Table 4. Appropriateness of statistical methods for predictors of mortality.

	2002–2006		2007–2011
	Prospective (N = 68)	Retrospective (N = 23)	Prospective (N = 62)	Retrospective (N = 36)
Statistical methods	No. (%)	No. (%)	No. (%)	No. (%)
Appropriate statistical analysis
Cox regression	53 (78)	15 (65)	48 (77)	15 (42)
Time dependent Cox regression	16 (24)	5 (22)	10 (16)	1 (3)
Cox regression analysis with frailty	1 (1)	0 (0)	3 (5)	3 (8)
Poisson regression	3 (4)	0 (0)	4 (6)	0 (0)
Sub-Optimal statistical analysis
Logistic regression	4 (6)	4 (17)	7 (11)	17 (47)
Unclear	1 (1)	2 (9)	2 (3)	3 (8)

Note: Totals in this table do not add up to the number of articles because some articles used more than one method in their analysis.