Doctors making decisions on flawed Covid-19 data

Researchers warn doctors are making decisions based on flawed Covid-19 data: Doctors may be making decisions about whether Covid-19 patients have the virus, need a ventilator, or remain in hospital, based on weak and over-optimistic evidence, Keele researchers have found.

Researchers from Keele University and institutions from across Europe, including the University of Oxford, have published the third iteration of a study which examines the quality of prediction models that are being proposed to inform the diagnosis and prognosis of patients with suspected Covid-19.

The study has reviewed 145 prediction models so far, with many being recommended for use in practice. The researchers, whose paper has been published in The British Medical Journal (BMJ), found that the data and methods used in these studies were potentially at high risk of bias, while some of the studies included recommendations that were questionable if put into practice.

The researchers warn that the potentially flawed models may result in doctors making inappropriate decisions about whether patients have the virus, need a ventilator, or should remain in hospital.

They found that all the studies had a high risk of bias. Some of the studies had a non-representative selection of patients, while others excluded patients who were still ill at the end of the studies. Others had poor statistical analysis.

To address the problem, the authors encourage data sharing across the world to examine the quality of models in more detail, and for new Covid-19 research studies to adhere to appropriate methodology standards and reporting guidelines.

Professor Richard Riley, Professor of biostatistics, said: “Researchers must stop churning out new models with inadequate data and methods. It is important to now share high-quality data that can be used to externally validate the performance of existing models. Until then, no models should be considered fit-for-purpose.”

Hippocratic Post
Latest posts by Hippocratic Post (see all)
Share:  

More in this category

Subscribe
Notify of
0 Comments
Inline Feedbacks
View all comments
0
Would love your thoughts, please comment.x
()
x