An assessment of the value of deep neural networks in genetic risk prediction for surgically relevant outcomes.

INTRODUCTION: Postoperative complications affect up to 15% of surgical patients constituting a major part of the overall disease burden in a modern healthcare system. While several surgical risk calculators have been developed, none have so far been shown to decrease the associated mortality and morbidity. Combining deep neural networks and genomics with the already established clinical predictors may hold promise for improvement. METHODS: The UK Biobank was utilized to build linear and deep learning models for the prediction of surgery relevant outcomes. An initial GWAS for the relevant outcomes was initially conducted to select the Single Nucleotide Polymorphisms for inclusion in the models. Model performance was assessed with Receiver Operator Characteristics of the Area Under the Curve and optimum precision and recall. Feature importance was assessed with SHapley Additive exPlanations. RESULTS: Models were generated for atrial fibrillation, venous thromboembolism and pneumonia as genetics only, clinical features only and a combined model. For venous thromboembolism, the ROC-AUCs were 60.1% [59.6%-60.4%], 63.4% [63.2%-63.4%] and 66.6% [66.2%-66.9%] for the linear models and 51.5% [49.4%-53.4%], 63.2% [61.2%-65.0%] and 62.6% [60.7%-64.5%] for the deep learning SNP, clinical and combined models, respectively. For atrial fibrillation, the ROC-AUCs were 60.3% [60.0%-60.4%], 78.7% [78.7%-78.7%] and 80.0% [79.9%-80.0%] for the linear models and 59.4% [58.2%-60.9%], 78.8% [77.8%-79.8%] and 79.8% [78.8%-80.9%] for the deep learning SNP, clinical and combined models, respectively. For pneumonia, the ROC-AUCs were 50.1% [49.6%-50.6%], 69.2% [69.1%-69.2%] and 68.4% [68.0%-68.5%] for the linear models and 51.0% [49.7%-52.4%], 69.7% [.5%-70.8%] and 69.7% [68.6%-70.8%] for the deep learning SNP, clinical and combined models, respectively. CONCLUSION: In this report we presented linear and deep learning predictive models for surgery relevant outcomes. Overall, predictability was similar between linear and deep learning models and inclusion of genetics seemed to improve accuracy.

Reinforcement learning-based anatomical maps for pancreas subregion and duct segmentation.

BACKGROUND: The pancreas is a complex abdominal organ with many anatomical variations, and therefore automated pancreas segmentation from medical images is a challenging application. PURPOSE: In this paper, we present a framework for segmenting individual pancreatic subregions and the pancreatic duct from three-dimensional (3D) computed tomography (CT) images. METHODS: A multiagent reinforcement learning (RL) network was used to detect landmarks of the head, neck, body, and tail of the pancreas, and landmarks along the pancreatic duct in a selected target CT image. Using the landmark detection results, an atlas of pancreases was nonrigidly registered to the target image, resulting in anatomical probability maps for the pancreatic subregions and duct. The probability maps were augmented with multilabel 3D U-Net architectures to obtain the final segmentation results. RESULTS: To evaluate the performance of our proposed framework, we computed the Dice similarity coefficient (DSC) between the predicted and ground truth manual segmentations on a database of 82 CT images with manually segmented pancreatic subregions and 37 CT images with manually segmented pancreatic ducts. For the four pancreatic subregions, the mean DSC improved from 0.38, 0.44, and 0.39 with standard 3D U-Net, Attention U-Net, and shifted windowing (Swin) U-Net architectures, to 0.51, 0.47, and 0.49, respectively, when utilizing the proposed RL-based framework. For the pancreatic duct, the RL-based framework achieved a mean DSC of 0.70, significantly outperforming the standard approaches and existing methods on different datasets. CONCLUSIONS: The resulting accuracy of the proposed RL-based segmentation framework demonstrates an improvement against segmentation with standard U-Net architectures.

Assessing the utility of natural language processing for detecting postoperative complications from free medical text.

BACKGROUND: Postoperative complication rates are often assessed through administrative data, although this method has proven to be imprecise. Recently, new developments in natural language processing have shown promise in detecting specific phenotypes from free medical text. Using the clinical challenge of extracting four specific and frequently undercoded postoperative complications (pneumonia, urinary tract infection, sepsis, and septic shock), it was hypothesized that natural language processing would capture postoperative complications on a par with human-level curation from electronic health record free medical text. METHODS: Electronic health record data were extracted for surgical cases (across 11 surgical sub-specialties) from 18 hospitals in the Capital and Zealand regions of Denmark that were performed between May 2016 and November 2021. The data set was split into training/validation/test sets (30.0%/48.0%/22.0%). Model performance was compared with administrative data and manual extraction of the test data set. RESULTS: Data were obtained for 17 486 surgical cases. Natural language processing achieved a receiver operating characteristic area under the curve of 0.989 for urinary tract infection, 0.993 for pneumonia, 0.992 for sepsis, and 0.998 for septic shock, whereas administrative data achieved a receiver operating characteristic area under the curve of 0.595 for urinary tract infection, 0.624 for pneumonia, 0.571 for sepsis, and 0.625 for septic shock. CONCLUSION: The natural language processing approach was able to capture complications with acceptable performance, which was superior to administrative data. In addition, the model performance approached that of manual curation and thereby offers a potential pathway for complete real-time coverage of postoperative complications across surgical procedures based on natural language processing assessment of electronic health record free medical text.

Have a vital end-user been overlooked? Developing a shared decision intervention for patients with potential pancreatic cancer regarding the choice of surgery.

Objective: To develop a patient decision aid facilitating shared decision making for patients with potential pancreatic cancer deciding about no treatment, surgical or medical treatment. Methods: Based on a user-centred design by Wittemann et al., we developed a shared decision making intervention in three phases: 1) Understanding decision needs 2) Development of a patient decision aid (PtDA) based on a generic template 3) Assessment of the intervention from interviews with patients (n = 11), relatives (n = 11), nurses (n = 4) and surgeons (n = 2) analysed with thematic analysis, and measuring patients' perceptions of choice of options with the Decisional Conflict Scale. Results: Results showed varying experiences with the use of the PtDA, with surgeons not finding PtDA useful as it was impractical and constraining with patients' conversations. There was no difference in patients' perceptions in choosing options for those being presented vs those patients not being presented for the PtDA. Conclusion: The format and structure of the PtDA was not feasible for the surgeons as fundamental users in the present clinic. Innovation: This study highlights the urgent need to consider clinical context before introducing a predefined tool and shows the importance of a multistakeholder approach. Research should focus on finding means to successful implement shared decision making.

Perioperative factors associated with survival following surgery for pancreatic cancer - a nationwide analysis of 473 cases from Denmark.

BACKGROUND: Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal cancers worldwide, with an overall 5-year survival rate of only 5%. The effect of perioperative treatment factors including duration of surgery, blood transfusions as well as choice of anesthesia and analgesia techniques on overall survival (OS) following pancreatic resections for PDAC, is currently not well known. We hypothesized that these perioperative factors might be associated with OS after pancreatic resections for PDAC. METHODS: This is a retrospective study from a nationwide cohort of patients who underwent surgery for PDAC in Denmark from 2011 to 2020. Kaplan-Meier 1, 2 and 5-year survival estimates were 73%, 49% and 22%, respectively. Data were obtained by joining the national Danish Pancreatic Cancer Database (DPCD) and the Danish Anaesthesia Database (DAD). Associations between the primary endpoint (OS) and perioperative factors including duration of surgery, type of anesthesia (intravenous, inhalation or mixed), use of epidural analgesia and perioperative blood transfusions were assessed using Hazard Ratios (HRs). These were calculated by Cox regression, controlling for relevant confounders identified through an assessment of the current literature. These included demographics, comorbidities, perioperative information, pre and postoperative chemotherapy, tumor staging and free resection margins. RESULTS: Overall, data from 473 resected PDAC patients were available. Multivariate Cox regression indicated that perioperative blood transfusions were associated with shorter OS (HR 2.53, p = 0.005), with survival estimates of 8.8% in transfused vs. 28.0% in non-transfused patients at 72 months after surgery. No statistically significant associations were identified for the duration of surgery or anesthesia/analgesia techniques. CONCLUSION: In this study, the use of perioperative blood transfusions was associated with shorter OS.

Assessing optimal methods for transferring machine learning models to low-volume and imbalanced clinical datasets: experiences from predicting outcomes of Danish trauma patients.

Introduction: Accurately predicting patient outcomes is crucial for improving healthcare delivery, but large-scale risk prediction models are often developed and tested on specific datasets where clinical parameters and outcomes may not fully reflect local clinical settings. Where this is the case, whether to opt for de-novo training of prediction models on local datasets, direct porting of externally trained models, or a transfer learning approach is not well studied, and constitutes the focus of this study. Using the clinical challenge of predicting mortality and hospital length of stay on a Danish trauma dataset, we hypothesized that a transfer learning approach of models trained on large external datasets would provide optimal prediction results compared to de-novo training on sparse but local datasets or directly porting externally trained models. Methods: Using an external dataset of trauma patients from the US Trauma Quality Improvement Program (TQIP) and a local dataset aggregated from the Danish Trauma Database (DTD) enriched with Electronic Health Record data, we tested a range of model-level approaches focused on predicting trauma mortality and hospital length of stay on DTD data. Modeling approaches included de-novo training of models on DTD data, direct porting of models trained on TQIP data to the DTD, and a transfer learning approach by training a model on TQIP data with subsequent transfer and retraining on DTD data. Furthermore, data-level approaches, including mixed dataset training and methods countering imbalanced outcomes (e.g., low mortality rates), were also tested. Results: Using a neural network trained on a mixed dataset consisting of a subset of TQIP and DTD, with class weighting and transfer learning (retraining on DTD), we achieved excellent results in predicting mortality, with a ROC-AUC of 0.988 and an F2-score of 0.866. The best-performing models for predicting long-term hospitalization were trained only on local data, achieving an ROC-AUC of 0.890 and an F1-score of 0.897, although only marginally better than alternative approaches. Conclusion: Our results suggest that when assessing the optimal modeling approach, it is important to have domain knowledge of how incidence rates and workflows compare between hospital systems and datasets where models are trained. Including data from other health-care systems is particularly beneficial when outcomes are suffering from class imbalance and low incidence. Scenarios where outcomes are not directly comparable are best addressed through either de-novo local training or a transfer learning approach.

Impact of para aortic lymph node removal on survival following resection for pancreatic adenocarcinoma.

INTRODUCTION: For PDAC patients undergoing resection, it remains unclear whether metastases to the paraaortic lymph nodes (PALN+) have any prognostic significance and whether metastases should lead to the operation not being carried out. Our hypothesis is that PALN + status would be associated with short overall survival (OS) compared with PALN-, but longer OS compared with patients undergoing surgical exploration only (EXP). METHODS: Patients with registered PALN removal from the nationwide Danish Pancreatic Cancer Database (DPCD) from May 1st 2011 to December 31st 2020 were assessed. A cohort of PDAC patients who only had explorative laparotomy due to non-resectable tumors were also included (EXP group). Survival analysis between groups were performed with cox-regression in a multivariate approach including relevant confounders. RESULTS: A total of 1758 patients were assessed, including 424 (24.1%) patients who only underwent explorative surgery leaving 1334 (75.8%) patients for further assessment. Of these 158 patients (11.8%) had selective PALN removal, of whom 19 patients (12.0%) had PALN+. Survival analyses indicated that explorative surgery was associated with significantly shorter OS compared with resection and PALN + status (Hazard Ratio 2.36, p < 0.001). No difference between PALN + and PALN- status could be demonstrated in resected patients after controlling for confounders. CONCLUSION: PALN + status in patients undergoing resection offer improved survival compared with EXP. PALN + should not be seen as a contraindication for curative intended resection.

Genetic risk factors for postoperative atrial fibrillation-a nationwide genome-wide association study (GWAS).

Background: Atrial fibrillation (AF) is a major cause of morbidity with a high prevalence among the elderly and has an established genetic disposition. Surgery is a well-known risk factor for AF; however, it is currently not recognized how much common genetic variants influence the postoperative risk. The purpose of this study was to identify Single Nucleotide Polymorphisms associated with postoperative AF. Methods: The UK Biobank was utilized to conduct a Genome-Wide Association Study (GWAS) to identify variants associated with AF after surgery. An initial discovery GWAS was performed in patients that had undergone surgery with subsequent replication in a unique non-surgical cohort. In the surgical cohort, cases were defined as newly diagnosed AF within 30 days after surgery. The threshold for significance was set at 5 × 10-8. Results: After quality control, 144,196 surgical patients with 254,068 SNPs were left for analysis. Two variants (rs17042171 (p = 4.86 × 10-15) and rs17042081 (p = 7.12 × 10-15)) near the PITX2-gene reached statistical significance. These variants were replicated in the non-surgical cohort (1.39 × 10-101 and 1.27 × 10-93, respectively). Several other loci were significantly associated with AF in the non-surgical cohort. Conclusion: In this GWAS-analysis of a large national biobank, we identified 2 variants that were significantly associated with postoperative AF. These variants were subsequently replicated in a unique non-surgical cohort. These findings bring new insight in the genetics of postoperative AF and may help identify at-risk patients and guide management.

Association of Trauma With Long-Term Risk of Death and Immune-Mediated or Cancer Disease in Same-Sex Twins.

Importance: Immediate consequences of trauma include a rapid and immense activation of the immune system, whereas long-term outcomes include premature death, physical disability, and reduced workability. Objective: To investigate if moderate to severe trauma is associated with long-term increased risk of death or immune-mediated or cancer disease. Design, Setting, and Participants: This registry-based, matched, co-twin control cohort study linked the Danish Twin Registry and the Danish National Patient Registry to identify twin pairs in which 1 twin had been exposed to severe trauma and the other twin had not from 1994 to 2018. The co-twin control design allowed for matching on genetic and environmental factors shared within twin pairs. Exposure: Twin pairs were included if 1 twin had been exposed to moderate to severe trauma and the other twin had not (ie, co-twin). Only twin pairs where both twins were alive 6 months after the trauma event were included. Main Outcome and Measure: Twin pairs were followed up from 6 months after trauma until 1 twin experienced the primary composite outcome of death or 1 of 24 predefined immune-mediated or cancer diseases or end of follow-up. Cox proportional hazards regression was used for intrapair analyses of the association between trauma and the primary outcome. Results: A total of 3776 twin pairs were included, and 2290 (61%) were disease free prior to outcome analysis and were eligible for the analysis of the primary outcome. The median (IQR) age was 36.4 (25.7-50.2) years. The median (IQR) follow-up time was 8.6 (3.8-14.5) years. Overall, 1268 twin pairs (55%) reached the primary outcome; the twin exposed to trauma was first to experience the outcome in 724 pairs (32%), whereas the co-twin was first in 544 pairs (24%). The hazard ratio for reaching the composite outcome was 1.33 (95% CI, 1.19-1.49) for twins exposed to trauma. Analyses of death or immune-mediated or cancer disease as separate outcomes provided hazard ratios of 1.91 (95% CI, 1.68-2.18) and 1.28 (95% CI, 1.14-1.44), respectively. Conclusion and Relevance: In this study, twins exposed to moderate to severe trauma had significantly increased risk of death or immune-mediated or cancer disease several years after trauma compared with their co-twins.

Assessing the utility of a sliding-windows deep neural network approach for risk prediction of trauma patients.

The risks of post trauma complications are regulated by the injury, comorbidities, and the clinical trajectories, yet prediction models are often limited to single time-point data. We hypothesize that deep learning prediction models can be used for risk prediction using additive data after trauma using a sliding windows approach. Using the American College of Surgeons Trauma Quality Improvement Program (ACS TQIP) database, we developed three deep neural network models, for sliding-windows risk prediction. Output variables included early- and late mortality and any of 17 complications. As patients moved through the treatment trajectories, performance metrics increased. Models predicted early- and late mortality with ROC AUCs ranging from 0.980 to 0.994 and 0.910 to 0.972, respectively. For the remaining 17 complications, the mean performance ranged from 0.829 to 0.912. In summary, the deep neural networks achieved excellent performance in the sliding windows risk stratification of trauma patients.

Assessing the utility of deep neural networks in detecting superficial surgical site infections from free text electronic health record data.

Background: High-quality outcomes data is crucial for continued surgical quality improvement. Outcomes are generally captured through structured administrative data or through manual curation of unstructured electronic health record (EHR) data. The aim of this study was to apply natural language processing (NLP) to chart notes in the EHR to accurately capture postoperative superficial surgical site infections (SSSIs). Methods: Deep Learning (DL) NLP models were trained on data from 389,865 surgical cases across all 11 hospitals in the Capital Region of Denmark. Surgical cases in the training dataset were performed between January 01st, 2017, and October 30th, 2021. We trained a forward reading and a backward reading universal language model on unlabeled postoperative chart notes recorded within 30 days of a surgical procedure. The two language models were subsequently finetuned on labeled data for the classification of SSSIs. Validation and testing were performed on surgical cases performed during the month of November 2021. We propose two different use cases: a stand-alone machine learning (SAM) pipeline and a human-in-the-loop (HITL) pipeline. Performances of both pipelines were compared to administrative data and to manual curation. Results: The models were trained on 3,983,864 unlabeled chart notes and finetuned on 1,231,656 labeled notes. Models had a test area under the receiver operating characteristic curves (ROC AUC) of 0.989 on individual chart notes and 0.980 on an aggregated case level. The SAM pipeline had a sensitivity of 0.604, a specificity of 0.996, a positive predictive value (PPV) of 0.763, and a negative predictive value (NPV) of 0.991. Prior to human review, the HITL pipeline had a sensitivity of 0.854, a specificity of 0.987, a PPV of 0.603, and a NPV of 0.997. Conclusion: The performance of the SAM pipeline was superior to administrative data, and significantly outperformed previously published results. The performance of the HITL pipeline approached that of manual curation.

Increased inflammatory markers in adult patients born with an atrial septal defect.

Patients with atrial septal defect (ASD) have higher mortality and higher risk of atrial fibrillation, heart failure, pneumonia, and stroke than the general population even if the ASD closes spontaneously in childhood. The reason for the long-term complications remains unknown. Since many of the complications can be linked up with alterations in inflammatory response, we speculate that inflammation may contribute to the association between ASD and morbidity and mortality. We investigated inflammatory activity in adults with an ASD compared with controls. We included 126 adults with an unrepaired ASD. A group of healthy controls were recruited as comparison group (n = 23). Serum samples were analyzed for 92 inflammation-related protein biomarkers using a proximity extension assay. A pathway enrichment analysis was performed using Reactome database. Out of 92 biomarkers, 73 were eligible for data analysis. Increased levels of 14 (19%) biomarkers were found in patients with open ASD and 24 (33%) biomarkers in patients with spontaneously closed defects compared with controls (p < 0.05). Multiple inflammatory pathways showed stronger enrichment in both patient groups when compared with controls. In conclusion, inflammatory activity is altered in adult patients with an unrepaired ASD compared with healthy controls. The increased inflammatory burden of patients with an unrepaired ASD may contribute to the development of morbidities.

A characterization of trauma laparotomies in a scandinavian setting: an observational study.

BACKGROUND: Despite treatment advances, trauma laparotomy continuous to be associated with significant morbidity and mortality. Most of the literature originates from high volume centers, whereas patient characteristics and outcomes in a Scandinavian setting is not well described. The objective of this study is to characterize treatments and outcomes of patients undergoing trauma laparotomy in a Scandinavian setting and compare this to international reports. METHODS: A retrospective study was performed in the Copenhagen University Hospital, Rigshospitalet (CUHR). All patients undergoing a trauma laparotomy within the first 24 h of admission between January 1st 2019 and December 31st 2020 were included. Collected data included demographics, trauma mechanism, injuries, procedures performed and outcomes. RESULTS: A total of 1713 trauma patients were admitted to CUHR of which 98 patients underwent trauma laparotomy. Penetrating trauma accounted for 16.6% of the trauma population and 66.3% of trauma laparotomies. Median time to surgery after arrival at the trauma center (TC) was 12 min for surgeries performed in the Emergency Department (ED) and 103 min for surgeries performed in the operating room (OR). A total of 14.3% of the procedures were performed in the ED. A damage control strategy (DCS) approach was chosen in 18.4% of cases. Our rate of negative laparotomies was 17.3%. We found a mortality rate of 8.2%. The total median length of stay was 6.1 days. CONCLUSION: The overall rates, findings, and outcomes of trauma laparotomies in this Danish cohort is comparable to reports from similar Western European trauma systems.

Pancreatic surgery during the COVID-19 pandemic 2020-2021: an observational cohort study from a third level referral center.

BACKGROUND: During the COVID pandemic there has been limited access to elective surgery including oncologic surgery in several countries world-wide. The aim of this study was to investigate if there was any lockdown effect on pancreatic surgery with special focus on malignant pancreatic and periampullary tumours. METHODS: Patients who underwent pancreatic surgery during the two Danish lockdown periods from 11. March 2020 and the following 12 months were compared with patients who were operated the preceding 3 years. Data on patients' characteristics, waiting time, operations, and clinical outcomes were evaluated. RESULTS: During lockdown and the previous three years the annual number of resections were 242, 232, 253, and 254, respectively (p = 0.851). Although the numbers were not significantly different, there were fluctuations in operations and waiting time during the lockdown. During the second outbreak of COVID October 2020 to March 2021 the overall median waiting time increased to 33 days (quartiles 26;39) compared to 23 (17;33) days during the first outbreak from March to May 2020 (p = 0.019). The same difference was seen for patients with malignant tumours, 30 (23;36) vs. 22 (18;30) months (p = 0.001). However, the fluctuations and waiting time during lockdown was like the preceding three years. Neither 30- nor 90-days mortality, length of stay, number of extended operations, and complications and tumour stage were significantly different from previous years. CONCLUSIONS: There were significant fluctuations in waiting time for operations during the lockdown, but these variations were not different from the preceding three years, wherefore other explanations than an impact from COVID are conceivable.

Agreement Between Standard and ICD-10-Based Injury Severity Scores.

INTRODUCTION: Injury Severity Score (ISS) is used to describe anatomical lesions. ISS is traditionally determined through medical record review (standard ISS), which requires specific training and may be time-consuming. An alternative way to obtain ISS is by use of ICD-9/10 injury diagnoses, and several conversion tools exist. We sought to evaluate the agreement between standard ISS and ISS obtained with two tools converting ICD-10 diagnoses. METHODS: Our cohort consisted of trauma patients ≥18 years admitted to Rigshospitalet between 1999 and 2016. The included patients had standard ISS recorded in the Trauma Audit and Research Network (TARN) database (ISS-TARN), and ICD-10 injury diagnoses for the trauma contact were recorded in the Danish National Patient Registry. We used the tools ICDPIC-R and ICD-AIS map to calculate ISS based on ICD-10 diagnoses. ICDPIC-R provided two ISSs: ISS-TQIP and ISS-NIS. The ICD-AIS map resulted in one ISS: ISS-map. The ISS-TARN was compared to the conversion tool ISSs using Bland-Altman plots. The agreement between ISS-TARN and the conversion tool ISSs for ISS above 15 was assessed using kappa statistics (κ). RESULTS: We included 4308 trauma patients. The median age was 44 years, 70% were male, and 92% had a blunt injury mechanism. The median ISS-TARN was 16 [IQR: 9-25], and the median conversion tool ISSs were 10 [2-25] (ISS-TQIP), 17 [5-26] (ISS-NIS), and 9 [4-16] (ISS-map). The Bland-Altman plots all showed increasing difference in ISS with increasing mean ISS. Bias ranged from -7.3 to 1.1 and limits of agreement ranged between -28.0 and 25.7. The agreement for ISS above 15 was fair to moderate (κ = 0.43 (ISS-TQIP), 0.44 (ISS-NIS), and 0.29 (ISS-map)). CONCLUSION: Using ICDPIC-R or ICD-AIS map to determine ISS is feasible, but limits of agreement were unacceptably wide. The agreement between ISS-TARN and ICDPIC-R was moderate for ISS above 15.

Long-Term Outcomes of Venous Resections in Pancreatic Ductal Adenocarcinoma Patients: A Nationwide Cohort Study.

To investigate whether pancreatic resections (PR) for pancreatic ductal adenocarcinoma (PDAC) is associated with worse survival when resection of the superior mesenteric vein/portal vein (SMV/PV) is required. Background: PR for PDAC with resection of the superior mesenteric vein/portal vein (SMV/PV, PR+V resection) may be associated with inferior overall survival (OS) compared with PR without the need for SMV/PV resection (PR-V). We hypothesized that PR+V results in lower OS compared with PR-V. Method: Retrospective study using data from the nationwide Danish Pancreatic Cancer Database from 2011 to 2020. Data on patients who underwent PR for PDAC were extracted. A group of PR patients found nonresectable on exploratory laparotomy (EXP) was also included. OS was assessed using Kaplan-Meier and Cox proportional hazards models adjusting for confounders (age, sex, R-resection level, chemotherapy, comorbidities, histology T and N classification, procedure subtype as well as tumor distance to the SMV/PV). Results: Overall, 2403 patients were identified. Six hundred two underwent exploration only (EXP group), whereas 412 underwent pancreatic resection with (PR+V group) and 1389 (PR-V) without SMV/PV resection. Five-year OS for the PR+V group was lower (20% vs 30%) compared with PR-V, although multivariate Cox proportional hazards modeling could not associate PR+V status with OS (Hazard ratio 1.11, P = 0.408). Conclusion: When correcting for confounders, PR+V was not associated with lower OS compared with PR-V.

An assessment of the effect of the genotype on postoperative venous thromboembolism risk in 140,831 surgical patients.

BACKGROUND: Postoperative Venous Thromboembolism Events (VTE) constitute a major source of morbidity and mortality after surgery. The aim of this study was to investigate whether commonly occurring Single Nucleotide Polymorphisms (SNPs) are associated with VTE in the surgical setting. METHODS: Retrospective study using data from the United Kingdom (UK) biobank, a genome biobank containing healthcare and genotyping data from more than 500.000 individuals. A cohort of 140,831patients with a registered surgical procedure was identified and used for a discovery genome wide association study (GWAS), with the remainder of the cohort (305,349 non-surgical patients) used as a replication cohort. Primary outcome was associations between SNPs and VTE within 30 days after a surgical procedure. Genome wide significance was set at p = 5 × 10-8. RESULTS: In the surgical (discovery) cohort, no SNPs reached genome wide significance. The VTE association of the top candidate SNP in the ABO gene rs505922 (p = 3.33 × 10-7), was replicated in the general (replication) cohort (p = 2.42 × 10-59). CONCLUSIONS: and Relevance: This study did not identify associations between SNPs and postoperative VTE events reaching genome-wide significance, although the VTE relevance of top candidates were demonstrated.

Postoperative complications: an observational study of trends in the United States from 2012 to 2018.

BACKGROUND: Postoperative complications continue to constitute a major issue for both the healthcare system and the individual patient and are associated with inferior outcomes and higher healthcare costs. The objective of this study was to evaluate the trends of postoperative complication rates over a 7-year period. METHODS: The NSQIP datasets from 2012 to 2018 were used to assess 30-day complication incidence rates including mortality rate following surgical procedures within ten surgical subspecialties. Multivariable logistic regression was used to associate complication rates with dataset year, while adjusting for relevant confounders. RESULTS: A total of 5,880,829 patients undergoing major surgery were included. Particularly the incidence rates of four complications were found to be decreasing: superficial SSI (1.9 to 1.3%), deep SSI (0.6 to 0.4%), urinary tract infection (1.6 to 1.2%) and patient unplanned return to the operating room (3.1 to 2.7%). Incidence rate for organ/space SSI exhibited an increase (1.1 to 1.5%). When adjusted, regression analyses indicated decreased odds ratios (OR) through the study period years for particularly deep SSI OR 0.92 [0.92-0.93], superficial SSI OR 0.94 [0.94-0.94] and acute renal failure OR 0.96 [0.95-0.96] as the predictor variable (study year) increased (p < 0.01). However, OR's for organ/space SSI 1.05 [1.05-1.06], myocardial infarction 1.01 [1.01-1.02] and sepsis 1.01 [1.01-1.02] increased slightly over time (all p < 0.01). CONCLUSIONS: Incidence rates for the complications exhibited a stable trend over the study period, with minor in or decreases observed.

Using machine learning for predicting intensive care unit resource use during the COVID-19 pandemic in Denmark.

The COVID-19 pandemic has put massive strains on hospitals, and tools to guide hospital planners in resource allocation during the ebbs and flows of the pandemic are urgently needed. We investigate whether machine learning (ML) can be used for predictions of intensive care requirements a fixed number of days into the future. Retrospective design where health Records from 42,526 SARS-CoV-2 positive patients in Denmark was extracted. Random Forest (RF) models were trained to predict risk of ICU admission and use of mechanical ventilation after n days (n = 1, 2, …, 15). An extended analysis was provided for n = 5 and n = 10. Models predicted n-day risk of ICU admission with an area under the receiver operator characteristic curve (ROC-AUC) between 0.981 and 0.995, and n-day risk of use of ventilation with an ROC-AUC between 0.982 and 0.997. The corresponding n-day forecasting models predicted the needed ICU capacity with a coefficient of determination (R2) between 0.334 and 0.989 and use of ventilation with an R2 between 0.446 and 0.973. The forecasting models performed worst, when forecasting many days into the future (for large n). For n = 5, ICU capacity was predicted with ROC-AUC 0.990 and R2 0.928, and use of ventilator was predicted with ROC-AUC 0.994 and R2 0.854. Random Forest-based modelling can be used for accurate n-day forecasting predictions of ICU resource requirements, when n is not too large.

Perioperative exercise training for patients with gastrointestinal cancer undergoing surgery: A systematic review and meta-analysis.

Exercise training is emerging as a supportive treatment strategy in surgical oncology, but its effects remain uncertain in patients with gastrointestinal cancer. The primary objective of this systematic review and meta-analysis was to evaluate the effects of perioperative exercise training on gastrointestinal cancer-specific mortality, recurrence, and surgical outcomes (postoperative complications, hospitalization, surgical stress) in patients with gastrointestinal cancer. Randomized or quasi-randomized controlled trials evaluating the effects of perioperative exercise training versus control in patients with GI cancer were eligible. MEDLINE, EMBASE, CENTRAL, CINAHL, PEDro, and SPORTDiscus were systematically searched on June 20, 2020. Data were synthesized using random-effects meta-analyses. Risk of bias was assessed using the Cochrane risk of bias tool 2, and the certainty of evidence was assessed using GRADE. Study selection, data extraction, risk of bias, and GRADE assessments were performed independently by two authors. Ten randomized controlled trials comprising 448 participants with gastrointestinal cancer were eligible. Meta-analyses indicated no statistical effects of exercise on postoperative complications (risk ratio: 1.11, 95% CI: 0.84; 1.47), readmissions (risk ratio: 2.76; 95% CI: 0.00, 9394.76), or postoperative length of stay (difference in means: -0.47, 95% CI: -17.2; 16.2 days). None of the eligible studies assessed gastrointestinal cancer-specific mortality or recurrence. Overall risk of bias was high or of some concerns in all studies, and the certainty of evidence was very low. The effects of perioperative exercise on cancer-specific and surgical outcomes are unknown in patients with gastrointestinal cancer due to lack of studies and very low certainty of evidence.