For years and considering the large increase in the number of publications, systematic reviews and meta-analyses have proven to be very useful methodologies for synthesizing the evidence available in a specific area of knowledge and, therefore, for the implementation of evidence-based practice. However, they also show some weak points, including the variable methodological quality or risk of bias of the primary studies that integrate them. This threat to the validity of research synthesis is especially relevant when systematic reviews and meta-analyses include studies of non-experimental designs, more prone to risk of bias. Added to this threat is the lack of a consensus tool to assess the risk of bias of non-experimental studies. Beyond the tools used to assess the risk of bias, there are also doubts about the choice of the most appropriate statistical procedures and techniques to incorporate the results of the evaluation of risk of bias in the reviews, as well as to assess the relationship between the measures of risk of bias and the effect size estimation of a meta-analysis. The general objective of this doctoral thesis is to carry out a comparative analysis of the different procedures proposed in the literature to incorporate the measures of the risk of bias in a research synthesis. For this purpose, first, the proposed techniques to incorporate the indicators of risk of bias in research synthesis were reviewed and classified. Subsequently, some of these techniques were applied to a meta-analysis of cohort studies, using different assessment tools and types of measures of risk of bias. Secondly, the need for more extensive information on the factors that may influence the relationship between the risk of bias and the results of the research synthesis, lead us to carry out a meta-review to know how is being assessed and incorporated the risk of bias in the research synthesis in health psychology field. The last step was to carry out an empirical study to compare three risk of bias assessment tools, applying them to primary studies of a published meta-analysis. The aim of this comparison was assessing the reliability, validity and suitability of the tools, as well as determining the possible differential effect of the results of risk of bias assessment obtained with the three tools on the meta-analysis results and conclusions. The work developed throughout this process has materialized in two publications which constitute this thesis by compendium. Once again, the worrisome results obtained from the meta-review, showed the lack of clear guidelines to evaluate and incorporate the risk of bias results in research synthesis. The results of risk of bias assessment showed great variability depending on the tool used and no significant associations between risk of bias and size of the effect were found. Considering the interpretation of the results, it is noted that advances in the field in which this doctoral thesis is framed go through new ways, such as living systematic reviews, the assessment of the quality of evidence when it comes from non-experimental or complex research, the implementation science or the quality of the report itself.
Evaluación e incorporación del riesgo de sesgo de estudios no experimentales en revisiones sistemáticas y metanàlisis
Oliveras Boté, I. (Author). 18 Dec 2018
Student thesis: Doctoral thesis
Student thesis: Doctoral thesis