Tuberculosis (TB) is caused by the Mycobacterium tuberculosis complex, which was first discovered in 1882 by Robert Koch¹. It usually affects the lungs (pulmonary TB), but can affect other sites as well (extrapulmonary TB). TB has been one of the most significant global pathogens in terms of human morbidity and mortality². In 2010, there were 8.8 million incident cases of TB and 1.45 million deaths from TB worldwide³. Although the absolute number of cases and the incidence has decreased since 2002, TB still poses a huge health burden and is a major global health problem³.

Although the prevalence of TB in Korea, based on chest radiography results, decreased from 5.1% in 1965 to 1.0% in 1995, 36,305 new TB patients were still registered on the Korean TB surveillance system (KTBS) in 2010, the highest incidence among all Organization for Economic Co-operation and Development (OECD) countries^4,5. The Korean National Tuberculosis Association conducted seven nationwide TB prevalence surveys every 5 years from 1965 to 1995⁶. A web-based surveillance system, the KTBS which is based on doctor notifications, was launched in 2000 and has been implemented across the country to replace the nationwide survey⁷. In Korea, doctors are required to notify the KTBS of all cases of active TB immediately after diagnosis and treatment in accordance with Article 4 of the Infectious Disease Preventive Law (notification by physicians) and Article 20 of the Tuberculosis Preventive Law (notifying duty of healthcare facilities) of Korea⁸.

KTBS data are used to analyze the burden of TB in Korea. Hence, the diagnostic accuracy of the KTBS data is important. Several previous studies have evaluated the accuracy of the TB surveillance system data. However, most of these reports have focused on the underestimation of TB incidence (due to the under-reporting of doctors) and emphasized the importance of notification^9-11. There have also been some studies that have focused on the overestimation of TB burden^12,13. These studies reported that notified TB cases were not always real TB patients. In accordance with this finding, sales of TB drugs are continuously decreasing in Korea (IMS PLUS, Korea, personal communication). To provide a more accurate measure of the incidence and characteristics of TB, "diagnosis changed" cases need to be excluded. The aim of our present study was to evaluate the diagnosis change rate, confounding diseases, and contributing factors to the notifications of TB cases to the KTBS at a single center (a private tertiary hospital).

A total of 1,273 patients notified from the Asan Medical Center in 2011 as TB cases on the KTBS were enrolled in this study. The electronic medical records and KTBS reporting data for these individuals were retrospectively analyzed.

Of the 1,273 patients enrolled in this study, 17 (1.3%) who were assigned to the "administrative error" group were excluded and a final cohort of 1,256 was included in the final analysis. Of these 1,256 patients, 130 (10.4%) were reported to have a treatment outcome of "diagnosis changed." After a review of medical records, an additional 29 patients (2.3%) were determined clinically to be "not TB" (nontuberculous mycobacterial [NTM] 7 patients, lung cancer 4 patients, Crohn's disease 2 patients, sarcoidosis 1 patient, Kikuchi disease 1 patient and others 14 patients). Consequently, the "diagnosis changed" group comprised 159 patients (12.7%), leaving 1,097 patients (87.3%) in the "diagnosis maintained" group (Figure 1).

The clinical characteristics of the notified TB patients are outlined in Table 1. The mean age at notification was 51.9±17.7 years in the "diagnosis maintained" group and 54.6±17.7 years in the "diagnosis changed" group. In these two groups, the BMI scores were 22.4±10.6 kg/m² and 22.3±3.7 kg/m², respectively. Smoking history, medication, and radiological findings did not differ between the groups. Malignancy was more frequent in the "diagnosis changed" group. The number of patients notified as pulmonary TB was 866 (68.9%), of whom 118 (13.6%) had a "diagnosis changed" outcome. Of the remaining 748 patients, 181 (24.2%) were acid-fast bacilli (AFB) smear-positive, 473 (63.2%) were M. tuberculosis culture positive, 229 (30.6%) were nucleic acid amplification (NAA) test positive, and 153 (20.5%) showed granulomas on histological examination. Patients with AFB smear-positive results have traditionally been considered to have TB. However, of the 217 patients with positive AFB smear and notified as pulmonary TB, only 83.4% had TB and others (15.7%) had NTM diseases in our study. The number of patients notified as extrapulmonary TB was 390 (31.1%), of whom 41 (10.5%) had a "diagnosis changed" outcome. Of the remaining 349 patients, 18 (5.2%) had AFB smear-positive results, 88 (25.2%) had M. tuberculosis culture positive results, 83 (23.8%) had positive NAA test results, and 176 (50.4%) showed granulomas on histological examination.

The common causes of "diagnosis changed" are described in Table 2 and include NTM disease (51.7%, 61 out of 118 patients), pneumonia (17.8%, 21), lung cancer (13.6%, 16), and interstitial lung disease (5.1%, 6) in patients notified as pulmonary TB. Other lung diseases were healed TB (4 patients), benign lung nodules (3), IgG4-related disease (2), Behcet's disease (1), destructive bronchiolitis (1), sclerosing hemangioma (1), and hemoptysis of unknown causes (2). In 41 patients notified as extrapulmonary TB, the causes were bacterial or viral meningitis (19.5%, 8 patients), Crohn's disease (12.2%, 5), and complicated parapneumonic effusion (9.8%, 4). Other extrapulmonary diseases were Kikuchi disease (2 patients), bladder tumor (2), cystitis (2), encephalitis (2), sarcoidosis (2), NTM infection (2), Behcet's disease (1), cytomegalovirus colitis (CMV) colitis (1), CMV ventriculoencephalitis (1), central nervous system aspergillosis (1), epidural abscess (1), hematuria (1), healed TB (1), lymphoma (1), pericarditis (1), Q fever (1), seminoma (1), and tenosynovitis (1).

The predictive factors for a "diagnosis changed" outcome are presented in Table 3. Being older than 35 years (odds ratio [OR], 2.18; p=0.037) and having a positive AFB smear (OR, 1.58; p=0.047) were positive predictors for a "diagnosis changed" outcome. A TB-related radiological finding (OR, 0.42; p<0.001) was a negative predictor in multivariate logistic regression analysis in patients notified as pulmonary TB. In patients notified as extrapulmonary TB, only the absence of TB-related radiological finding (OR, 5.25; p<0.001) was a significant predictor for a "diagnosis changed" outcome.

In the current TB notification system, both underestimation (due to under-reporting) and overestimation (due to "diagnosis change" cases) are possible, but in our study we could not estimate which one is bigger than the other. The most important result of our study is that more than 10% of patients notified as TB cases did not in fact have TB. Given that the estimated incidence of TB in Korea is made using the annual notification rate, this incidence may well be an overestimate.

From 1965 to 1995, the prevalence of TB in Korea was estimated through a nationwide representative sampling survey every 5 years. As this prevalence has decreased, the TB burden has been followed since 2000 using the KTBS notification system. However, TB notifications have been considered more likely to be an underestimate of the true incidence of this disease because of a lack of voluntary reporting by doctors^17,18. On the other hand, because TB is still prevalent in Korea, many doctors diagnose it on the basis of a single chest radiograph alone, without bacteriological confirmation. In addition, many doctors commence anti-TB medication without receiving a confirmation of TB. According to Korean Infectious Disease Preventive Law, doctors should notify the KTBS of any new cases as soon as anti-TB medication is prescribed, irrespective of any bacteriological confirmation. This procedure may result in an overestimation of TB notification. In our present study, about 12.5% of the patients from our hospital notified as TB cases (13.6% in pulmonary TB and 10.5% in extrapulmonary TB) later became "diagnosis changed". Although 29 patients were not notified as "diagnosis changed," these 29 patients should have been notified as "diagnosis changed" case at that time. Hence, we enrolled these 29 patients as "diagnosis changed" cases. If we excluded these 29 patients from "diagnosis changed" case, "diagnosis changed" rate would be 10.4% (130/1,256) in patients notified as TB. We suggest that "diagnosis changed" TB cases should be excluded when reporting the annual TB notification rate in Korea.

In patients notified as pulmonary TB cases on the database, predictive factors for a "diagnosis changed" outcome were found to be an age older than 35 years and a positive AFB smear at diagnosis. The most common cause of "diagnosis change" was found to be NTM pulmonary infection. The chance of developing NTM disease increases with age in AFB smear-positive patients¹⁹. In Korea, patients with AFB smear-positive specimens have traditionally been considered to have TB, and anti-TB treatment is promptly administered in these cases²⁰. We should recognize that more than 15% of sputum AFB smear-positive patients notified as pulmonary TB did not have TB but had NTM disease in our study. Considering the trend towards increased levels of NTM disease, it is expected that the mistaken notification of NTM disease as TB will increase. Hence, it is important to be able to rapidly distinguish between M. tuberculosis and NTM by conducting an NAA test in AFB-positive specimens. We contend that this recommendation should be incorporated into current Korean TB guidelines.

In our present analyses, the proportion of extrapulmonary TB (31.1%) was relatively high compared with that found in a previous study⁵. This may be due to the inherent characteristics of a given tertiary care hospital. Because the diagnosis of extrapulmonary TB is more difficult than that of pulmonary TB, more patients with suspected extrapulmonary TB tend to visit tertiary care hospitals. Due to the diagnostic difficulty associated with extrapulmonary TB, we expected that the "diagnosis changed" rate would be higher for extrapulmonary TB than pulmonary TB. However, in reality the "diagnosis changed" rate was slightly, though not significantly, higher for pulmonary TB (13.6% vs. 10.5%, respectively). This result may be due to the finding that the exclusion of TB diagnosis is more difficult in instances of extrapulmonary TB than in pulmonary TB.

This study has several limitations to note. First, due to the retrospective nature of our study and the fact that it involves data from a single tertiary care center, it is more difficult to generalize and apply the results to other hospitals in Korea. However, an overestimation of TB is highly probable in small-scale hospitals because of a lower rate of bacteriological confirmation for TB diagnosis²¹. Second, because we only analyzed patients notified as TB cases, other TB patients who were not notified as TB may have been overlooked in our analysis, leading to an underestimation of the TB burden. Third, a high proportion of TB patients (278 patients, 25.3%) were clinically diagnosed and the accuracy of the TB diagnosis could not be confirmed in these cases. However, this is a common challenge with TB diagnoses. Finally, due to the increased NTM infection trend found in this study, as well as more widely in Korea, an ironic situation has developed whereby a positive sputum AFB stain, which is also used to diagnose NTM, has become a significant predictive factor for a "diagnosis changed" outcome in patients notified as pulmonary TB cases. However, this would not be the situation in countries where the NTM infection rate is low.

In conclusion, TB notifications may be an overestimate of the true incidence of this disease in Korea, particularly because of the increased frequency of NTM infections. Therefore, in order to estimate, more accurately, the incidence of TB in Korea, it would be desirable in the future to exclude patients with a "diagnosis changed" outcome from the total number of patients notified as TB.