Investigation and Evaluation of Systematic Reviews of Prediction Models Published in Chinese Journals: Methodological and Reporting Quality
WANG Ziyi,LU Cuncun,HUANG Jiayi,ZHANG Jinglei,SHANG Wenru,CUI Lu,LIU Wendi,DENG Xiuxiu,ZHAO Xiaoxiao,YANG Kehu,LI Xiuxia
DOI: https://doi.org/10.12290/xhyxzz.2023-0418
2023-01-01
Abstract:ObjectiveTo analyze the methodological and reporting quality of systematic reviews of prediction models published in Chinese journals, with the aim of providing reference for enhancing the overall quality of Chinese systematic reviews of prediction models.MethodsWe searched the CNKI, WanFang Data, CBM, and VIP databases for Chinese systematic reviews of prediction models from inception to July 20, 2023. After two independent reviewers screened literature and extracted data, the AMSTAR(A Measurement Tool to Assess Systematic Reviews) and PRISMA 2020(Preferred Reporting Items for Systematic reviews and Meta-Analyses 2020) tools were used to assess the methodological and reporting quality of the included reviews.ResultsA total of 55 systematic reviews published between 2015 and 2023 were included, 12 of which were meta-analysis. The reviews covered various topics, mainly including cardiovascular diseases, stroke, and diabetes. The identified systematic reviews exhibited obvious deficiencies: items 1, 4, 5, 6, and 10 of AMSTAR showed poor methodological quality, and items 7, 10a, 12, 13a-f, 14, 15, 16a-b, 17, 20b-d, 21, 22, 23d, 24a-c, 25 and 26 of PRISMA 2020 needed improvement in reporting quality. Furthermore, a moderate positive correlation (r=0.58, P < 0.001) was observed between the methodological and reporting quality. Multiple linear regression analysis revealed that a greater number of pages, more recent publications, and funding support were associated with higher methodological quality (P < 0.05). Similarly, a 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).ConclusionThe methodological and reporting quality of existing systematic reviews of prediction models published in Chinese journals is relatively poor and demands improvement.