Preoperative Prediction of Postoperative Infections Using Machine Learning and Electronic Health Record Data.

OBJECTIVE: To estimate preoperative risk of postoperative infections using structured electronic health record (EHR) data. BACKGROUND: Surveillance and reporting of postoperative infections is primarily done through costly, labor-intensive manual chart reviews on a small sample of patients. Automated methods using statistical models applied to postoperative EHR data have shown promise to augment manual review as they can cover all operations in a timely manner. However, there are no specific models for risk-adjusting infectious complication rates using EHR data. METHODS: Preoperative EHR data from 30,639 patients (2013-2019) were linked to the American College of Surgeons National Surgical Quality Improvement Program preoperative data and postoperative infection outcomes data from 5 hospitals in the University of Colorado Health System. EHR data included diagnoses, procedures, operative variables, patient characteristics, and medications. Lasso and the knockoff filter were used to perform controlled variable selection. Outcomes included surgical site infection, urinary tract infection, sepsis/septic shock, and pneumonia up to 30 days postoperatively. RESULTS: Among >15,000 candidate predictors, 7 were chosen for the surgical site infection model and 6 for each of the urinary tract infection, sepsis, and pneumonia models. Important variables included preoperative presence of the specific outcome, wound classification, comorbidities, and American Society of Anesthesiologists physical status classification. The area under the receiver operating characteristic curve for each model ranged from 0.73 to 0.89. CONCLUSIONS: Parsimonious preoperative models for predicting postoperative infection risk using EHR data were developed and showed comparable performance to existing American College of Surgeons National Surgical Quality Improvement Program risk models that use manual chart review. These models can be used to estimate risk-adjusted postoperative infection rates applied to large volumes of EHR data in a timely manner.

Pulmonary metastasectomy is associated with survival after lung-only recurrence in pancreatic cancer.

BACKGROUND: After surgical resection of pancreatic ductal adenocarcinoma, 14% of patients have lung-only recurrence. We hypothesize that in patients with isolated lung metastases from pancreatic ductal adenocarcinoma, pulmonary metastasectomy offers a survival benefit with minimal additional morbidity after resection. METHODS: This was a single-institution, retrospective study of patients who underwent definitive resection of pancreatic ductal adenocarcinoma and later developed isolated lung metastases between 2009 and 2021. Patients were included if they carried a diagnosis of pancreatic ductal adenocarcinoma, underwent pancreatic resection with curative intent, and subsequently developed lung metastases. Patients were excluded if they developed multiple sites of recurrence. RESULTS: We identified 39 patients with pancreatic ductal adenocarcinoma and isolated lung metastases, 14 of whom underwent pulmonary metastasectomy. During the study period, 31 (79%) patients died. Across all patients, there was an overall survival of 45.9 months, a disease-free interval of 22.8 months, and survival after recurrence of 22.5 months. Survival after recurrence was significantly longer in patients who underwent pulmonary metastasectomy than those who did not (30.8 months vs 18.6 months, P < .01). There was no difference in overall survival between groups. However, patients who underwent pulmonary metastasectomy were significantly more likely to be alive 3 years after their diagnosis (100.0% vs 64%, P = .02) and 2 years after recurrence (79% vs 32%, P < .01) than those in who did not undergo pulmonary metastasectomy. No mortalities occurred related to pulmonary metastasectomy, and procedure-related morbidity was 7%. CONCLUSION: Patients who underwent pulmonary metastasectomy for isolated pulmonary pancreatic ductal adenocarcinoma metastases had significantly longer survival after recurrence and clinically meaningful survival benefit with minimal additional morbidity after pulmonary resection.

Effect of Present at Time of Surgery on Unadjusted and Risk-Adjusted Postoperative Complication Rate.

BACKGROUND: Present at the time of surgery (PATOS) is an important measure to collect in postoperative complication surveillance systems because it may affect a patient's risk of a subsequent complication and the estimation of postoperative complication rates attributed to the healthcare system. The American College of Surgeons (ACS) NSQIP started collecting PATOS data for 8 postoperative complications in 2011, but no one has used these data to quantify how this may affect unadjusted and risk-adjusted postoperative complication rates. STUDY DESIGN: This study was a retrospective observational study of the ACS NSQIP database from 2012 to 2018. PATOS data were analyzed for the 8 postoperative complications of superficial, deep, and organ space surgical site infection; pneumonia; urinary tract infection; ventilator dependence; sepsis; and septic shock. Unadjusted postoperative complication rates were compared ignoring PATOS vs taking PATOS into account. Observed to expected ratios over time were also compared by calculating expected values using multiple logistic regression analyses with complication as the dependent variable and the 28 nonlaboratory preoperative variables in the ACS NSQIP database as the independent variables. RESULTS: In 5,777,108 patients, observed event rates for each outcome were reduced by between 6.1% (superficial surgical site infection) and 52.5% (sepsis) when PATOS was taken into account. The observed to expected ratios were similar each year for all outcomes, except for sepsis and septic shock in the early years. CONCLUSIONS: Taking PATOS into account is important for reporting unadjusted event rates. The effect varied by type of complication-lowest for superficial surgical site infection and highest for sepsis and septic shock. Taking PATOS into account was less important for risk-adjusted outcomes (observed to expected ratios), except for sepsis and septic shock.

Development and validation of models for detection of postoperative infections using structured electronic health records data and machine learning.

BACKGROUND: Postoperative infections constitute more than half of all postoperative complications. Surveillance of these complications is primarily done through manual chart review, which is time consuming, expensive, and typically only covers 10% to 15% of all operations. Automated surveillance would permit the timely evaluation of and reporting of all operations. METHODS: The goal of this study was to develop and validate parsimonious, interpretable models for conducting surveillance of postoperative infections using structured electronic health records data. This was a retrospective study using 30,639 unique operations from 5 major hospitals between 2013 and 2019. Structured electronic health records data were linked to postoperative outcomes data from the American College of Surgeons National Surgical Quality Improvement Program. Predictors from the electronic health records included diagnoses, procedures, and medications. Infectious complications included surgical site infection, urinary tract infection, sepsis, and pneumonia within 30 days of surgery. The knockoff filter, a penalized regression technique that controls type I error, was applied for variable selection. Models were validated in a chronological held-out dataset. RESULTS: Seven percent of patients experienced at least one type of postoperative infection. Models selected contained between 4 and 8 variables and achieved >0.91 area under the receiver operating characteristic curve, >81% specificity, >87% sensitivity, >99% negative predictive value, and 10% to 15% positive predictive value in a held-out test dataset. CONCLUSION: Surveillance and reporting of postoperative infection rates can be implemented for all operations with high accuracy using electronic health records data and simple linear regression models.

Development and validation of a model for surveillance of postoperative bleeding complications using structured electronic health records data.

BACKGROUND: Postoperative bleeding complications surveillance is done primarily through manual chart review. The purpose of this study was to develop and validate a detection model for postoperative bleeding complications using structured electronic health records data. METHODS: Patients who underwent operations at 1 of 5 hospitals within our local health system between 2013 and 2019 and whose complications were reported by the American College of Surgeons National Surgical Quality Improvement Program were included. Electronic health records data were linked to American College of Surgeons National Surgical Quality Improvement Program data using personal health identifiers. Electronic health records predictors included diagnosis codes mapped to PheCodes, procedure names, and medications within 30 days after surgery. We defined bleeding events as the transfusion of red blood cell components within 30 days after surgery. The knockoff filter and the lasso were used to develop a model in a training set of operations from January 2013 to March 2017. Performance of each model was tested in a held-out data set of patients who underwent operations from March 2017 to October 2019. RESULTS: A total of 30,639 patients were included; 1,112 patients (3.6%) had a bleeding event. Eight predictor variables were selected by the knockoff filter. When applied to the test set, specificity was 94%, sensitivity was 94%, area under the curve was 0.97, and accuracy was 93%. Calibration was consistent in lower predicted risk patients, whereas the model slightly overpredicted risk in high-risk patients. CONCLUSION: We created a parsimonious, accurate model for identifying patients with bleeding complications. This model can be used to augment manual chart review for surveillance and reporting of perioperative bleeding complications, enabling inclusion of all surgeries in quality improvement efforts.

Comparing Statistical Tests for Differential Network Analysis of Gene Modules.

Genes often work together to perform complex biological processes, and "networks" provide a versatile framework for representing the interactions between multiple genes. Differential network analysis (DiNA) quantifies how this network structure differs between two or more groups/phenotypes (e.g., disease subjects and healthy controls), with the goal of determining whether differences in network structure can help explain differences between phenotypes. In this paper, we focus on gene co-expression networks, although in principle, the methods studied can be used for DiNA for other types of features (e.g., metabolome, epigenome, microbiome, proteome, etc.). Three common applications of DiNA involve (1) testing whether the connections to a single gene differ between groups, (2) testing whether the connection between a pair of genes differs between groups, or (3) testing whether the connections within a "module" (a subset of 3 or more genes) differs between groups. This article focuses on the latter, as there is a lack of studies comparing statistical methods for identifying differentially co-expressed modules (DCMs). Through extensive simulations, we compare several previously proposed test statistics and a new p-norm difference test (PND). We demonstrate that the true positive rate of the proposed PND test is competitive with and often higher than the other methods, while controlling the false positive rate. The R package discoMod (differentially co-expressed modules) implements the proposed method and provides a full pipeline for identifying DCMs: clustering tools to derive gene modules, tests to identify DCMs, and methods for visualizing the results.