Identifying empirical studies for mixed studies reviews: The mixed filter and the automated text classifier

Abstract

Mixed studies reviews include empirical studies with diverse designs (qualitative, quantitative and mixed methods). To make the process of identifying relevant empirical studies for such reviews more efficient, we developed a mixed filter that included different keywords and subject headings for quantitative (e.g., cohort study), qualitative (e.g., focus group), and mixed methods studies. It was tested for six journals from three disciplines. We measured precision (proportion of retrieved documents being relevant), sensitivity (proportion of relevant documents retrieved), and specificity (proportion of non-relevant documents not retrieved). Records were coded before applying the filter and compared with retrieved records, and descriptive statistics were performed, suggesting the mixed filter has high sensitivity, but lower precision and specificity (close to 50%). Next, based on the success of the filter, we developed an automated text classification system that can automatically select empirical studies in order to facilitate systematic mixed studies reviews. Several algorithms were trained and validated with 8,050 database records that were previously manually categorized. Decision trees had the best results and surpassed the accuracy of the filter by 30% when using full-text documents. This algorithm was then adapted into an online format that can be used by researchers to analyze their bibliography and categorize records into “empirical” and “nonempirical”.