Methodological issues in research on drug-related admissions: A meta-epidemiological review with focus on causality assessments

Abstract

Aim: To investigate methodological aspects potentially related to the diverging scientific literature on the prevalence of drug-related hospitalisations, focusing on causality assessments.

Methods: Original studies contributing data to a recent meta-analysis were reviewed. Methodological aspects, in particular those related to causality assessments, were extracted and compiled.

Results: Thirteen studies provided data on the prevalence of drug-related admissions. Seven studies focused on adverse drug reactions (prevalences 1.3-10%), and six studies used the broader concept of drug-related problems (prevalences 4.5-41%). In 10 out of 13 studies, causality between the drug and the specified problem was assessed. One study required a probable causal relationship; the remaining studies merely required a possible causal relationship. Five studies assessed the association between the problem assumed to be related to drug therapy and the admission, at one end requiring the former to be demonstrated as the underlying cause and, at the other, merely requiring a temporal relationship between drug intake and admission. Three out of eight studies involving multiple assessors for all/some cases reported the inter-rater agreement, ranging from none to almost perfect. Physicians were involved in the assessments in five studies, reporting prevalences of 3.2% to 4.5%, while studies without such medical input reported prevalences of 8.8% to 41%.

Conclusions: This review illustrates that methodological issues contribute to the diverse literature on drug-related admissions. We provide suggestions for harmonisation of research, including explicitly assessing the drug-problem-admission relationships from a medical perspective, focusing on problems where the drug treatment is the probable culprit.

Keywords: adverse drug reaction; assessment; causality; drug-related admission; methodology; pharmacoepidemiology.