Identifying direct risk factors in UK Biobank with simultaneous Bayesian-frequentist model-averaged hypothesis testing using Doublethink

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 2, 2024

Abstract Big data approaches to discovering non-genetic risk factors have lagged behind genome-wide association studies that routinely uncover novel genetic for diverse diseases. Instead, epidemiology typically focuses on candidate factors. Since modern biobanks contain thousands of potential factors, may introduce bias, inadequately control multiple testing, and miss important signals. Bayesian model averaging offers a solution, but classical statistics predominates, perhaps because concern the prior unduly influences results. Here we show simultaneous frequentist discovery direct is possible via model-averaged hypothesis testing approach large samples called ‘Doublethink’. Doublethink produces interchangeable posterior odds p -values false rate (FDR) familywise error (FWER). We implement in R apply it discover COVID-19 hospitalization 2020 among 1,912 variables UK Biobank. find nine exposome-wide significant at 9% FDR 0.05% FWER. These include several commonly reported (e.g. age, sex, obesity) exclude others diabetes, cardiovascular disease, hypertension) which might be mediated through measuring general comorbidity numbers medications). identify effects infrequently (psychiatric disorders, infection, dementia aging), how groups correlated useful alternative pre-analysis variable selection. discuss impact limitations joint Bayesian-frequentist inference, mutual insights afforded into long-standing differences statistical scientific discovery.

Language: Английский

---

C. Everett Koop Healthcare System for Biosecurity and Defense

Springer eBooks, Journal Year: 2024, Volume and Issue: unknown, P. 165 - 192

Published: Jan. 1, 2024

Language: Английский

---

What is Occluding Our Understanding of Retinal Vein Occlusion?

Ophthalmology and Therapy, Journal Year: 2024, Volume and Issue: 13(12), P. 3025 - 3034

Published: Oct. 10, 2024

Language: Английский

---

Predicting Post-Liver Transplant Outcomes in Patients with Acute-on-Chronic Liver Failure using Expert-Augmented Machine Learning

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2023, Volume and Issue: unknown

Published: March 5, 2023

Abstract Background Liver transplantation (LT) is a treatment for acute-on-chronic liver failure (ACLF) but up to 40% mortality post-LT has been reported. Existing models in ACLF have limited by small samples. In this study, we developed novel Expert-Augmented Machine Learning (EAML) model predict outcomes. Methods We identified patients the University of California Health Data Warehouse (UCHDW). used EAML, which uses RuleFit machine learning (ML) algorithm extract rules from decision-trees that are then evaluated human experts, compared EAML/RuleFit’s performances versus other popular models. Results 1,384 patients. For death at one-year: areas-under-the-receiver-operating characteristic curve (AUROCs) were 0.707 (Confidence Interval [CI] 0.625-0.793) EAML and 0.719 (CI 0.640-0.800) RuleFit. 90-days: AUROCs 0.678 0.581-0.776) 0.615-0.800) pairwise comparisons, EAML/RuleFit outperformed cross-sectional Divergences between experts ML rankings revealed biases artifacts underlying data. Conclusions Significant discrepancies occurred biomarkers clinical practice. may serve as method ML-guided hypothesis generation further research.

Language: Английский

---

Sex hormones and the total testosterone:estradiol ratio as predictors of severe acute respiratory syndrome coronavirus 2 infection in hospitalized men

Andrology, Journal Year: 2024, Volume and Issue: 12(6), P. 1381 - 1388

Published: Jan. 11, 2024

The predictive ability of the early determination sex steroids and total testosterone:estradiol ratio for risk severe coronavirus disease 2019 or potential existence a biological gradient in this relationship has not been evaluated.

Language: Английский

---