Using meta-analytic priors to incorporate external information for study evaluation

by   Thilo Welz, et al.

Background: The COVID-19 pandemic has had a profound impact on health, everyday life and economics around the world. An important complication that can arise in connection with a COVID-19 infection is acute kidney injury. A recent observational cohort study of COVID-19 patients treated at multiple sites of a tertiary care center in Berlin, Germany identified risk factors for the development of (severe) acute kidney injury. Since inferring results from a single study can be tricky, we validate these findings and potentially adjust results by including external information from other studies on acute kidney injury and COVID-19. Methods: We synthesize the results of the main study with other trials via a Bayesian meta-analysis. The external information is used to construct a predictive distribution and to derive posterior estimates for the study of interest. We focus on various important potential risk factors for acute kidney injury development such as mechanical ventilation, use of vasopressors, hypertension, obesity, diabetes, gender and smoking. Results: Our results show that depending on the degree of heterogeneity in the data the estimated effect sizes may be refined considerably with inclusion of external data. Our findings confirm that mechanical ventilation and use of vasopressors are important risk factors for the development of acute kidney injury in COVID-19 patients. Hypertension also appears to be a risk factor that should not be ignored. Shrinkage weights depended to a large extent on the estimated heterogeneity in the model. Conclusions: Our work shows how external information can be used to adjust the results from a primary study, using a Bayesian meta-analytic approach. How much information is borrowed from external studies will depend on the degree of heterogeneity present in the model.


page 1

page 2

page 3

page 4


Bayesian inference for asymptomatic COVID-19 infection rates

To strengthen inferences meta analyses are commonly used to summarize in...

Deep Learning Prediction of Severe Health Risks for Pediatric COVID-19 Patients with a Large Feature Set in 2021 BARDA Data Challenge

Most children infected with COVID-19 have no or mild symptoms and can re...

Bounds for the weight of external data in shrinkage estimation

Dynamical borrowing of information may be facilitated via shrinkage esti...

Covid-19 risk factors: Statistical learning from German healthcare claims data

We analyse prior risk factors for severe, critical or fatal courses of C...

metamedian: An R package for meta-analyzing studies reporting medians

When performing an aggregate data meta-analysis of a continuous outcome,...

Methods for the inclusion of real world evidence in network meta-analysis

Background: Network Meta-Analysis (NMA) is a key component of submission...

Please sign up or login with your details

Forgot password? Click here to reset