Assessing the validity of tuberculosis surveillance data in California

BACKGROUND: The Centers for Disease Control and Prevention (CDC) convened a workgroup to revise the tuberculosis (TB) case report in the United States of America (U.S.). The group proposed substantial revisions. Study objectives were to systematically assess the validity and completeness of reported TB case surveillance data in California and to inform TB case report revision process. METHODS: A sample of 594 cases was retrospectively selected from the cohort of all TB cases reported during 6/1/96-5/31/97 to the State TB Registry. Cases, stratified by treatment outcome, were randomly sampled within each outcome category. Data for 53 variables were abstracted from each case's public health medical record and compared to data recorded on the TB case report. Using the medical record as the "gold standard," estimates were developed for 1) concordance, sensitivity, and positive predictive value of reported data for categorical variables; 2) the absolute mean difference between the two information source for date variables; and 3) the completeness of data on the case report and in medical record. RESULTS: At least 90% of the values for 35 (79.5%) categorical variables submitted on the TB case report form were identical to values in the medical record. Concordance between data on the case report and medical record was lower for the remaining nine (20.5%) categorical variables: status of abnormal chest x-ray (46.8%); directly observed therapy (48.6%); smear result for tissue or body fluid other than sputum (49.2%); type(s) of tissue or body fluid for smears and cultures other than sputum (76.4% and 73.9% respectively); provider type (73.4%); occupation (84.4%); sputum culture conversion (85.4%); and sputum smear result (89.6%). Case report data were more complete than data in the medical record; 2.9% versus 9.8% of data were missing/unknown, respectively. CONCLUSION: For most variables examined on the TB case report, data validity was excellent, indicating a robust surveillance system. However, lower data quality was noted for a small number of variables primarily impacting treatment adherence, including assessment and planning; advocacy; allocation and garnering of resources; and research. The study provides compelling evidence supporting the CDC workgroup's proposed revisions to the TB case report.