External validation of multivariable prediction models: a systematic review of methodological conduct and reporting

Table 3 Handling of missing data ^‡

	Single development & validation articles^§(n = 33)		Separate development & validation articles
	Development cohort	Validation cohort	Development paper	Validation paper^**
	Development cohort	Validation cohort	(n = 66)	(n = 45)
Studies with no mention of missing data	13 (39)	21 (64)	30 (45)	21 (47)
Studies reporting number of participants with missing data	Information not extracted	5 (15)	Information not extracted	18 (40)
Studies reporting number of missing values for each predictor	Information not extracted	3 (9)	Information not extracted	5 (11)
Studies carrying out complete-case analysis^††	26 (79)	30 (91)	43 (65)	20 (44)
Studies explicitly mentioning carrying out multiple imputation	Information not extracted	2 (6)	Information not extracted	7 (16)

^‡Percentages are in given parentheses.
^§Articles that developed a new model and also evaluated the performance on a separate dataset.
^**Articles that only described the evaluation of a previously published prediction model.
^††In the absence of clear reporting, those studies that did not mention how missing data were handled were assumed to have conducted a complete-case analysis.

ISSN: 1471-2288