Skip to main content

Table 3 Comparison of C-index and net reclassification index (NRI) across models. Models are predicting cardiovascular disease mortality in the testing dataset of the National Health and Nutrition Examination Survey

From: Use of biomarkers of metals to improve prediction performance of cardiovascular disease mortality

Model

C-index

Continuous NRI (95% CI)

# of Predictorsa

Cox Proportional Hazards Model

 Model 1: Traditional Predictorsb

0.845

 

10

 Model 2: + Blood Metalsc

0.847

0.23 (0.09, 0.38)

13

 Model 3: + Urinary Metalsd

0.843

0.21 (0.07, 0.35)

28

 Model 4: + Metals quadratic/interaction termse

0.773

0.35 (0.21, 0.48)

177

Elastic-Net

 Model 5: Traditional Predictorsb

0.828

 

10

 Model 6: + Blood Metalsc

0.830

0.66 (0.52, 0.80)

12

 Model 7: + Urinary Metalsd

0.830

0.51 (0.37, 0.64)

14

 Model 8: + Metals quadratic/interaction termse

0.830

0.76 (0.62, 0.90)

13

Survival Random Forest

 Model 9: Traditional Predictorsb

0.833

 

10

 Model 10: + Blood Metalsc

0.826

-0.18 (-0.32, -0.04)

13

 Model 11: + Urinary Metalsd

0.826

-0.49 (-0.63, -0.35)

28

  1. a Number of Predictors: refers to the total number of variables included in the model
  2. b Traditional predictors include age, race/ethnicity, smoking status, systolic blood pressure, total cholesterol, high density lipoprotein cholesterol, body mass index, hypertension status, and diabetes status
  3. c Blood metals include lead, mercury, and cadmium
  4. d Urinary metals include cesium, molybdenum, thallium, cobalt, barium, lead, cadmium, uranium, tungsten, antimony, mercury, arsenic, dimethylarsinic acid, and arsenobetaine
  5. e Metals quadratic/interaction terms indicate all possible quadratic and pairwise interactions between metals listed above