Pulmonary tuberculosis among patients hospitalised with community ...

7 downloads 0 Views 283KB Size Report
COMMUNITY-ACQUIRED PNEUMONIA (CAP), one of the most common infectious diseases in hu- mans, is usually caused by Streptococcus pneumoniae,.
INT J TUBERC LUNG DIS 17(12):1626–1631 © 2013 The Union http://dx.doi.org/10.5588/ijtld.13.0183

Pulmonary tuberculosis among patients hospitalised with community-acquired pneumonia in a tuberculosis-prevalent area S-B. Chon,* T. S. Kim,† W. S. Oh,† S-J. Lee,† S. S. Han,† W. J. Kim† * Department of Emergency Medicine, Seoul National University Hospital, Seoul, † Department of Internal Medicine, Kangwon National University School of Medicine, Chuncheon, Korea SUMMARY SETTING:

A suburban teaching hospital in a tuberculosis (TB) prevalent area. O B J E C T I V E S : To evaluate the proportion of pulmonary TB among patients hospitalised with suspected community-acquired pneumonia (CAP), and to develop a diagnostic index for identifying TB among these patients. D E S I G N : TB cases confirmed using 1) sputum culture, or 2) both sputum acid-fast bacilli smear and polymerase chain reaction for Mycobacterium tuberculosis, were compared with non-tuberculous CAP by demographic, clinical, laboratory and radiographic information. Using multiple logistic regression analysis, risk factors for TB were identified. A diagnostic index was developed by summing up their simplified regression coefficients. Its performance was checked using c-statistic.

R E S U LT S : TB was the second leading cause of CAP (37/528, 7.0%). Risk factors were initial symptoms >7 days, serum albumin 20 breaths/min) or tachycardia (⩾100 beats/min), or crackles or other auscultatory evidence of pulmonary consolidation.21 To be included, patients also had to have undergone sputum AFB smear, culture or polymerase chain reaction (PCR) for M. tuberculosis to reveal the presence or absence of TB. Patients already on anti-tuberculosis treatment at the time of initial presentation or those who had been hospitalised for >72 h within the previous 30 days were excluded. Data collection and definition To evaluate the causative micro-organisms of CAP, microbiological data were collected whenever feasible. Serological tests for atypical pathogens such as Mycoplasma, Chlamydia and Legionella were not performed. Blood and sputum cultures had been routinely performed for CAP patients. Urinary pneumococcal antigen tests and real-time PCR for influenza virus were checked if indicated by the clinician in charge. Patients who showed positive results from 1) sputum culture, or 2) both sputum AFB smear and PCR for M. tuberculosis, were defined as the TB group. The remaining patients were designated the nontuberculous (non-TB) group. Demographic, clinical, laboratory and radiographic data were obtained for all patients. All of the variables derived from the Pneumonia Severity Index or CURB-65 (confusion of new onset, urea > 7 mmol/l, respiratory rate ⩾30 breaths per minute, blood pres-

1627

sure < 90 mmHg systolic or diastolic blood pressure ⩽ 60 mmHg, age ⩾ 65 years) were gathered as part of clinical and laboratory data.22,23 In addition, whether or not the patient was underweight, based on the Asian criterion of body mass index (BMI) < 18.5 kg/m2, was noted.24 Comorbidities included pulmonary disease (chronic obstructive pulmonary disease or asthma), cardiovascular disease (congestive, ischaemic or congenital heart disease), renal disease (nephrotic syndrome or chronic renal failure), hepatic disease (liver cirrhosis or chronic active hepatitis), stroke and malignancy. Radiographic variables included location (upper, middle, lower lobes), pattern of infiltrates (consolidation, cavitary, nodular or reticular) and presence or absence of pleural effusion. To eliminate information bias, all of the data were collected without knowledge of sputum AFB smear, culture or PCR results. The clinical courses of the TB and non-TB groups after admission were compared in terms of the use of empirical antibiotics, duration of fever after initiation of antibiotic agents and implementation of respiratory isolation. Statistical analysis In the univariate analysis, the TB and non-TB groups were compared using the variables mentioned above. Pearson’s χ2 or Fisher’s exact test was used for categorical variables, and Student’s t-test was used for continuous variables. Thereafter, the variables chosen from the univariate analysis (P < 0.20), together with those of clinical relevance (sex, age, history of TB), were included in a multiple logistic regression model. Levels of significance of 0.05 for exclusion were required in this step. The goodness-of-fit of the model was evaluated using the Hosmer-Lemeshow test. A diagnostic index was developed by assigning the nearest whole number points to all of the chosen variables in proportion to their regression coefficients. With this index applied to the whole data set, the optimal cut-off point was determined by drawing a receiver operating characteristic (ROC) curve. For comparison, the same procedure was followed with the original Moran’s criteria, which was developed to identify TB among CAP patients at emergency departments with a low prevalence of TB.19 It suggests TB if only one of the following seven risk factors is present: history of TB, immigrant from an endemic area, homelessness, incarceration, weight loss, apical infiltrate and cavity on CXR. Of these, ‘immigrant from an endemic area’ was meaningless in this study population, as Korea is an endemic area. For comparison, we therefore used the other six risk factors as ‘modified’ Moran’s criteria. As a surrogate for internal validation while adjusting the model parameters for potential over-fitting, the 95% confidence interval (CI) of the area under

1628

The International Journal of Tuberculosis and Lung Disease

Figure 1 Inclusion and exclusion of cases with CAP. CAP = community-acquired pneumonia; TB = tuberculosis; AFB = acid-fast bacilli; PCR = polymerase chain reaction.

the ROC curve (c-statistic) was then calculated by bootstrapping with 1000 replications. A two-tailed P value of