Investigation and Evaluation of Systematic Reviews of Prediction Models Published in Chinese Journals: Methodological and Reporting Quality

Objective To analyze the methodological and reporting quality of systematic reviews of prediction models published in Chinese journals, with the aim of providing references for enhancing the quality of Chinese systematic reviews of prediction models. Methods Chinese systematic reviews of prediction models were electronically searched in CNKI, WanFang Data, CBM, and VIP databases from inception to July 20, 2023. Two independent reviewers screened literature, extracted data, and used the AMSTAR and PRISMA 2020 tools to assess methodological and reporting quality of the included reviews. Results A total of 55 systematic reviews published between 2015 and 2023 were included, with 12 of them being meta-analyses. These reviews encompassed a range of topics, with a primary focus on cardiovascular diseases, stroke, and diabetes. The identified systematic reviews exhibited obvious deficiencies in some areas, including items 1, 4, 5, 6, and 10 of AMSTAR, as well as items 7, 10a, 12, 13a-f, 14, 15, 16a-b, 17, 20b-d, 21, 22, 23d, 24a-c, 25 and 26 of PRISMA 2020. Furthermore, a moderate positive correlation (r = 0.58, P < 0.001) was observed between the methodological and reporting quality. Multiple linear regression analyses revealed: greater number of pages, more recent publications, and funding support were associated with higher methodological quality (P < 0.05). Similarly, greater number of pages, more recent publications, qualitative systematic reviews, and funding support were associated with higher reporting quality, but the number of authors showed a negative association (P < 0.05). Conclusions The current systematic reviews of prediction models published in Chinese journals require enhancement in both methodological and reporting quality.