The inefficiency of literature searches is an old informatics problem. There has been much progress in facilitating literature searches, such as the registration of trials and observational studies, the proposal of optimal search strategies for various fields, and the proposal of search filters, but the efficiency of literature searches has not markedly improved. The goal of this project is to strengthen the scientific basis of the CoCites method, a novel citation-based search method for scientific literature that we recently developed. The method aims to find articles that are similar to one or more `known' articles in two consecutive searches: 1) finding articles that are cited together with the known articles by ranking their co-citation frequencies; and 2) finding backward and forward citations by ranking their citation frequencies. Articles with frequencies above the selection thresholds are screened for similarity. In two pilot studies, where we aimed to reproduce the literature searches of 52 published meta-analyses, we showed that articles included in each meta-analysis ranked high on the list of (co-)citation frequencies. The CoCites method was more efficient than keyword searches and able to retrieve 80% of the studies included in the meta-analyses. The proposed project aims to refine the CoCites method and investigates its application in various fields. Similar as in the pilot studies, we will reproduce the literature searches of published meta-analyses, which allows to calculate the sensitivity and efficiency of the method. To investigate the method in relation to citation practices and study characteristics, we will create a database in which we document the characteristics of 250 published meta-analyses as well as characteristics of each of their included studies. The cocitations will be downloaded from Web of Science. We will write a macro in Microsoft Excel to automatically process the downloaded files and produce the co-citation and citation rankings. We will apply the method to all or selections of the 250 meta-analyses, where we investigate the sensitivity and efficiency while: varying the selection threshold of the first search (Aim 1); adding a third search of co-citations, direct citations or both (Aim 2); and varying the selection and number of `known' articles (Aim 4). We will investigate the impact of network and study characteristics on the sensitivity and efficiency using regression analysis (Aim 5). To investigate whether the results of the meta-analyses change when studies are not found, we repeat the meta-analyses excluding the `missed' studies (Aim 3). We will apply the method to meta-analyses in five fields that differ in citation practice (Aim 5). Finally, we will prospectively apply the method to five then-ongoing systematic reviews conducted at the Community Guide of the Centers for Disease Control and Prevention.
Finding eligible studies for meta-analysis and systematic reviews relies on keyword-based searching as the gold standard, despite its inefficiency. This project investigates a novel search method that ranks articles on their degree of co-citation with one or more known articles before reviewing their eligibility. This method, applied over time, will strengthen the connections between articles, promote the accumulation of evidence, and facilitate evidence synthesis.
