When I use a word . . . . Too much healthcare—observational studies

Abstract

Observational studies generally provide less reliable evidence than well designed randomised trials and systematic reviews and meta-analyses of such trials. Their results can be affected by confounders, colliders, and a wide range of biases, and they are likely to overestimate effect sizes in studies of benefits. However, they do have their uses and are likely to give different types of information. Systematic reviews and meta-analyses of observational studies are generally discouraged, except in some cases, including studies of children, pregnant women, and older people, and in other cases when randomised trials are not possible or to inform the design of such trials. In studies of adverse drug reactions, observational studies may suggest associations when none exists. Over-reliance on observational studies of poor quality during covid-19 has led to misleading results.

When I use a word . . . . Too little healthcare—coroners’ concerns

“…The topics have included overdetection,4 overdefinition,5 overdiagnosis,6 technology,7 overconfidence,8 underconfidence,9 self-assessment,10 incompetence,11 biomarkers,12 monitoring,13 and observational studies 14. Other topics await exploration.…”

STROBE and pre-registration of observational studies

“…Aronson rightly acknowledges that observational studies only generate conjectures and warns that residual confounding can be quite difficult to fix 1. But the issue is deeper than this.…”