Objective: To identify the main determinants of cellular atypia detection in the cervical screening program in the state of Rio de Janeiro, Brazil, using data from the Cervical Cancer Information System SISCOLO.
Methods: A random sample of 65 535 Pap smears performed in 2007 was obtained from SISCOLO. This sample was used to produce a logistic regression model to identify variables that impact the process of detecting cellular atypia. A ROC curve was used to define the most suitable cutoff point to classify the presence or absence of atypia. A sensitivity analysis was performed to assess the impact on the model of factors related to the organization of the service model.
Results: The variables of impact were "reference laboratory," which reflects laboratory production scale; "presence of cellular elements representative of the transformation zone," which reflects the quality of the sampling; "immature squamous metaplasia," "presence of other benign cellular changes," and "absence of typical vaginal microorganisms." Each increase of 1 year in age was associated with a 1.7% reduction in the chance of detecting atypia. The ROC curve defined a cutoff of 4.5%, which allowed the maximization of the model's sensitivity (73.0%) and specificity (66.8%). Sensitivity analysis indicated an increase of 46.4% in the frequency of atypia following a simulated increase in the number of samples analyzed in the excellence (42.9%) and in the presence of cellular elements representative of the transformation zone (43.0%).
Conclusions: The model revealed that the detection of atypical cells is strongly influenced by organizational factors such as adequate sample collection and laboratory size. Because these factors can be changed by adequate management practices, the proposed model may be an important tool to improve cervical screening programs.