Multivariable Logistic Regression
To adjust for confounders a multivariable logistic regression analysis was used. The model estimated the likelihood and statistical significance of a discordant referral as a function of the cervical cytology while controlling for clinic type, race, age group, and time between policy and referral to the VCU Department of Obstetrics and Gynecology. We used the c-statistic to evaluate the strength of the model’s ability in predicting similar outcomes to those observed in the study population.