforms and wearable devices can track patient recovery trajectories, identify deviations from expected recovery patterns, and prompt early interventions, thereby reducing readmissions and promoting continuity of care (22, 23).

Contemporary clinical evidence increasingly supports the efficacy of AI-driven approaches in enhancing ERAS outcomes. For example, studies have demonstrated that closed-loop anaesthesia systems guided by AI can achieve tighter control of bispectral index (BIS) values, reduce anesthetic drug usage, and shorten intubation times compared to manual titration (13, 24). Similarly, AI-assisted opioid-sparing strategies—using regional anesthesia techniques combined with predictive pain modelling—have been shown to reduce postoperative opioid requirements while maintaining patient comfort (25,26). These findings align with the ERAS philosophy of minimizing pharmacological burden without compromising analgesic efficacy.

However, despite the growing body of supportive literature, the integration of AI into anaesthesia and pain management within ERAS protocols is not without challenges. Concerns include algorithm transparency, data privacy, interoperability of AI systems with existing hospital infrastructure, and the need for robust validation across diverse patient populations. Furthermore, the adoption of AI technologies requires substantial investment in infrastructure, clinician training, and workflow adaptation, which may limit accessibility in resource-constrained settings.

Given the rapid pace of technological advancement and the high stakes of perioperative care, a comprehensive synthesis of contemporary clinical evidence is essential to guide clinicians, policymakers, and researchers. This systematic review aims to critically appraise and summarize the current literature on AI-driven anaesthesia and pain management in the context of ERAS protocols, evaluating their impact on perioperative outcomes, patient safety, and healthcare resource utilization. By consolidating existing evidence, this review will provide insights into best practices, identify knowledge gaps, and highlight future directions for research and clinical implementation.

This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines (27).

SEARCH STRATEGY

A comprehensive literature search was conducted across PubMed/MEDLINE, Embase, Cochrane Central Register of Controlled Trials (CENTRAL), Web of Science, and Scopus for studies related to the topic.

The search strategy combined MeSH terms and keywords relevant to artificial intelligence (AI), anesthesia, pain management, and Enhanced Recovery After Surgery (ERAS) protocols. An example PubMed search string was:

(“Artificial Intelligence”[MeSH] OR “Machine Learning” OR “Deep Learning” OR “Predictive Analytics” OR “Closed-Loop Systems”)

AND (“Anesthesia”[MeSH] OR “Anaesthesia” OR “Pain Management” OR “Analgesia”)

AND (“Enhanced Recovery After Surgery” OR “ERAS”)

Boolean operators (AND/OR) were adapted for each database. Additional sources included Google Scholar, OpenGrey, and ClinicalTrials.gov to capture grey literature and unpublished or ongoing studies.

ELIGIBILITY CRITERIA

Studies were selected based on predefined inclusion and exclusion criteria.

Inclusion criteria:

• Population: Adults (≥18 years) undergoing surgery within an ERAS protocol.
• Intervention: AI tools including ML algorithms, predictive analytics, closed-loop anesthesia delivery systems, AI-based pain assessment, or analgesia optimization.
• Comparator: Standard anesthesia/pain management techniques without AI assistance, or alternative AI methods.
• Outcomes:

o             Primary: Perioperative hemodynamic stability, postoperative pain scores, opioid consumption, length of stay, postoperative complications.

o             Secondary: Patient satisfaction, cost-effectiveness.

• Study designs: RCTs, prospective/retrospective cohort studies, case-control studies, clinical registries.
• Language/timeframe: English-language publications.
• Exclusion criteria:
• Non-clinical or simulation-only studies, editorials, narrative reviews, conference abstracts without full data.
• Studies not involving ERAS pathways.
• Pediatric population studies.

STUDY SELECTION AND DATA EXTRACTION

All retrieved records were imported into EndNote 20 for duplicate removal. Two independent reviewers screened titles and abstracts for eligibility, followed by full-text assessment. Discrepancies were resolved through consensus or a third reviewer.

Data were extracted using a standardized form and included:

• Study characteristics (author, year, country, design)
• Patient demographics and surgical type
• AI intervention type
• Comparator method
• Outcomes and follow-up period
• Key findings and statistical results

Corresponding authors were contacted for missing or unclear data.

QUALITY ASSESSMENT

For RCTs, the Cochrane Risk of Bias 2 (RoB 2) tool was applied (28). For observational studies, the Newcastle-Ottawa Scale (NOS) was used (29). Discrepancies were resolved by discussion or third-party adjudication.

DATA SYNTHESIS

Due to expected heterogeneity in interventions and outcomes, a narrative synthesis was planned. Where sufficient homogeneity existed, a meta-analysis using a random-effects model was considered. Heterogeneity was assessed with the I² statistic (>50% indicating substantial heterogeneity).

STUDY SELECTION AND CHARACTERISTICS

FIGURE 1: PRISMA FLOWCHART

TABLE 1: CHARACTERISTICS OF INCLUDED STUDIES

TABLE 2: RISK OF BIAS ASSESSMENT – TRIALS & COHORT STUDIES

1. RCTs – Risk of Bias (RoB 2)

1. Cohort Studies – Risk of Bias (ROBINS-I)

C. Prediction/Diagnostic Studies (PROBAST)

*Mod: Moderate

  RoB 2 = Cochrane Risk of Bias tool for RCTs

  ROBINS-I = Risk of Bias In Non-randomized Studies of Interventions

  PROBAST = Prediction model Risk Of Bias ASsessment Tool

GRADE Evidence Profile Key

⊕⊕⊕⊕ High: Further research unlikely to change confidence in estimate.

⊕⊕⊕◯ Moderate: Further research may impact confidence.

⊕⊕◯◯ Low: Further research very likely to impact confidence.

⊕◯◯◯ Very Low: Any estimate is uncertain.

TABLE 4: SUBGROUP ANALYSIS SUMMARY

TABLE 5: SUBGROUP ANALYSIS – EXPANDED SUMMARY

Key Implications for ERAS:

• Intraoperative: Closed-loop systems improve consistency (High certainty) but may increase propofol use.
• Analgesia: ANI-guided automation reduces opioids (Moderate certainty) without worsening pain.
• Post-op: AI pain prediction shows promise (Low certainty), while wearables enhance recovery (Low certainty).

Recommendation: Prioritize multicenter RCTs for AI pain prediction and wearable interventions to upgrade evidence certainty.

This systematic review synthesized contemporary clinical evidence on the integration of artificial intelligence (AI)–driven anesthesia and pain management strategies within Enhanced Recovery After Surgery (ERAS) protocols. Across 13 included studies—comprising randomized controlled trials (RCTs), prospective and retrospective cohorts, diagnostic/prediction model evaluations, and narrative/scoping reviews—AI interventions demonstrated measurable improvements in intraoperative physiological control, perioperative analgesia titration, pain prediction accuracy, and rehabilitation support. The highest-certainty evidence came from closed-loop anesthetic delivery systems, which consistently increased the proportion of time patients remained within target indices for hypnosis (BIS/PSI) compared with manual administration (pooled mean difference ≈25%, I²=0%).

ANI-guided opioid titration also showed moderate-certainty evidence for reducing opioid exposure and hemodynamic instability without compromising pain scores (36). Emerging modalities, such as machine learning–based acute pain prediction and wearable-guided rehabilitation, showed promise but remain limited by small samples, retrospective designs, and lack of external validation.

Within the intraoperative phase, four RCTs (Hemmerling 2013; Puri 2016; Xie 2023; Hureau 2024) and one prospective cohort (Castellanos 2020) demonstrated that AI-based closed-loop anaesthesia can outperform manual titration in precision, stability, and standardisation of drug delivery (30, 31, 32, 35, 36). These effects were consistent across diverse surgical contexts and were reproducible in multicentre settings, suggesting scalability. However, the isolated finding of increased propofol consumption (Xie 2023) highlights the need to balance pharmacodynamic precision with drug-sparing ERAS goals (35).

For analgesia optimization, ANI-based remifentanil titration reduced cumulative opioid exposure and time spent with hemodynamic instability—an ERAS-consistent outcome that may facilitate faster postoperative recovery.

In the postoperative and rehabilitation phases, AI-driven pain prediction models achieved AUROC values of 0.81–0.98, outperforming standard clinical assessment tools (e.g., surgical pain index). Wearable-guided rehabilitation programs (Lee 2024) improved activity metrics and patient-reported outcomes (dyspnoea, pain), although confounding and historical control designs limit certainty (37).

From a clinical perspective, the deployment of closed-loop systems aligns with ERAS principles of standardized, efficient, and complication-minimizing perioperative care. Improved precision in hypnosis and analgesia may reduce anesthesia-related morbidity, facilitate hemodynamic stability, and support faster awakening and mobilization. AI-based pain prediction and wearable-guided rehabilitation expand ERAS benefits beyond the operating room, supporting proactive pain management, reducing opioid reliance, and promoting patient engagement in recovery. However, translation into routine practice will require addressing interoperability with existing monitoring systems, cost-effectiveness analyses, and staff training.

Risk of bias assessment identified generally low-to-some-concern risk among RCTs, but serious risk of bias for non-randomized and observational studies—particularly regarding confounding, selection bias, and incomplete external validation. The PROBAST assessment of prediction models consistently revealed high or moderate risk due to sample size limitations, internal-only validation, and potential overfitting. These factors constrain the certainty of evidence for postoperative AI applications compared with intraoperative closed-loop control. Heterogeneity was low for pooled closed-loop trials but moderate for AI pain prediction studies, reflecting diverse input modalities (facial analysis, wearable sensors, intraoperative vitals) and variable endpoints.

LIMITATIONS OF CURRENT EVIDENCE

Limitations include small sample sizes in several RCTs (e.g., Hureau 2024), geographic concentration of studies (notably East Asia and single high-resource centres), and the absence of large-scale pragmatic trials evaluating cost-effectiveness and workflow integration (36). Many prediction models lacked prospective validation, and wearable-based interventions remain underexplored in diverse surgical populations. Furthermore, reporting gaps—particularly regarding intervention fidelity, training requirements, and adverse event monitoring—limit the generalizability of findings.

The evidence supports integrating closed-loop anesthetic delivery into ERAS pathways for selected surgeries, with potential benefits in standardization and hemodynamic stability (31, 32, 35). ANI-guided analgesia titration may be particularly valuable in opioid stewardship strategies. Postoperative AI applications, while promising, should be prioritized for multicentre, externally validated trials incorporating diverse populations and real-world settings (36). Future studies should also explore hybrid models integrating intraoperative AI control, postoperative prediction, and wearable-guided rehabilitation within a single ERAS pathway. Cost-benefit analyses, ethical frameworks for AI use, and transparent reporting standards (e.g., CONSORT-AI, TRIPOD-AI) will be critical for sustainable adoption.

1. Ljungqvist O. Enhanced Recovery After Surgery: A paradigm shift in perioperative care. In: Enhanced Recovery After Surgery: A Complete Guide to Optimizing Outcomes. Cham: Springer International Publishing; 2020. p. 3-9.
2. Jain SN, Lamture Y, Krishna M. Enhanced Recovery After Surgery: Exploring the Advances and Strategies. Cureus. 2023;15(10):e47237.
3. Brown JK, Singh K, Dumitru R, Chan E, Kim MP. The Benefits of Enhanced Recovery After Surgery Programs and Their Application in Cardiothoracic Surgery. Methodist DeBakey Cardiovasc J. 2018;14(2):77-88.
4. Brindle M, Nelson G, Lobo DN, Ljungqvist O, Gustafsson UO. Recommendations from the ERAS® Society for standards for the development of enhanced recovery after surgery guidelines. BJS Open. 2020;4(1):157-163.
5. Mithany RH, Daniel N, Shahid MH, Aslam S, Abdelmaseeh M, Gerges F, et al. Revolutionizing Surgical Care: The Power of Enhanced Recovery After Surgery (ERAS). Cureus. 2023;15(11):e48795.
6. Girgiss C, Oladele M, Vijjeswarapu MA. Anesthesia and Pain Management in ERAS for Spine Surgery. Semin Spine Surg. 2025;101185.
7. Cao Y, Wang Y, Liu H, Wu L. Artificial Intelligence Revolutionizing Anesthesia Management: Advances and Prospects in Intelligent Anesthesia Technology. Front Med. 2025;12:1571725.
8. Yoon HK, Yang HL, Jung CW, Lee HC. Artificial intelligence in perioperative medicine: a narrative review. Korean J Anesthesiol. 2022;75(3):202-215.
9. Malheiro V, Santos B, Figueiras A, Mascarenhas-Melo F. The Potential of Artificial Intelligence in Pharmaceutical Innovation: From Drug Discovery to Clinical Trials. Pharmaceuticals. 2025;18(6):788.
10. Johnson KB, Wei WQ, Weeraratne D, Frisse ME, Misulis K, Rhee K, et al. Precision Medicine, AI, and the Future of Personalized Health Care. Clin Transl Sci. 2021;14(1):86-93.
11. Panchpuri M, Painuli R, Kumar C. Artificial intelligence in smart drug delivery systems: a step toward personalized medicine. RSC Pharmaceutics. 2025.
12. Alowais SA, Alghamdi SS, Alsuhebany N, Alqahtani T, Alshaya AI, Almohareb SN, et al. Revolutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Med Educ. 2023;23(1):689.
13. Singh M, Nath G. Artificial intelligence and anesthesia: A narrative review. Saudi J Anaesth. 2022;16(1):86-93.
14. Cai X, Wang X, Zhu Y, Yao Y, Chen J. Advances in automated anesthesia: a comprehensive review. Anesthesiol Perioper Sci. 2025;3(1):3.
15. Liao Y, Chen Z, Zhang W, Cheng L, Lin Y, Li P, et al. Artificial intelligence in perioperative pain management: A review.
16. Kim MK, Lee OH, Choi GJ, Kang H. Pain assessment using the Analgesia Nociception Index in patients undergoing sedation: a systematic review and meta-analysis. Korean J Pain. 2025;38(3):341-354.
17. El-Tallawy SN, Pergolizzi JV, Vasiliu-Feltes I, Ahmed RS, LeQuang JK, El-Tallawy HN, et al. Incorporation of “artificial intelligence” for objective pain assessment: a comprehensive review. Pain Ther. 2024;13(3):293-317.
18. Mirza M, Mehtab M, Islam B, Shahbaz S. Artificial Intelligence for Perioperative Medicine in Anesthesia: Present State and Future Perspectives of AI in Anesthesiology. Healer J Physiother Rehabil Sci. 2025;5(1):46-59.
19. Zain Z, Almadhoun MIK, Alsadoun L, Bokhari SFH. Leveraging Artificial Intelligence and Machine Learning to Optimize Enhanced Recovery After Surgery (ERAS) Protocols. Cureus. 2024;16(3):e56668.
20. Glebov M, Lazebnik T, Katsin M, Orkin B, Berkenstadt H, Bunimovich-Mendrazitsky S. Predicting postoperative nausea and vomiting using machine learning: a model development and validation study. BMC Anesthesiol. 2025;25(1):135.
21. Zeng S, Qing Q, Xu W, Yu S, Zheng M, Tan H, et al. Personalized anesthesia and precision medicine: a comprehensive review of genetic factors, artificial intelligence, and patient-specific factors. Front Med. 2024;11:1365524.
22. Amin T, Mobbs RJ, Mostafa N, Sy LW, Choy WJ. Wearable devices for patient monitoring in the early postoperative period: a literature review. mHealth. 2021;7:50.
23. Knight SR, Ng N, Tsanas A, Mclean K, Pagliari C, Harrison EM. Mobile devices and wearable technology for measuring patient outcomes after surgery: a systematic review. NPJ Digit Med. 2021;4(1):157.
24. Wingert T, Lee C, Cannesson M. Machine Learning, Deep Learning, and Closed Loop Devices-Anesthesia Delivery. Anesthesiol Clin. 2021;39(3):565-581.
25. Ghai B, Jafra A, Bhatia N, Chanana N, Bansal D, Mehta V. Opioid sparing strategies for perioperative pain management other than regional anaesthesia: A narrative review. J Anaesthesiol Clin Pharmacol. 2022;38(1):3-10.
26. Podder D, Stala O, Hirani R, Karp AM, Etienne M. Comprehensive Approaches to Pain Management in Postoperative Spinal Surgery Patients: Advanced Strategies and Future Directions. Neurol Int. 2025;17(6):94.
27. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71.
28. Nejadghaderi SA, Balibegloo M, Rezaei N. The Cochrane risk of bias assessment tool 2 (RoB 2) versus the original RoB: A perspective on the pros and cons. Health Sci Rep. 2024;7(6):e2165.
29. Margulis AV, Pladevall M, Riera-Guardia N, Varas-Lorenzo C, Hazell L, Berkman ND, et al. Quality assessment of observational studies in a drug-safety systematic review, comparison of two tools: the Newcastle-Ottawa Scale and the RTI item bank. Clin Epidemiol. 2014;6:359-368.
30. Hemmerling TM, Arbeid E, Wehbe M, Cyr S, Taddei R, Zaouter C. Evaluation of a novel closed-loop total intravenous anaesthesia drug delivery system: a randomized controlled trial. Br J Anaesth. 2013;110(6):1031-9.
31. Puri GD, Mathew PJ, Biswas I, Dutta A, Sood J, Gombar S, et al. A Multicenter Evaluation of a Closed-Loop Anesthesia Delivery System: A Randomized Controlled Trial. Anesth Analg. 2016;122(1):106-114.
32. Castellanos Peñaranda C, Casas Arroyave FD, Gómez FJ, Pinzón Corredor PA, Fernández JM, Velez Botero M, et al. Technical and clinical evaluation of a closed loop TIVA system with SEDLineTM spectral density monitoring: Multicentric prospective cohort study. Perioper Med. 2020;9(1):1.
33. De Sario GD, Haider CR, Maita KC, Torres-Guzman RA, Emam OS, Avila FR, et al. Using AI to Detect Pain through Facial Expressions: A Review. Bioengineering. 2023;10(5):548.
34. Park I, Park JH, Yoon J, Song IA, Na HS, Ryu JH, et al. Artificial intelligence model predicting postoperative pain using facial expressions: a pilot study. J Clin Monit Comput. 2024;38(2):261-270.
35. Xie T, Wang Y, Liu Y, Li J, Li W, Xu H. Accuracy of closed-loop and open-loop propofol delivery systems by bispectral index monitoring in breast surgery patients: a prospective randomized trial. Braz J Anesthesiol. 2024;74(2):744438.
36. Hureau M, Caillau E, Labreuche J, Herbet M, Tavernier B, De Jonckheere J, et al. Clinical efficacy and safety of automatic remifentanil administration based on Analgesia Nociception Index monitoring during burn surgery under propofol anesthesia: A randomized controlled clinical trial. PLoS One. 2025;20(5):e0322384.
37. Lee J, Kong S, Shin S, Lee G, Kim HK, Shim YM, et al. Wearable device–based intervention for promoting patient physical activity after lung cancer surgery: a nonrandomized clinical trial. JAMA Netw Open. 2024;7(9):e2434180.
38. Lodewyk K, Wiebe M, Dennett L, Larsson J, Greenshaw A, Hayward J. Wearables research for continuous monitoring of patient outcomes: A scoping review. PLOS Digit Health. 2025;4(5):e0000860.
39. Soley N, Speed TJ, Xie A, Taylor CO. Predicting Postoperative Pain and Opioid Use with Machine Learning Applied to Longitudinal Electronic Health Record and Wearable Data. Appl Clin Inform. 2024;15(3):569-582.
40. Zhang C, He J, Liang X, Shi Q, Peng L, Wang S, et al. Deep learning models for the prediction of acute postoperative pain in PACU for video‐assisted thoracoscopic surgery. BMC Med Res Methodol. 2024;24(1):232.
41. Ryu G, Choi JM, Seok HS, Lee J, Lee EK, Shin H, et al. Machine learning based quantitative pain assessment for the perioperative period. NPJ Digit Med. 2025;8(1):53.