Slow convergence: Career impediments to interdisciplinary biomedical research

Interdisciplinary research is seen as critical for solving society's grand challenges but it is not clear that the outcomes of interdisciplinary researchers reflect that urgency. These code combine data (available from other sources) to analyze career outcomes of interdisciplinarians. Specifically, they use data from three major sources of data: publication data from the Author-ity disambiguation of PubMed (available for free from: https://databank.illinois.edu/datasets/IDB-2273402); measures of interdisciplinarity of publications generated from the Web of Science, and data on dissertations from ProQuest's Dissertations and Theses database (available for cost from https://clarivate.com/webofsciencegroup/solutions/web-of-science/). The Author-ity data are used to identify unique authors. These are complemented with Authorlink, which links the publication clusters in Author-ity to ProQuest, to determine the dissertation field and year of the highest degree for researchers. ProQuest data is used first to identify individuals who obtained a PhD degree and whose dissertation main subject is related to a biomedical field to ensure that their papers are more likely to be covered by PubMed.

Once linked to a dissertation, PubMed IDs are used to obtain bibliographic data on PubMed publications from the Clarivate Analytics Web of Science (WoS). For each publication WoS data are used to calculate the interdisciplinarity score and track the number of citations received. Finally, the linked PubMed/ProQuest/WoS dataset is used to select the sample. The sample includes researchers meeting three criteria: (a) they published at least one paper in the "dissertation stage" of their career, defined as the period between 5 years prior to and 1 year after obtaining their degree (to allow for a 1-year publication lag); (b) their first and last degrees in Authorlink are no more than 8 years apart (to reduce name disambiguation errors); and (c) at least one of their papers have an interdisciplinarity score. (An eight-year window between multiple degrees that appear in the ProQuest database was chosen to allow for a 5 or 6 year PhD process plus 2 or 3 years between the previous dissertation and starting a new degree.) Finally, career year (or career age) for researchers was defined as the number of years elapsed since graduation, and career length as the career year at the time of their last publication.

A publication is considered to be an author's last and their publishing career to have ended if there has been no publishing activity for at least five years. This also means that completed career lengths can only identified for researchers who stopped being active prior to 2014. In the analysis all the publications in the "dissertation stage" are bundled identified as being published at career age 0. The final dataset includes 154,021 researchers who received a PhD in a biomedical field between 1970 and 2013, and 2,612,553 unique papers between 1970 and 2018. We use papers five years after the last graduation in our sample, to measure both citation accumulation and the gap in publishing that marks the end of a researcher's career.

Interdisciplinarity is measured using is an article-level classification based on the disciplinary composition of references using the WoS journal-based classification as a starting point and iterative refinement. To measure interdisciplinarity that crosses boundaries among broad research areas, roughly 250 WoS subject categories were grouped into 14 broad disciplines. Papers were assigned to a single category based on the discipline (one of the 14 broad research areas) that was most frequently found among the references. Interdisciplinarity for an article was defined as the share the fraction of references that does not belong to the primary discipline, which can take a value 0≤I<1 (i.e., between 0 and 100%). Article-level interdisciplinarity scores were used to calculate interdisciplinarity scores for the researchers. For each researcher, initial interdisciplinarity was calculated as the median interdisciplinarity score of all the papers they published in their "dissertation stage," as defined above. We relate the interdisciplinarity of advisees to that of advisors. In this analysis advisee publications with their advisors are excluded. Interdisciplinarity of researchers after the dissertation stage was calculated for each year by taking the mean of interdisciplinarity scores of publications they published in each year.

Code

• Readme
• Python Code
• Stata Code
• All code in single file (zip)