Provider- and Facility-Level Variation in Precancerous Cervical Biopsy Diagnoses

Abstract

Objectives: Reproducibility of cervical biopsy diagnoses is low and may vary based on where the diagnostic test is performed and by whom. Our objective was to measure multilevel variation in diagnoses across colposcopists, pathologists, and laboratory facilities.

Methods: We cross-sectionally examined variation in cervical biopsy diagnoses within the 5 sites of the Population-Based Research Optimizing Screening through Personalized Regimens (PROSPR I) consortium within levels defined by colposcopists, pathologists, and laboratory facilities. Patients aged 18 to 65 years with a colposcopy with biopsy performed were included, with diagnoses categorized as normal, cervical intraepithelial neoplasia grade 1 (CIN1), grade 2 (CIN2), and grade 3 (CIN3). Using Markov Chain Monte-Carlo methods, we fit mixed-effects logistic regression models for biopsy diagnoses and presented median odds ratios (MORs), which reflect the variability within each level. Median odds ratios can be interpreted as the average increased odds a patient would have for a given outcome (e.g., CIN2 or CIN3 vs normal or CIN1) when switching to a provider with higher odds of diagnosing that outcome. The MOR is always 1 or greater, and a value of 1 indicates no variation in outcome for that level, with higher values indicating greater variation.

Results: A total of 130,110 patients were included who received care across 82 laboratory facilities, 2,620 colposcopists, and 489 pathologists. Substantial variation in biopsy diagnoses was found at each level, with the most occurring between laboratory facilities, followed by pathologists and colposcopists. Substantial variation in biopsy diagnoses of CIN2 or CIN3 (vs normal or CIN1) was present between laboratory facilities (MOR: 1.26; 95% credible interval = 1.19-1.36).

Conclusions: Improving consistency in cervical biopsy diagnoses is needed to reduce underdiagnosis, overdiagnosis, and unnecessary treatment resulting from variation in cervical biopsy diagnoses.

Figure 1.. Distribution of percent of cervical biopsy diagnoses with more severe result by colposcopist, pathologist, and laboratory facility.

Box plots are depicted showing the distribution of the more severe cervical biopsy diagnoses within individual units corresponding with provider-specific levels defined by colposcopists, pathologists, and laboratory facilities. The left panel displays the outcome of CIN2 or CIN3 vs. Normal or CIN1, the middle panel displays CIN2 vs. CIN1, and the right panel displays CIN3 vs. CIN2. The lower and upper hinges (bottom and top of the box) correspond to first and third quartiles.

Figure 2.. Median odds ratio (MOR) and 95% credible interval (CI) from full models by cervical biopsy diagnosis category and levels defined by colposcopists, pathologists, and laboratory facilities.

The left panel displays the outcome of cervical intraepithelial neoplasia grade 2 (CIN2) or grade 3 (CIN3) vs. Normal or grade 1 (CIN1), the middle panel displays CIN2 vs. CIN1, and the right panel displays CIN3 vs. CIN2. The MOR is a measure of variation between units within a given level. It is the median value of the ratio of the predicted odds of outcome of interest for patients (with otherwise equivalent covariates) that are randomly selected from different units within a given level. For example, the first MOR in the left panel represents the increased odds (of receiving a diagnosis of CIN2 or CIN 3 vs. Normal or CIN1) a patient would have when switching from a colposcopist with lower odds of diagnosing CIN2 or CIN3 to a different colposcopist with higher odds of diagnosing CIN2 or CIN3.