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.

One-year outcomes in atrial fibrillation presenting during infections: a nationwide registry-based study.

AIMS: Thromboprophylaxis guidelines for patients with concurrent atrial fibrillation (AF) during infections are unclear and not supported by data. We compared 1-year outcomes in patients with infection-related AF and infection without AF. METHODS AND RESULTS: By crosslinking Danish nationwide registry data, AF naïve patients admitted with infection (1996-2016) were identified. Those with AF during the infection (infection-related AF) were matched 1:3 according to age, sex, type of infection, and year with patients with infection without AF. Outcomes (AF, thromboembolic events) were assessed by multivariable Cox regression. The study population comprised 30 307 patients with infection-related AF and 90 912 patients with infection without AF [median age 79 years (interquartile range 71-86), 47.6% males in both groups]. The 1-year absolute risk of AF and thromboembolic events were 36.4% and 7.6%, respectively (infection-related AF) and 1.9% and 4.4%, respectively (infection without AF). In the multivariable analyses, infection-related AF was associated with an increased long-term risk of AF and thromboembolic events compared with infection without AF: hazard ratio (HR) 25.98, 95% confidence interval (CI) 24.64-27.39 for AF and HR 2.10, 95% CI 1.98-2.22 for thromboembolic events. Further, differences in risks existed across different subtypes of infections. CONCLUSION: During the first year after discharge, 36% of patients with infection-related AF had a new hospital contact with AF. Infection-related AF was associated with increased risk of thromboembolic events compared with infection without AF and our results suggest that AF related to infection may merit treatment and follow-up similar to that of AF not related to infection.

Secondary stroke prophylaxis in atrial fibrillation patients with chronic kidney disease: a nationwide cohort study.

AIMS: Oral anticoagulation (OAC) therapy as secondary stroke prophylaxis in atrial fibrillation (AF) patients with chronic kidney disease (CKD) remains unexplored and poses a clinical treatment dilemma. We assessed the long-term risk of thromboembolic events according to post-stroke OAC therapy in AF patients with CKD after their first ischaemic stroke. METHODS AND RESULTS: We identified Danish AF patients with CKD who presented with first-time ischaemic stroke from 2005 to 2014. Chronic kidney disease was defined as a diagnosis code for CKD before baseline, defined as 100 days after stroke discharge. Post-stroke antithrombotic therapy (OAC therapy and antiplatelet therapy) was identified from prescription claims from discharge to baseline. Cumulative incidences and adjusted hazard ratios (HRs) of thromboembolic events according to post-stroke OAC therapy were examined. Of 1252 AF patients with CKD presenting with ischaemic stroke, 631 (50.4%) patients were on OAC therapy and 621 (49.6%) were on antiplatelet therapy alone at baseline [median age 76 (interquartile range, IQR 71-83) and 80 (IQR 72-86), respectively]. The median follow-up period was 1.9 years (IQR 0.8-3.6). Cumulative incidence rates of thromboembolic events and bleeding showed no significant difference between those on OAC therapy and antiplatelet therapy. The results from the multivariable analysis revealed similar results: thromboembolic risk was not modified by OAC treatment [adjusted HR 0.89, 95% confidence interval (CI) 0.73-1.09] nor was the risk of bleeding (adjusted HR 0.88, 95% CI 0.67-1.17). CONCLUSION: Oral anticoagulation in patients with CKD and prior stroke was not associated with a reduced risk of recurrent thromboembolic events compared with antiplatelet therapy.