ABSTRACT
Background and Aims:
Perioperative anaesthesia is progressively integrating decision support systems based on artificial intelligence (AI). Safe AI use requires not only technical validation but also ethical preparedness and trust to uphold patient autonomy, informed consent, and clinician oversight. This study evaluated ethical readiness and trust towards AI integration among anaesthesia professionals using a cross-sectional survey.
Methods:
Anaesthesia professionals across Indian government, private, and corporate hospitals completed a structured, content-validated, and pilot-tested self-administered questionnaire assessing five domains. AI Ethical Trust and Readiness Index (AETRI) was derived as a composite score. Group differences, associations, and predictors were analysed using Mann–Whitney U/Kruskal–Wallis tests, Pearson's correlation, and multiple linear regression, respectively.
Results:
A total of 255 anaesthesia professionals participated, including 44.7% junior residents, 36.9% senior residents, 11.0% consultants, and 7.4% faculty. Most (69.4%) worked in government hospitals. The mean ± standard deviation domain scores were: Knowledge (K) 0.73 ± 0.21, Utility and Trust (UT) 3.41 ± 0.56, Ethical Readiness (ER) 3.62 ± 0.61, Transparency and Control (TC) 3.66 ± 0.64, and Institutional Support (IS) 1.86 ± 0.58. The mean AETRI was 57.2 ± 9.43, indicating moderate readiness. Knowledge varied by designation (P = 0.029), with higher scores among faculty and consultants. Faculty demonstrated higher AETRI than residents (P = 0.011). ER and TC were the main predictors of AETRI (R2 = 0.52).
Conclusion:
Anaesthesia professionals demonstrated moderate AI knowledge and trust but limited institutional readiness. Ethical readiness and transparency, rather than technical knowledge alone, are the primary determinants of clinician preparedness for AI integration.
Keywords: Anaesthesiology, artificial intelligence, clinical decision support systems, ethics, perioperative care
INTRODUCTION
Artificial intelligence (AI) is increasingly being used in perioperative practice, including hypotension prediction and anaesthesia decision-support. Early studies suggest potential benefit in haemodynamic management, although broader validation using patient-centred outcomes is still needed.[1,2] As AI moves into clinical practice, anaesthesiologists need to interpret and critically assess its recommendations. However, many AI models functioning as “black box” systems with limited explainability can make unexpected outputs difficult to understand and justify, creating ethical and professional challenges.[3] Uncertainty about disclosing AI use may affect patient autonomy and informed consent,[4] while unclear responsibility can lead to overreliance on AI or its complete rejection.[5] Although clinicians remain cautiously optimistic, concerns persist about data quality, bias, transparency, medicolegal issues, and skill erosion.[6] Reviews in perioperative and critical care settings similarly emphasise the need for continued human oversight, highlighting gaps in ethical governance, data protection, fairness, and accountability even when AI performs well in trials.[7] However, evidence on how anaesthesia professionals, particularly in low- and middle-income settings, perceive and ethically engage with AI remains limited. Existing research has focussed mainly on AI literacy rather than ethical readiness, which includes judging contextual appropriateness, overriding AI outputs, documenting its use, and communicating risks to patients.[8] Variability in digital systems, training programmes, infrastructure, and regulatory frameworks may further influence India’s readiness for AI-enabled perioperative care.
Prior research on AI in anaesthesia has largely focussed on technical performance or general attitudes, with limited evidence on ethical preparedness. In India, diverse healthcare systems, variable digital infrastructure, and evolving regulations emphasise the need to understand clinicians’ trust, preparedness, and support requirements. This study addresses these gaps through AI Ethical Trust and Readiness Index (AETRI). It assesses anaesthesia professionals’ knowledge, attitude, trust, transparency, ethical readiness, and institutional support regarding AI-assisted decision-making. It also examines differences across professional and institution groups, relationship with AETRI, and perceived barriers to AI adoption.
METHODS
The study is a descriptive, cross-sectional, questionnaire-based survey of anaesthesia professionals in India. The analyses were confined to explaining patterns and associations; no causal or deterministic relationships were inferred given the cross-sectional design. The institutional ethics committee provided the ethical clearance for the study (AIIMSA-4746/2025). The study employed a non-probability accessibility sampling approach with elements of snowball dissemination using institutional mailing lists, professional networks, and social media groups. The survey link was disseminated through institutional mailing lists, departmental contacts, professional anaesthesia networks, and anaesthesia-focussed social media groups, including WhatsApp-based academic discussion forums. Because the survey link was openly forwardable across multiple digital platforms, the exact number of institutions, groups, or individuals exposed to the survey could not be reliably quantified. Consequently, a conventional denominator and response rate could not be calculated. To minimise duplicate responses, only one submission per Google account was permitted, sign-in was required for survey submission, and a response to all items was required for submissions. Email identifiers were not exported with responses. Internet Protocol (IP) addresses were not collected. Aggregate metadata were screened for obvious duplicate entries, and no apparent duplicate submissions were identified. All responses were collected anonymously in accordance with institutional ethical standards from September 2025 to the end of December 2025.
The study included anaesthesia professionals (senior residents, consultants, faculty, and junior residents with at least 6 months of supervised clinical anaesthesia experience) who were currently involved in perioperative care, teaching, or decision-making. Only respondents who could understand the English language were eligible. “Prior AI exposure” was defined a priori as one or more of the following: a) Formal training (structured workshop/course/certificate), b) Clinical use (AI-driven monitoring/decision support), c) Academic exposure (research/teaching), and d) General-use tools (e.g., chat-based, analytic tools, large language models). Undergraduate students, interns, non-anaesthesia specialists, and individuals exclusively in administrative roles without active clinical responsibilities were not eligible to participate. In the absence of prior Indian data on ethical readiness towards AI in anaesthesia, we assumed a conservative expected proportion of 50%. We calculated the sample size using the OpenEpi online sample size calculator (version 3) for single proportions. With a 95% confidence level, 10% absolute precision, and a design effect of 1.0, the software estimated a minimum sample size of 97. To account for potential incomplete or unusable responses, we inflated the target sample size to approximately 100–120 participants. Through broad dissemination, we obtained 255 complete post-pilot responses, increasing the precision of estimates.
A content-validated and pilot-tested self-administered questionnaire was developed based on literature on AI in anaesthesia and critical care, ethical frameworks for medical AI (including autonomy, beneficence, non-maleficence, and justice),[9] World Health Organization (WHO) digital health ethics guidance,[10] and existing surveys assessing clinician attitudes towards AI in perioperative medicine.[6] The items were organised into five domains: Knowledge (K), capturing conceptual understanding of AI tools; Utility and Trust (UT), assessing perceived usefulness, reliability, and willingness to incorporate AI outputs; Ethical Readiness (ER), evaluating comfort with explaining AI use, assessing fairness, and balancing AI recommendations with clinical judgment; Transparency and Control (TC), addressing concerns about “black-box’’ behaviour, explainability, and clinician override; and Institutional Support (IS), reflecting training availability, access to AI platforms, legal clarity, and departmental culture. As ethical readiness for AI-assisted anaesthesia is considered multidimensional, including individual knowledge, trust, ethical judgement, supervisory control, and institutional preparedness, these domains were prioritised. Response formats included dichotomous items, multiple-choice items, and 5-point Likert scales (strongly disagree to agree strongly).
A multidisciplinary panel of senior anaesthesiologists and AI researchers assessed content validity for relevance, clarity, and representativeness. The questionnaire was pilot-tested in 55 participants to ensure clarity and reasonable completion time (≤ 15 min). Cronbach’s alpha was 0.53 for UT, 0.50 for ER, and 0.69 for TC, indicating limited to moderate internal consistency. UT and ER addressed broader concepts of trust and ethical readiness, which likely contributed to the lower alpha values. Cronbach’s alpha was not assessed for the K and IS domain because of their response formats. The pilot data were excluded from the final analysis. The questionnaire was considered content-validated and pilot-tested but not fully psychometrically validated. Given the exploratory nature of the instrument and the multidimensional constructs assessed, this pilot-testing provided an appropriate initial assessment of its internal consistency.
Anonymised data were exported from the Google Forms to Microsoft Excel and analysed using Jamovi (version 2.6). Continuous variables were summarised as mean ± standard deviation (SD) or median [interquartile range (IQR)] based on the Shapiro–Wilk test, and categorical variables as frequencies and percentages. K was calculated as the mean of four binary items (0–1), while scores for UT, ER, TC and IS were derived from their respective 5-point Likert items. These domains were combined into the AETRI and expressed as a percentage (0–100%). Participants could select multiple barriers and enablers from a list of relevant potential factors. Group differences were assessed using the Mann–Whitney U or Kruskal–Wallis tests, as appropriate. The associations among domains were examined using Pearson’s correlation coefficient, and independent predictors of AETRI were identified using multiple linear regression. Statistical significance was set at P < 0.05.
RESULTS
A total of 255 anaesthesia professionals completed the survey. The mean age of the respondents was 31.4 ± 6.3 years, and 53.7% were female. The cohort comprised junior residents (44.7%), senior residents (36.9%), consultants (11.0%), and faculty (7.4%). Most respondents worked in government hospitals (69.4%), followed by private (17.6%) and corporate/academic institutions (13.0%). 61.2% of participants showed prior AI exposure [Table 1]. Participants revealed moderate to high scores in most domains, except IS, which was lowest across the cohort. The composite AETRI mean score was 57.2 ± 9.43, indicating moderate overall readiness [Table 2].
Table 1.
Baseline characteristics of respondents
| Variable | Summary |
|---|---|
| Age (years) | 31.4±6.3 (range 22–52) |
| Gender | Male 118 (46.3%), Female 137 (53.7%) |
| Designation | Junior resident 44.7%, Senior resident 36.9%, Consultant 11.0%, Faculty 7.4% |
| Experience (years) | Mean 4.9±5.2 |
| Institution type | Government 69.4%, Private 17.6%, Corporate/Academic 13.0% |
| Prior AI exposure | Yes 61.2%, No 38.8% |
Data in Mean±SD (standard deviation). AI: Artificial intelligence
Table 2.
Domain-wise scores of AI readiness and trust
| Domain | Scores | Scores |
|---|---|---|
| Knowledge (K) | 0.73±0.21 | 0.75 [0.50–1.00] |
| Utility and Trust (UT) | 3.41±0.56 | 3.40 [3.00–3.80] |
| Ethical Readiness (ER) | 3.62±0.61 | 3.60 [3.20–4.00] |
| Transparency and Control (TC) | 3.66±0.64 | 3.75 [3.25–4.25] |
| Institutional Support (IS) | 1.86±0.58 | 1.75 [1.50–2.25] |
Data are expressed in Mean±SD (standard deviation), and Median [IQR Interquartile Range]. AI: Artificial intelligence
Mann–Whitney U tests presented no statistically significant differences in any domain scores or in AETRI between participants with prior AI exposure and those without. Kruskal–Wallis test showed a significant difference in K scores across designation groups (P = 0.029), with higher scores among faculty and consultants [Table 3].
Table 3.
Comparison of domain scores by designation (Kruskal–Wallis test)
| Domain | JR (n=114) | SR (n=94) | Consultant (n=28) | Faculty (n=19) | P |
|---|---|---|---|---|---|
| Knowledge | 0.68±0.22 | 0.81±0.19 | 0.85±0.18 | 0.91±0.16 | 0.029 |
| Utility and Trust | 3.34±0.60 | 3.40±0.54 | 3.46±0.62 | 3.58±0.55 | 0.705 |
| Ethical Readiness | 3.56±0.58 | 3.73±0.59 | 3.68±0.60 | 3.85±0.65 | 0.229 |
| Transparency and Control | 3.58±0.71 | 3.72±0.70 | 3.66±0.67 | 3.90±0.69 | 0.182 |
| Institutional Support | 1.77±0.58 | 1.84±0.59 | 1.91±0.65 | 1.95±0.57 | 0.655 |
Data are presented as mean±SD. Between-group comparisons were performed using the Kruskal–Wallis test. A two-tailed P<0.05 was considered statistically significant. JR: Junior resident; SR: Senior resident; SD: Standard deviation
Kruskal–Wallis test showed no significant differences across government, private, and corporate/academic institutions for any domain scores or AETRI. Pearson's correlations revealed a moderate positive correlation between ER and TC, and ER and UT and weak but significant correlations between K and the other domains, and institutional support showed weak positive correlations [Figure 1].
Figure 1.
Inter-domain correlation heatmap. Matrix-style plot with correlation coefficients (Pearson's r)
The predictors account for 52% of the variance in AETRI, according to the model’s R² = 0.52 [Table 4]. A categorical analysis of a few variables demonstrated common themes: Barriers included limited access to AI tools, lack of training, and medico-legal uncertainty; enablers included structured training, clear guidelines, and institutional support [Figure 2].
Table 4.
Multivariable regression model predicting AETRI
| Predictor Variable | β (Standardised) | P |
|---|---|---|
| Ethical Readiness | 0.44 | <0.001 |
| Transparency and Control | 0.31 | 0.004 |
| Utility and Trust | 0.12 | 0.132 |
| Knowledge | 0.09 | 0.218 |
| Institutional Support | 0.07 | 0.266 |
| Model summary: R² = 0.52, P<0.001 |
β denotes standardised regression coefficients. Multicollinearity was assessed prior to model fitting and was within acceptable limits. Model fit is expressed as R². All tests were two-tailed, with P<0.05 considered statistically significant. AETRI: Artificial intelligence ethical trust and readiness index
Figure 2.
Perceived barriers to AI adoption among anaesthesia professionals. AI: Artificial intelligence
DISCUSSION
With applications expanding from automated monitoring to haemodynamic prediction, image analysis, and clinical decision support, AI is gradually driving perioperative practice.[11,12,13,14] When technologies begin to direct intraoperative management, successful integration requires not only technical reliability but also clinical preparedness, ethical awareness, and institutional readiness.[15,16,17] Here, we underscored prevailing insights and structural barriers by providing a distinctive assessment of ethical trust and readiness for AI adoption among Indian anaesthesia professionals. The majority of participants expressed favourable opinions and a reasonable understanding of AI’s potential, which is in line with international data demonstrating interest in predictive algorithms and AI-assisted monitoring.[6,16,18] However, prior AI exposure did not influence knowledge or AETRI scores. This finding may reflect the nature of current ‘AI exposure’ which may represent superficial or non-clinical interactions such as chat-based tools or passive conference exposure.[3,19] Such exposure does not necessarily translate into clinical understanding, ethical reasoning, or decision-making competence. Therefore, to improve readiness, structured, context-specific training is likely needed rather than passive exposure. Readiness appears to depend less on familiarity and more on structured training. The high perceived UT in our cohort mirrors international findings that clinicians believe AI can reduce workload and augment vigilance.[16,18]
A key finding of this study is that ER and TC were the main predictors of AETRI, suggesting that readiness is more strongly shaped by clinicians’ confidence in supervising AI recommendations than by baseline knowledge. ER encompasses fairness, accountability, bias, and shared decision-making, issues repeatedly emphasised in recent reviews of AI in anaesthesia.[20,21] Transparency relates to explainability, interpretability, and the ability to identify algorithmic errors. Studies have shown that opaque “black-box’’ systems undermine clinician trust, hinder risk communication, and complicate informed consent.[4,22,23] Longstanding calls for explainable AI (XAI), clear override mechanisms, and transparent documentation pathways to minimise automation bias are supported by the predictive power of ER and TC.[5,21,22]
The IS domain had the lowest score in the current survey, indicating systemic barriers to AI adoption. Participants noted concerns about a lack of medico-legal clarity, inadequate training pathways, and limited access to validated tools; these issues were echoed in reviews conducted across multiple health systems.[15,24] When AI is used to support clinical decisions, even high-income regions acknowledge the uncertainty surrounding governance, operational responsibility, and liability.[5,22,23] These difficulties are intensified in low- and middle-income environments by variable digital infrastructure and evolving legal frameworks.[9,10] In our study, faculty had higher knowledge and AETRI scores, possibly reflecting their greater clinical experience and exposure to medico-legal and academic discussions concerning AI. Studies from Canada, Europe, the United States, and the Middle East have shown similar designation-linked disparities.[6,16,18] The predominance of residents and the mean age of 31.4 ± 6.3 years reflect that the greater participation of younger professionals may partly be related to higher engagement with the online platforms through which the questionnaire was circulated; however, this was not specifically assessed. The lack of variations among institution types implies that barriers to AI readiness are not sector-specific but rather widespread.
The strong correlation between ER and Transparency (r=0.47) indicates that clinicians who value ethical governance also prioritise explainability. This is consistent with the behavioural science literature that demonstrates how transparency builds trust, supports accountability, and facilitates the acceptance of AI-derived recommendations.[17,22,25] The fact that UT and ethical domains have moderate correlations highlights that trust in AI is fundamentally ethical rather than just technical.
Overall, the findings of our study support the global evidence that safe AI integration requires much more than just clinical interest. When AI is used to guide clinical decisions, reviews often highlight the risks of biased outputs, challenges with informed consent, and persistent uncertainties about accountability and liability.[20,21,22,23,24,26] Complexity is increased in India due to variations in infrastructure, documentation practices, and regulatory clarity. These findings support the need for national guidance, institutional governance frameworks, and ethics-focussed training to ensure the safe and responsible use of AI in perioperative care.[5,16,17,20,21,27,28] While the knowledge, attitudes, and ethical concerns regarding AI have been addressed in previous studies, the novelty of our study focusses on ethical readiness, transparency, clinical judgment, and institutional support as distinct dimensions among Indian anaesthesia professionals. It has highlighted the domain-specific gaps, providing a practical baseline for assessing preparedness for integrating AI-assisted tools into anaesthesia practice.
The survey findings have several limitations that should be addressed. First, selection bias may have been introduced by using a non-probability sampling technique with online distribution, which would have favoured participation by individuals with greater interest in or familiarity with AI. It was not possible to determine the precise number of people who had access to the survey because it used open online dissemination with forwardable links across social media groups and professional networks. Therefore, a conventional response rate and formal assessment of non-response bias were not feasible. The limited representation of senior practitioners may not adequately reflect their views. Although the survey was disseminated nationally, it was not geographically stratified; therefore, balanced representation across regions and cities could not be assumed. The cohort featured a higher proportion of respondents from government institutions, which might have reduced the representation of viewpoints from corporate and private healthcare settings. The results are based on self-reported responses from participants, which may have been influenced by personal bias or a propensity to provide socially acceptable answers. Although the questionnaire underwent content validation, pilot testing, and exploratory internal consistency assessment, comprehensive psychometric evaluation including construct validity assessment, factor analysis, and test–retest reliability was not undertaken. Therefore, the instrument should be viewed as exploratory and may require further validation before larger longitudinal or multicentre deployment. Finally, variations in infrastructure, training systems, and regulatory frameworks across regions may influence the applicability of our results to different healthcare settings.
CONCLUSION
In conclusion, this study suggests that anaesthesia professionals in India are broadly receptive to AI-assisted care in perioperative practice but report only moderate overall readiness and limited institutional support. Ethical Readiness and Transparency and Control were the main predictors of AETRI, highlighting the importance of clinician oversight, explainability, and ethical judgement alongside technical competence. However, given the cross-sectional design, and limited psychometric validation of the questionnaire, these associations should be interpreted cautiously and should not be considered causal. Further multicentre studies using robustly validated instruments and longitudinal or implementation-based approaches are needed to determine whether structured ethics-focused training, clearer institutional governance, and defined accountability mechanisms can translate into greater readiness for safe and responsible AI use in anaesthesia.
Author contributions
TJ, RS, and PG conceptualised the study and designed the study protocol. TJ and KA collected the data and prepared the initial manuscript draft. RS and PG analysed the data and critically revised the manuscript. SSN and PS contributed to the final review of the manuscript. All authors approved the final version for publication.
Presentation at conferences/CMEs and abstract publication
This study has not been presented at any conference or continuing medical education programme, and its abstract has not been published elsewhere.
Study data availability
De-identified data may be requested with reasonable justification from the authors (email to the corresponding author) and shall be shared after approval as per the authors’ Institution policy.
Declaration of AI use
The authors have used AI-based language model solely for grammatical correction, proofreading, and stylistic polishing of the manuscript’s text. All study design, data collection, statistical modelling, and clinical interpretation were independently elaborated and executed by the authors without algorithmic assistance.
Declaration of use of permitted tools
No copyrighted instrument, scale, figure, or other tool requiring permission or licensing was used.
Conflicts of interest
There are no conflicts of interest.
Acknowledgement
Nil.
Funding Statement
Nil.
REFERENCES
- 1.Wijnberge M, Geerts BF, Hol L, Lemmers N, Mulder MP, Berge P, et al. Effect of a machine learning-derived early warning system for intraoperative hypotension vs standard care on depth and duration of intraoperative hypotension during elective non-cardiac surgery: The HYPE randomised clinical trial. JAMA. 2020;323:1052–60. doi: 10.1001/jama.2020.0592. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Schuurmans J, Rellum SR, Schenk J, van der Ster BJP, van der Ven WH, Geerts BF, et al. Effect of a machine learning-derived early warning tool with treatment protocol on hypotension during cardiac surgery and ICU stay: The Hypotension Prediction 2 (HYPE-2) randomised clinical trial. Crit Care Med. 2025;53:e328–40. doi: 10.1097/CCM.0000000000006518. [DOI] [PubMed] [Google Scholar]
- 3.Ocampo Osorio F, Alzate-Ricaurte S, Mejia Vallecilla TE, Cruz-Suarez GA. The anaesthesiologist's guide to critically assessing machine-learning research: A narrative review. BMC Anesthesiol. 2024;24:452. doi: 10.1186/s12871-024-02840-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Park HJ. Patient perspectives on informed consent for medical AI: A web-based experiment. Digit Health. 2024;10:20552076241247938. doi: 10.1177/20552076241247938. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Price WN, 2nd, Gerke S, Cohen IG. Liability for use of artificial intelligence in medicine. In: Solaiman B, Bin Khalifa H, Cohen IG, editors. Research Handbook on Health, AI and the Law. Cheltenham (UK): Edward Elgar Publishing; 2024. pp. p. 150–66. [Google Scholar]
- 6.Henckert D, Malorgio A, Schweiger G, Raimann FJ, Piekarski F, Zacharowski K, et al. Attitudes of anaesthesiologists toward artificial intelligence in anaesthesia: A multicenter, mixed qualitative-quantitative study. J Clin Med. 2023;12:2096. doi: 10.3390/jcm12062096. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Sarkhosh M, Kiaei MM, Aligholizadeh M, Sangi S, Akbarpour P, Jalalkamali E. Ethical implications of artificial intelligence in anaesthesiology: A scoping review. Arch Anesthesiol Crit Care. 2025;11((Suppl 1)):749–56. [Google Scholar]
- 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. doi: 10.1097/ALN.0000000000002960. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Beauchamp T, Childress J. Principles of biomedical ethics: Marking its fortieth anniversary. Am J Bioeth. 2019;19:9–12. doi: 10.1080/15265161.2019.1665402. [DOI] [PubMed] [Google Scholar]
- 10.World Health Organization. Ethics and Governance of Artificial Intelligence for Health. Geneva: World Health Organization; 2021. [[Last accessed on 2026 Jan 14]]. Available from: https://www.who.int/publications/i/item/9789240029200 . [Google Scholar]
- 11.Lai CJ, Cheng YJ, Han YY, Hsiao PN, Lin PL, Chiu CT, et al. Hypotension prediction index for prevention of intraoperative hypotension in patients undergoing general anaesthesia: A randomised controlled trial. Perioper Med (Lond) 2024;13:57. doi: 10.1186/s13741-024-00414-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kouz K, Monge García MI, Cerutti E, Lisanti I, Draisci G, Frassanito L, et al. Intraoperative hypotension when using hypotension prediction index software during major non-cardiac surgery: A European multicentre prospective observational registry (EU HYPROTECT) BJA Open. 2023;6:100140. doi: 10.1016/j.bjao.2023.100140. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Côté CD, Kim PJ. Artificial intelligence in anesthesiology: Moving into the future. Univ Tor Med J. 2019;96:33–6. [Google Scholar]
- 14.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. doi: 10.1007/s10877-023-01088-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Kumar A, Taneja A, Singh YP, Pratap Singh G, Jain S, Meena S, et al. Artificial intelligence in critical care: Promise, peril, and the path forward. Cureus. 2025;17:e85323. doi: 10.7759/cureus.85323. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Dost A, Alaraj R, Mayet R, Agrawal DK. Reshaping anaesthesia with artificial intelligence: From concept to reality. Anesth Crit Care. 2025;7:77–90. [PMC free article] [PubMed] [Google Scholar]
- 17.Ennab M, Mcheick H. Enhancing interpretability and accuracy of AI models in healthcare: A comprehensive review on challenges and future directions. Front Robot AI. 2024;11:1444763. doi: 10.3389/frobt.2024.1444763. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Morone G, De Angelis L, Martino Cinnera A, Carbonetti R, Bisirri A, Ciancarelli I, et al. Artificial intelligence in clinical medicine: A state-of-the-art overview of systematic reviews with methodological recommendations for improved reporting. Front Digit Health. 2025;7:1550731. doi: 10.3389/fdgth.2025.1550731. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Sidey-Gibbons JA, Sidey-Gibbons CJ. Machine learning in medicine: A practical introduction. BMC Med Res Methodol. 2019;19:64. doi: 10.1186/s12874-019-0681-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Ye J. Patient safety of perioperative medication through the lens of digital health and artificial intelligence. JMIR Perioper Med. 2023;6:e34453. doi: 10.2196/34453. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Amann J, Blasimme A, Vayena E, Frey D, Madai VI. Explainability for artificial intelligence in healthcare: A multidisciplinary perspective. BMC Med Inform Decis Mak. 2020;20:310. doi: 10.1186/s12911-020-01332-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Ghassemi M, Oakden-Rayner L, Beam AL. The false hope of current approaches to explainable artificial intelligence in healthcare. Lancet Digit Health. 2021;3:e745–50. doi: 10.1016/S2589-7500(21)00208-9. [DOI] [PubMed] [Google Scholar]
- 23.Hamet P, Tremblay J. Artificial intelligence in medicine. Metabolism. 2017;69S:S36–40. doi: 10.1016/j.metabol.2017.01.011. [DOI] [PubMed] [Google Scholar]
- 24.Leslie D. Understanding Artificial Intelligence Ethics and Safety: A Guide for the Responsible Design and Implementation of AI Systems in the Public Sector. London: The Alan Turing Institute; 2019. [Google Scholar]
- 25.Trolice MP, Curchoe C, Quaas AM. Artificial intelligence-the future is now. J Assist Reprod Genet. 2021;38:1607–12. doi: 10.1007/s10815-021-02272-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Gondode P, Duggal S, Mahor V. Artificial intelligence hallucinations in anaesthesia: Causes, consequences and countermeasures. Indian J Anaesth. 2024;68:658–61. doi: 10.4103/ija.ija_203_24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Kurdi MS, Gupta B. Mirror, mirror on the wall, can artificial intelligence predict it all? Indian J Anaesth. 2026;70:512–5. doi: 10.4103/ija.ija_523_26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Myatra SN, Jagiasi BG, Singh NP, Divatia JV. Role of artificial intelligence in haemodynamic monitoring. Indian J Anaesth. 2024;68:93–9. doi: 10.4103/ija.ija_1260_23. [DOI] [PMC free article] [PubMed] [Google Scholar]