Deployable machine learning-based decision support system for tracheostomy in acute burn patients

Abstract

Background: Airway obstruction is a common emergency in acute burns with high mortality. Tracheostomy is the most effective method to keep patency of airway and start mechanical ventilation. However, the indication of tracheostomy is challenging and controversial. We aimed to develop and validate a deployable machine learning (ML)-based decision support system to predict the necessity of tracheostomy for acute burn patients. Methods: We enrolled 1011 burn patients from Southwest Hospital (2018–20) for model development and feature selection. The final model was validated on an independent internal cross-temporal cohort (2021, n = 274) and an external cross-institutional cohort (Second Affiliated Hospital of Zhejiang University School of Medicine 2020–21, n = 376). To improve the model’s deployment and interpretability, an ML-based nomogram, an online calculator, and an abbreviated scale were constructed and validated. Results: The optimal model was the eXtreme Gradient Boosting classifier (XGB), which achieved an AUROC of 0.973 and AUPRC of 0.879 in training dataset, and AUROCs of greater than 0.95 in both cross-temporal and cross-institutional validation. Moreover, it kept stable discriminatory ability in validation subgroups stratified by sex, age, burn area, and inhalation injury (AUROC ranging 0.903–0.990). The analysis of calibration curve, decision curve, and score distribution proved the feasibility and reliability of the ML-based nomogram, abbreviated scale (BETS), and online calculator. Conclusions: The developed system has strong predictive ability and generalizability in cross-temporal and cross-institutional evaluations. The nomogram, online calculator, and abbreviated scale based on ML show comparable prediction performance and can be deployed in broader application scenarios, especially in resource-limited clinical environments.

Keywords

Burns / Tracheostomy / Decision support system / Artificial intelligence / Machine learning

References

[1] Legrand M, Depret F, Mallet V. Management of Burns. N Engl J Med 2019; 381: 1188-9. https://doi.org/10.1056/NEJMc1909342.

[2] WHO. Burns. World Health Organization. Available from: https://www.who.int/news-room/fact-sheets/detail/burns.

[3] Jeschke MG, van Baar ME, Choudhry MA, Chung KK, Gibran NS, Logsetty SJNRDP. Burn injury. Nat Rev Dis Primers 2020; 6: 11-25. https://doi.org/10.1038/s41572-020-0145-5.

[4] Peck M, Pressman MA. The correlation between burn mortality rates from fire and flame and economic status of countries. Burns 2013; 39: 1054-9. https://doi.org/10.1016/j.burns.2013.04.010.

[5] Edgar DW, Fish JS, Gomez M, Wood FM. Local and systemic treatments for acute edema after burn injury: a systematic review of the literature. J Burn Care Res 2011; 32: 334-47. https://doi.org/10.1097/BCR.0b013e31820ab019.

[6] Luo G, Peng Y, Yuan Z, Liu Y, Cheng W, Huang Y. et al. Inhalation injury in Southwest China-the evolution of care. Burns 2010; 36: 506-10. https://doi.org/10.1016/j.burns.2009.07.008.

[7] Janik S, Grasl S, Yildiz E, Besser G, Kliman J, Hacker P. et al. A new nomogram to predict the need for tracheostomy in burned patients. Eur Arch Otorrinolaringol 2021; 278: 3479-88. https://doi.org/10.1007/s00405-020-06541-3.

[8] Trouillet JL, Collange O, Belafia F, Blot F, Capellier G, Cesareo E. et al. Tracheotomy in the intensive care unit: guidelines from a French expert panel. Ann Intensive Care 2018; 8: 37. https://doi.org/10.1186/s13613-018-0381-y.

[9] Madsen KR, Guldager H, Rewers M, Weber SO, Købke-Jacobsen K, White J,et al. Danish guidelines 2015 for percutaneous dilatational tracheostomy in the intensive care unit. Dan Med J 2015; 62: C5042.

[10] De Leyn P, Bedert L, Delcroix M, Depuydt P, Lauwers G, Sokolov Y. et al. Tracheotomy: clinical review and guidelines. Eur J Cardiothorac Surg 2007; 32: 412-21. https://doi.org/10.1016/j.ejcts.2007.05.018.

[11] Society., B.a.T.B.o.C.G, Ming J, Lei P, Duan JL, Tan JH, Lou HP. et al. National experts consensus on tracheotomy and intubation for burn patients (2018 version). Chinese journal of burns 2018; 34: 782-5.

[12] Liu L, Xie JF, Wu WJ, Chen H, Li SS, He HL. et al. A simple nomogram for predicting failure of non-invasive respiratory strategies in adults with COVID-19: a retrospective multicentre study. Lancet Digital Health 2021; 3: E166-74. https://doi.org/10.1016/S2589-7500(20)30316-2.

[13] Hiremath A, Shiradkar R, Fu PF, Mahran A, Rastinehad AR, Tewari A. et al. An integrated nomogram combining deep learning, prostate imaging-reporting and data system (PI-RADS) scoring, and clinical variables for identification of clinically significant prostate cancer on biparametric MRI: a retrospective multicentre study. Lancet Digital Health 2021; 3: E445-54. https://doi.org/10.1016/S2589-7500(21)00082-0.

[14] Yadaw AS, Li YC, Bose S, Iyengar R, Bunyavanich S, Pandey G. Clinical features of COVID-19 mortality: development and validation of a clinical prediction model. Lancet Digital Health 2020; 2: E516-25. https://doi.org/10.1016/S2589-7500(20)30217-X.

[15] Moura FSE, Amin K, Ekwobi C. Artificial intelligence in the management and treatment of burns: a systematic review. Burns Trauma 2021; 9: tkab022. https://doi.org/10.1093/burnst/tkab022.

[16] Huang S, Dang J, Sheckter CC, Yenikomshian HA, Gillenwater J. A systematic review of machine learning and automation in burn wound evaluation: a promising but developing frontier. Burns 2021; 47: 1691-704. https://doi.org/10.1016/j.burns.2021.07.007.

[17] Parreco J, Hidalgo A, Parks JJ, Kozol R, Rattan R. Using artificial intelligence to predict prolonged mechanical ventilation and tracheostomy placement. J Surg Res 2018; 228: 179-87. https://doi.org/10.1016/j.jss.2018.03.028.

[18] Takhar A, Surda P, Ahmad I, Amin N, Arora A, Camporota L. et al. Timing of tracheostomy for prolonged respiratory wean in critically ill coronavirus disease 2019 patients: a machine learning approach. Crit Care Explor 2020; 2: e0279. https://doi.org/10.1097/CCE.0000000000000279.

[19] Rathee G, Garg S, Kaddoum G, Wu Y, Jayakody DNK, Alamri A. ANN assisted-IoT enabled COVID-19 patient monitoring. IEEE Access 2021; 9: 42483-92. https://doi.org/10.1109/ACCESS.2021.3064826.

[20] Schwartz JM, Moy AJ, Rossetti SC, Elhadad N, Cato KD. Clinician involvement in research on machine learning-based predictive clinical decision support for the hospital setting: a scoping review. J Am Med Inform Assoc 2021; 28: 653-63. https://doi.org/10.1093/jamia/ocaa296.

[21] Adlung L, Cohen Y, Mor U, Elinav E. Machine learning in clinical decision making. Med 2021; 2: 642-65. https://doi.org/10.1016/j.medj.2021.04.006.

[22] Vasey B, Ursprung S, Beddoe B, Taylor EH, Marlow N, Bilbro N. et al. Association of Clinician Diagnostic Performance with machine learning-based decision support systems: a systematic review. JAMA Netw Open 2021; 4: e211276. https://doi.org/10.1001/jamanetworkopen.2021.1276.

[23] Noble WS.What is a support vector machine? Nat Biotechnol 2006; 24: 1565-7. https://doi.org/10.1038/nbt1206-1565.

[24] Breiman L. Random forests. Mach Learn 2001; 45: 5-32. https://doi.org/10.1023/A:1010933404324.

[25] Chen TQ, Guestrin C. XGBoost: A Scalable Tree Boosting System. In: Kdd’16: Proceedings of the 22nd Acm Sigkdd International Conference on Knowledge Discovery and Data Mining. p. 785-94. 2016. https://doi.org/10.1145/2939672.2939785.

[26] Ke GL, Meng Q, Finley T, Wang TF, Chen W, Ma WD. et al. LightGBM: A Highly Efficient Gradient Boosting Decision Tree. 31st Conference on in Neural Information Processing Systems (Nips 2017), 2017, 30.

[27] Prokhorenkova L, Gusev G, Vorobev A, Dorogush AV, Gulin A. CatBoost: unbiased boosting with categorical features.32nd Conference on Neural Information Processing Systems (NeurIPS 2018), 2018, 31.

[28] Pudil P, Novoviˇcová J, Kittler J. Floating search methods in feature selection. Pattern Recogn Lett 1994; 15: 1119-25. https://doi.org/10.1016/0167-8655(94)90127-9.

[29] Alba AC, Agoritsas T, Walsh M, Hanna S, Iorio A, Devereaux PJ. et al. Discrimination and calibration of clinical prediction models: Users’ guides to the medical literature. JAMA 2017; 318: 1377-84. https://doi.org/10.1001/jama.2017.12126.

[30] Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Mak 2006; 26: 565-74. https://doi.org/10.1177/0272989X06295361.

[31] Kerr KF, Brown MD, Zhu KH, Janes H. Assessing the clinical impact of risk prediction models with decision curves: guidance for correct interpretation and appropriate use. J Clin Oncol 2016; 34: 2534. https://doi.org/10.1200/JCO.2015.65.5654.

[32] Vickers AJ, Van Calster B, Steyerberg EW. Net benefit approaches to the evaluation of prediction models, molecular markers, and diagnostic tests. BMJ 2016; 352: i6. https://doi.org/10.1136/bmj.i6.

[33] Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O. et al. Scikit-learn: machine learning in python. J Mach Learn Res 2011; 12: 2825-30.

[34] Jr F EH.Rms: regression Modeling Strategies. Springer Cham, 2017. https://doi.org/10.1007/978-3-319-19425-7.

[35] Brown M. Rmda: risk model decision analysis. 2018. Available from: https://cran.r-project.org/web/packages/rmda/index.html.

[36] Tagami T, Matsui H, Fushimi K, Yasunaga H. Validation of the prognostic burn index: a nationwide retrospective study. Burns 2015; 41: 1169-75. https://doi.org/10.1016/j.burns.2015.02.017.

[37] Tan J, Zhou J, Li N, Yuan L, Luo G. A new resuscitation formula based on burn index provides more reliable prediction for fluid requirement in adult major burn patients. J Burn Care Res 2021; 42: 962-7. https://doi.org/10.1093/jbcr/irab013.

[38] Sheppard N, Hemington-Gorse S, Shelley OP, Philp B, Dziewulski PJB. Prognostic scoring systems in burns: a review. Burns 2011; 37: 1288-95. https://doi.org/10.1016/j.burns.2011.07.017.

[39] Hosokawa K, Nishimura M, Egi M, Vincent JL. Timing of tracheotomy in ICU patients: a systematic review of randomized controlled trials.Crit Care 2015; 19: 424. https://doi.org/10.1186/s13054-015-1138-8.

[40] Wang D, Wang S, Song Y, Wang H, Zhang A, Wu L. et al. Predictors and outcomes of postoperative tracheostomy in patients undergoing acute type a aortic dissection surgery. BMC Cardiovasc Disord 2022; 22: 94. https://doi.org/10.1186/s12872-022-02538-4.

[41] Xu S, Wang K, Liu K, Liu Y, Huang Y, Zhang Y. et al. Predictive nomogram for the necessity of tracheotomy during oral and oropharyngeal cancer surgery. Laryngoscope 2021; 131: E1489-e1495. https://doi.org/10.1002/lary.29121.

[42] Chen XY, Chen Y, Lin N, Chen JY, Ding CY, Kang DZ. et al. A nomogram for predicting the need of postoperative tracheostomy in patients with aneurysmal subarachnoid Hemorrhage. Front Neurol 2021; 12: 711468. https://doi.org/10.3389/fneur.2021.711468.

[43] Woodson LC, Talon M, Traber DL, Herndon DN. Chapter 19-Diagnosis and treatment of inhalation injury. In Herndon DN (ed).Total Burn Care (Fourth Edition). London: W.B. Saunders, 2012. 229-237.e223.

[44] Velamuri SR, Ali Y, Lanfranco J, Gupta P, Hill DM. Inhalation injury, respiratory failure, and ventilator support in acute burn care. Clin Plast Surg 2024; 51: 221-32. https://doi.org/10.1016/j.cps.2023.11.001.

[45] Nourigheimasi S, Yazdani E, Ghaedi A, Khanzadeh M, Lucke-Wold B, Dioso E. et al. Association of inflammatory biomarkers with overall survival in burn patients: a systematic review and meta-analysis. BMC Emerg Med 2024; 24: 76. https://doi.org/10.1186/s12873-024-00988-x.

[46] Li H, Yao Z, Tan J, Zhou J, Li Y, Wu J. et al. Epidemiology and outcome analysis of 6325 burn patients: a five-year retrospective study in a major burn center in Southwest China. Sci Rep 2017; 7: 46066. https://doi.org/10.1038/srep46066.

[47] Heng X, Cai P, Yuan Z, Peng Y, Luo G, Li H. Efficacy and safety of extracorporeal membrane oxygenation for burn patients: a comprehensive systematic review and meta-analysis. Burns. Trauma 2023; 11: tkac056. https://doi.org/10.1093/burnst/tkac056.

[48] Aggarwal S, Smailes S, Dziewulski P. Tracheostomy in burns patients revisited. Burns 2009; 35: 962-6. https://doi.org/10.1016/j.burns.2009.03.005.

[49] Cartotto R, Burmeister DM, Kubasiak JC. Burn shock and resuscitation: review and state of the science. J Burn Care Res 2022; 43: 567-85. https://doi.org/10.1093/jbcr/irac025.

[50] Agarwal V, O’Neill PJ, Cotton BA, Pun BT, Haney S, Thompson J. et al. Prevalence and risk factors for development of delirium in burn intensive care unit patients. J Burn Care Res 2010; 31: 706-15. https://doi.org/10.1097/BCR.0b013e3181eebee9.

[51] Oto J, Li Q, Kimball WR, Wang J, Sabouri AS, Harrell PG. et al. Continuous positive airway pressure and ventilation are more effective with a nasal mask than a full face mask in unconscious subjects: a randomized controlled trial. Crit Care 2013; 17: R300. https://doi.org/10.1186/cc13169.

[52] Moshrefi S, Sheckter CC, Shepard K, Pereira C, Davis DJ, Karanas Y. et al. Preventing unnecessary intubations: a 5-year regional burn Center experience using flexible Fiberoptic laryngoscopy for airway evaluation in patients with suspected inhalation or airway injury. J Burn Care Res 2019; 40: 341-6. https://doi.org/10.1093/jbcr/irz016.

[53] Carboni Bisso I, Ruiz V, Huespe IA, Rosciani F, Cantos J, Lockhart C. et al. Bronchoscopy-guided percutaneous tracheostomy during the COVID-19 pandemic. Surgery 2023; 173: 944-9. https://doi.org/10.1016/j.surg.2022.12.010.