Purpose:

In 2014, Korea ranked as the first among the Organization for Economic Cooperation and Development countries on the prevalence, incidence, and mortality of pulmonary tuberculosis (TB). The prevalence of TB among the homeless was 6.4% in the United State and 7.1% in South Korea. The aim of this study is to develop predicting indicators of TB by analyzing homeless people who visit the public hospital emergency department (ED). 

Method:

We analyzed 7,500 homeless individuals who visited a public hospital ED between January 1, 2001 and May 31, 2014. A total of 4,552 patients were included, and of these, 145 homeless patients were infected with TB. We conducted univariate and multivariate analysis of clinical variables obtained from the initial check list and later lab analysis, and made a scoring system by weighing each variable. Then applying this scoring system, the area under the receiver (AUC) operating characteristic curve (ROC) was calculated. 

Results:

The prevalence of TB was 3.2%. The initial meaningful predictor variables were as follows: Being homeless, abnormal heart rate, abnormal respiratory rate, no alcohol intake, hypoalbuminemia, and CRP elevation. The AUC of ROC curve from these predictor variables were 0.815. 

Conclusion:

We developed a novel scoring system to screen TB patients in a vulnerable social group who visit the ED. We can detect potential TB patients early and effectively control TB, preventing the spread of TB. Prospective internal and external validation is necessary by using the scoring system of TB among the homeless.