Skip to main content
. 2020 Feb 10;15(2):e0228570. doi: 10.1371/journal.pone.0228570

Table 4. Multivariable multinomial logistic regression of factors associated with diabetes and prediabetes (N = 168).

Characteristics RRR (95% Confidence Interval)a
Diabetes P value Prediabetes P value
Number (%) 33 (19.64) 17 (10.12)
Age (n = 168) 1.03 (0.98, 1.08) 0.301 1.06 (1.01, 1.13) 0.027
Waist circumference (inch) 1.07 (0.95, 1.21) 0.264 1.17 (1.01, 1.34) 0.031
Abdominal obesity
  No 1 1
  Yes 8.29 (1.46, 47.04) 0.017 0.86 (0.23, 3.25) 0.822
  Family history of diabetes
  No 1 1
  Yes 5.48 (1.70, 17.64) 0.004 1.61 (0.50, 5.19) 0.425
Hypertension status
  No 1 1
  Yes 9.66 (2.84, 32.79) <0.001 0.83 (0.20, 3.46) 0.804
Self-perceived on poor diet
  No 1 1
  Yes 3.00 (0.91, 9.87) 0.070 4.69 (1.14, 19.19) 0.032

aRRR: Relative risk ratio.

Linearity of age and waist circumference were checked and reported to be linear.

No interaction was found.

Overall fitness reported to be Hosmer-Lemeshow test (logit function 1, p = 0.897; logit function 2: p = 0.134); overall correctly classified percentage ((logit function 1, 88.0%; logit function 2: 82.6%); area under the ROC curve (logit function 1, 75.8%; logit function 2: 79.9%).

Regression diagnostic for individual fitness was performed, and no removal of individual case.

The percent changes in regression coefficient less than 20%.