A Coverage Break in Open Bibliometric Data: How Records Outside Crossref and DataCite Disappeared from OpenAlex After 2021, and Why National Indicators for Iran Are Affected

Purpose. OpenAlex is increasingly used for national and institutional science indicators, but its coverage is not constant over time. We test whether coverage of scholarly works changed abruptly around 2022, for which languages and countries, and whether the pattern is related to how works are identified (DOI presence and DOI registration agency). Iran is examined in detail because its scholarly infrastructure relies on a registration agency outside Crossref. Design and methods. We queried the OpenAlex API (retrieved 1–2 October 2026) for yearly counts of works by language (15 languages), by author country (18 countries) and by DOI presence (2015–2025), sampled 3,000 DOI-bearing works each for Persian, Korean, Japanese and Turkish (2019–2021) and identified the DOI registration agency of the observed prefixes through the DOI resolver, and counted works for 32 prefixes by year (2019–2024). Findings. Works in Persian fell from 44,935 in 2020 to 3,342 in 2022 (ratio 0.07), Korean from 91,147 to 2,395 (0.03) and Japanese by 72% (0.28), whereas Arabic (1.01), English (0.95), Indonesian (0.97) and Portuguese (0.86) were stable. Works without a DOI fell in 17 of 18 countries (median ratio 2022/2021 about 0.2; Iran 0.07, Indonesia 0.04), while DOI-bearing works did not (median 1.07), with Iran (0.80), Russia (0.89) and Brazil (0.93) the exceptions. In the Persian DOI sample, 90.5% of DOIs were registered with mEDRA (5.1% Crossref); eight mEDRA prefixes lost a median 93% of works between 2020 and 2023, three KISTI prefixes 100%, one JaLC prefix 99% and one ISTIC prefix 92%, against a median 10% for 14 Crossref prefixes. Iran's all-field output fell 29.5% from 2021 to 2022 (−38,860 works) when Türkiye, Egypt, Pakistan and Saudi Arabia grew by 7–16%, and its share of world output fell from 12.3 to 9.1 per mille; Persian-language works account for two thirds of the fall. Originality. The study documents and characterises a data-infrastructure discontinuity that is invisible in annual totals unless disaggregated, links it to DOI registration agencies, and provides a simple diagnostic and recommendations for bibliometric studies of countries whose literature is registered outside Crossref.