Doctors, nurses and medical professionals all over the world are working long hours diagnosing and treating COVID-19 coronavirus patients. Their choices are being made based on rapidly developing scientific evidence. Could that evidence essentially be “bad intel?”
If the researchers at Keele University, the University of Oxford and other universities and institutions across Europe are right, many front line medical personnel are making treatment decisions based on “weak and over-optimistic evidence.” Their claim comes after a review of 27 research studies which recommend models for the diagnosis and prognosis of COVID-19 patients.
With health care systems across the globe under great strain and stretched to the limits for equipment, medications and personnel, the researchers fear that these potentially flawed diagnosis and treatment models will make things worse. Doctors may make incorrect decisions about whether a patient has COVID-19, whether patients should be hospitalized and if or when to use a ventilator.
In their findings, published in the British Medical Journal (BMJ,) the researchers claim that the methods and data used in the studies they reviewed had a high risk of bias. Some of the studies made recommendations that may be questionable if put into widespread practice.
With nearly 1.6 million cases and over 95,000 deaths worldwide as of this writing, doctors are under a lot of pressure to detect and diagnose infected patients, give prognoses for those cases and recommend and start treatment.
Of the studies reviewed by the researchers, 25 used data from Chinese cases, one used data from Italy and the data for the other was internationally sourced. Collection dates ran from December 8, 2019 to March 15, 2020.
A high risk of bias was found in every study. Some simply had poor quality statistical analysis. Some used a non-representative selection of patients and others left out patients who remained ill at the conclusion of the studies.
Clinical data from COVID-19 patients is scarce and these studies were all executed under serious time constraints, since the information was needed to inform medical decision-making as fast as possible. The researchers recognized those facts.
But the flaws discovered in their reviews lead them to believe that the biases and errors may lead to worse medical decisions, not better. Since these studies and their findings are already in use, the sooner the biases are corrected, the better.
The researchers believe more rigorous prediction models and better diagnostic and treatment protocols can be delivered to medical professionals. They believe that if researchers share anonymized, high-quality patient data through a clearing house organization, that collaborative effort will speed the improvement of diagnosing and treating COVID-19.
Richard Riley, professor of biostatistics at Keele University, said:”Doctors should be making decisions based on robust evidence. Unfortunately, current studies fall short in providing robust models for diagnosis or prognosis of Covid-19.
“Researchers around the world should unite in sharing their data immediately to improve the quality of future studies in this area. Health professionals and methodologists must work together, to pool high-quality and representative Covid-19 data that is then analyzed using appropriate statistical methods.”
Dr. Kym Snell, Lecturer in Biostatistics at Keele University, said: “Many researchers are focusing their efforts on trying to assist healthcare professionals in this very difficult time—this should be applauded. However, we still need to ensure that models for the diagnosis and prognosis of patients with Covid-19 are properly tested.
“We will continue to critically appraise new models so that healthcare professionals have an up-to-date summary of the evidence.”
Laure Wynants, Assistant Professor of Epidemiology at Maastricht University, which led the research, said:”This research has shown how important it is for researchers working in this field to share good quality evidence. This helps quickly identify models that work well and can potentially support decision-making in clinical practice.”
Researchers often say “garbage in, garbage out” as it relates to data. In this case, there may have been quality data going in (or maybe not,) but the frantic pace of research in the effort to stem the spread of COVID-19 may have led to some dangerous biases. With lives on the line, nobody on the medical front lines can afford that.
Keep the faith and keep after it!
Related Content –
Stanford Study Says COVID-19 Coronavirus Patients Often Have Secondary Respiratory Illness
COVID-19 Coronavirus May Spread More Easily Than We First Thought
For Some COVID-19 Coronavirus Patients, The Virus Lingers After Symptoms Disappear
Journal Reference – Laure Wynants et al. Prediction models for diagnosis and prognosis of covid-19 infection: systematic review and critical appraisal, BMJ (2020). DOI: 10.1136/bmj.m1328