1. Introduction
2. Methods
2.1. Protocol, Registration, and Reporting
2.2. Search Strategy
2.3. Eligibility Criteria
2.4. Data Extraction
2.5. Risk of Bias Assessment
2.6. Statistical Analysis
3. Results
| Review | Scope and Population | Prediction Target | Eligible Models | Included Studies | Measures Synthesized | Statistical Approach | Main Remaining Gaps |
|---|---|---|---|---|---|---|---|
| Mohammadi 2024 (23) | Any surgery; intraoperative hypotension | Hypotension at any intraoperative time, mostly minutes-ahead | HPI and other AI models | 43 (5 overlap with present review) | HPI clinical outcomes; AUROC of non-HPI models | Univariate random effects | PIH not separated; heterogeneous horizons; no accuracy pairs |
| Mehta 2024 (24) | Perioperative care | Clinical and process outcomes of ML-guided care | ML-augmented interventions | Interventional studies | Intervention effects | Meta-analysis of interventions | Not a prediction-accuracy synthesis |
| Meng 2026 (25) | Adults under GA; published after our search | PIH with explicit window | Any prediction model, including conventional regression | 17 (8 overlap); both VitalDB reports included | AUC; pooled ORs of predictors | Univariate random effects; best model per study | No pooling of sensitivity/specificity; overlapping data; predictor timing not examined |
| Present review | Adults under GA | PIH with explicit definition and window | ML, including data-driven logistic regression | 10 reports (8 independent cohorts) | AUC; joint sensitivity and specificity; predictive values | restricted maximum-likelihood with HKSJ CI and prediction interval; consistent AUC SE; bivariate model; optimism and timing analyses | Small number of cohorts; no external validation in primary studies |
3.1. Study Selection Process
3.2. Characteristics of the Included Studies
| Study | Country; design; data source (period) | Population / surgery (age, y) | N (cohort) | PIH | PIH definition (blood pressure, mm Hg); outcome window (seconds or minutes) | Evaluation set used (n; events) |
|---|---|---|---|---|---|---|
| Lin 2011 (43) | Taiwan; retrospective; single center (2007 - 2008) | Patients under GA with IV induction (age range not stated b); mixed surgery | 1,311 | 482 (36.8) | SBP < 90 mm Hg or SBP fall > 30% from baseline; induction to surgical stimulation | Temporal validation (294; 106) |
| Kang 2020 (22) c | South Korea; retrospective; single center (Oct 2018–May 2019) | Adults; laparoscopic cholecystectomy under TIVA | 222 | 126 (56.8) | SBP < 90 or MBP < 65 mm Hg (or vasopressor use); tracheal intubation to incision | Repeated 4-fold CV (222; 126) |
| Lee 2020 (44) c | South Korea; retrospective; single center (Oct 2018–Jul 2019) | Adults; laparoscopic cholecystectomy under GA | 282 | 151 (53.5) | SBP < 90 or MBP < 65 mm Hg; tracheal intubation to incision (prediction 1 min before intubation) | Repeated 10-fold CV (282; 151) |
| Li 2021 (20) | China; retrospective; single center (2007 - 2018) | Adults; cardiac surgery | 3,030 | 1,578 (52.1) | SBP < 90 or MBP < 65 mm Hg; induction to start of surgery | Held-out test set (909; 490) d |
| Yoshimura 2022 (41) | Japan; retrospective; single center (2014 - 2019) | Adults with preoperative echocardiography; non-cardiac, non-obstetric surgery | 1,956 | 670 (34.3) | Any MAP < 55 mm Hg; induction to start of surgery | Held-out test set (587; ≈200) d |
| Park 2025 (4)§ | South Korea; retrospective; open registry VitalDB (2016 - 2017) | Adults with arterial line; non-cardiac surgery | 3,669 | 2,321 (63.3) | Sustained MAP < 65 mm Hg (≥ 6 s); 20 min after induction or incision | Held-out test set (1, 101, 672) |
| Chen 2025 (11)§ | China (Korean data); retrospective; open registry VitalDB | Non-cardiac surgery (age restriction not reported) | 5,406 | 921 (17.0) | Any MAP < 55 mm Hg; induction to start of surgery | Held-out test set (1,622; ≈276) d |
| Bian 2025 (45) | China; prospective cohort; single center (Dec 2022–Aug 2023) | Adults ≥ 65 y; elective non-cardiac surgery | 938 | 481 (51.3) | MAP ≤ 65 mm Hg or MAP fall > 30% from ward baseline; 15 min after induction or incision | Temporal validation (281; 142) |
| Katsin 2025 (21) | Israel; retrospective; single center (2018 - 2023) | Adults; non-obstetric inpatient surgery with IV induction | 20,309 | 4,948 (24.4) | Any MAP < 55 mm Hg; 10 min after completion of induction | 5-fold CV / 20% hold-out, unclear (4,062; ≈991) d |
| Sarullo 2025 (42) | USA; retrospective; two-center registry (2005 - 2017) | Adults; non-cardiac surgery under GA | 201,000 | NR ("relatively even") | MAP < 65 mm Hg; first 10 min after induction | 5-fold CV (201,000; NR) |
3.3. Risk of Bias
| Study | Selected Model (Family) | Predictor Timing b | AUC (95% CI), as Reported | Sensitivity | Specificity | Threshold Selection | Calibration / DCA | Overall ROB |
|---|---|---|---|---|---|---|---|---|
| Lin 2011 (43) | ANN (neural network (NN)) | Tier 2 | 0.893 (NR) | 0.764 | 0.856 | NR | H-L test, plot / No | High |
| Kang 2020 (22) | Random forest (tree) | Tier 3 | 0.842 (0.736 - 0.948) | 0.837 | 0.745 c | NR | No / No | High |
| Lee 2020 (44) | Random forest (tree) | Tier 3 | NR | 0.784 | 0.711 c | NR | No / No | High |
| Li 2021 (20) | Random forest (tree) | Tier 2 | 0.843 (0.808 - 0.877) | 0.788 | 0.856 | NR | No / No | High |
| Yoshimura 2022 (41) | Deep NN | Tier 1 | 0.72 (0.67 - 0.76) | 0.650 | 0.700 | Youden index | No / Yes | High |
| Park 2025 (4) | XGBoost (tree) | Tier 3 | 0.772 (0.771 - 0.774) d | 0.688 | 0.714 | Fixed 0.5 | No / No | High |
| Chen 2025 (11) | Logistic regression (regression) | Tier 1 | 0.74 (0.71 - 0.77) | 0.700 | 0.68 c | Youden index | Curve, Brier / Yes | Unclear |
| Bian 2025 (45) | Logistic regression (regression) | Tier 1 | 0.697 (0.635 - 0.759) | 0.634 | 0.698 | NR | Curve, H-L, Brier / Yes | High |
| Katsin 2025 (21) | Random forest (tree) | Tier 2 | 0.732 (NR) | 0.760 | NR e | NR | No / No | High |
| Sarullo 2025 (42) | XGBoost (tree) | Tier 2 | 0.697 (NR) | NR | NR | NA | No / No | High |




