高宇晨
中国医学科学院阜外医院 麻醉科
Background:To derive and validate a machine learning (ML) prediction model of acute kidney injury (AKI) that could be used for AKI surveillance and management to improve clinical outcomes.Methods:This retrospective cohort study was conducted in Fuwai Hospital, including patients aged 18 years and above undergoing cardiac surgery admitted between January 1, 2017, and December 31, 2018. Seventy percent of the observations were randomly selected for training and the remaining 30% for testing. The demographics, comorbidities, laboratory examination parameters, and operation details were used to construct a prediction model for AKI by logistic regression and eXtreme gradient boosting (Xgboost). The discrimination of each model was assessed on the test cohort by the area under the receiver operator characteristic (AUROC) curve, while calibration was performed by the calibration plot.Results:A total of 15,880 patients were enrolled in this study, and 4845 (30.5%) had developed AKI. Xgboost model had the higher discriminative ability compared with logistic regression (AUROC, 0.849 [95% CI, 0.837-0.861] vs 0.803[95% CI 0.790-0.817], P<0.001) in the test dataset. The estimated glomerular filtration (eGFR) and creatine on intensive care unit (ICU) arrival are the two most important prediction parameters. A SHAP summary plot was used to illustrate the effects of the top 15 features attributed to the Xgboost model.Conclusion:ML models can provide clinical decision support to determine which patients should focus on perioperative preventive treatment to preemptively reduce acute kidney injury by predicting which patients are not at risk.
Clinical epidemiology 2023
BACKGROUND:Our previous showed that a blood management program in the cardiopulmonary bypass (CPB) department, reduced red blood cell (RBC) transfusion and complications, but assessing transfusion practice solely based on transfusion rates was insufficient. This study aimed to design a risk stratification score to predict perioperative RBC transfusion to guide targeted measures for on-pump cardiac surgery patients.STUDY DESIGN AND METHODS:We analyzed data from 42,435 adult cardiac patients. Eight predictors were entered into the final model including age, sex, anemia, New York Heart Association classification, body surface area, cardiac surgery history, emergency surgery, and surgery type. We then simplified the score to an integer-based system. The area under the receiver operating characteristic curve (AUC), Hosmer-Lemeshow goodness-of-fit test, and a calibration curve were used for its performance test. The score was compared to existing scores.RESULTS:The final score included eight predictors. The AUC for the model was 0.77 (95% CI, 0.76-0.77) and 0.77 (95% CI, 0.76-0.78) in the training and test set, respectively. The calibration curves showed a good fit. The risk score was finally grouped into low-risk (score of 0-13 points), medium-risk (14-19 points), and high-risk (more than 19 points). The score had better predictive power compared to the other two existing risk scores.DISCUSSION:We developed an effective risk stratification score with eight variables to predict perioperative RBC transfusion for on-pump cardiac surgery. It assists perfusionists in proactively preparing blood conservation measures for high-risk patients before surgery.
Transfusion 2023
BACKGROUND:Prevention, screening, and early treatment are the aims of postoperative delirium management. The scoring system is an objective and effective tool to stratify potential delirium risk for patients undergoing cardiac surgery.METHODS:Patients who underwent cardiac surgery between January 1, 2012, and January 1, 2019, were enrolled in our retrospective study. The patients were divided into a derivation cohort (n = 45,744) and a validation cohort (n = 11,436). The AD predictive systems were formulated using multivariate logistic regression analysis at three time points: preoperation, ICU admittance, and 24 h after ICU admittance.RESULTS:The prevalence of AD after cardiac surgery in the whole cohort was 3.6% (2,085/57,180). The dynamic scoring system included preoperative LVEF ≤ 45%, serum creatinine > 100 µmol/L, emergency surgery, coronary artery disease, hemorrhage volume > 600 mL, intraoperative platelet or plasma use, and postoperative LVEF ≤ 45%. The area under the receiver operating characteristic curve (AUC) values for AD prediction were 0.68 (preoperative), 0.74 (on the day of ICU admission), and 0.75 (postoperative). The Hosmer‒Lemeshow test indicated that the calibration of the preoperative prediction model was poor (P = 0.01), whereas that of the pre- and intraoperative prediction model (P = 0.49) and the pre, intra- and postoperative prediction model (P = 0.35) was good.CONCLUSIONS:Using perioperative data, we developed a dynamic scoring system for predicting the risk of AD following cardiac surgery. The dynamic scoring system may improve the early recognition of and the interventions for AD.
Journal of cardiothoracic surgery 2023
OBJECTIVES:The existing literature has shown conflicting results regarding the association between preoperative statin exposure and the risk of postoperative cardiac surgery-associated acute kidney injury (CSA-AKI).DESIGN:A single-center retrospective observational study.SETTING:A single, large, tertiary care center.PARTICIPANTS:Adult patients undergoing open cardiac surgery between January 1, 2012 and January 1, 2019.INTERVENTIONS:AKI was defined using the Kidney Disease: Improving Global Outcomes criteria. A multivariate logistic regression analysis and propensity score-matched analysis were used to study the association.MEASUREMENTS AND MAIN RESULTS:A total of 58,399 patient charts were retrospectively reviewed. The preoperative statin exposure cohort had a lower prevalence of all stages of CSA-AKI (30.7% v 36.3%, p < 0.001) and stage 3 CSA-AKI (0.9% v 2.1%, p < 0.001). After adjusting for confounding factors, preoperative statin exposure was a protective factor against all stages of postoperative CSA-AKI (odds ratio [OR], 0.885, 95% confidence interval [CI], 0.852-0.920, p < 0.001) and stage 3 CSA-AKI in adults (OR, 0.671, 95% CI, 0.567-0.795, p < 0.001). A propensity score-matched analysis showed that the preoperative statin exposure cohort had a lower risk of all stages of postoperative CSA-AKI (30.7% v 35.3%, p < 0.001) and stage 3 CSA-AKI (0.9% v 2.2%, p < 0.001) than the control cohort.CONCLUSIONS:Preoperative statin exposure was associated with all stages of postoperative CSA-AKI and stage 3 CSA-AKI.
Journal of cardiothoracic and vascular anesthesia 2022
BACKGROUND:Cardiopulmonary resuscitation (CPR) is an important technique of first aid. It is necessary to be popularized. Large-scale offline training has been affected after the outbreak of Coronavirus disease 2019 (COVID-19). Online training will be the future trend, but the quality of online assessment is unclear. This study aims to compare online and offline evaluations of CPR quality using digital simulator and specialist scoring methods.METHODS:Forty-eight out of 108 contestants who participated in the second Chinese National CPR Skill Competition held in 2020 were included in this study. The competition comprised two stages. In the preliminary online competition, the contestants practiced on the digital simulator while the specialist teams scored live videos. The final competition was held offline, and consisted of live simulator scoring and specialist scoring. The grades of the simulator and specialists in different stages were compared.RESULTS:There was no statistical significance for simulator grades between online and offline competition(37.7 ± 2.0 vs. 36.4 ± 3.4, p = 0.169). For specialists' grades, the video scores were lower than live scores (55.0 ± 1.4 vs. 57.2 ± 1.7, p < 0.001).CONCLUSION:Simulator scoring provided better reliability than specialist scoring in the online evaluation of CPR quality. However, the simulator could only collect quantified data. Specialist scoring is necessary in conjunction with online tests to provide a comprehensive evaluation. A complete and standardized CPR quality evaluation system can be established by combining simulator and specialist contributions.
BMC emergency medicine 2022
Objectives:Postoperative major bleeding is a common problem in patients undergoing cardiac surgery and is associated with poor outcomes. We evaluated the performance of machine learning (ML) methods to predict postoperative major bleeding.Methods:A total of 1,045 patients who underwent isolated coronary artery bypass graft surgery (CABG) were enrolled. Their datasets were assigned randomly to training (70%) or a testing set (30%). The primary outcome was major bleeding defined as the universal definition of perioperative bleeding (UDPB) classes 3-4. We constructed a reference logistic regression (LR) model using known predictors. We also developed several modern ML algorithms. In the test set, we compared the area under the receiver operating characteristic curves (AUCs) of these ML algorithms with the reference LR model results, and the TRUST and WILL-BLEED risk score. Calibration analysis was undertaken using the calibration belt method.Results:The prevalence of postoperative major bleeding was 7.1% (74/1,045). For major bleeds, the conditional inference random forest (CIRF) model showed the highest AUC [0.831 (0.732-0.930)], and the stochastic gradient boosting (SGBT) and random forest models demonstrated the next best results [0.820 (0.742-0.899) and 0.810 (0.719-0.902)]. The AUCs of all ML models were higher than [0.629 (0.517-0.641) and 0.557 (0.449-0.665)], as achieved by TRUST and WILL-BLEED, respectively.Conclusion:ML methods successfully predicted major bleeding after cardiac surgery, with greater performance compared with previous scoring models. Modern ML models may enhance the identification of high-risk major bleeding subpopulations.
Frontiers in cardiovascular medicine 2022
BACKGROUND:Hospital-acquired infection (HAI) after cardiac surgery is a common clinical concern associated with adverse prognosis and mortality. The objective of this study is to determine the prevalence of HAI and its associated risk factors in elderly patients following cardiac surgery and to build a nomogram as a predictive model.METHODS:We developed and internally validated a predictive model from a retrospective cohort of 6405 patients aged ≥70 years, who were admitted to our hospital and underwent cardiac surgery. The primary outcome was HAI. Multivariable logistic regression analysis was used to identify independent factors significantly associated with HAI. The performance of the established nomogram was assessed by calibration, discrimination, and clinical utility. Internal validation was achieved by bootstrap sampling with 1000 repetitions to reduce the overfit bias.RESULTS:Independent factors derived from the multivariable analysis to predict HAI were smoking, myocardial infarction, cardiopulmonary bypass use, intraoperative erythrocytes transfusion, extended preoperative hospitalization days and prolonged duration of mechanical ventilation postoperatively. The derivation model showed good discrimination, with a C-index of 0.706 [95% confidence interval 0.671-0.740], and good calibration [Hosmer-Lemeshow test P = 0.139]. Internal validation also maintained optimal discrimination and calibration. The decision curve analysis revealed that the nomogram was clinically useful.CONCLUSIONS:We developed a predictive nomogram for postoperative HAIs based on routinely available data. This predictive tool may enable clinicians to achieve better perioperative management for elderly patients undergoing cardiac surgery but still requires further external validation.
Clinical interventions in aging 2022
Background and Purpose:Chronic postoperative pain (CPSP) after cardiac surgery can cause severe health problems. As demonstrated in noncardiac surgeries, preoperative chronic pain can potentially lead to CPSP. However, the association between preoperative chronic pain and CPSP over follow-up in cardiac surgical settings in the context of sex differences is still lacking. This observational study aims to explore the role and sex differences of preoperative chronic pain in the occurrence and development of long-term CPSP and CPSP-related complications after cardiac surgery.Patients and Methods:This observational study enrolled 495 patients (35.3% women) who underwent cardiac surgery via median sternotomy in March 2019. Validated questionnaires were delivered to assess preoperative chronic pain and moderate to severe CPSP at 3 and 24 months following surgical procedures. The secondary outcomes included the occurrence of moderate to severe chronic pruritus, sleep disturbance, and daily activities interference at follow-up. Multivariable logistic regression was employed.Results:Of 495 patients analyzed, the incidences of preoperative chronic pain (29.7% versus 20.6%) and moderate to severe CPSP (14.8% versus 8.1%) were both higher in females than males. Female sex (P = 0.048) and preoperative chronic pain (P = 0.008) were identified as significant risk factors for CPSP occurrence. However, preoperative chronic pain contributed significantly to CPSP (P = 0.008), sleep disturbance (P =0.047), and daily activities interference (P =0.019) in females, but not in males.Conclusion:The 2-year prevalence of moderate to severe CPSP after cardiac surgery was 10.5%. Compared to males, females are more susceptible to CPSP and pain-related outcomes in the long term. In addition, preoperative chronic pain was associated with a higher risk of CPSP in females but not in males.
Journal of pain research 2022
BACKGROUND:Previous study found that C-reactive protein (CRP) can predict bleeding after on-pump CABG. To evaluate whether preoperative C-reactive protein (CRP) can be a novel marker of postoperative bleeding in patients having off-pump coronary artery bypass grafting (CABG).METHODS:This is a retrospective cohort study. Multiple variable regression analyses were performed. 537 patients undergoing off-pump isolated primary CABG at Fuwai Hospital from September 2017 to July 2018 were recorded. The primary endpoint was bleeding volume within 24 h after surgery.RESULTS:Data of 537 patients undergoing off-pump isolated primary CABG at Fuwai Hospital were recorded. The correlations between bleeding volume within 24 h after surgery and preoperative data were analyzed with univariate and multivariate linear regression. Much more preoperative CRP concentration (B = -0.089, P < 0.05) was associated with less postoperative bleeding volume and fibrinogen (B = 0.594, p < 0.001).CONCLUSIONS:Preoperative CRP concentration is independently correlated with the postoperative volume of bleeding within 24 h. CRP may become a novel coagulation index in coronary artery atherosclerotic disease.
Journal of cardiothoracic surgery 2022
BACKGROUND:Patients with heart failure who undergo cardiac surgery have increased long-term mortality in which acute kidney injury (AKI) plays a role. However, little is known about whether the incidence of AKI differs according to stratified left ventricular ejection fraction (LVEF).OBJECTIVES:To assess the risks of mild AKI and moderate to severe AKI postcardiac surgery among patients with heart failure.DESIGN:Retrospective cohort analysis of patient data. Ejection fractions were categorised as LVEF less than 40%, heart failure with reduced ejection fraction (HFrEF); LVEF 40 to 49%, heart failure with mid-range ejection fraction (HFmrEF); and LVEF at least 50%, heart failure with preserved ejection fraction (HFpEF).PATIENTS AND SETTINGS:Patients who underwent cardiac surgery from 2012 to 2019 in Fuwai Hospital, Beijing, China, were consecutively enrolled.MAIN OUTCOME MEASURES:The primary endpoint was postoperative AKI staged either as mild AKI or moderate to severe AKI. The secondary outcome was the peri-operative composite adverse event of dialysis support, tracheotomy, intrasurgical and postsurgical mechanical cardiac support and in-hospital mortality. This study also assessed chronic renal dysfunction at follow-up.RESULTS:Of the 54 696 included patients, 18.9% presented with heart failure. Among these with HFpEF, HFmrEF and HFrEF, the incidence of postoperative mild AKI was 37.0, 33.4 and 37.6%, respectively. Patients with HFpEF and HFmrEF were characterised by numerically greater prevalence of moderate to severe AKI than HFrEF (8.5 vs. 9.1 vs. 5.8%). HFrEF and HFmrEF patients had comparable risks for mild AKI relative to HFpEF patients, odds ratio (OR) 0.885; 95% confidence interval CI 0.763 to 1.027 for HFmrEF vs. HFpEF; OR 1.083; 95% CI 0.933 to 1.256 for HFrEF vs. HFpEF. Patients with HFmrEF were more at risk for moderate to severe AKI than patients with HFpEF (OR, 1.368; 95% CI 1.066 to 1.742), but HFrEF and HFpEF did not differ significantly (OR 1.012; 95% CI 0.752 to 1.346). An increasing number of noncardiac comorbidities led to a higher risk of mild AKI and moderate to severe AKI in patients with heart failure; and its effect on AKI was almost equal among the three heart failure strata. The incidence of postoperative composite adverse outcome increased in a graded manner from HFpEF to HFmrEF to HFrEF. Information on the creatine concentrations at 3 months postoperatively and longer were retained for 5200 out of 10 347 (50.6%) heart failure patients in our charts.The AKI severity and the presence of HFmrEF contributed substantially to the development of renal dysfunction over a median [IQR] follow-up of 10 months [4.0 to 21.0].CONCLUSIONS:Initiative programmes aimed at patients with HFrEF to prevent moderate to severe AKI and chronic kidney dysfunction should also include patients with HFmrEF.
European journal of anaesthesiology 2022
