Open Access and Citation Impact in STEM Research: OpenAlex Bibliometric Dataset, 2016–2025

The population of each field × access-status stratum was first determined, followed by proportional stratified sampling with a maximum target of 500,000 articles per year. The sample allocated to each stratum was proportional to its representation in the eligible STEM literature for that year. Where fewer than 500,000 eligible articles were available, the available population determined the sample size.

Duplicate records were removed using unique OpenAlex work identifiers. The final dataset contained 4,984,351 articles. Retrieved information included publication year, primary OpenAlex classification, field, domain, OA status, and total citation count.

For the primary analysis, publications were grouped as Open Access or Closed Access. Citation impact was normalized to account for differences in citation patterns across fields and publication years. For each article, field- and year-normalized citation impact (NCI) was calculated as: NCI_i = C_i / Mean(C_field,year) where C_i is the article's total citation count and Mean(C_field,year) is the mean citation count of sampled articles from the same field and publication year.

An NCI of 1.0 therefore represents the average citation impact for that field and year. For each publication year, mean NCI was calculated separately for OA and closed-access articles. The annual Open Access

Citation

Advantage was calculated as: OA

Citation

Advantage (%) = [(Mean NCI_OA / Mean NCI_Closed) − 1] × 100 Positive values indicate higher normalized citation impact for OA articles, while negative values indicate higher impact for closed-access articles. Bootstrap resampling was used to estimate 95% confidence intervals. Mean total citations were also calculated separately for OA and closed-access articles by publication year.