The heterogeneous methods used to conduct systematic reviews of prevalence studies in adult populations: a cross-sectional study

Article type
Authors
Buitrago-Garcia D1, Robles Rodriguez WG2, Eslava-Schmalbach J3, Salanti G4, Low N5
1Institute of Social and Preventive Medicine, Graduate School of Health Sciences, University of Bern
2Maestría en Epidemiología Clínica, Fundación Universitaria de Ciencias de la Salud (FUCS)
3Facultad de Medicina, Universidad Nacional de Colombia
4Institute of Social and Preventive Medicine, University of Bern
5Institute of Social and Preventive Medicine, University of Bern
Abstract
Background:Systematic reviews (SRs) of prevalence can provide information about the prevalence of a condition or disease, identify differences in prevalence between populations, inform stakeholders and clinicians about the burden of a disease, and assess the impact of interventions on the prevalence over time and space. However, few resources have been invested so far into understanding and enhancing the methods underpinning systematic reviews of prevalence studies. In particular, empirical evidence about systematic reviews of prevalence studies is scarce.

Objectives:To examine the methodological characteristics of published systematic reviews of prevalence studies in adult populations.

Methods:We conducted a systematic search in the following databases: MEDLINE-Ovid, Embase-Ovid, CINAHL and LILACS. We included systematic reviews of prevalence studies related to medical conditions in adult populations from 1/2010 until 9/2019. Data were collected using a 30-item form. We performed a descriptive analysis using Stata16®

Results:A total of 870 systematic reviews were included. There was an eightfold increment in the number of publications over the years (2% in 2010 to 17% in 2019). Most of them aimed to answer questions related to psychiatric conditions (15%), cardiovascular conditions (10%), respiratory conditions (5%), diabetes (5%), and neurological conditions (5%). The median of included studies was 23 (IQR 13-42). The quality of the included studies was assessed in 61% of SRs; we identified 38 different tools used to perform this step. It is noteworthy that 15% of authors chose to design their own tool or to adapt an existing tool for their systematic review. Meta-analysis was done in 61% of the included SRs, and in 45% of them, the authors did not mention in detail the method used. In the cases where it was mentioned, the most frequently used transformation was the Double Arcsine, followed by DerSimonian and Laird's random-effects synthesis model. Heterogeneity was assessed in 93% of meta-analyses, mainly statistical heterogeneity using the I2 statistic.

Conclusions:The increase in the publications of SRs of prevalence makes evident the importance of these studies, which are relevant for a variety of populations and conditions. However, due to the high heterogeneity in the methods used. Although, there are methodological proposals for developing SRs of prevalence, there is a need for stablishing a consensus in order to remove biases and support the publication of high quality evidence.

Patient or healthcare consumer involvement: Many decisions made in public health and clinical practice are based on the evidence gathered from systematic reviews. In many cases, the only data available comes from prevalence studies. Due to the heterogeneity in the development of these studies, there is a potential risk of introducing bias in the reported evidence.