Skip to main content

Table 3 Documentation of multivariable regression, calibration, discrimination and reclassification

From: A methodology review on the incremental prognostic value of computed tomography biomarkers in addition to Framingham risk score in predicting cardiovascular disease: the use of association, discrimination and reclassification

Part 1. Documentation of multivariable regression (n = 35)

No.

(%)

a. Information on whether additional predictor is significant at <.05 level

24

68.6

b. Results of a test that penalises for the inclusion of additional predictor

8

22.9

Appropriate documentation (1a or 1b)

26

74.3

Part 2. Documentation of AUC in ROC analysis (n = 33)

  

a. Described method used to compare ROC curves

13

39.4

b. Presented the AUC values with and without the additional predictor

31

93.9

c. Presented CIs of AUC values with and without additional predictor

9

27.3

d. Presented P value for comparison

26

78.8

f. Availability or enable calculation of Δ AUC CIs

30

90.9

Appropriate documentation 1 (2a and 2b and [2c or 2d])

11

33.3

Appropriate documentation 2 (2a and 2b and [2c or {2d or 2f}])

12

36.4

Part 3. Documentation of calibration (n = 35)

  

Documentation of Hosmer-Lemeshaw test (n = 7) or Schoenfeld residuals (n = 1)

8

22.9

Part 4. Documentation of reclassification analysis (n = 35)

  

Report using table or text

  

Not reported

19

54.3

Partial

5

14.3

Complete

11

31.4

Standard of reporting of reclassification analysis (n = 16)

  

a. Use of standard categories of risk

11

68.8

b. Justified use of other categories of risk

15

93.8

c. Reported the number of patients changing categories

9

56.3

Appropriate documentation ([4a or 4b] and 4c)

9

56.3

Inadequate

7

43.8

Part 5. Documentation of NRI (n = 23)

  

Type of NRIs

  

Continuous/ category-free NRI

4

17.4

Categorical NRI

16

69.6

Reported both continuous & categorical NRIs

1

4.3

Reported relative NRI

1

4.3

Unclear

3

13.0

Standard of reporting of categorical NRI (n = 16)

  

a. Report censor handling

5

31.3

b. No extrapolation

7

43.8

c. Categorical NRI reference available

14

87.5

d. Justification of risk categories

14

87.5

e. Report NRI components

5

31.3

f. Availability of reclassification table showing event and non-event

8

50.0

g. Reclassification table enables the calculation of NRI components

7

43.8

h. Combined NRI reported as a sum not a percentage

8

50.0

i. The proportion of correctly reclassified subjects available

7

43.8

j. Reported NRI not used to construct strong summary

5

31.3

Adequate reporting of categorical NRI (> 5 items listed 5a–j)a

11

68.8

  1. aThe threshold is the median number of items reported in a skewed sample