Application of machine learning to predict postoperative gastrointestinal bleed in bariatric surgery.
Artificial intelligence
Bariatric surgery
MBSAQIP
Machine learning
Postoperative bleeding
Risk calculator
Journal
Surgical endoscopy
ISSN: 1432-2218
Titre abrégé: Surg Endosc
Pays: Germany
ID NLM: 8806653
Informations de publication
Date de publication:
09 2023
09 2023
Historique:
received:
02
04
2023
accepted:
20
05
2023
medline:
31
8
2023
pubmed:
14
6
2023
entrez:
13
6
2023
Statut:
ppublish
Résumé
Postoperative gastrointestinal bleeding (GIB) is a rare but serious complication of bariatric surgery. The recent rise in extended venous thromboembolism regimens as well as outpatient bariatric surgery may increase the risk of postoperative GIB or lead to delay in diagnosis. This study seeks to use machine learning (ML) to create a model that predicts postoperative GIB to aid surgeon decision-making and improve patient counseling for postoperative bleeds. The Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program (MBSAQIP) database was used to train and validate three types of ML methods: random forest (RF), gradient boosting (XGB), and deep neural networks (NN), and compare them with logistic regression (LR) regarding postoperative GIB. The dataset was split using fivefold cross-validation into training and validation sets, in an 80/20 ratio. The performance of the models was assessed using area under the receiver operating characteristic curve (AUROC) and compared with the DeLong test. Variables with the strongest effect were identified using Shapley additive explanations (SHAP). The study included 159,959 patients. Postoperative GIB was identified in 632 (0.4%) patients. The three ML methods, RF (AUROC 0.764), XGB (AUROC 0.746), and NN (AUROC 0.741) all outperformed LR (AUROC 0.709). The best ML method, RF, was able to predict postoperative GIB with a specificity and sensitivity of 70.0% and 75.4%, respectively. Using DeLong testing, the difference between RF and LR was determined to be significant with p < 0.01. Type of bariatric surgery, pre-op hematocrit, age, duration of procedure, and pre-op creatinine were the 5 most important features identified by ML retrospectively. We have developed a ML model that outperformed LR in predicting postoperative GIB. Using ML models for risk prediction can be a helpful tool for both surgeons and patients undergoing bariatric procedures but more interpretable models are needed.
Sections du résumé
BACKGROUND
Postoperative gastrointestinal bleeding (GIB) is a rare but serious complication of bariatric surgery. The recent rise in extended venous thromboembolism regimens as well as outpatient bariatric surgery may increase the risk of postoperative GIB or lead to delay in diagnosis. This study seeks to use machine learning (ML) to create a model that predicts postoperative GIB to aid surgeon decision-making and improve patient counseling for postoperative bleeds.
METHODS
The Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program (MBSAQIP) database was used to train and validate three types of ML methods: random forest (RF), gradient boosting (XGB), and deep neural networks (NN), and compare them with logistic regression (LR) regarding postoperative GIB. The dataset was split using fivefold cross-validation into training and validation sets, in an 80/20 ratio. The performance of the models was assessed using area under the receiver operating characteristic curve (AUROC) and compared with the DeLong test. Variables with the strongest effect were identified using Shapley additive explanations (SHAP).
RESULTS
The study included 159,959 patients. Postoperative GIB was identified in 632 (0.4%) patients. The three ML methods, RF (AUROC 0.764), XGB (AUROC 0.746), and NN (AUROC 0.741) all outperformed LR (AUROC 0.709). The best ML method, RF, was able to predict postoperative GIB with a specificity and sensitivity of 70.0% and 75.4%, respectively. Using DeLong testing, the difference between RF and LR was determined to be significant with p < 0.01. Type of bariatric surgery, pre-op hematocrit, age, duration of procedure, and pre-op creatinine were the 5 most important features identified by ML retrospectively.
CONCLUSIONS
We have developed a ML model that outperformed LR in predicting postoperative GIB. Using ML models for risk prediction can be a helpful tool for both surgeons and patients undergoing bariatric procedures but more interpretable models are needed.
Identifiants
pubmed: 37311893
doi: 10.1007/s00464-023-10156-0
pii: 10.1007/s00464-023-10156-0
doi:
Types de publication
Journal Article
Research Support, Non-U.S. Gov't
Langues
eng
Sous-ensembles de citation
IM
Pagination
7121-7127Informations de copyright
© 2023. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.
Références
Alalwan AA, Friedman J, Park H, Segal R, Brumback BA, Hartzema AG (2021) US national trends in bariatric surgery: a decade of study. Surgery 170(1):13–17
doi: 10.1016/j.surg.2021.02.002
pubmed: 33714616
Aminian A, Andalib A, Khorgami Z, Cetin D, Burguera B, Bartholomew J, Brethauer SA, Schauer PR (2017) Who should get extended thromboprophylaxis after bariatric surgery? Ann Surg 265(1):143–150
doi: 10.1097/SLA.0000000000001686
pubmed: 28009739
Dreifuss NH, Xie J, Schlottmann F, Cubisino A, Baz C, Vanetta C, Mangano A, Bianco FM, Gangemi A, Masrur MA (2022) Risk factors for readmission after same-day discharge sleeve gastrectomy: a metabolic and bariatric surgery accreditation and quality improvement program database analysis. Obes Surg 32(4):962–969
doi: 10.1007/s11695-022-05919-y
pubmed: 35060023
pmcid: 8773397
Dreifuss NH, Vanetta C, Schlottmann F, Cubisino A, Mangano A, Baz C, Bianco FM, Gangemi A, Hassan C, Masrur MA (2022) Is same-day discharge after Roux-en-Y gastric bypass safe? A metabolic and bariatric surgery accreditation and quality improvement program database analysis. Obes Surg 32(12):3900–3907
doi: 10.1007/s11695-022-06303-6
pubmed: 36194348
pmcid: 9531221
Choi YY, Bae J, Hur KY (2020) Risk factors of postoperative bleeding after bariatric surgery: a systematic review and meta-analysis. Surg Obes Relat Dis 16:560–568
Golzarand M, Toolabi K, Parsaei R (2022) Prediction factors of early postoperative bleeding after bariatric surgery. Obes Surg 32:1–8
doi: 10.1007/s11695-022-06059-z
pubmed: 35474043
Straatman J, Verhaak T, Demirkiran A, Harlaar NJ, Cense HA, Jonker FHW (2022) Dutch Audit for Treatment of Obesity (DATO) Research Group Risk factors for postoperative bleeding in bariatric surgery. Surg Obes Relat Dis 18(8):1057–1065
doi: 10.1016/j.soard.2022.05.010
pubmed: 35764472
Goodfellow I, Bengio Y, Courville A (2016) Deep learning. MIT Press, Cambridge
Cheung AT, Tanaka KA, Cohn LH, Mihaljevic T, Aklog L, Chenoweth S et al (2020) Perioperative prediction of transfusion requirements: a machine learning algorithm derived from a multicenter, retrospective cohort. Anesth Analg 131:342–351
Nudel J, Bishara AM, de Geus SWL, Patil P, Srinivasan J, Hess DT, Woodson J (2021) Development and validation of machine learning models to predict gastrointestinal leak and venous thromboembolism after weight loss surgery: an analysis of the MBSAQIP database. Surg Endosc 35(1):182–191
doi: 10.1007/s00464-020-07378-x
pubmed: 31953733
Moons KGM, Altman DG, Reitsma JB et al (2015) Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): explanation and elaboration. Ann Intern Med 162(1):W1–W73
doi: 10.7326/M14-0698
pubmed: 25560730
Géron A (2019) Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow: Concepts, tools, and techniques to build intelligent systems. O’Reilly Media
Lecun Y, Bengio Y, Hinton G (2015) Deep learning. Nature 521:436–444
doi: 10.1038/nature14539
pubmed: 26017442
DeLong ER, DeLong DM, Clarke-Pearson DL (1988) Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 44:837
doi: 10.2307/2531595
pubmed: 3203132
Lundberg SM, Allen PG, Lee S-I (2017) A unified approach to interpreting model predictions. Adv Neural Inf Process 30:4768–4777
Pollard TJ, Johnson AEW, Raffa JD, Mark RG (2018) tableone: an open source Python package for producing summary statistics for research papers. JAMIA Open 1:26
doi: 10.1093/jamiaopen/ooy012
pubmed: 31984317
pmcid: 6951995
Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, Blondel M, Prettenhofer P, Weiss R, Dubourg V, Vanderplas J, Passos A, Cournapeau D, Brucher M, Perrot M, Duchesnay E (2011) Scikit-learn: machine learning in Python. J Mach Learn Res 12:2825–2830
Chen T, Guestrin C (2016) XGBoost: a scalable tree boosting system. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, New York, NY, USA, pp 785–794
Chollet F et al (2015) Keras. https://github.com/fchollet/keras
Grieco A, Huffman KM, Cohen ME, Hall BL, Morton JM, Ko CY (2021) The metabolic and bariatric surgery accreditation and quality improvement program bariatric surgical risk/benefit calculator: 30-day risk. Surg Obes Relat Dis 17(6):1117–1124
doi: 10.1016/j.soard.2021.02.005
pubmed: 33773930
Cao Y, Montgomery S, Ottosson J, Näslund E (2020) Stenberg E (2020) Deep learning neural networks to predict serious complications after bariatric surgery: analysis of Scandinavian obesity surgery registry data. JMIR Med Inform 8(5):e15992
doi: 10.2196/15992
pubmed: 32383681
pmcid: 7244994
Young AJ, Hare A, Subramanian M, Weaver JL, Kaufman E, Sims C (2021) Using machine learning to make predictions in patients who fall. J Surg Res 257:118–127
doi: 10.1016/j.jss.2020.07.047
pubmed: 32823009
Gao J, Merchant AM (2021) A machine learning approach in predicting mortality following emergency general surgery. Am Surg 87(9):1379–1385
doi: 10.1177/00031348211038568
pubmed: 34378431
El Hechi MW, Maurer LR, Levine J, Zhuo D, El Moheb M, Velmahos GC, Dunn J, Bertsimas D, Kaafarani HM (2021) Validation of the artificial intelligence-based predictive optimal trees in emergency surgery risk (POTTER) calculator in emergency general surgery and emergency laparotomy patients. J Am Coll Surg 232(6):912–919
doi: 10.1016/j.jamcollsurg.2021.02.009
pubmed: 33705983
Choi J, Anderson T, Tennakoon L, Spain DA, Forrester JD (2022) Explainable machine learning to bring database to the bedside: development and validation of the TROUT (Trauma fRailty OUTcomes) Index, a point-of-care tool to prognosticate outcomes after traumatic injury based on frailty. Ann Surg. https://doi.org/10.1097/SLA.0000000000005649
doi: 10.1097/SLA.0000000000005649
pubmed: 36538632
Wijnberge M, Geerts BF, Hol L, Lemmers N, Mulder MP, Berge P, Schenk J, Terwindt LE, Hollmann MW, Vlaar AP, Veelo DP (2020) Effect of a machine learning-derived early warning system for intraoperative hypotension vs standard care on depth and duration of intraoperative hypotension during elective noncardiac surgery: the HYPE Randomized Clinical Trial. JAMA 323(11):1052–1060
doi: 10.1001/jama.2020.0592
pubmed: 32065827
pmcid: 7078808
D'Ascenzo F, De Filippo O, Gallone G, Mittone G, Deriu MA, Iannaccone M, Ariza-Solé A, Liebetrau C, Manzano-Fernández S, Quadri G, Kinnaird T, Campo G, Simao Henriques JP, Hughes JM, Dominguez-Rodriguez A, Aldinucci M, Morbiducci U, Patti G, Raposeiras-Roubin S, Abu-Assi E, De Ferrari GM, PRAISE study group (2021) Machine learning-based prediction of adverse events following an acute coronary syndrome (PRAISE): a modelling study of pooled datasets. Lancet 397(10270):199–207.
Bertsimas D, Dunn J (2017) Optimal classification trees. Mach Learn 106:1039–1082
doi: 10.1007/s10994-017-5633-9
Rudin C, Ustun B (2018) Optimized scoring systems: Toward trust in machine learning for healthcare and criminal justice. Interfaces 48(5):449–466
doi: 10.1287/inte.2018.0957