Method of predictor selection, stratified by whether prediction was the primary or secondary study aim.

<p>Numbers are column percentages, with absolute numbers in parentheses. Impact and external validation studies (<i>n</i> = 6) were excluded from this table as these issues are not applicable for these type of studies. Hence, <i>n</i> = 65.</p>a<p>More than one method may be used within a study; percentages do not add up to 100%.</p>b<p>Percentage (number) of studies that reported the applied method for selecting which predictors were included in the multivariable analyses, if it was not based on statistical analysis (i.e., univariable predictor–outcome associations).</p>c<p>Predictor inclusion in multivariable model was pre-specified, as the specific aim was to quantify the added value of a new predictor to existing predictors.</p>d<p>For example, systolic and diasystolic blood pressure combined to mean blood pressure.</p>e<p>For the items below, two percentages are given. The first percentage includes all studies (i.e., 48 predictor finding studies, 17 model development studies, or 65 total); the second is the percentage of all studies that applied some type of predictor selection in the multivariable analysis (35 predictor finding studies, 11 model development studies, and 46 total; the excluded studies did not apply any predictor selection in the multivariable analysis but simply pre-specified the final model).