Background and Objective: Prognostic and diagnostic risk or prediction models are frequently encountered in the medical literature. Various recommendations exist of good clinical prediction research. We investigated the methodological quality of studies on the development, validation or implementation of prediction models. Methods: We searched PUBMED using a validated search strategy to select prediction studies in six general journals (Annals of Internal Medicine, British Medical Journal, Lancet, New England Journal of Medicine, PLos Medicine, and Journal of the American Medical Association). Studies were included based on pre-defined inclusion criteria. We used an exhaustive item list to score the quality of the papers, based on recent recommendations for multivariable prediction research. Two reviewers independently scored the studies, and a third in case of doubt. Results: The search strategy revealed 347 hits. At the time of this abstract, we finished three journals and retrieved 29 papers for full text review. 26 of these described 59 developed models. Only three (10%) studies involved a validation (of a previous developed model) or the quantification of a model’s impact on patient outcome. Of the 26 studies, five had both an etiological and prognostic aim. Even though they may require a different design and analyses, two (40%) gave information of the missing value per predictor, and none reported on the number of patients with missing values; all etiologic-prognostic studies reported on loss-to-follow-up. The predictive accuracy of the prediction models was reported in 13 of the 26 studies. Only 6 of the 26 studies used any form of internal or external validation and in 26 (100%) the number of events per studied predictor was not mentioned or to low for ≥1 of the presented model(s); both aspects commonly lead to overfitted models. Conclusion: Despite various recent recommendations for conducting studies on diagnostic and prognostic prediction modeling, the vast majority does not follow these guidelines.