Risk factors associated with loss to follow-up from tuberculosis treatment in Tajikistan: a case-control study

Setting

Tajikistan notified to WHO 7035 TB cases in 2011 and 6508 in 2012, of whom 604 and 694 were confirmed MDR cases. The WHO estimated incidence rate in Tajikistan was 87 per 100,000 in 2016. In 2016 a total of 6232 TB patients were registered for treatment in Tajikistan [19].

Design

We chose a matched case-control design for efficiency, since the proportion of patients LTFU is reported to be only 4%. Patients registered for treatment who became LTFU (as defined by WHO [1]) were defined as cases. Patients diagnosed with TB who did not start treatment are also part of the official WHO definition, but could not be included since they were not registered as such. For each case, two successfully treated (ST) patients following them in the same register were chosen as controls, in effect matching them only on health facility and time of registration (controls had date of registration near that of a case) rather than on the basis of individual characteristics. If any of them was not ST, the next ST patient on the list was chosen as a control.

Study population

The study population consisted only of TB patients registered for first line treatment in 2011 and 2012, who were marked as defaulters (now called LTFU) or successfully treated in the TB registers. In Tajikistan patients who migrated or moved are defined as LTFU, except when further treatment in a different location is confirmed. Therefore migrated and moved patients marked as LTFU were included as cases in the study. It should be noted that patients who move AND report in another clinic for continuation of treatment, were registered as ‘transfer out’ and have been excluded from this study. About 1% of patients had a treatment outcome registered as ‘transfer out’ in 2011.

Sampling method and size

Multistage purposeful sampling was used to maximize the number of LTFU patients found visiting a minimum number of facilities. Patients were selected from TB registers located in regional TB centres. Tajikistan has 5 administrative areas: Sughd, Khatlon, region under republican sub-ordination, Dushanbe and Gorno-Badakhshan Autonomous region (GBAO), with together 59 districts. We selected all districts that each had at least five LTFU DS-TB patients registered for treatment in 2011 [18]. These were 19 districts within 4 administrative areas (only 1 patient LTFU was registered in GBAO in 2011 and this whole region was excluded).

Data collection and validation

Data collection took place in all 34 TB centres between May and June 2014 in the 19 selected districts. Information was taken from TB registers and individual patient records, while facility medical staff were consulted when additional information was needed or when information in registers contradicted patient records.¹ The risk factors measured in the study were those registered in the patient records. Factors such as homelessness and alcoholism had no systematic definition in the patient records, and were based on the observations made by TB clinicians.

An electronic data collection form was developed using the Russian translation of EpiData [20] to obtain information from the TB registers TB03/TB0 3 U, TB01/TB0 1 U. [http://www.who.int/tb/dots/r_and_r_forms/en/], patient records and laboratory registers from TB centres. The tool enabled the direct entry of data extracted from paper documents in an electronic database. To ensure quality, the data entry form was tested in a pilot by two of the researchers, who also participated during the actual data collection to enter data.

Data have been collected by trained study clinicians and epidemiologists in 4 teams of 2 persons each. During the course of data collection the teams held contact via phone in order to solve problems concerning the use of different codes and other queries, ensuring consistency between the teams.

We also asked medical staff (the treating clinician and nurse, whoever knew most about the patient) what was the reason for LTFU according to the health care provider, for each individual patient LTFU.

Data analysis

For data cleaning and analysis the database was exported from EpiData into SPSS 21 for Windows (SPSS, Inc., Chicago, USA). First univariate conditional logistic regressions were run to analyse associations between collected variables and LTFU as dependent variable. Since each case had more than one control the Cox regression command was used to perform conditional logistic regression. Variables with significant associations in the univariate models were included in a multivariate conditional logistic regression model using the likelihood ratio with backward selection of independent variables. 95% confidence intervals (CI) were calculated for all variables included in the univariate and multivariate regressions.

Permission and informed consent

Official permission to conduct this study was issued by the National Ethics Committee of the Ministry of Health (MoH) Tajikistan, and all facility directors gave permission to participate prior to data collection. Data from TB registers, patient medical records and laboratory registers could only be matched by patient names, which were deleted immediately after linking was completed.

Selection and characteristics of study population

The majority (69%) of ST patients were smear-negative at end of treatment, the others had unknown smear status. Among those LTFU, the proportion of those who were lost to follow up from TB treatment during the intensive treatment phase (34%) was lower than during the continuation phase (59%), the remainder was unknown.

Factors associated with loss to follow-up

Treatment phase during LTFU

LTFU occurred more often in the continution phase (58.8%) than in the intensive phase of treatment (34.2%). Conversely, an older study (2007) on default in Uzbekistan found that the majority of LTFU cases interrupted treatment in the intensive phase (62%) [10]. At the time of our study, Tajikistan was more advanced in decentralizing services to primary health care facilities than was Uzbekistan; therefore, patients in Tajikistan may have stayed on treatment longer.

Risk factors

We found the main risk factor for LTFU in Tajikistan was migration to another country. To our knowledge, this has not been reported before, except in Moldova, where spending more than three months abroad within the past year was a risk factor for LTFU. For example a similar study in Uzbekistan did not find this [10]. Many Tajiks migrate to Russia for work. Often when during migration they are diagnosed with TB, they are sent home for treatment. TB treatment in Tajikistan means a substantial drop in financial support for families due to lost wages. Although TB treatment and diagnosis should be free of charge, certain diagnostic procedures and additional medication accompanying TB treatment often have to be paid out of pocket. This is a major burden for patients and their families and increases the motivation to leave the country for work again even if patients have not successfully completed treatment [21].

In contrast to the situation in Central Asia, studies globally found a relationship between being an immigrant coming from another country and LTFU [13, 23–28]. Most of those studies were conducted in European Union (EU) member states or USA where over a third of the reported TB cases are immigrants. In combination with our study, this shows that migration - regardless whether out of or into a country or within a country- may inhibit completion of TB treatment.

For the variable patient category, being previously treated showed a significant positive association with LTFU, as was found elsewhere [14, 15, 17, 26]. This means that patients who already received TB treatment were more likely to be LTFU from subsequent treatments, including patients with a relapse, former treatment failure, after LTFU, and transferred patients.

Our finding that side effects were associated with loss to follow-up, was also found by others [15]. Most studies do not report on an association between LTFU and side effects, possibly since side-effects are often not recorded in a standard way in TB records [4]. An association with LTFU is expected for individuals taking MDR-TB drugs since second line drugs are more toxic and more likely to cause severe side effects.

Many studies have found other factors significantly associated with LTFU, such as alcohol use and drug addiction [12, 14, 16, 22, 29–31]. We could not confirm these, possibly since the factor migration was so strong in our study population. Another reason may be that these factors may not have been noted accurately since alcohol and drug abuse are not socially acceptable in the local culture. Even though every patient diagnosed with TB is tested for HIV as of 2010, many physicians do not note this in the TB registers [32].

Another risk factor that was found to be significant in many other studies but not significant at all in ours was hospitalization versus ambulatory treatment [10, 30, 33]. Hospitalization during intensive phase is common practice in Tajikistan. Possibly the result of the analysis is related to a clear pre-selection of patients to be hospitalized in Tajikistan since 2011–2012. Alternatively, we may not have measured the actual risk factor, which is: being released from hospitalization to ambulatory care without proper follow-up [10, 30].

Reasons for LTFU given by medical staff

The factors mentioned by the clinical staff, such as stigma, family problems and patient refusing to take more treatment, were within FSU countries also found in Uzbekistan and Armenia [9, 10, 15]. We also attempted in-depth interviews in a later cohort that was LTFU and could interview only 20 of the patients loss to follow-up. Half of those patients still had TB and 6/20 thought they had completed treatment. Although these numbers low, they indicate that there may be communication issues between medical staff and patients on treatment completion.

Strength and limitations

The strength of this study was that it was the first in its kind reported from Tajikistan, and the only recent one from the central Asian region, included a large number of patients and is largely generalizable to the country. A limitation was that only variables available in medical records could be analyzed. Further drug sensitivity testing was not standard yet at time of the study. Lastly, it is possible that answers from medical staff on reasons for LTFU were influenced by recall bias.