Skip to main content

Table 1 Distribution of predictor values and outcome, and predictor effects in multivariable logistic regression models

From: Prediction models for clustered data: comparison of a random intercept and standard regression model

 

Distribution (N = 1642)

Interquartile range per anesthesiologist *

Standard model Beta (95% CI) †

Random intercept model Beta (95% CI) †

Fixed effect

    

Female gender

882 (54%)

46%-63%

0.75 (0.53 – 0.97)

0.78 (0.54 – 1.01)

Age, years ‡

49 (16.5)

46-51

-0.008 (-0.014 – -0.001)

-0.009 (-0.016 – -0.003)

History of PONV or motion sickness

530 (32%)

30%-36%

0.41 (0.19 – 0.64)

0.42 (0.18 – 0.65)

Current smoking

510 (31%)

20%-33%

-0.43 (-0.66 – -0.19)

-0.45 (-0.69 – -0.20)

Abdominal or middle ear surgery

209 (13%)

4%-16%

0.62 (0.32 – 0.93)

0.57 (0.24 – 0.89)

Volatile anesthetics

844 (51%)

33%-74%

-0.03 (-0.24 – 0.18)

0.19 (-0.09 – 0.46)

Intercept

616 (38%) §

29%-47% §

-0.65 (-1.05 – -0.25)

-0.68 (-1.16 – -0.19)

Random effect

    

Intercept variance ††

-

-

-

0.15 (0.07 – 0.33)

  1. * interquartile ranges of the predictor distributions per anesthesiologist.
  2. † logistic regression coefficients with 95 percent confidence intervals. Intervals of the random intercept model were based on the t-distribution.
  3. ‡ mean (standard deviation).
  4. § number (percentage) of patients with PONV, rather than the intercept is reported.
  5. PONV = post operative nausea and vomiting.
  6. †† non-parametric confidence interval of the random intercept variance was obtained from a cluster bootstrap procedure.