Creating order out of chaos
A guide to systematic literature reviews of predictive models.
Predictive models are becoming increasingly important in clinical practice. Unfortunately, research into these models is often not reproducible and their use unclear. Systematic literature reviews are needed to assess and summarize evidence about models in a specific clinical area.
In her thesis, Anneke Damen presents guidelines for systematically summarizing literature. The guidelines show that there is an abundance of models for cardiovascular risk prediction. The majority of these models consist of a similar set of predictors, including age, sex, smoking, diabetes, blood pressure and cholesterol values in the blood.
Anneke tells us, "There is a proliferation of new predictive models – there are now some 363 – while only one is used in Dutch practice. This is a waste of research time. Research should focus on the evaluation and improvement of existing models. A researcher tests the model on their own dataset. If it does not work, the existing predictive model can be adapted."
In her research, she shows that predictive models used in the United States overestimate the risk of cardiovascular disease. She also demonstrates that incomplete reporting leads to a shortage of key information needed to make the developed models practicable.
Anneke has developed a number of formulas that quantitatively summarize the results. "In this way I'm creating a little order out of the research literature chaos."
Based on the results of her research, Anneke wants to offer better education to (bio)medical students about the development and reporting of predictive models. In addition, she wants to make online educational resources available to researchers who make a systematic literature review of predictive models.