Clinical profiles of early and tuberculosis-related mortality in South Korea between 2015 and 2017: a cross-sectional study

An estimated 1.3 million people died due to tuberculosis (TB) in 2017, making TB one of the leading causes of death due to an infectious agent worldwide [1]. The World Health Organization’s (WHO) End TB target is a 95% reduction in the number of deaths due to active TB between 2015 and 2035 [2]. In South Korea, the TB rate has decreased significantly, with a 5.2% annual reduction in the incidence of newly reported TB cases from 2011 to 2016; however, South Korea has the highest TB incidence and mortality rates among the high-income countries [3]. In 2016, the total number of reported cases was 39,245, with an incidence rate of 78.8 persons per 100,000; the mortality rate was 5.1 persons per 100,000 [4]. As South Korea becomes an older-aged society, TB mortality and incidence are rapidly increasing among those over 60 years old; this is a huge obstacle to national TB control [5].

A review [6] of the risk factors associated with death during anti-TB treatment, which include human immunodeficiency virus (HIV) positivity, old age, comorbidities, and use of alcohol and drugs, indicates that there are differences in risk factors among regions with low and high incidence of TB. Due to a low prevalence of HIV infection [7, 8] and an intermediate TB burden, South Korea requires a different strategy to control TB mortality. In addition, the causes of TB mortality may differ depending on the phase of anti-TB treatment; however, there are only few studies investigating early deaths, defined as death occurring within the first 2 months of anti-TB treatment [9‐12]. TB mortality is a key indicator for the national TB control program in South Korea. Further studies using nationwide data are required to better understand TB mortality, hence leading to opportunities for public health intervention that can decrease TB mortality and improve treatment outcomes.

With the introduction of the nationwide health insurance system in 1989, TB control in South Korea began to transition from a public health center-based program to a private hospital-based program [13]. In 2011, a public–private mix (PPM) collaboration model was implemented as a national TB control strategy. We collected data of TB mortality cases at PPM-participating hospitals for monitoring and evaluation. The purpose of this study was to assess early and TB-related mortality during anti-TB treatment and describe the associated clinical characteristics.

Of 5595 TB patients with a notifiable death, we excluded patients with anti-TB treatment duration of over 1 year (n = 121), patients with drug-resistant TB (n = 400), patients with miliary TB (n = 135), patients with extra-pulmonary TB (n = 826), patients not initially receiving the standard regimen (n = 370), and those with misdiagnosis and missing data (n = 8). Ultimately, 3735 patients with pulmonary TB were included in this study (Fig. 1). There were 944 (25.3%) TB-related deaths and 2791 (74.7%) non-TB-related deaths. The proportion of early TB-related deaths was significantly higher than that of early non-TB-related deaths (82.7% (781/944) vs. 63.2% (1764/2791), P = 0.000).

The baselines characteristics of 3735 enrolled male and female patients (Tables 1 and 2). There were 2541 (68.0%) male patients, and 2935 (78.6%) mortality cases were observed in elderly patients over 65 years of age. The mean age of female patients was significantly higher than that of male patients (78.9 ± 11.7 vs. 72.1 ± 13.0, P = 0.000). The proportion of individuals with body mass index < 18.5 kg/m² was similar between male and female patients. The proportion of prior TB history in male patients was significantly higher than that in female patients (21.5% vs. 10.9%, P = 0.000). Male patients were more likely to have chronic lung disease (8.0% vs. 4.7%, P = 0.000) and malignancy (24.8% vs. 12.7%, P = 0.000), and female patients had higher frequency of cardiovascular disease (8.1% vs. 13.1%, P = 0.000). The proportion of positive acid-fast bacilli (AFB) smear and culture test result were 46.9 and 69.7%, respectively.

The cumulative number of deaths within 30 and 60 days were 1993 (53.4%) and 2545 (68.1%), respectively (Fig. 2). The median interval between diagnosis and death among all patients was 26.0 days (interquartile range 6.0–81.0 days). Among 383 (10.3%) patients who did not receive anti-TB treatment, 187 (5.0%) patients were diagnosed post-mortem. Thirty-eight (1.0%) patients died on the first day of treatment.

In multivariate analysis, current smoker (P = 0.023), body mass index less than 18.5 kg/m² (P = 0.000), prior TB history (P = 0.002), no reported illness (P = 0.026), cavitary (P = 0.001) and bilateral disease (P = 0.000) on chest X-ray, positive AFB smear results (P = 0.000), cough (P = 0.002), and hemoptysis (P = 0.007) were significantly associated with TB-related death (Table 3). Cardiovascular disease, malignancy, and chest pain were associated with non-TB-related death. Calibration of this predictive model was good, as shown by a Hosmer-Lemeshow test (P = 0.590). Among 2791 patients who died of non-TB-related causes, acute respiratory failure (18.9%) was the most frequent cause of death, followed by pneumonia (18.7%) and malignant neoplasm (17.4%) (Table 4).

In comparison with individuals with late death, individuals with early death had significantly higher proportion of body mass index < 18.5 kg/m² (P = 0.003), no reported illness (P = 0.005), bilateral disease on chest X-ray (P = 0.001), positive AFB smear result (P = 0.001), and dyspnea (P = 0.000) (Table 5). Current smoker status (P = 0.039) and malignancy (P = 0.000) were associated with late death. Model adequacy was checked using a Hosmer-Lemeshow test (P = 0.057).

After identifying variables significantly associated with each predefined subsets of TB mortality, we plotted a log-log graph describing factors related to early TB-related and late non-TB-related deaths (Fig. 3). The factors in the right upper quadrant, such as low body mass index, no reported illness, bilateral disease on chest X-ray, and positive AFB smear result, were associated with both TB-related and early death. In contrast, malignancy was associated with both non-TB-related death and late death.

This large cross-sectional study assessed 3735 TB mortality cases that occurred during anti-TB treatment among adult patients with pulmonary TB in South Korea. According to our study, three of every four deaths occurring during anti-TB treatment were attributed to non-TB-related causes such as acute respiratory failure and malignancy. We also observed a high number of TB mortality cases among elderly patients in this study. South Korea has become an aged society, and the epidemiologic pattern within the past few decades has changed from infectious to chronic non-communicable diseases that cause individuals (particularly elderly individuals) to be at a greater risk of developing TB [5]. More attentions to this elderly population are needed, and accordingly, South Korea is currently preparing a comprehensive management plan for its elderly population, which is a key vulnerable group [3].

There are several factors that contribute to the enormous TB death toll worldwide [17]. Although global initiatives address the importance of malnutrition [18] and smoking [19], which have been clearly linked with excess TB mortality, such key determinants of TB mortality are still under-emphasized at the country level. A weak and underinvested public health system, especially in low-income countries, is another important issue, and has led to a suboptimal cascade of care for TB patients [20]. Since 2011, South Korea has increased budgets and strengthened patient management policies together with a PPM collaboration model, which led to significant decreases in TB incidence [3]. However, a high and stagnant TB mortality rate poses a great challenge to the national TB control program in South Korea. Our study findings may allow for planning of effective interventions to strengthen patient-centered care and improve patient survival.

As illustrated in the results, low body mass index, no reported illness, bilateral disease on chest X-ray, and positive AFB smear result were significantly associated with early and TB-related death. Because we aimed to identify mortality cases that could be controlled and prevented through appropriate management, we excluded TB cases presumed to be hard to treat, such as cases involving immunocompromised host, drug resistance, and disseminated disease. Cases of spinal or miliary disease reflect compromised immunity and are difficult to manage even with proper intervention. Patients with drug resistance are prone to poor medication adherence and severe adverse reactions related to non-TB-related death [21]. Based on our results, we can propose that early interventions should target patients with severe TB disease and malnutrition, in order to improve TB mortality.

Proper diagnosis and rapid treatment of patients with infectious TB remains the cornerstone of a TB control program [22]; however, 10% of patients enrolled in our study died without receiving anti-TB drugs, and 5% of patients were diagnosed post-mortem. TB control is also complicated by post-mortem diagnosis, as these patients likely had multiple encounters with healthcare facilities while infected with unrecognized TB [23]. Automated nucleic acid amplification tests such as Xpert MTB/RIF assay are increasingly deployed in many countries as the point-of-care diagnostic test for TB [24]. Recently, South Korea has expanded the coverage of national health insurance for the Xpert MTB/RIF assay. Elderly patients who cannot produce adequate sputum specimen would benefit from the implementation of such a sensitive tool, which could reduce missed or delayed diagnosis.

Due to the difficulties inherent to differentiating TB-related death from non-TB-related death, only a small number of published studies have examined risk factors for TB-related death. A Taiwanese study [25] found that extrapulmonary, miliary, and pneumonic radiographic patterns were independent risk factors for TB-related death when compared to survivors. A single-center study from South Korea [11] suggested that early TB-related deaths were mainly attributable to delayed diagnosis. Another recent study that evaluated TB mortality revealed that an underlying chronic condition, lower hemoglobin level, and acute respiratory failure were independent risk factors for TB-related death among non-elderly patients with miliary TB [12]. In order to meet the target goal of reducing TB mortality in South Korea, clinicians should be educated on the identification of preventable risk factors for TB-related death. In addition, South Korea estimates for TB mortality are based on National Statistical Office data. These data reflect only TB-related death, because the cause of death on the death certificate is used to identify the mortality case in accordance with the WHO recommendations [26]. Accuracy and utility of cause of death data from death certificates are uncertain and often questionable. The current study is valuable in that it is the first attempt to assess TB mortality using a large nationwide database in South Korea. An accurate estimate of mortality enables a clearer understanding of the burden and cost of TB and can inform national TB control strategies to improve mortality [27].

Our study has several limitations. First, we only described and compared characteristics of predefined subsets of TB mortality cases, and data of TB survivors were unavailable for the statistical analysis. The WHO states that for country-specific purposes, deaths due to TB and deaths due to other causes can be separated in the treatment outcomes section [16]. In South Korea, both TB-related and non-TB-related deaths are reported in the KNTSS. Although our results cannot be extrapolated to the general TB population, they remain valuable for the purpose of assessing the national TB status and planning necessary interventions. Second, those who visited PPM hospitals before death were included for this analysis, which also limits the generalizability. More information regarding the characteristics and treatment outcomes of TB patients managed in non-PPM hospitals are essential to reducing overall TB mortality. Third, we extracted data regarding the causes of death from case report forms completed by TB nurses at PPM hospitals, who identify the causes by reviewing and comparing two available sources (medical charts and death certificates). However, in routine clinical practice, it is hard to identify the correct mode of death, especially for elderly individuals. We could not implement a robust verification method to validate the information regarding the mode of death, because of the large cross-sectional design of the study. This may have caused a potential misclassification bias. In addition, cases of out-of-hospital mortality are collected by interviewing the guardian. The mode of death was unknown or forgotten in 10% of cases and resulted in the underestimation of TB-related death. Fourth, because clinical data for this analysis were collected from a database designed for TB surveillance, we could not obtain all clinical data related to death, such as laboratory findings and vital signs. A detailed history of TB diagnosis and treatment were not captured. For example, non-adherence to anti-TB drugs and delays to diagnosis and treatment were known to be associated with TB death [28]. Because adverse reactions to anti-TB drugs are distinct concerns for elderly patients with TB [29], further analysis investigating adverse drug reactions and mortality could be helpful to elucidate the high proportion of mortality among the elderly population. Lastly, because of the cross-sectional study design and unavailability of data regarding TB survivors, we could not measure outcomes over time or apply time-to-event analysis, which might have provided a more complex understanding of TB mortality.

A high proportion of TB death was observed among elderly patients, indicating that special attention to this population is necessary. Our results revealed that most of the deaths during anti-TB treatment were attributed to non-TB related causes, such as acute respiratory failure, malignancy, and cardiovascular diseases. Although a timely diagnosis may not prevent most non-TB related deaths, early treatment could reduce transmission risk, which is valuable in the public health perspective. Many TB-related deaths occurred during the intensive phase, particularly within the first month. Further studies identifying risk factors for different causes of TB death at different phases of anti-TB treatment are warranted for early identification of risk factors and targeted intervention, which may reduce early TB-related deaths. Any such intervention requires the collaboration of public and private sectors to maximize its effect.