BACKGROUND: Connecticut conducts surveillance of high-grade cervical neoplasia to monitor human papillomavirus vaccine impact. Cases entered in the surveillance database meet the case definition of having pathologic diagnoses of cervical intraepithelial neoplasia grades 2 and 3, or adenocarcinoma in situ. For state-mandated reporting, laboratories may send results from searching their records for International Classification of Diseases(ICD) codes known to be associated with these lesions. Because this search also yields pathologic diagnoses not meeting the surveillance case definition that need to be screened out by surveillance staff, we evaluated its performance at Connecticut’s highest-volume reporting laboratory that uses ICD code searching (Laboratory A).
METHODS: Reports from Laboratory A during 2014 were compared to the surveillance database of confirmed cases, the gold standard. We calculated positive predictive value (PPV) as the percentage of laboratory-reported cases found in the database. We identified missed cases by reviewing medical records for 25% of patients with pathologic results reported during the first quarter of 2014, who were not in the database. We calculated sensitivity of laboratory reporting as the number of first quarter cases entered in the database divided by the sum of that number and four times the number of missed cases identified.
RESULTS: During 2014, Laboratory A identified 2,782 pathologic results based on ICD codes, of which 901 (PPV 32%) entered the surveillance database. For the first quarter, 244 of 583 reported results entered the surveillance database; review of 65 medical records revealed three unreported cases (95% sensitivity).
CONCLUSIONS: Sensitivity is high, but PPV for Laboratory A’s ICD code search is low, leading to reporting of confidential information for non-case patients and additional reviews for surveillance staff.