Skip to main content

Table 4 Predictive power of models with established health utilization predictors without and with SNPs: results of the 10-fold cross validation

From: Association of PNPLA3 rs738409 and TM6SF2 rs58542926 with health services utilization in a population-based study

Statistical evaluation criteria

Annual outpatient visits

Hospitalization

Annual inpatient days

Median (Nagelkerke R2)

   

 Model 2

22.62

1.61

0.0175

 Model 2 + PNPLA3 rs738409

22.96

1.74

0.0138

 Model 2 + TM6SF2 rs58542926

20.33

1.56

0.0218

 Model 2 + PNPLA3 rs738409 + TM6SF2 rs58542926

22.73

1.99

0.0268

Median (root mean squared error)

 

 

 Model 2

8.70

 

8.111

 Model 2 + PNPLA3 rs738409

8.78

 

7.121

 Model 2 + TM6SF2 rs58542926

8.69

 

7.422

 Model 2 + PNPLA3 rs738409 + TM6SF2 rs58542926

8.68

 

7.174

Median (absolute prediction error)

 

 

 Model 2

5.89

 

3.605

 Model 2 + PNPLA3 rs738409

5.97

 

3.548

 Model 2 + TM6SF2 rs58542926

6.01

 

3.565

 Model 2 + PNPLA3 rs738409 + TM6SF2 rs58542926

6.07

 

3.594

Area under ROC curve (p-value for comparison to model 2)

 

 Model 2

 

0.605

 

 Model 2 + PNPLA3 rs738409

 

0.605 (0.846)

 

 Model 2 + TM6SF2 rs58542926

 

0.606 (0.868)

 

 Model 2 + PNPLA3 rs738409 + TM6SF2 rs58542926

 

0.606 (0.965)

 
  1. Model 2: age, sex, school education, income, alcohol intake, smoking status, eGFR, Functional Comorbidity Index
  2. ROC receiver operating characteristic