Sample size requirements are not being considered in studies developing prediction models for binary outcomes: a systematic review

Table 3 Summary of sample size available and used, and the number of candidate predictors and minimum number of candidate predictor parameters for the development of the prediction model

	Number of studies (%)*	Median [p25-p75]	Range
Total available data**	118 (99%)	669 [283–3013]	37 to 345,718
Total available events**	112 (94%)	137 [55–411]	8 to 48,262
Total used data***	119 (100%)	603 [220–2236]	37 to 242,003
Total used events***	97 (82%)	100 [44–243]	8 to 33,784
Total number of candidate predictors	115 (97%)	19 [13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28]	5 to 838
Minimum number of candidate predictor parameters****	110 (92%)	23 [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31]	7 to 179

*number of studies with sample size information available
**total available refers to the maximum available sample size before any potential data splitting or handling of missing data
***total used refers to the sample size actually used to develop the model, after any potential data splitting or handling of missing data
***minimum number of candidate predictor parameters were calculated by the research team using the reported candidate predictors. This is the minimum number as it does not account for any additional parameters that may be introduced through handling of nonlinear predictors and additional categorisation

ISSN: 1471-2288