Reporting quality and risk of bias in systematic reviews of clinical prediction models: a cross-sectional overview of 1004 studies.

This review aims to assess the condition and primary factors influencing reporting quality and bias in systematic reviews of clinical prediction models. We conducted a critical review through three databases (PubMed, Embase, Cochrane Library) from inception to January 7, 2023. Screening, data extraction, and assessment were conducted independently by pairs. Employing TRIPOD-SRMA and ROBIS to evaluate the reporting and methodological quality. Data were summarized through descriptive statistics, visual analysis, and multiple linear regression analysis. 1004 reviews were included. Reporting quality was poor. The average main text quality score was 16.4 ± 4.1 out of 33 points, and the average abstract quality score was 5.2 ± 1.7 out of 12 points. Year, economic level, purpose of the model, impact factors, and types of systematic reviews were associated with the quality of abstracts or the main text. Merely 0.2% (2/1004) of the reviews assessed were classified as having low risk of bias. Systematic reviews of clinical prediction models demonstrate inadequate reporting and significant risk of bias. Optimizing research design, enforcing stricter reporting guidelines, promoting protocol pre-registration, and enhancing journal review processes to boost reliability and transparency are warranted.