Table 2.

The Relationship Between Unhealthy Days and Risk Behaviors: a Multilevel Ordinal Regression Model20 (n = 3738)

ML EstimateStandard ErrortTest ScorePAdjusted Odds Ratio (95% CI)
Threshold 2*1.79190.042342.37<.0001
Intercept0.61370.12454.93<.00011.00 (1.00)
Patient risk behaviors
    Smoking0.41480.07595.47<.00011.51 (1.30, 1.76)
    Diet Score0.09660.01287.56<.00011.10 (1.07, 1.13)
    Exercise MET minutes-0.03750.0102-3.70.00050.96 (0.94, 0.98)
Patient demographics
    Age-0.01050.0022-4.83<.00010.99 (0.99, 0.99)
    Female0.33670.07414.55<.00011.40 (1.21, 1.62)
    African-American§-0.13150.0916-1.44.15590.88 (0.73, 1.05)
    Hispanic§-0.23180.1084-2.14.03630.79 (0.64, 0.98)
    Low income0.58670.07597.73<.00011.80 (1.55, 2.09)
Practice characteristics
    Registry-0.20450.1142-1.79.07820.82 (0.65, 1.02)
    University owned0.43890.14912.94.00451.55 (1.16, 2.08)
    Hospital owned-0.02140.1056-0.20.83970.98 (0.80, 1.20)
    Other0.48490.16972.86.00581.62 (1.16, 2.26)
  • Bolded values signify P < 0.05.

  • * With intercept plus coefficients, model estimates log odds (ML estimate) of having 0 unhealthy days as opposed to 1–30 unhealthy days. The addition of threshold 2 to the log odds estimates log odds of having 0–13 unhealthy days compared to 14–30 unhealthy days.

  • Ranging 0–14, where 0 is best and 14 is worst diet.

  • Centered at the mean of the sample.

  • § Compared with non-Hispanic whites.

  • Compared with private clinician-owned.