Table II.

Multiple logistic regression model for predictors of positive culture

Model outcome and predictor variables∗	Bootstrapped parameter estimates
	Mean estimate	SD	95% CI for estimate	Mean OR	95% CI for OR	Standardized estimate	Mean AUC
Positive culture (positive vs. negative)							0.6468
Intercept	–0.9999	0.4565	–1.9938, –0.1910			0
Sex (male vs. female)	0.9239	0.5324	–0.0369, 2.0254	2.5191	0.9637, 7.5791	0.2098
History of ipsilateral shoulder infection (yes vs. no)	3.0423	0.6175	1.7135, 4.2081	20.9533	5.5483, 67.2287	0.4233

SD, standard deviation; OR, odds ratio; AUC, area under the curve.

The parameter estimates were based on 10,000 bootstrap samples of the logistic regression model, with penalized maximum likelihood estimation along with Firth's bias correction. The mean and variance were estimated on the logarithmic scale and represent log odds; 95% CI for the mean parameter estimate. For the 95% CI for estimate that does not contain zero (0), the respective mean parameter estimate is statistically significant at alpha = 0.05. Observed sample, N = 83.

^∗Predictor variables for the model were selected from a pool of 10 potential predictor variables via SAS's PROC HPREDUCE to perform supervised variable selection and then implemented in the context of a penalized logistic regression model that was based on 10,000 bootstrap samples.