tetano
Editor, Senior Moderator
J Med Virol
. 2020 Aug 4.
doi: 10.1002/jmv.26390. Online ahead of print.
Statistical issues in the development a COVID-19 prediction models
Gary S Collins[SUP] 1 [/SUP], Jack Wilkinson[SUP] 2 [/SUP]
Affiliations
Abstract
Clinical prediction models to aid diagnosis, assess disease severity or prognosis have enormous potential to aid clinical decision making during the covid-19 pandemic. A living systematic review has, so far, identified 145 covid-19 prediction models published (or preprinted) between 03-January-2020 and 05-May-2020. Despite the considerable interest in developing covid-19 prediction models, the review concluded that all models to date, with no exception, are at high risk of bias with concerns related to data quality, flaws in the statistical analysis and poor reporting, and none are recommended for use. Disappointingly, the recent study by Yang and colleagues describing the development of a prediction model to identify covid-19 patients with severe disease, is no different. The study has failed to report important information needed to judge the study findings, but has numerous methodological concerns in design and analysis that deserve highlighting. This article is protected by copyright. All rights reserved.
Keywords: Biostatistics & Bioinformatics; Coronavirus < Virus classification; Epidemiology.
. 2020 Aug 4.
doi: 10.1002/jmv.26390. Online ahead of print.
Statistical issues in the development a COVID-19 prediction models
Gary S Collins[SUP] 1 [/SUP], Jack Wilkinson[SUP] 2 [/SUP]
Affiliations
- PMID: 32749704
- DOI: 10.1002/jmv.26390
Abstract
Clinical prediction models to aid diagnosis, assess disease severity or prognosis have enormous potential to aid clinical decision making during the covid-19 pandemic. A living systematic review has, so far, identified 145 covid-19 prediction models published (or preprinted) between 03-January-2020 and 05-May-2020. Despite the considerable interest in developing covid-19 prediction models, the review concluded that all models to date, with no exception, are at high risk of bias with concerns related to data quality, flaws in the statistical analysis and poor reporting, and none are recommended for use. Disappointingly, the recent study by Yang and colleagues describing the development of a prediction model to identify covid-19 patients with severe disease, is no different. The study has failed to report important information needed to judge the study findings, but has numerous methodological concerns in design and analysis that deserve highlighting. This article is protected by copyright. All rights reserved.
Keywords: Biostatistics & Bioinformatics; Coronavirus < Virus classification; Epidemiology.