Predictors of Follow-up of Atypical and ASCUS Papanicolaou Test Results in a High-Risk Population
Methods
Setting and Population
Data were collected in 7 community health centers serving a low-income, predominantly minority population. The centers represent diversity with regard to practice type (medicine/pediatrics/obstetrics model or family medicine model) and size (20,000-80,000 visits/year). The subjects were women with newly identified atypia or ASCUS Pap test results in 1996; women for whom the index Pap test represented persistent atypia or ASCUS were excluded, as were women with a history of CIN or squamous intraepithelial lesion (SIL). Those who were known to be infected with HIV represent a special category19,20 that may have been managed differently at the clinician’s discretion; therefore, they were also excluded from the analysis.
Data Collection
After approval by the appropriate institutional review boards, women with atypical results in 1996 were identified by 2 mechanisms: site logs maintained for clinical tracking, and because all sites exclusively use the same laboratory, the central cytology database. Pap test results described as atypia or ASCUS were included, as these represent different terminology for the same spectrum of cytologic abnormalities with continued use of the older term by some cytopathologists. Chart abstraction by a trained research assistant collected information regarding: (1) complicating medical conditions (immunosuppression, history of cancer, substance abuse, pregnancy); (2) history of cervical dysplasia; (3) primary care provider and number of visits to site; (4) the provider’s management plan; (5) all subsequent Pap test and colposcopy results; (6) evidence of communication of the management plan to the subject; and (7) demographics.
Accuracy of chart abstraction was confirmed by the lead author rescreening 20% of the records. The data were entered into a Paradox database (Borland Software Corporation, Scotts Valley, Calif), and statistical analysis was performed using SAS software (SAS Institute, Inc, Cary, NC) and Stata software (Stata Corporation, College Station, Tex).
Analysis
The goal of our analysis was to determine the factors associated with follow-up. We defined 3 mutually exclusive levels of adherence to the CPG: (1) complete adherence (colposcopy done if recommended or 3 repeat tests approximately every 6 months until 3 are normal or colposcopy if there is persistent abnormality); (2) moderate adherence (1 or 2 follow-up tests after the index Pap test); and (3) low adherence (no follow-up cytology or histology, or abnormality on repeat Pap test without colposcopy.)
Univariate analyses were performed relating the level of follow-up to the description of the abnormality (atypia vs ASCUS), patient demographics, and practice and provider characteristics (eg, training, availability of colposcopy on site, and plan of care). In analyses where cell size fell below 5, Fisher exact test or Yates correction (for analyses with row or column size greater than 2) replaced the Pearson chi-square. We performed multivariate analysis using a hierarchical model with patients nested within sites, and ordinal logistic regression to account for the 3 ordered levels of adherence to the CPG. Variables significant at P less than .2 in the univariate analyses were included in the multivariate model to ensure inclusion of potentially important variables. Also, insurance status was included, because others have found significant associations for this variable.21-23 The provider’s management plan was not included as a variable in the regression model, because it is not sufficiently independent of the adherence levels.
Because subjects were clustered within 7 sites (not a simple random sample), we treated site as a random effect with patients nested within the site. An ordered logistic regression model was fit by maximum likelihood estimation, using the program GLLAMM6 in Stata version 6.0 The fit of the ordered logistic regression model to the data was tested using the method of Harrell and colleagues.24 Briefly, this method compares observed mean values of the model’s linear combination of the predictor variables to expected values under the assumption that the model (and in particular the proportional odds ratio assumption) is correctly specified.
Results
of the 570 women identified with an atypical or ASCUS Pap test result in 1996, 554 charts were located and reviewed. The charts of 16 individuals could not be located for review. Some women (n=69) identified had a previous atypical or ASCUS Pap test result. The sample of women with no previous ASCUS or atypical result, excluding those with documented HIV infection (n=19) and those with a history of CIN or SIL (n=79), consisted of 387 subjects. The mean age of the subjects at the time of the index Pap test was 32 years (standard deviation=11.7). More than half of the subjects (58.4%) were Hispanic. The majority of subjects had Medicaid (73.1%). The majority of index Pap result abnormalities were described as atypia (61.5%), with the remainder ASCUS (38.5%). A small percentage of women (8.8%) were pregnant at the time of the index Pap test. The subjects were approximately evenly split between receiving care in medicine/pediatrics/obstetrics model sites and family medicine model sites. Colposcopy was provided on site (at least 2 times per week) in 4 of the centers that served 68.7% of the subjects.
