Data published today in The BMJ show that prediction models for the diagnosis and prognosis of the novel coronavirus (COVID-19) may be based on weak evidence from biased and unreliable studies.
A team of European researchers reviewed 31 prediction models from 27 studies in published and preprint reports. The models predicted existing COVID-19 infection, future complications for those already diagnosed or which individuals are at high risk for COVID-19 in the general population. Most of the studies used data from COVID-19 cases in China.
The researchers concluded that all of the studies had a high risk for bias because they had a nonrepresentative selection of control patients, excluded patients who were still ill at the end of the study and had poor statistical analysis.
Reporting quality also varied significantly, with most studies not including a description of the study population or the intended use of the models.