Background NICE guidelines for TB diagnosis recommend that sputum is obtained for culture for all suspected cases of pulmonary TB, and biopsies for all cases extrapulmonary TB. As results can take 6 weeks, treatment initiation decisions are frequently made without microbiological confirmation.

Aim This study set out to examine the accuracy of clinical diagnoses in a high incidence area, and the basis for these decisions.

Methods The data entered onto the national TB database was used to obtain a list of patients for whom no culture results had been recorded. Clinic letters, laboratory records and imaging were examined to determine whether samples had been sent for culture, how diagnoses were made in the event of negative results, and if alternative diagnoses were concluded.

Results Of 323 patients on the database, 7% had no samples sent for culture. There were 109 culture negative patients, of whom 13 (4% all cases) had alternative diagnoses. A combination of relevant history and imaging was the most commonly used method of diagnosis when culture was negative (47%). Histology was used in 17% patients and Mantoux or IGRA testing supported initiating treatment in 39% cases. The database was missing positive culture results for 102 patients, of which four were MDR TB.

Conclusions In this study, we found accurate initial clinical diagnoses, with only 4% patients subsequently obtaining alternative diagnoses. Most diagnoses were made on the basis of relevant history and imaging. Of concern are the 7% patients for whom tissue was never sent for culture. This is likely to be an underestimate when including all patients initially suspected of TB, raising the possibility of missed diagnoses. The utility of Mantoux and IGRA testing in active disease is now disputed. It is hoped with inter-specialty education regarding the importance of culture and futility of immunological based assays, the proportion of patients with suspected TB who have sputum or tissue sent for culture increases. Accurate recording of MDR-TB on the national TB database needs to be improved, to enable efficient monitoring of intervention programmes.