Abstract B94: The effect of clinic on identifying predictors of HPV co-testing with Pap in routine cervical cancer screening

Abstract

Abstract Background: Human papillomavirus (HPV) co-testing with Pap is recommended for cervical cancer screening among women ≥ 30 years. However, little has been reported regarding actual provider use of HPV co-testing with Pap in clinical practice. Purpose: We sought to identify predictors of HPV co-testing with Pap among women ≥ 30 years by utilizing laboratory and administrative data from the Johns Hopkins Hospital Division of Cytopathology in Baltimore, MD. Methods: We analyzed 51,027 Pap specimen records and 4,148 HPV test records among 30,272 women ≥ 30 years in 88 clinics from February 2004 to December 2007. The outcome was HPV co-testing with Pap among women ≥ 30 years, and covariates included patient race, age, insurance type, ZIP code median household income, and calendar year. We constructed population-average logistic regression models. These included robust standard error estimation for clinic clustering and, at times, a fixed-effects estimator for within-clinic effects. We also stratified our analyses to five individual clinics. Results: The overall proportion of HPV co-testing with Pap was 8.1%. Notably, 12.8% of Pap specimens of black women received HPV co-testing compared to 5.2% for specimens of white women and 6.2% for those of women of other races (p<0.01). Bivariate predictors of HPV co-testing with Pap among women ≥ 30 years included race, age, insurance type, ZIP code median household income, and calendar year. Most of these associations, however, could be explained by clinic-level differences in HPV co-testing. For example, the bivariate odds ratio for black race, OR 2.69 (95% CI 1.21–5.98), was modified to OR 0.92 (95% CI 0.82–1.03) by a fixed-effects estimator for clinic. Using five representative large clinics, clinic-stratified analyses did not support an independent association with HPV co-testing for race, insurance type, or ZIP code median household income, although an increase in HPV co-testing over time was observed in most clinics. Conclusions: Under a population-average public health approach, the predictors of HPV co-testing with Pap varied between the pooled analyses and clinic stratified analyses. Using fixed-effects estimation to account for between-clinics effects in the pooled analyses provided further evidence for the substantial influence of clinics on the findings. This research suggests that the predictors of HPV co-testing with Pap among women ≥ 30 years likely lay at the provider and/or clinic level(s). Citation Information: Cancer Epidemiol Biomarkers Prev 2011;20(10 Suppl):B94.