Abstract
Background
When pulmonary complications occur, postlobectomy patients have a higher mortality
rate, increased length of stay, and higher readmission rates. Because of a lack of
high-quality consolidated clinical data, it is challenging to assess and recognize
at-risk thoracic patients to avoid respiratory failure and standardize outcome measures.
Methods
The National (Nationwide) Inpatient Sample for 2015 was used to establish our model.
We identified 417 respiratory failure from a total of 4,062 patients who underwent
pulmonary lobectomy. Risk factors for respiratory failure were identified, analyzed,
and used in novel machine learning models to predict respiratory failure.
Results
Factors that contributed to increased odds of respiratory failure, such as preexisting
chronic diseases, and intraoperative and postoperative events during hospitalization
were identified. Two machine learning-based prediction models were generated and optimized
by the knowledge accrued from the clinical course of postlobectomy patients. The first
model, with high accuracy and specificity, is suited for performance evaluation, and
the second model, with high sensitivity, is suited for clinical decision making.
Conclusion
We identified risk factors for respiratory failure after lobectomy and introduced
2 machine learning-based techniques to predict respiratory failure for quality review
and clinical decision-making settings. Such techniques can be used to not only provide
targeted support but also standardize quality peer review measures.
To read this article in full you will need to make a payment
Purchase one-time access:
Academic & Personal: 24 hour online accessCorporate R&D Professionals: 24 hour online accessOne-time access price info
- For academic or personal research use, select 'Academic and Personal'
- For corporate R&D use, select 'Corporate R&D Professionals'
Subscribe:
Subscribe to SurgeryAlready a print subscriber? Claim online access
Already an online subscriber? Sign in
Register: Create an account
Institutional Access: Sign in to ScienceDirect
References
- Can postoperative pulmonary complications be objectively evaluated?.Postoperat Care Thoracic Surg. 2017; : 43-59
- Patients experiencing early acute respiratory failure have high postoperative mortality after pneumonectomy.J Thorac Cardiovasc Surg. 2018; 156: 2368-2376
- Postoperative pneumonia after major lung resection.Yearbook Surg. 2007; 2007: 375-376
- Predicting pulmonary complications after pneumonectomy for lung cancer.Eur J Cardiothorac Surg. 2003; 23: 201-208
- Pulmonary complications following lung resection: a comprehensive analysis of incidence and possible risk factors.Chest. 2000; 118: 1263-1270
- Thoracoscopic surgery in elderly lung cancer patients.Crit Rev Oncol Hematol. 2004; 49: 165-171
- Anesthetic considerations for lung resection: preoperative assessment, intraoperative challenges and postoperative analgesia.Ann Transl Med. 2019; 7: 356
- Intensive care unit (ICU) readmission after major lung resection: prevalence, patterns, and mortality.Thorac Cancer. 2017; 8: 33-39
- Ultrasound and other innovations for fluid management in the ICU.Surg Clin North Am. 2017; 97: 1323-1337
- Perioperative risk factors associated with postoperative unplanned intubation after lung resection.J Cardiothorac Vasc Anesth. 2018; 32: 1739-1746
- Machine learning for healthcare: on the verge of a major shift in healthcare epidemiology.Clin Infect Dis. 2018; 66: 149-153
- Machine learning approach for breast cancer prognosis prediction.Computat Model Bioeng Bioinform. 2020; : 41-68
- Machine learning applications in cancer prognosis and prediction.Comput Struct Biotechnol J. 2015; 13: 8-17
- A machine learning approach for early prediction of breast cancer.Int J Eng Comput Sci. 2017; 6
- Machine learning for predicting outcomes in trauma.Shock. 2017; 48: 504-510
- A comparison of a machine learning model with EuroSCORE II in predicting mortality after elective cardiac surgery: a decision curve Analysis.PLoS One. 2017; 12e0169772
- School of Computer Engineering, Suranaree University of Technology (SUT), et al. Post-operative life expectancy of lung cancer patients predicted by Bayesian network model.Int J Machine Learn Comput. 2018; 8: 280-285
- HCUP National Inpatient Sample (NIS). Healthcare Cost and Utilization Project (HCUP).www.hcup-us.ahrq.gov/nisoverview.jspDate accessed: April 8, 2020
- Big data and machine learning in health care.JAMA. 2018; 319: 1317-1318
- Random forests.Machine Learn. 2001; 45: 5-32
- SMOTE: synthetic minority over-sampling technique.J Artificial Intell Res. 2002; 16: 321-357
- Hyperparameter search in machine learning.arXiv [csLG]. 2015;http://arxiv.org/abs/1502.02127Date accessed: January 19, 2020
- Corrgrams: exploratory displays for correlation matrices.Am Stat. 2002; 56: 316-324
- A graphical display of large correlation matrices.Am Stat. 1996; 50: 178
Rokach L, Maimon O. Clustering methods. In: Maimon O, Rokach L, eds: Data Mining and Knowledge Discovery Handbook Boston, MA: Springer:321-352.
- Sparse inverse covariance estimation with the graphical lasso.Biostatistics. 2008; 9: 432-441
- High-dimensional graphs and variable selection with the Lasso.Ann Stat. 2006; 34: 1436-1462
- Tableone: an open source Python package for producing summary statistics for research papers.JAMIA Open. 2018; 1: 26-31
- Package “compareGroups.”.https://cran.r-project.org/web/packages/compareGroups/compareGroups.pdfDate accessed: December 26, 2019
- Package “corrplot.”.https://cran.r-project.org/web/packages/corrplot/corrplot.pdfDate accessed: December 26, 2019
- Package “survival.”.https://cran.r-project.org/web/packages/survival/survival.pdfDate accessed: December 26, 2019
- Package “qgraph.”.https://cran.r-project.org/web/packages/qgraph/qgraph.pdfDate accessed: December 26, 2019
- Learning from Imbalanced Data Sets.Springer, Boston, MA2018
- Noninvasive ventilation reduces mortality in acute respiratory failure following lung resection.Am J Respir Crit Care Med. 2001; 164: 1231-1235
- Acute lung injury and acute respiratory distress syndrome after pulmonary resection.Semin Cardiothorac Vasc Anesth. 2004; 8: 317-334
- Pulmonary function tests do not predict pulmonary complications after thoracoscopic lobectomy.Ann Thorac Surg. 2010; 89: 1044-1052
- Incidence and management of post-lobectomy and pneumonectomy bronchopleural fistula.Lung. 2016; 194: 299-305
- National perioperative outcomes of pulmonary lobectomy for cancer in the obese patient: a propensity score matched analysis.J Thorac Cardiovasc Surg. 2013; 145: 1312-1318
- Modifiable risk factors for patients undergoing lung cancer surgery and their optimization: a review.J Thorac Dis. 2018; 10: S3761-S3772
- Preoperative evaluation for lung resection surgery.Curr Anesthesiol Rep. 2020; 10: 176-184
- Lung resection surgery: can we prevent unplanned intubations?.J Cardiothorac Vasc Anesth. 2018; 32: 1747-1749
- Minimal alteration of pulmonary function after lobectomy in lung cancer patients with chronic obstructive pulmonary disease.Ann Thorac Surg. 2003; 76 (discussion 362): 356-361
- Postoperative pulmonary function changes according to the resected lobe: a 1-year follow-up study of lobectomized patients.J Thorac Dis. 2018; 10: 6891-6902
- Comparison of two national databases for general thoracic surgery.Ann Thorac Surg. 2015; 100 (discussion 1161-1162): 1155-1161
- Introduction to the STS National Database Series: outcomes analysis, quality improvement, and patient safety.Ann Thorac Surg. 2015; 100: 1992-2000
- Adherence to methodological standards in research using the national inpatient sample.JAMA. 2017; 318: 2011-2018
- Evaluation of random forest and ensemble methods at predicting complications following cardiac surgery.Artificial Intell Med. 2019; : 376-385
Article info
Publication history
Published online: July 15, 2020
Accepted:
May 22,
2020
Identification
Copyright
© 2020 Elsevier Inc. All rights reserved.
ScienceDirect
Access this article on ScienceDirectLinked Article
- Corrigendum to “Predicting respiratory failure after pulmonary lobectomy using machine learning techniques”SurgeryVol. 169Issue 4
- PreviewWe discovered an error in a line of the code used in the experimentation of the 2 methods presented in Fig 3.1 The error is in the section where the prediction models are being performed and introduces a “leak” between the training set and testing set. This error exists for both models presented. The implication of this error is that the predictive ability of the 2 models is overestimated. After correcting the error, the area under the receiver operating characteristic of the models diminishes by ~10%.
- Full-Text
- Preview
