Skip to main content
. 2009 Mar 15;6(2):A48.

Table 3.

Logistic Regression Models for Factors Associated With Self-Reported Physician-Diagnosed Prediabetes in Adults Participating in Diabetes Awareness Program, New York State, 2006

Variable Model With Demographic Controlsa Best Fit Modelb

OR (95% CI) P value OR (95% CI) P value
Family history of diabetes
Yes 3.56 (2.61-4.85) <.001 3.65 (2.70-4.92) <.001
No 1 [Reference] 1 [Reference]
Body mass index, kg/m2
≥25 2.78 (1.94-3.97) <.001 2.79 (1.96-3.97) <.001
<25 1 [Reference] 1 [Reference]
Age, y
≥45 3.14 (2.12-4.64) <.001 2.77 (1.91-4.00) < .001
<45 1 [Reference] 1 [Reference]
Health care coverage
Yes 2.32 (1.28-4.22) .006 2.38 (1.35-4.18) .003
No 1 [Reference]  1 [Reference]
Racec
Non-Hispanic white 0.96 (0.96-1.77) .90
Non-Hispanic black 0.44 (0.20-0.99) .048
Hispanic 0.55 (0.22-1.34) .19
American Indian 3.63 (1.63-9.09) .001
All other 1 [Reference]
Sexc
Male 1.22 (0.86-1.73) .66
Female 1 [Reference]
a

The Hosmer-Lemeshow goodness-of-fit test was used to assess overall fit (χ28 = 10.9, P = .21).

b

The Hosmer-Lemeshow goodness-of-fit test was used to assess overall fit (χ27 = 2.2, P = .95).

c

Race and sex were removed in the best fit model through the backward stepwise selection method. These variables were found to be nonsignificant based on the probability of a likeihood-ratio statistic.