Interdisciplinary research attracts greater attention from policy documents: evidence from COVID-19

The paper investigates whether interdisciplinary research receives greater attention from policy documents than unidisciplinary work. Motivated by the growing emphasis on demonstrating both scientific and societal impacts of research, and by policy and funding bodies’ encouragement of interdisciplinarity, the study focuses on the previously underexplored link between interdisciplinarity and policy attention. Interdisciplinarity is framed relative to multidisciplinarity and transdisciplinarity, emphasizing integration of knowledge across disciplines to address complex problems. The authors argue that attention from policy documents signals potential influence on policy formulation and implementation, reflecting the translation of scientific knowledge into societal action. Using COVID-19 publications—an area marked by rapid research growth and extensive policy citing—the study examines how varying degrees and dimensions of interdisciplinarity relate to policy citations, aiming to fill a gap in the literature and inform researchers and policymakers about pathways to policy impact.

The related work discusses quantitative measures of interdisciplinarity and diversity, breaking diversity into three components—variety (number of categories), balance (evenness across categories), and disparity (cognitive distance among categories). Prior indicators integrating these dimensions include Rao–Stirling and DIV, while the authors choose to analyze both single components and composite metrics to minimize information loss. Studies on interdisciplinarity and scientific attention (often citation-based) have reported mixed findings, potentially due to different measurement approaches and disciplinary contexts. Research on societal attention has explored links between interdisciplinarity and public engagement or local problem-orientation. However, a clear gap remains regarding whether interdisciplinary research garners more attention from policy documents. The authors propose measuring policy attention analogously to scientific attention by using citations from policy documents to scientific publications as indicators.

Data sources: The study draws on OpenAlex (successor to Microsoft Academic Graph) for publications and metadata, and Overton for policy document citations. COVID-19 topic scope is defined for 2020-01-01 to 2021-12-31. Overton retrieval (topic “covid-*”) yields 32,379 DOIs and associated policy citation counts; 31,105 of these match records in OpenAlex. COVID-19-related publications from OpenAlex are combined with Overton-matched records, producing 489,361 unique publications. Filtering to journal articles with recorded references results in 175,950 records; further exclusions (e.g., reference count < 3, missing variables) yield a regression dataset of 159,957 publications. Interdisciplinarity measures: Based on references’ field categories (OpenAlex concepts), the authors build a 292×292 citation matrix M across fields and compute disparity using cosine-based distances between field vectors. They operationalize three single-dimension indicators: variety (breadth of reference categories), balance (Gini-based evenness across categories), and disparity (cognitive distance). They also compute composite indicators Rao–Stirling (RS) and DIV. Dependent variable: policy_cited is a binary variable indicating whether a publication is cited by at least one policy document. Controls: team_size (number of authors), scientific_citations (citations from scientific literature), references_count, and journal_impact_factor (year-specific journal impact factor). Field and time fixed effects are included (discipline and month identifiers within 2020–2021). A linear probability model is used for estimation. Coarsened Exact Matching (CEM): For robustness, publications are classified as interdisciplinary (Inter=1) or unidisciplinary (Inter=0) using two thresholds: (1) median-based—Inter=1 if both RS and DIV exceed their medians (yielding 64,608 interdisciplinary and 93,450 unidisciplinary records); (2) quartile-based—Inter=1 if both RS and DIV are in the top 25% and Inter=0 if both are in the bottom 25% (27,670 interdisciplinary and 29,569 unidisciplinary). CEM is performed in Stata 17, matching on team_size, publication month, and Level 0 field. Post-matching regressions replicate the main models to assess robustness.

Descriptive context: In the regression dataset (n=159,957), mean policy_cited=0.142 (SD=0.349). Mean values: variety=0.653, disparity=0.520, balance=0.015, RS=0.776, DIV=0.033. Publications and policy attention are unevenly distributed across fields; medicine dominates counts, while engineering/technology and humanities/arts receive fewer policy citations on average. Binning analyses: Across fields, average policy citations generally rise with higher interdisciplinarity (RS and DIV), except natural sciences exhibit a decrease-then-increase pattern. Main regressions (Table 4): - Single-dimension model: variety positively associated with policy citation (coef≈1.065, t≈17.0); disparity positive (coef≈0.041, t≈5.7); balance negative (coef≈−0.504, t≈−8.2). - Composite indicators: RS positive (coef≈0.082, t≈13.1); DIV positive (coef≈2.000, t≈18.7). Controls: scientific_citations positive; team_size positive; references_count mixed small effects; journal_impact_factor small negative. Models include field and time fixed effects; R² ≈ 0.089–0.091. Robustness via CEM (Table 5): Results remain directionally consistent. Median-threshold CEM regressions show positive effects for variety (≈0.856, t≈14.6), disparity (≈0.040, t≈5.6), balance (≈0.023, t≈3.6), RS (≈0.096, t≈15.6), and DIV (reported as significant). Quartile-threshold CEM regressions similarly show positive, significant coefficients for RS (≈0.090) and DIV (≈1.734). Observations after matching remain large (e.g., ≈156–159k), with modest R² (≈0.029–0.037), indicating robust associations. Field-specific regressions (visualized in Fig. 3): Variety consistently promotes policy citation across all fields. Disparity effects vary by field (positive, negative, or non-significant). Balance is positive in engineering and technology, negative in natural sciences and humanities/arts, and non-significant elsewhere. RS and DIV are generally positive except in natural sciences and humanities/arts where effects are weaker or non-significant. Overall: Interdisciplinarity is positively correlated with the likelihood of being cited by policy documents across most fields, with variety showing the strongest and most consistent positive association.

The findings directly address the research question, showing that more interdisciplinary COVID-19 publications are more likely to be cited in policy documents. This suggests that interdisciplinarity enhances the translation of scientific knowledge into policy-relevant outputs. The consistent positive associations for composite measures (RS, DIV) and the especially strong and universal effect of variety imply that spanning more disciplinary categories in references is salient to policy audiences. Field-level heterogeneity indicates that the role of balance and disparity depends on disciplinary norms and policy needs, while the generally positive results across robustness checks (CEM) strengthen causal plausibility by mitigating confounding. These insights are relevant to researchers seeking policy impact and to policymakers aiming to identify research with actionable, integrative knowledge.

This study fills a gap by systematically linking interdisciplinarity to attention from policy documents, using a large COVID-19 dataset from OpenAlex and Overton. It shows that interdisciplinary research—especially with higher variety—attracts more policy citations, and that composite measures (RS, DIV) are robustly positive predictors even after matching on observables. The work underscores interdisciplinarity’s role in facilitating knowledge translation into policy. Future research could generalize beyond COVID-19, refine outcome measures (e.g., using policy citation counts or weighting by policy influence), improve interdisciplinarity measurement, and incorporate additional controls (e.g., funding, institutional prestige, open access) to sharpen causal inference.

- Topic specificity: The focus on COVID-19 limits generalizability to other domains and time periods. - Outcome granularity: The binary dependent variable (any policy citation) is coarse; using counts or weighted policy impacts could provide finer insight. - Measurement uncertainty: Interdisciplinarity lacks a universally accepted metric; RS and DIV are common but not definitive. - Controls and data constraints: Potential omitted variables (e.g., funding status, open access, institutional reputation) were not included due to data limitations. - Field/time classification inconsistencies: Variations in field resolution and monthly bins may introduce measurement noise.