BackgroundTuberculosis (TB) surveillance systems have some pitfalls outside of a National Tuberculosis Program and lack of efficient surveillance hampers accurate epidemiological quantification of TB burden.In the present study we assessed the quality of surveillance at the University Hospital in Pisa (UHP), Italy, and TB incidence rates over a ten year period (1999–2008).MethodsAssessment of underreporting was done by record-linkage from two sources: databases of TB diagnoses performed in the UHP and the Italian Infectious Disease Surveillance (IIDS) system. Two different databases were examined: a) TB diagnoses reported in the Hospital Discharge Records (HDR) from three Units of UHP (Respiratory Pathophysiology, Pulmonology and Infectious Diseases Units) (TB database A); b) TB diagnoses reported in HDR of all Units of UHP plus TB positive cases obtained by the Laboratory Register (LR) of UHP (TB database B). For the TB database A, the accuracy of TB diagnosis in HDR was assessed by direct examination of the Clinical Record Forms of the cases. For the TB database B, clinical and population data were described, as well as the trend of incidence and underreporting over 10 yrs.ResultsIn the first study 293 patients were found: 80 patients (27%) with a confirmed TB diagnosis were underreported, 39 of them were microbiologically confirmed. Underreporting was related to age (Reported vs Non Reported, mean age: 49.27 ± 20 vs 55 ± 19, p < 0.005 ), diagnosis (smear positive vs negative cases 18.7 vs 81.2%, p = 0.001), microbiological confirmation (49% vs 51%, p < 0.05), X-ray findings (cavitary vs non-cavitary cases: 12.5 vs 87.5%, p = 0.001) but not to nationality.In the second study, 666 patients were found. Mean underreporting rate was 69.4% and decreased over time (68% in 1999, 48% in 2008). Newly diagnosed TB cases were also found to decrease in number whereas immigration rate increased. Underreporting was related to nationality (Immigrants vs Italians: 18% vs 68%, p < 0.001), diagnosis (microbiological confirmation: 25% vs 75%, p < 0.01), kind of hospital regimen (hospitalized patients vs Day Hospital: 70% vs 16%, p < 0.001), and position of TB code in the HDR (TB code in first position vs in the following position: 39,5% vs 45% p < 0.001).ConclusionsTB is underreported in Pisa, particularly in older patients and those without microbiological confirmation. The TB code in first position of HDR seems fairly accurate in confirming TB diagnosis.