Audit of pulmonary tuberculosis management in a hospital with high tuberculosis prevalence: a tool for quality improvement in tuberculosis control

Abstract number: P803

Hites M., Melot C., Van Den Wijngaert S., Dediste A., De Wit S., Gerard M.

Objectives: To avoid nosocomial TB transmission, infection-control measures, using administrative controls, engineering controls and personal respiratory protection are recommended by the CDC. An audit on the implementation of these administrative controls was performed at Centre Hospitalier Universitaire Saint Pierre (CHUSP), a public hospital in Brussels, Belgium, with high incidence of TB admissions. In countries with low TB incidence like Belgium, many doctors lack experience in recognition of RTB.

Methods: A retrospective observational study on all patients admitted to CHUSP, with suspicion of active respiratory TB (RTB) or with positive cultures for M. tuberculosis from a respiratory specimen, between January 1, 2002 and December 31, 2006. The use of airborne precautions (AP) was evaluated for each hospitalisation episode (HE). A decision tree for predicting active RTB based on 4 criteria (upper lung infiltrate at the chest X-Ray, fever, weight loss and CD4 count), elaborated by El-Solh and al. (Am J Respir Crit Care Med 1997; 155: 1711–16), was tested. Median TB management intervals (25th and 75th inter-quartiles) were calculated. Intervals exceeding 24 hours were considered to be a delay.

Figure: % of HE with delays in diagnosis and treatment in active RTB (n = 374).

Results: The incidence of active RTB from 2002 to 2006 was 4.25/1000 admissions. The cohort consisted of 971 HE. The ratio of active RTB to number of AP HE was 374/962. Among the 374 HE with active RTB, there were 92 AP failures (51 delays in TB suspicion, and 41 AP stopped prematurely). Theoretical delays in TB suspicion (10/364: 2.75%) generated with the decision tree were significantly inferior to the observed delays (46/364: 12.64%, p: 0.008). The median TB suspicion, treatment and overall management interval was: 0 (0–6) days, 3 (1–6) days and 4 days (1–7 days), respectively [figure: % of patients with delays in diagnosis and treatment in confirmed RTB (n = 374)].

Conclusions: APs are not correctly implemented in 25% of patients diagnosed with active RTB at CHUSP. Although overall TB management intervals seem reasonable, they are under-estimated, as some patients with active RTB are referred to CHUSP by other institutions. Results will be presented by distinguishing HE referred from other institutions to those HE entirely managed by CHUSP medical fellows. The retrospective utilisation of a decision tree has increased sensibility of RTB detection from 84% to 97%. Prospective evaluation of the tree is required to test specificity.