Does poor methodological quality of prediction modeling studies translate to poor model performance? An illustration in traumatic brain injury

Table 3 The median AUC at development and external validation and the absolute and percentage change between development AUC and validation AUC stratified by risk of bias (RoB) of model development studies based on the original PROBAST

	N	Median AUC at development (N = 10) [IQR]	Median AUC at external validation (N = 245) [IQR]	Median delta AUC [IQR]	Median AUC change in percentage [IQR]
Low RoB	139	0.78 [0.77, 0.79]	0.80 [0.76, 0.84]	0.02 [− 0.01, 0.06]	8% [− 4, 21]
High RoB	45	0.86 [0.84, 0.86]	0.79 [0.69, 0.84]	− 0.06 [− 0.16, − 0.01]	− 18% [− 43, − 2]
Unclear RoB	61	0.83 [0.81, 0.86]	0.83 [0.77, 0.88]	0.00 [− 0.06, 0.04]	0.0% [− 19, 10]

ISSN: 2397-7523