OBJECTIVES: Meta-analysis is a statistical approach to develop a systematic overview of the effects of a therapeutic treatment. It is the highest type of evidence for medical guidelines and health economic prioritization. It is time consuming and only provides a static ‘here and now’ overview, which must be repeated. We wanted to explore how many parts of the meta-analysis could be optimized with an improved information technology (IT) system.

METHODS: The process of meta-analysis was divided into 4 phases: Data import, Screening of articles, Extraction of data and Statistical analysis. For each phase we tested which parts of the process could be eliminated through improved IT software and Artificial intelligence (AI).

RESULTS: In the data import phase we could directly connect to literature search engines such as Pubmed and Embase. The Screening phase was optimized with dynamic back loops, so the search string would automatically be updated. In the extraction phase, AI was used to extract data directly from the articles. For statistical analysis, the system was connected directly to the statistical software ‘R’. By incorporating all phases into one IT-system i.e. linking the calculation of the statistical forest plot to the search string, we discovered an additional optimization of the meta-analysis process. It was possible to play around with the forest plot without losing track of the initial search string, e.g. it opened for exploration of the direct impact on the forest-plot of deleting an article or of limiting the analysis by age or gender.

CONCLUSIONS: We optimized all 4 phases of a meta-analysis. The time reduced on developing meta-analysis will ensure better decision making and prioritization. However, the essential optimization is perhaps the connectivity with the search engines, changing the work with meta-analysis from a static picture of effect and side-effect to constantly updated evidence.