Improving perioperative management to reduce postoperative acute kidney injury and long-term renal risk

改善围手术期管理以减少术后急性肾损伤和长期肾脏风险

• 批准号: 10475332
• 负责人: Minjae Kim
• 金额: $ 10万
• 依托单位: COLUMBIA UNIVERSITY HEALTH SCIENCES
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-09-25 至 2023-05-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10475332
• 关键词: Abdominal Aortic Aneurysm; Academic Medical Centers; Acute Renal Failure with Renal Papillary Necrosis; Acute respiratory failure; Address; Age; Algorithms; American College of Surgeons; Anesthesiology; Anesthetics; Antifibrinolytic Agents; Assessment tool; Award; Bibliography; Bioinformatics; Biometry; Blood Pressure; Blood Transfusion; Burn injury; Caring; Characteristics; Child; Cholecystectomy; Chronic Kidney Failure; Clinical; Clinical Practice Guideline; Cohort Analysis; Complex; Consensus; Data; Databases; Detection; Development; Electronic Health Record; End stage renal failure; Epidemiology; Face; Foundations; Heart Rate; Hepatectomy; Hepatobiliary; Heterogeneity; Hip Fractures; Historical Cohort Studies; Hospitalization; Hospitals; Hypertension; Hypotension; Individual; Infant; Institutes; Institution; Intervention; Intra-abdominal; Intubation; Kidney; Kidney Diseases; Knowledge; Link; Liquid substance; Logistic Regressions; Machine Learning; Maps; Measures; Medicine; Methods; Modeling; Morbidity - disease rate; Nephrology; Operative Surgical Procedures; Outcome; Pathway interactions; Patient risk; Patients; Pattern; Performance; Perioperative; Perioperative Care; Perioperative complication; Pharmaceutical Preparations; Phase; Play; Postoperative Complications; Postoperative Period; Prevention; Procedures; Recovery; Reducing Agents; Renal function; Research; Risk; Risk Assessment; Risk Factors; Role; Sepsis; Severities; Spinal Fusion; Surgical complication; System; Techniques; Testing; The Sun; Time; Trees; Validation; Variant; Vasoconstrictor Agents; Vertebral column; Walkers; Work; adverse outcome; base; cohort; comorbidity; cost; demographics; design; effective intervention; effective therapy; experience; hemodynamics; high risk; hospital readmission; improved; individual patient; machine learning method; modifiable risk; mortality; mortality risk; multidisciplinary; neural information processing; neural network; novel; perioperative mortality; personalized risk prediction; postoperative delirium; predictive modeling; prevent; programs; random forest; repaired; risk minimization; risk prediction; risk prediction model; risk stratification; sex; tool; tool development; working group

Postoperative acute kidney injury (AKI) is a major cause of morbidity and mortality in patients who undergo intraabdominal surgery procedures. AKI leads to increased risks of adverse outcomes such as hospital readmission and progression to chronic kidney disease, to reduced short- and long-term survival, and to greater costs to both patients and hospital systems. There are no effective treatments to prevent postoperative AKI, and as such, appropriate perioperative management is critical for mitigating its risk. However, there is an incomplete understanding of the ways in which patient and perioperative management factors interact to influence the risk of postoperative AKI. Traditional methods only identify generic, global risk factors and only quantify the average effects of each factor on AKI risk. Using an institutional cohort of intraabdominal surgery patients containing detailed preoperative (e.g., patient demographics, risk factors, comorbidities) and intraoperative data (e.g., hemodynamic variables, medications, fluid management), we will utilize a novel interpretable machine learning framework (SHAP [Shapley Additive exPlanations]) to model postoperative AKI risk. Machine learning provides the ability to model complex relationships among predictor variables, and SHAP quantifies the contribution to AKI risk for each variable in individual patients. We will compare this approach to traditional methods, comparing both predictive performance and the clinical understanding provided by the models. The knowledge gained will support the development and validation of novel, interpretable models in a large, multi-institution cohort and analyses to identify modifiable risk factors at all phases of the perioperative period so that personalized perioperative management strategies can be developed to reduce the risk of postoperative AKI and long-term renal outcomes. SPECIFIC AIMS Postoperative acute kidney injury (AKI) is a common and major cause of morbidity and mortality in surgical patients. ADDIN EN.CITE 1 Postoperative AKI is associated with a 4-fold increased risk of mortality ADDIN EN.CITE 2 and increased costs (ranging from $12,100 to $47,700 per patient). ADDIN EN.CITE 3 AKI patients face greater long-term risks of developing chronic kidney disease (CKD), end stage renal disease (ESRD), and associated morbidity and mortality. ADDIN EN.CITE 4 Despite the severe ramifications of postoperative AKI, there are no effective strategies to identify individuals with an elevated risk or inform efforts to prevent its occurrence.5 Optimization of perioperative management may provide the greatest opportunity to fine-tune strategies to minimize AKI risk, as it is a volatile period with an abundance of potential opportunities for intervention. ADDIN EN.CITE 6,7 However, the complex relationships between perioperative factors and the risk of adverse renal outcomes have made it difficult to develop effective tools to guide perioperative patient management towards reducing AKI risk. Traditional risk stratification measures are limited in that they identify global risk factors but do not allow for personalized risk assessment at the level of an individual patient. We propose to use SHAP (Shapley Additive exPlanations), ADDIN EN.CITE 8-10 a novel framework that leverages the ability to fit complex models using machine learning approaches while providing interpretable explanations for individual patients. We have extensive experience characterizing the epidemiology of postoperative AKI ADDIN EN.CITE 2,6,11,12 and the development of risk stratification tools for postoperative AKI ADDIN EN.CITE 13,14 and other surgical complications. ADDIN EN.CITE 15-21 Through the KL2 award from the Columbia University Medical Center (CUMC) Irving Institute (CTSA), we assembled a cohort of intraabdominal surgery patients at CUMC, linking extensive electronic health record data to American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) data. We found that the addition of detailed intraoperative data improved the risk prediction model for postoperative AKI containing only preoperative data ADDIN EN.CITE 7 and evaluated the ability of machine learning to incorporate preoperative and intraoperative data to predict mortality. The proposed project will utilize our CUMC/ACS NSQIP cohort (N=3834) to refine the development of interpretable machine learning models and demonstrate clinical utility to identify modifiable perioperative factors to reduce postoperative AKI and long-term renal risk. To achieve this, we plan to: Develop and validate personalized risk prediction models for postoperative AKI using interpretable machine learning. We will use both preoperative (e.g., age, sex, comorbidities, surgical procedure) and detailed intraoperative data (e.g., fluid types and volumes, medications [e.g., vasopressors], hemodynamic measures [e.g., heart rate, blood pressure]) to develop machine learning models for predicting the development of postoperative AKI, defined by the consensus KDIGO criteria.22 Machine learning provides the ability to fit complex models that account for nonlinear relationships among predictors, and various algorithms will be tested (e.g., logistic regression, neural networks, random forest). The SHAP framework quantifies the contribution of each variable to AKI risk (both increases and decreases) for individual patients. We will compare performance to traditional approaches (i.e., parsimonious risk factor model using traditional logistic regression). We hypothesize that the machine learning/SHAP approach will have improved performance and provide a more robust understanding of critical risk factors and the heterogeneity of their effects on AKI risk among patients. This data will serve as the foundation to support the development and validation of novel, interpretable risk prediction models at all stages of the perioperative period (i.e., prior to surgery, during surgery, and immediately after surgery) in a large, multi-institutional cohort at 3 major academic centers. The knowledge gained will identify modifiable risk factors at all phases of the perioperative period so that personalized perioperative management strategies can be developed to reduce the risk of postoperative AKI and long-term renal outcomes. Bibliography 1. Gumbert SD, Kork F, Jackson ML, Vanga N, Ghebremichael SJ, Wang CY, Eltzschig HK. Perioperative Acute Kidney Injury. Anesthesiology. 2020;132(1):180-204. 2. Kim M, Brady JE, Li G. Variations in the risk of acute kidney injury across intraabdominal surgery procedures. Anesth Analg. 2014;119(5):1121-32. 3. Hobson C, Ozrazgat-Baslanti T, Kuxhausen A, Thottakkara P, Efron PA, Moore FA, Moldawer LL, Segal MS, Bihorac A. Cost and Mortality Associated With Postoperative Acute Kidney Injury. Ann Surg. 2015;261(6):1207-14. PMCID: PMC4247993. 4. Bihorac A, Yavas S, Subbiah S, Hobson CE, Schold JD, Gabrielli A, Layon AJ, Segal MS. Long-term risk of mortality and acute kidney injury during hospitalization after major surgery. Ann Surg. 2009;249(5):851-8. 5. Zacharias M, Mugawar M, Herbison GP, Walker RJ, Hovhannisyan K, Sivalingam P, Conlon NP. Interventions for protecting renal function in the perioperative period. Cochrane Database Syst Rev. 2013;9(9):CD003590. 6. Mathis MR, Naik BI, Freundlich RE, Shanks AM, Heung M, Kim M, Burns ML, Colquhoun DA, Rangrass G, Janda A, Engoren MC, Saager L, Tremper KK, Kheterpal S, Aziz MF, Coffman T, Durieux ME, Levy WJ, Schonberger RB, Soto R, Wilczak J, Berman MF, Berris J, Biggs DA, Coles P, Craft RM, Cummings KC, Ellis TA, 2nd, Fleishut PM, Helsten DL, Jameson LC, van Klei WA, Kooij F, LaGorio J, Lins S, Miller SA, Molina S, Nair B, Paganelli WC, Peterson W, Tom S, Wanderer JP, Wedeven C, Multicenter Perioperative Outcomes Group I. Preoperative Risk and the Association between Hypotension and Postoperative Acute Kidney Injury. Anesthesiology. 2020;132(3):461-75. PMCID: PMC7015776. 7. Kim M, Li G, Mohan S, Turnbull ZA, Kiran RP, Li G. Intraoperative Data Enhance the Detection of High-Risk Acute Kidney Injury Patients When Added to a Baseline Prediction Model. Anesth Analg. 2021;132(2):430-41. PMCID: PMC7855510. 8. Lundberg SM, Erion G, Chen H, DeGrave A, Prutkin JM, Nair B, Katz R, Himmelfarb J, Bansal N, Lee SI. From Local Explanations to Global Understanding with Explainable AI for Trees. Nat Mach Intell. 2020;2(1):56-67. PMCID: PMC7326367. 9. Lundberg SM, Lee SI. A Unified Approach to Interpreting Model Predictions. Advances in Neural Information Processing Systems: Curran Associates.; 2017. p. 4765-74. 10. Lundberg SM, Nair B, Vavilala MS, Horibe M, Eisses MJ, Adams T, Liston DE, Low DK, Newman SF, Kim J, Lee SI. Explainable machine-learning predictions for the prevention of hypoxaemia during surgery. Nat Biomed Eng. 2018;2(10):749-60. PMCID: PMC6467492. 11. Kim M, Brady JE, Li G. Anesthetic technique and acute kidney injury in endovascular abdominal aortic aneurysm repair. J Cardiothorac Vasc Anesth. 2014;28(3):572-8. 12. Kim M, Brady JE, Li G. Interaction Effects of Acute Kidney Injury, Acute Respiratory Failure, and Sepsis on 30-Day Postoperative Mortality in Patients Undergoing High-Risk Intraabdominal General Surgical Procedures. Anesth Analg. 2015;121(6):1536-46. 13. Kim M, Wall MM, Kiran RP, Li G. Latent class analysis stratifies mortality risk in patients developing acute kidney injury after high-risk intraabdominal general surgery: a historical cohort study. Can J Anaesth. 2019;66(1):36-47. PMCID: PMC6370047. 14. Kim M, Kiran RP, Li G. Acute kidney injury after hepatectomy can be reasonably predicted after surgery. J Hepatobiliary Pancreat Sci. 2019;26(4):144-53. PMCID: PMC6453720. 15. Burke J, Rattan R, Sedighim S, Kim M. A Simple Risk Score to Predict Clavien-Dindo Grade IV and V Complications After Non-elective Cholecystectomy. J Gastrointest Surg. 2021;25(1):201-10. PMCID: PMC7415492. 16. Kim EM, Li G, Kim M. Development of a Risk Score to Predict Postoperative Delirium in Patients With Hip Fracture. Anesth Analg. 2020;130(1):79-86. PMCID: PMC6917900. 17. Eisler LD, Lenke LG, Sun LS, Li G, Kim M. Do Antifibrinolytic Agents Reduce the Risk of Blood Transfusion in Children Undergoing Spinal Fusion?: A Propensity Score-matched Comparison Using a National Database. Spine (Phila Pa 1976). 2020;45(15):1055-61. 18. Eisler LD, Hua M, Li G, Sun LS, Kim M. A Multivariable Model Predictive of Unplanned Postoperative Intubation in Infant Surgical Patients. Anesth Analg. 2019;129(6):1645-52. PMCID: PMC6894615. 19. Kim M, Wall MM, Li G. Applying Latent Class Analysis to Risk Stratification for Perioperative Mortality in Patients Undergoing Intraabdominal General Surgery. Anesth Analg. 2016;123(1):193-205. 20. Kim M, Li G. Two-way Interaction Effects of Perioperative Complications on 30-Day Mortality in General Surgery. World J Surg. 2018;42(1):2-11. 21. Kim M, Wall MM, Li G. Risk Stratification for Major Postoperative Complications in Patients Undergoing Intra-abdominal General Surgery Using Latent Class Analysis. Anesth Analg. 2018;126(3):848-57. 22. National Kidney Foundation: Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. KDIGO clinical practice guideline for acute kidney injury. Kidney Int Suppl. 2012;2(1):19-36. 2. Kim M, Brady JE, Li G. Variations in the risk of acute kidney injury across intraabdominal surgery procedures. Anesth Analg. 2014;119(5):1121-32. 3. Hobson C, Ozrazgat-Baslanti T, Kuxhausen A, Thottakkara P, Efron PA, Moore FA, Moldawer LL, Segal MS, Bihorac A. Cost and Mortality Associated With Postoperative Acute Kidney Injury. Ann Surg. 2015;261(6):1207-14. PMCID: PMC4247993. 4. Bihorac A, Yavas S, Subbiah S, Hobson CE, Schold JD, Gabrielli A, Layon AJ, Segal MS. Long-term risk of mortality and acute kidney injury during hospitalization after major surgery. Ann Surg. 2009;249(5):851-8. 5. Zacharias M, Mugawar M, Herbison GP, Walker RJ, Hovhannisyan K, Sivalingam P, Conlon NP. Interventions for protecting renal function in the perioperative period. Cochrane Database Syst Rev. 2013;9(9):CD003590. 6. Mathis MR, Naik BI, Freundlich RE, Shanks AM, Heung M, Kim M, Burns ML, Colquhoun DA, Rangrass G, Janda A, Engoren MC, Saager L, Tremper KK, Kheterpal S, Aziz MF, Coffman T, Durieux ME, Levy WJ, Schonberger RB, Soto R, Wilczak J, Berman MF, Berris J, Biggs DA, Coles P, Craft RM, Cummings KC, Ellis TA, 2nd, Fleishut PM, Helsten DL, Jameson LC, van Klei WA, Kooij F, LaGorio J, Lins S, Miller SA, Molina S, Nair B, Paganelli WC, Peterson W, Tom S, Wanderer JP, Wedeven C, Multicenter Perioperative Outcomes Group I. Preoperative Risk and the Association between Hypotension and Postoperative Acute Kidney Injury. Anesthesiology. 2020;132(3):461-75. PMCID: PMC7015776. 7. Kim M, Li G, Mohan S, Turnbull ZA, Kiran RP, Li G. Intraoperative Data Enhance the Detection of High-Risk Acute Kidney Injury Patients When Added to a Baseline Prediction Model. Anesth Analg. 2021;132(2):430-41. PMCID: PMC7855510. 8. Lundberg SM, Erion G, Chen H, DeGrave A, Prutkin JM, Nair B, Katz R, Himmelfarb J, Bansal N, Lee SI. From Local Explanations to Global Understanding with Explainable AI for Trees. Nat Mach Intell. 2020;2(1):56-67. PMCID: PMC7326367. 9. Lundberg SM, Lee SI. A Unified Approach to Interpreting Model Predictions. Advances in Neural Information Processing Systems: Curran Associates.; 2017. p. 4765-74. 10. Lundberg SM, Nair B, Vavilala MS, Horibe M, Eisses MJ, Adams T, Liston DE, Low DK, Newman SF, Kim J, Lee SI. Explainable machine-learning predictions for the prevention of hypoxaemia during surgery. Nat Biomed Eng. 2018;2(10):749-60. PMCID: PMC6467492. 11. Kim M, Brady JE, Li G. Anesthetic technique and acute kidney injury in endovascular abdominal aortic aneurysm repair. J Cardiothorac Vasc Anesth. 2014;28(3):572-8. 12. Kim M, Brady JE, Li G. Interaction Effects of Acute Kidney Injury, Acute Respiratory Failure, and Sepsis on 30-Day Postoperative Mortality in Patients Undergoing High-Risk Intraabdominal General Surgical Procedures. Anesth Analg. 2015;121(6):1536-46. 13. Kim M, Wall MM, Kiran RP, Li G. Latent class analysis stratifies mortality risk in patients developing acute kidney injury after high-risk intraabdominal general surgery: a historical cohort study. Can J Anaesth. 2019;66(1):36-47. PMCID: PMC6370047. 14. Kim M, Kiran RP, Li G. Acute kidney injury after hepatectomy can be reasonably predicted after surgery. J Hepatobiliary Pancreat Sci. 2019;26(4):144-53. PMCID: PMC6453720. 15. Burke J, Rattan R, Sedighim S, Kim M. A Simple Risk Score to Predict Clavien-Dindo Grade IV and V Complications After Non-elective Cholecystectomy. J Gastrointest Surg. 2021;25(1):201-10. PMCID: PMC7415492. 16. Kim EM, Li G, Kim M. Development of a Risk Score to Predict Postoperative Delirium in Patients With Hip Fracture. Anesth Analg. 2020;130(1):79-86. PMCID: PMC6917900. 17. Eisler LD, Lenke LG, Sun LS, Li G, Kim M. Do Antifibrinolytic Agents Reduce the Risk of Blood Transfusion in Children Undergoing Spinal Fusion?: A Propensity Score-matched Comparison Using a National Database. Spine (Phila Pa 1976). 2020;45(15):1055-61. 18. Eisler LD, Hua M, Li G, Sun LS, Kim M. A Multivariable Model Predictive of Unplanned Postoperative Intubation in Infant Surgical Patients. Anesth Analg. 2019;129(6):1645-52. PMCID: PMC6894615. 19. Kim M, Wall MM, Li G. Applying Latent Class Analysis to Risk Stratification for Perioperative Mortality in Patients Undergoing Intraabdominal General Surgery. Anesth Analg. 2016;123(1):193-205. 20. Kim M, Li G. Two-way Interaction Effects of Perioperative Complications on 30-Day Mortality in General Surgery. World J Surg. 2018;42(1):2-11. 21. Kim M, Wall MM, Li G. Risk Stratification for Major Postoperative Complications in Patients Undergoing Intra-abdominal General Surgery Using Latent Class Analysis. Anesth Analg. 2018;126(3):848-57. 22. National Kidney Foundation: Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. KDIGO clinical practice guideline for acute kidney injury. Kidney Int Suppl. 2012;2(1):19-36.