Trends in tuberculosis notification and mortality and factors associated with treatment outcomes in Serbia, 2005 to 2015

Background Previously a country with medium tuberculosis (TB) burden, Serbia almost reached a low TB burden during the period 2005 to 2015. Aim The aim of this study was to analyse the trends in notification rates and treatment success rates as well as to identify predictors of treatment outcomes. Methods We performed a trend analysis and logistic regression analysis of 17,441 TB cases registered from 2005 to 2015 in all health facilities in Serbia, to identify predictors of treatment success, loss to follow-up and mortality. Results From 2005 to 2015, TB notification rate and mortality in Serbia decreased but treatment success remained below the global target. Loss to follow-up was associated with retreatment (odds ratio (OR) = 2.38; 95% confidence interval (CI): 2.08–2.77), male sex (OR = 1.57; 95% CI: 1.39–1.79), age younger than 65 years (OR = 1.37; 95% CI: 1.20–1.51), lower education level (OR = 2.57; 95% CI: 1.74–3.80) and pulmonary TB (OR = 1.28; 95% CI: 1.06–1.56). Deaths were more frequent in retreatment cases (OR = 1.39; 95% CI: 1.12–1.61), male patients (OR = 1.34; 95% CI: 1.19–1.52), those 65 years and older (OR = 4.34; 95% CI: 4.00–5.00), those with lower education level (OR = 1.63; 95% CI: 1.14–2.33) and pulmonary TB (OR = 2.24; 95% CI: 1.78–2.83). Conclusions Special interventions should be implemented to address groups at risk of poor treatment outcome.


Introduction
Tuberculosis (TB) is still one of the major public health problems affecting more than 10 million people and causing 1.6 million deaths worldwide [1]. Since 1994, milestones of TB prevention and control have been the three World Health Organization (WHO) strategies Directly observed treatment short course (DOTS), Stop TB and End TB [2,3]. Implementing the WHO strategies, many countries established between 2000 and 2014 the basic requirements for providing high-quality TB diagnosis and treatment [3]. Those strategies focused on case notification and monitoring of treatment outcome as the essential measures to evaluate the effectiveness of interventions and identify potential gaps in TB control [4]. The global target is to reach a treatment success rate (TSR) of at least 85% [2]. However, the WHO's post-2015 End TB Strategy aims for at least 90% [3]. TB disproportionately affects different segments of the population. Higher male-to-female ratio was reported in some studies and surveillance data as well as higher TB notification rate and lower mortality rate among males [1,5,6]. Changes in the occurrence of the disease by age, sex, education, anatomical site of the disease, history of previous treatment, HIV status and drug resistance are also important indicators of TB control programmes [7][8][9][10][11].
In the period 2005-15, Serbia almost became a country with a low TB burden, reducing TB notification rate from 32 per 100,000 in 2005 to 13 per 100,000 in 2015 by implementing the WHO DOTS and Stop TB strategies in the National Tuberculosis Programme with financial support from the Global Fund to Fight AIDS, Tuberculosis and Malaria (GFATM). An analysis of trends and differences in TB notification and treatment outcomes by sex, age, education, occupation, TB localisation and history of previous treatment has not yet been performed in Serbia. It may help improve our understanding of the performance of the TB control programme. Previous studies in Serbia have focused on trends in TB incidence and mortality and characteristics of TB among the elderly population [12][13][14]. Table 1 Characteristics    However, these studies did not analyse factors associated with successful and unsuccessful treatment outcome, which could provide useful evidence for targeted evidence-based interventions. Therefore, the objective of this study was to assess the trends in TB notification, TB mortality rate and treatment success and to identify factors associated with treatment success, loss to follow-up and mortality among TB patients notified in Serbia in a period of 11 years.

Methods
We performed retrospective trend analysis and analysis of treatment outcomes of all (n = 17,441) TB cases notified in Serbia in the period 2005-15 from all health facilities, based on the electronic data collection of the Ministry of Health and Institute of Public Health of Serbia. The electronic case-based data collection system in Serbia was introduced in 2005 with the support of the GFATM in line with the WHO TB notification form. Standard indicators from the WHO checklist Standards and benchmarks for TB surveillance and vital registration systems were used to assess internal consistency of the data: changes over time in TB notification rate, the ratio between the number of notified pulmonary and extra-pulmonary TB cases, and the male-to-female ratio of TB cases [15].
To describe TB epidemiology in the country, we analysed notified TB cases by age, sex, education, occupation, anatomical site of the disease, history of previous treatment, HIV status (recorded since 2010) and multidrug resistance. Treatment results included: cured, treatment completed, defaulted (lost to follow-up), died, failed and not evaluated [16,17]. National TB control guidelines for Serbia, in line with WHO guidelines, were used for TB diagnosis and case definition [16].

Tuberculosis diagnosis
TB diagnosis was based on clinical and radiological findings and was confirmed bacteriologically and/or histologically. Laboratory confirmation was performed by detection of Mycobacterium tuberculosis complex from a clinical specimen, either by culture or by a newer molecular technique [18].

Case definition and classification
Pulmonary tuberculosis (PTB) refers to any bacteriologically confirmed or clinically diagnosed case of TB involving the lung parenchyma or the tracheobronchial tree. PTB also includes cases that affect the lung with additional extra-pulmonary manifestations.
Extra-pulmonary tuberculosis (EPTB) refers to any bacteriologically confirmed or clinically diagnosed case of TB involving organs other than the lungs, e.g. pleura, lymph nodes, abdomen, genitourinary tract, skin, joints and bones or meninges.
New patients are defined as those who have never been treated for TB or have taken anti-TB drugs for less than 1 month.
Previously treated patients are defined as those who have received 1 month or more of anti-TB drugs in the past.
Multidrug resistance refers to resistance to at least isoniazid and rifampicin together.
Treatment outcomes were categorised according to the latest WHO definitions and reporting framework for tuberculosis [11].

Statistical analysis
Descriptive statistics of trends of TB notification, TB mortality rates and treatment outcomes during the 11-year study period were performed for all PTB and EPTB cases. Population data for each year were obtained from the National Statistical Office [18]. We analysed trends of case notification by age, sex, education, occupation, anatomical site of the disease, history of previous treatment, HIV status and first-line drug susceptibility.
A logistic regression analysis was used to identify factors associated with treatment success, loss to followup and mortality. We used three separate models. The variables age, sex, education, occupation, anatomical site of the disease, history of previous treatment, HIV status and first-line drug susceptibility test (DST) were entered in each univariate logistic regression analysis (ULRA) model. For the first model, the outcome was dichotomised as treatment success (cured or completed) vs unsuccessful treatment (failed, died, lost to follow-up or not evaluated). For the second model, the outcome was dichotomised as lost to follow-up vs follow-up (treatment success, failed, died, or not The number of laboratory-confirmed MDR TB cases enrolled in second-line TB treatment. The difference between registered patients and patients enrolled in treatment is due to backlog of patients from previous years. evaluated). For the third model, the outcome was dichotomised as died vs not died (treatment success, failed, lost to follow-up or not evaluated).
The variables that showed statistically significant association at p value ≤ 0.05 in ULRA were entered in multivariable logistic regression analysis (MLRA) models. Data were analysed using the Statistical Package for Social Sciences (IBM SPSS) version 24 and p values ≤ 0.05 were considered statistically significant.

Ethical statement
This study was approved by the Ethics Committee of the Institute of Public Health of Serbia (No 24/2016). Personal identifiers of TB cases were coded to maintain the confidentiality of the patient information before analysis, and patient records or information were anonymised and de-identified before analysis.

Internal consistency indicators
The mean percentage of annual changes of TB notification rate from 2005 to 2015 was 8.25%. During the same period, the ratio of nationally notified PTB vs EPTB cases ranged from 6.04 to 8.09, with a slight decrease in PTB cases from 2013 to 2015. The maleto-female ratio of notified TB cases was generally consistent (range: 1.40-1.62), but there was a notable increase in the proportion of male cases over time.

Trends in treatment outcomes
Among the 17,441 notified cases, treatment outcome was available for 17,388. Over the 11-year period, there were no significant changes in the treatment outcomes of drug-susceptible TB (   MDR TB cases were included in each of the three univariate models. The variables that showed a statistically significant association at p value ≤ 0.05 in ULRA were entered in multivariable models. MDR TB was not statistically significantly associated with treatment success (model 1), loss to follow-up (model 2) or mortality (model 3) and was therefore not included in the models of MLRA.

Factors associated with treatment outcomes
In the multivariable logistic regression analysis (

Discussion
From 2005 to 2015 in Serbia, annual changes in the TB notification rate were less than 10%. The ratio of PTB to EPTB cases notified nationally ranged from 6.04 to 8.09, with a slight decrease in PTB cases from 2013 to 2015, suggesting that these data were internally consistent [16]. The male-to-female ratio increasing over time may suggest women were less likely in the later years than in the earlier years to seek healthcare and/ or were underdiagnosed when they did seek care. In addition, it may indicate that epidemiologically, there was a real increase in men (or decrease in women) who developed TB [20].
Over the 11-year study period, we found a decrease in TB notification and mortality rates and a stable trend in treatment outcomes and notification rates per age category, education, occupation, anatomical site of the disease and history of previous treatment. The dominant proportion of male TB cases corresponds to the global surveillance data [1] and surveillance data for Europe [5], with the male-to-female ratio ranging from 1.5∶1 to 2∶1.
We found that the proportion of PTB among TB patients in our study was 20% higher than the proportion of PTB among TB patients globally [1] and in the European Union/European Economic Area (EU/EEA) [5]. Solovic et al. reported that the percentage of EPTB cases in the EU/EEA) in 2011 ranged from 4% to 48% [21]. The differences may be explained by different risk factors for EPTB or by diagnostic challenges as EPTB can manifest with a variety of symptoms. Low levels of suspicion of clinicians as well as difficulties in obtaining adequate samples for confirmation are also reported as a challenge in EPTB diagnosis [21]. In addition, some of the reasons for higher proportion of PTB in Serbia compared with EU/EEA countries could be due to low HIV prevalence in Serbia [22] and to the dominant proportion of men among TB patients as there is evidence that the incidence of EPTB is likely to be higher among women than men [23].
We found that the proportion of previously treated cases was 20 percentage points lower in Serbia than globally [1]. This could indicate good implementation of the DOTS strategy, good case management, low burden of drug-resistant TB, good adherence to treatment or good access to and availability of TB services [2]. We assume that reliability of the data can be considered satisfactory because TB case definitions are clearly laid out in the guidelines for TB doctors as a part of routine TB training in the country and are consistent with the WHO guidelines. Firstly, TB cases are usually classified at the moment of filling out the TB notification form, by a physician who diagnosed TB based on medical investigations and medical records. Data from the TB notification form are further validated at the regional level by epidemiologists and pulmonologists before they are entered into a case-based electronic data collection system. A third round of data quality control is usually performed at the central (national) level in the form of automatic checks for duplicates, missing records and consistency during the process of data cleaning before data analysis; the central level receives copies of all case-based TB notification forms.
Educational status of TB patients in our study was lower than in the general population in Serbia [24]. There were no data on educational status of TB patients in Europe. Nevertheless, a large body of scientific evidence supports education as an important determinant of health and indicates that lower education is associated with poor health, stress and lower self-esteem [25]. The majority of patients in our study were employed. In most European countries, there is a greater share of unemployment among TB patients [26].
The proportion of HIV-positive TB cases was consistent with reports from the Balkan countries (excluding Romania with significantly higher TB/HIV prevalence) [1,27] and lower than in other EU/EEA countries [5]. Studies conducted in Africa and Asia found a much higher prevalence of HIV infection among TB patients (30-38%) [28,29]. The slight increase in the proportion of HIV-positive TB cases observed in Serbia was due to implementation of the WHO policy on collaborative TB/ HIV activities and increased HIV testing of TB patients. However, among drug users, homeless people, prisoners and alcoholics, the proportion of TB cases with known HIV status is still very low in Serbia compared with EU/EEA countries (68%) because HIV testing is not routinely offered to all TB patients. There is a need for continuous training of all medical and non-medical workers who deal with TB and HIV infection and for a joint TB/HIV communication and social mobilisation strategy [5,30]. Implementation of the WHO Stop TB strategy resulted in an increased DST coverage, in line with the EU countries [5]. Given that the latest WHO End TB strategy target is 100% DST coverage, this is a programmatic challenge to be addressed in the coming years. We identified very small percentages of notified TB cases among children (0.8-2% of notified cases per year were children aged 0-14 years) in relation to WHO standards and benchmarks for childhood TB (5-15%) for middle-income countries where Serbia belongs [20]. This result pointed to under-reporting because of limited sensitivity of surveillance system as well as limited capacities of the TB programme to diagnose childhood TB.
The treatment success rate in Serbia was higher than the EU treatment success rate [5] but still below the global target of 90% [3]. The minor improvements during the study period could be attributed to better overall case management (early diagnosis and treatment, reduction in loss to follow-up and mortality). Other possible reasons could be community-based patient support interventions which improved access to and use of TB control services and decreased the proportion of non-evaluated cases compared with previous years [3].
We identified previous TB treatment, male sex, lower level of education and PTB as independent predictors of loss to follow-up and mortality, consistent with many other studies [29,31]. Some studies identified unemployment [32] and migration [33] as predictors of loss to follow-up caused by poor adherence to treatment.
Although there were no foreign-born cases registered in our study, there is a possibility of under-reporting because of the lack of a sensitive surveillance system. Authors from the United Kingdom accentuated the public health implications of patients lost to follow-up, who may contribute to ongoing disease transmission and the risk of drug resistance [32]. Patients with severe forms of extra-pulmonary disease may be less likely to become lost, because they are likely to be in closer contact with the healthcare provider and may be hospitalised [34]. Weaknesses in the health system, such as lack of updated clinical guidelines, non-compliance of healthcare professionals with the guidelines, interruptions in drug supply or poor training, supervision and organisation of TB programmes, have also been identified as factors contributing to the poor management of patients, resulting in treatment interruptions and loss to follow-up [2,23].
We found age 65 years and above to be an independent predictor of TB mortality. Many studies reported higher mortality rates among elderly people [4,12]. Co-morbidities among elderly people influence the reactivation of previously acquired TB disease by reducing the immunity. Moreover, immunocompromised patients are more likely to develop severe disseminated forms of TB disease and adverse drug reactions [12,34]. Delay in diagnosis and treatment among older age groups was also identified, which could increase the risk of death. In addition, we identified age below 65 years as predictor of loss to follow-up. Patients younger than 65 years are more likely to be immunocompetent, with fewer co-morbidities. Loss to follow-up among this group of patients may be related to social barriers and perception of stigma. TB is known to be a highly stigmatised disease and many studies have identified stigma as an obstacle in cooperation and communication with the health service, resulting in delayed seeking of medical help, and an obstacle during treatment, leading to interruptions, failure, recurrence and drug resistance [34,35].

Limitations
We could not distinguish between deaths that were due to TB or deaths that were due to other diseases during TB treatment because this information was not available. It was not possible to investigate other factors associated with TB treatment outcomes, such as socioeconomic factors, tobacco use, drug or alcohol abuse, other co-morbidities or homelessness. Data on DST results were missing for half of the cases, some of whom may have had MDR TB. Despite its limitations, our study provides evidence on the risk factors associated with TB treatment outcomes in Serbia. Further analyses are needed to explore further associations.

Conclusions
In order to improve performance of the TB programme in Serbia, it is necessary to address groups at risk of poor treatment outcomes. To prevent loss to follow-up and mortality and to achieve better adherence to treatment, case management has to be improved towards providing holistic patient-centred inter-sectoral interventions and support during treatment. It is important to update the national TB prevention and control strategy and to revise existing guidelines in line with the available scientific evidence, so as to increase coverage of DST and HIV testing and improve early detection of HIV co-infection and drug resistance, the main obstacles in modern TB control.