OBJECTIVES: Generating scientific evidence that can inform healthcare decisions and improve health outcomes is a top priority for medical affairs teams. To maximise the value of the evidence, it is crucial that it is disseminated to the right audience and foster engagement. This study aims to predict the engagement of publications in both the scientific and social media communities. It also seeks to identify the drivers behind this engagement, to inform medical affairs teams on effective communication strategies.

METHODS: We defined scores to measure publication impact in social media and scientific community. Social media engagement was quantified using the number of tweets, likes, and followers. The impact on the scientific community was defined as the number of citations, weighted by the Scimago Journal Impact Factor (SJR). We retrieved oncology publications across multiple sources, leveraging public and IQVIA proprietary data. We generated features capturing study characteristics, scientific evidence and communication strategies used to disseminate findings.

RESULTS: The models yielded good performance, encompassing the reference baseline prediction. SJR was the most influential predictor, where higher values correlated with greater engagement. In social media, the journal’s number of twitter followers and average citations per document were other important drivers. Average citations were also the second most important driver for impact in the scientific community, followed by the study sample size.

CONCLUSIONS: Machine learning methods can be leveraged to predict engagement of studies in social media and scientific community, providing insights into what drives engagement. Our findings align with expectations that higher impact factors journals lead to higher engagement, supporting the fundamental validity of our model. The approach can help to validate the impact of publication decisions and expanded to include additional social media platforms.