Work

The science of prevalence: development of an evidence-based tool to assess the risk of bias in prevalence studies.

Year: 2024
Type: dissertation
Abstract: Prevalence is a core concept in epidemiology, defined as the proportion of people with specific characteristics (diseases, risk factors, or protective factors) at a certain point or during a certain period. The importance of prevalence might have been undervalued because calculating it or designing prevalence studies are perceived to be straightforward. Furthermore, research on the methodological aspects on prevalence estimates has not been exhaustive, compared to other study designs such as randomised controlled trials. The same potential underestimation and lack of research apply to systematic reviews of prevalence, despite their increase in recent years. During the COVID-19 pandemic, prevalence gained relevance because prevalence data were essential for many research questions related to topics such as seroprevalence or the prevalence of specific symptoms and comorbidities of people with SARS-CoV-2 in different settings. The work presented in this thesis aims to provide knowledge about the epidemiology of prevalence studies, and biases that affect them. The thesis contains seven chapters: an introduction; four chapters that present the projects conducted during my doctoral studies, which, at the time of this writing, either have been published or are currently being prepared for submission; and finally, a discussion of the main findings and implications of the doctoral project. This thesis also includes two additional projects that contributed to broadening my skills as an epidemiologist, even though they did not constitute the core of my doctoral project. In Chapter 1, I comprehensively introduce prevalence and relevant uses of prevalence studies. Afterwards, I present systematic reviews, and the relevance of risk of bias assessment followed by systematic reviews of observational studies and systematic reviews of prevalence. Then, I present the current guidance available to conduct systematic reviews of prevalence. Chapter 2 is a communication published in the BMJ Open. This publication approaches the basic concepts of prevalence and the biases that affect prevalence estimates. Each scenario is illustrated with an example of a study conducted on COVID-19. This communication highlights how prevalence became more relevant during the pandemic. Chapter 3 presents a meta-epidemiological study of systematic reviews of prevalence conducted on adult populations from 2010 to 2020, where I included 1172 publications. I describe their characteristics and compliance with the PRISMA 2009 checklist and assessed the relationship between compliance with the PRISMA checklist and study level variables in a linear regression analysis. Compliance with the PRISMA checklist is still suboptimal for some items, specifically in the methods section. The year of publication, the number of authors who published the review, publishing in an open-access journal and reporting the use of a guideline were associated with completeness of reporting to the PRISMA 2009. This manuscript will be submitted soon. Chapter 4 presents two iterations of a living systematic review about the proportion of people with asymptomatic SARS-CoV-2. The first included 80 studies, and it found that the pooled proportion of people with a truly asymptomatic infection was 20%, 95% CI (17-25), with a prediction interval of 3% to 67%. In the next version, the eligibility criteria became more strict, and were met by 146 studies. A pooled proportion of asymptomatic people with SARS-CoV-2 was not presented, due to the high heterogeneity of the studies. A meta-regression analysis did not identify the characteristics that were associated with heterogeneity. Chapter 5 presents the development of a tool to assess bias in the prevalence of mental health studies conducted during the first wave of the COVID-19 pandemic. The proposed tool approached selection and information bias with three questions: one about the sample's representativeness regarding the target population, the second about the representativeness of the respondents, and the third one dealt with the measurement of the condition. Agreement for the three items was 83%, 90% and 93%, respectively. The weighted kappa scores were 0.63, 95% CI (0.54 - 0.73), 0.71, 95% CI (0.67 - 0.85), and 0.32, 95% CI (–0.04 - 0.63). Chapter 6 discusses the findings of the studies, the current stage of an evidence-based tool to assess the risk of bias for any prevalence study, the study’s strengths and limitations, and future questions. This thesis provides a conceptual framework that portrays the bias affecting prevalence studies. I show how systematic reviews of prevalence have been conducted, which aspects need improvement, and the need for new methodologies. Furthermore, I present a living systematic review that used the conceptual framework to answer questions relevant to the COVID-19 pandemic. I conclude with a risk of bias tool that assesses the bias of prevalence studies on mental health. This tool forms the basis for future development of a domain-based tool to assess the biases in prevalence studies.
Author: Buitrago Garcia, Diana
Language: English
Open Access status: green
OpenAlex ID: W7043132407
Cites:
Cited by:
Related to: 9