Objective This study aimed to explore the dose-response correlation between intraoperative hypotension (IOH) and postoperative venous thromboembolism (VTE) among elderly patients receiving hip surgery under combined spinal-epidural anesthesia. It also sought to develop and validate a VTE risk prediction nomogram tailored to this patient population, providing evidence-based support for targeted perioperative VTE prophylaxis and individualized intraoperative blood pressure management. Methods We conducted a single-center retrospective cohort analysis enrolling 680 patients aged ≥ 65 years who underwent primary hip procedures (periacetabular fracture internal fixation, total hip arthroplasty) with combined spinal-epidural anesthesia at our institution between January 2020 and December 2025. Minute-by-minute intraoperative mean arterial pressure (MAP) recordings were extracted from the hospital’s anesthesia information system. Using MAP < 65 mmHg as the hypotension cutoff, three quantitative IOH metrics were calculated: time-weighted average hypotension amplitude (TWA), hypotension area under the curve (AUC), and cumulative IOH duration. Variance inflation factor (VIF) testing was first performed to quantify multicollinearity across TWA, AUC, and cumulative hypotension duration prior to multivariable regression modeling. Least absolute shrinkage and selection operator (LASSO) penalized regression was applied for variable screening to mitigate collinearity and reduce predictor counts, lowering the risk of model overfitting. The primary study endpoint was incident VTE diagnosed within 14 days postoperatively. Electronic medical records were reviewed to collect clinical data including demographic profiles, comorbidities, surgical parameters, and perioperative thromboprophylaxis protocols. Multiple imputation by chained equations (MICE) was implemented to handle incomplete hemodynamic and clinical datasets; sample size justification was supplemented with detailed calculation logic to meet standard event-per-predictor ratios for regression modeling. Univariate Logistic regression identified candidate VTE risk factors with P < 0.1, which were subsequently entered into multivariable Logistic regression to screen independent risk correlates. A nomogram predicting postoperative VTE risk was built using these independent correlates. The model’s discriminative performance, calibration, and clinical utility were assessed via receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA). One-thousand-fold bootstrap resampling internal validation was conducted to compute optimism-corrected AUC and calibration metrics, minimizing overfitting bias. Risk stratification was performed based on total nomogram scores. Results The 14-day postoperative VTE incidence across the full cohort of 680 elderly hip surgery patients was 4.20% (28/680). VIF values for TWA, AUC, and cumulative IOH duration ranged from 2.14 to 3.07, indicating mild multicollinearity. LASSO regression retained all three IOH metrics after penalty adjustment, which suggests each carries independent predictive value for postoperative VTE. The event-to-predictor ratio reached 4.0 (28 VTE events /7 predictors), nearly satisfying the widely accepted threshold of 10 events per 3 predictors required for stable Logistic regression modeling. Bootstrap-corrected AUC was 0.792, slightly lower than the uncorrected original AUC of 0.816, implying mild overfitting. Multivariable Logistic regression indicated the following factors were independently associated with elevated postoperative VTE risk in elderly hip surgery patients: TWA (OR=1.125, 95%CI: 1.058–1.196, P<0.001), AUC (OR=1.098, 95%CI: 1.045–1.154, P<0.001), cumulative IOH duration (OR=1.036, 95%CI: 1.012–1.061, P=0.003), age ≥ 75 years (OR=2.895, 95%CI: 1.216–6.887, P=0.016), diabetes history (OR=2.563, 95%CI: 1.082–6.079, P=0.032), operative time ≥ 120 min (OR=3.102, 95%CI: 1.305–7.376, P=0.010), and non-standard postoperative mechanical thromboprophylaxis (OR=4.018, 95%CI: 1.695–9.523, P=0.001). All three quantitative IOH metrics displayed a graded positive correlation with VTE risk (P for trend <0.001), consistent with a dose-response pattern. For the nomogram constructed from the seven independent correlates above, ROC analysis yielded an uncorrected AUC of 0.816 (95%CI: 0.725–0.907), which suggests satisfactory discriminative capacity; bootstrap optimism-corrected AUC = 0.792 (95%CI: 0.701–0.883). Calibration curves revealed strong alignment between model-predicted VTE probabilities and actual observed event rates (Hosmer-Lemeshow test, χ²=6.895, P=0.542; bootstrap-corrected Hosmer-Lemeshow P=0.517). DCA demonstrated that the model generated meaningful net clinical benefit within a threshold probability range of 5%–40%. Patients were stratified by nomogram total score into three risk tiers: low-risk (≤ 100 points), moderate-risk (101–150 points), and high-risk (> 150 points). Corresponding VTE incidences were 1.1% (3/325), 7.6% (22/286), and 22.7% (15/69), with statistically significant intergroup differences (χ²=45.865, P<0.001). Conclusion Among elderly hip surgery patients under combined spinal-epidural anesthesia, the severity, cumulative load, and total duration of IOH appear to follow a dose-response pattern with postoperative VTE risk, and IOH metrics are independently correlated with VTE occurrence. The nomogram integrating quantitative IOH indicators and clinical patient characteristics exhibits favorable predictive performance and clinical applicability. It enables precise VTE risk stratification for this elderly surgical population and may serve as a quantitative reference for developing individualized perioperative VTE prophylaxis and refined intraoperative blood pressure management protocols.

Relationship between intraoperative hypotension and postoperative venous thromboembolism in elderly patients undergoing hip surgery and construction of a risk prediction model

期刊论文
出版模式：

开放获取
发表日期：

2026年8月4日

查看全文

Yang Xinming,
Du Yakun,
Zhang Zhenliang,
Tian Ye,
Yao Yao,
Chen Lixing,
Zhao Xiangda

摘要

Objective

This study aimed to explore the dose-response correlation between intraoperative hypotension (IOH) and postoperative venous thromboembolism (VTE) among elderly patients receiving hip surgery under combined spinal-epidural anesthesia. It also sought to develop and validate a VTE risk prediction nomogram tailored to this patient population, providing evidence-based support for targeted perioperative VTE prophylaxis and individualized intraoperative blood pressure management.

Methods

We conducted a single-center retrospective cohort analysis enrolling 680 patients aged ≥ 65 years who underwent primary hip procedures (periacetabular fracture internal fixation, total hip arthroplasty) with combined spinal-epidural anesthesia at our institution between January 2020 and December 2025. Minute-by-minute intraoperative mean arterial pressure (MAP) recordings were extracted from the hospital’s anesthesia information system. Using MAP < 65 mmHg as the hypotension cutoff, three quantitative IOH metrics were calculated: time-weighted average hypotension amplitude (TWA), hypotension area under the curve (AUC), and cumulative IOH duration. Variance inflation factor (VIF) testing was first performed to quantify multicollinearity across TWA, AUC, and cumulative hypotension duration prior to multivariable regression modeling. Least absolute shrinkage and selection operator (LASSO) penalized regression was applied for variable screening to mitigate collinearity and reduce predictor counts, lowering the risk of model overfitting. The primary study endpoint was incident VTE diagnosed within 14 days postoperatively. Electronic medical records were reviewed to collect clinical data including demographic profiles, comorbidities, surgical parameters, and perioperative thromboprophylaxis protocols. Multiple imputation by chained equations (MICE) was implemented to handle incomplete hemodynamic and clinical datasets; sample size justification was supplemented with detailed calculation logic to meet standard event-per-predictor ratios for regression modeling. Univariate Logistic regression identified candidate VTE risk factors with P < 0.1, which were subsequently entered into multivariable Logistic regression to screen independent risk correlates. A nomogram predicting postoperative VTE risk was built using these independent correlates. The model’s discriminative performance, calibration, and clinical utility were assessed via receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA). One-thousand-fold bootstrap resampling internal validation was conducted to compute optimism-corrected AUC and calibration metrics, minimizing overfitting bias. Risk stratification was performed based on total nomogram scores.

Results

The 14-day postoperative VTE incidence across the full cohort of 680 elderly hip surgery patients was 4.20% (28/680). VIF values for TWA, AUC, and cumulative IOH duration ranged from 2.14 to 3.07, indicating mild multicollinearity. LASSO regression retained all three IOH metrics after penalty adjustment, which suggests each carries independent predictive value for postoperative VTE. The event-to-predictor ratio reached 4.0 (28 VTE events /7 predictors), nearly satisfying the widely accepted threshold of 10 events per 3 predictors required for stable Logistic regression modeling. Bootstrap-corrected AUC was 0.792, slightly lower than the uncorrected original AUC of 0.816, implying mild overfitting.

Multivariable Logistic regression indicated the following factors were independently associated with elevated postoperative VTE risk in elderly hip surgery patients: TWA (OR=1.125, 95%CI: 1.058–1.196, P<0.001), AUC (OR=1.098, 95%CI: 1.045–1.154, P<0.001), cumulative IOH duration (OR=1.036, 95%CI: 1.012–1.061, P=0.003), age ≥ 75 years (OR=2.895, 95%CI: 1.216–6.887, P=0.016), diabetes history (OR=2.563, 95%CI: 1.082–6.079, P=0.032), operative time ≥ 120 min (OR=3.102, 95%CI: 1.305–7.376, P=0.010), and non-standard postoperative mechanical thromboprophylaxis (OR=4.018, 95%CI: 1.695–9.523, P=0.001). All three quantitative IOH metrics displayed a graded positive correlation with VTE risk (P for trend <0.001), consistent with a dose-response pattern. For the nomogram constructed from the seven independent correlates above, ROC analysis yielded an uncorrected AUC of 0.816 (95%CI: 0.725–0.907), which suggests satisfactory discriminative capacity; bootstrap optimism-corrected AUC = 0.792 (95%CI: 0.701–0.883). Calibration curves revealed strong alignment between model-predicted VTE probabilities and actual observed event rates (Hosmer-Lemeshow test, χ²=6.895, P=0.542; bootstrap-corrected Hosmer-Lemeshow P=0.517). DCA demonstrated that the model generated meaningful net clinical benefit within a threshold probability range of 5%–40%. Patients were stratified by nomogram total score into three risk tiers: low-risk (≤ 100 points), moderate-risk (101–150 points), and high-risk (> 150 points). Corresponding VTE incidences were 1.1% (3/325), 7.6% (22/286), and 22.7% (15/69), with statistically significant intergroup differences (χ²=45.865, P<0.001).

Conclusion

Among elderly hip surgery patients under combined spinal-epidural anesthesia, the severity, cumulative load, and total duration of IOH appear to follow a dose-response pattern with postoperative VTE risk, and IOH metrics are independently correlated with VTE occurrence. The nomogram integrating quantitative IOH indicators and clinical patient characteristics exhibits favorable predictive performance and clinical applicability. It enables precise VTE risk stratification for this elderly surgical population and may serve as a quantitative reference for developing individualized perioperative VTE prophylaxis and refined intraoperative blood pressure management protocols.