In 2008, there were an estimated 8.9-9.9 million incident cases of TB, 9.6-13.3 million prevalent cases of TB, an estimated 1.3 million (range, 1.1-1.7 million) deaths, including 0.5 million (range, 0.45-0.62 million) deaths among women, occurred among HIV-negative incident cases of TB, and there were an estimated 0.5 million deaths among incident TB cases who were HIV-positive [1]. In 2005, 15.4 million TB cases were reported globally, among 8.8 million people new TB cases, 3.9 million were smear positive, and 1.7 million people died of TB in the same year [2]. Ninety eight percent of TB deaths occur in the developing countries and if left unchecked, with in 20 years, TB will kill about 35 million people [2]. WHO developed the DOTS strategy as the internationally recommended approach to TB control in the mid-1990s. DOTS is also the foundation of the Stop TB Strategy, which was launched by WHO in 2006 to guide TB control efforts during the period 2006-2015. The start of WHO efforts to systematically monitor progress in TB control on an annual basis in 1995 coincided with global promotion and expansion of the DOTS strategy [1].

In Ethiopia, TB has long been recognized as a major public health problem since the 1950s [3] and the country has been implementing the WHO recommended DOTS (Directly Observed Treatment Short-course) strategy since 1992[4]. At present, TB control strategy in Ethiopia relies on WHO recommended Stop TB Strategy and it has been implemented in the country since 2006[5].According to the 2008 WHO TB report [4], Ethiopia ranks 7th in the list of the world's 22 high burden countries for TB with incidence estimated at 379/100,000 for all forms of TB and 168/100,000 for smear positive TB. According to the Ministry of Health hospital statistics data, TB mortality rate is estimated at 84 per 100,000 populations per year [3].

TB mortality has been variously attributed to a dramatic rise in drug resistance [6] and to deteriorations in TB services [7, 8] and the risk of dying could be reduced through improved care [9, 10]. So to avert the problem of TB mortality, policy makers should devise a plan, with clear reflection on the magnitude of mortality and associated risks factors of TB, to improve the quality and standard of TB care. The objective of this paper was to identify the magnitude and risk factors associated with death among a cohort of TB patients treated under DOTS programme in Addis Ababa, Ethiopia.

WHO defines TB mortality as the number of TB cases dying during treatment, regardless of the cause [13]. The mortality in our study was higher than in china [14] but lower than in those reported in previous studies including 24% in Baltimore City, USA [15], 14% in Vaud County, Switzerland [16] and similar to the mortality found in Addis (4%) and to the National (3%) [12]. The Mean and median duration from the onset of treatment to death in the current study were similar to a study conducted in south India, where by 94-97% of patients were surviving at the end of seven months of anti-TB treatment, irrespective of TB categories [17]. A study conducted in Chennai city [18] showed less survival rate (91%) compared to our study. This difference may be due to long follow up period in Chennai study and different number of study subjects. In other studies, 6% of the 676 patients died during the course of treatment; all patients who died had pulmonary TB [19]. Of the 39 patients who died, 65.0% died within 2 months of the start of treatment [19] and in other study, the survival probabilities were 97.5%, after the first month of treatment, 96% after the 3 months, 95% after 6 months and 93% after 1 year [20]. Unlike to our study, other studies have also found that a significant proportion of patients died in the early stages of treatment [21, 22]. Early mortality reflects advanced disease and could be attributed to delayed treatment and late diagnosis [22, 23]. In Vietnam study, similar to our finding, among 27 deceased patients the median time interval from treatment initiation to death was 8 months (IQR 5 to 11 months) [24]. Other studies demonstrated high risk of death during the first 2 months of anti-tuberculosis therapy [25–27]. Other study similar to our study showed the risk of mortality increased substantially among all subgroups from the second six months to the final eight months; hence it is likely that these deaths were due to drug sensitive or drug-resistant relapse [18].

In our study, general mortality rate per 100 PYO was 6.3/100(6.3%) per annum for the cohort which is similar to other studies. The general mortality rate for rural south India cohort was 61.0/1000 person-years or 6.1% per annum[28], which was not different from the mortality rate of 6% per annum reported for the Chennai urban cohort for the year 2000[18] but higher than a study in southern Ethiopia which shows mortality per 100 PYO was 2.5% per annum [29].

In our study baseline body weight was significantly associated with higher death. Similar finding was observed in south India, in which body weight at initiation of anti-TB treatment (<35 kgs) was a significant risk factors of death during anti-tuberculosis treatment period [17, 19]. Other study also reported that among persons who were underweight at diagnosis, weight gain of 5% or less after two months of treatment was associated with an increased risk of relapse [30]. Similarly, low BMI was significant risk factor for mortality in patients with miliary tuberculosis (MTB) and lower BMI was found in patients who were died [31]. It may be assumed that malnutrition would be a clinical finding reflecting the severity of pulmonary TB (PTB) and that the determinant of in-hospital mortality would be due to PTB severity rather than malnutrition itself [32]. Underweight was associated with a 10-fold increase in TB mortality [33].

In our study treatment category was associated with higher death particularly previously treated patients. The characteristics of the TB patients associated with mortality during anti-TB treatment was being in retreatment case [34].Previously reported risk factors of mortality among tuberculosis patients include advanced human immunodeficiency virus (HIV) infection [35, 36], multidrug resistance [36–38], and irregular treatment [39]. Similarly, multidrug resistance may explain the high combined failure and death rate of 55% among patients with transfer-out [24]. But in one study the mortality amongst the patients with previously treated TB was not statistically different [40]. This might be due to drug resistance cases at hospital level whereby most patients transferred or associated immune suppression at the beginning of the treatment and HIV prevalence may therefore be higher among patients in retreatment cases compared to patients with reported treatment success or these patients may not recover during treatment and may be transferred without well established referral system.

Our study is not without limitations. The data were extracted from medical records of those already visited and registered at the respective health facilities; it may be subjected to selection bias. In many tuberculosis patients, multiple causes of death may act simultaneously, so the cause of death may not be determined accurately [41, 42]. In Norway this misclassification was shown to be considerable [43]. In the present study, the analyses included all deaths, irrespective of the cause of death, so misclassification of the cause of death might not have major influence on the results and as the nature of secondary data, some of the data are incomplete to gather the necessary variables such as HIV infection, patient's compliance, use of Directed Observed Treatment, use of ART, CPT and other prophylaxis, co-morbidity, drug resistance pattern. It is possible that some patients hidden their history of previous anti-tuberculosis treatment at the time of diagnosis and were inappropriately classified under different category. Despite the above limitations, this study was conducted at health centers that represent different level of potential health service coverage in sub-cities, which in turn reflect the TB patient under DOTS program at primary health care level in Addis Ababa and inclusion of several sites for comparison, and to get higher sample size to elucidate on the matter.

Higher death rate was noted in patients with low baseline weights and patients who are previously treated cases. Additionally, the death rate was varied across year of enrollment. Therefore, reinforcing the existing DOTS program and nutritional support for underweight patients for better clinical and treatment outcome. Strengthen the recording system for completeness of the registry which is a common problem in practice and conducting further study with different design and area is needed to assess the TB mortality situation.