Untreated depression and tuberculosis treatment outcomes, quality of life and disability, Ethiopia

Abstract Objective To investigate the association between comorbid depression and tuberculosis treatment outcomes, quality of life and disability in Ethiopia. Methods The study involved 648 consecutive adults treated for tuberculosis at 14 primary health-care facilities. All were assessed at treatment initiation (i.e. baseline) and after 2 and 6 months. We defined probable depression as a score of 10 or above on the nine-item Patient Health Questionnaire. Data on treatment default, failure and success and on death were obtained from tuberculosis registers. Quality of life was assessed using a visual analogue scale and we calculated disability scores using the World Health Organization’s Disability Assessment Scale. Using multivariate Poisson regression analysis, we estimated the association between probable depression at baseline and treatment outcomes and death. Results Untreated depression at baseline was independently associated with tuberculosis treatment default (adjusted risk ratio, aRR: 9.09; 95% confidence interval, CI: 6.72 to 12.30), death (aRR: 2.99; 95% CI: 1.54 to 5.78), greater disability (β: 0.83; 95% CI: 0.67 to 0.99) and poorer quality of life (β: −0.07; 95% CI: −0.07 to −0.06) at 6 months. Participants with probable depression had a lower mean quality-of-life score than those without (5.0 versus 6.0, respectively; P < 0.001) and a higher median disability score (22.0 versus 14.0, respectively; P < 0.001) at 6 months. Conclusion Untreated depression in people with tuberculosis was associated with worse treatment outcomes, poorer quality of life and greater disability. Health workers should be given the support needed to provide depression care for people with tuberculosis.


Introduction
Tuberculosis is the principal cause of death due to infectious disease worldwide; 1 it accounts for 2.0% of the global disease burden, as measured in disability-adjusted life-years. 2 In Ethiopia, tuberculosis is the fourth highest contributor to the disease burden. 3 The World Health Organization's (WHO's) End-TB Strategy, launched in 2015, aims to achieve a treatment success rate of 90% by 2030 in all people with tuberculosis, including those with multidrug-resistant disease. 1 People with tuberculosis often suffer from depression, [4][5][6] which can reduce the likelihood of successful tuberculosis treatment, 7 impair functioning 8 and decrease quality of life. 9 Systematic reviews have shown that depression is associated with poor medication adherence in people with human immunodeficiency virus (HIV) infections and acquired immune deficiency syndrome (AIDS). 10 Moreover, in chronic noncommunicable diseases, depression has been observed to lead to poor treatment adherence and to lower immunity through neuroendocrine and behavioural mechanisms. 11,12 These mechanism may also have a detrimental effect on responses to tuberculosis treatment.
Evidence on the impact of comorbid depression in tuberculosis is scarce. 7 Although a few studies have assessed the association between depression and adherence to antituberculosis treatment, they were limited by small sample sizes of less than 70 patients. One study did analyse the relationship between depression and death or treatment discontinuation in people with tuberculosis, 13 but we were unable to identify any study that disaggregated these outcomes. Other studies assessed disability 14 and quality of life 15 cross-sectionally or investigated changes from baseline in these variables after tuberculosis treatment. 8 However, they did not evaluate the impact of comorbid depression, which is known to be an important cause of disability and poor quality of life in people with chronic disorders. 11 Although global plans to end tuberculosis stress that both a patient-centred approach and social support are important for maximizing the treatment success rate, specific recommendations for people with comorbid depression is lacking. 1 In addition, in low-income countries like Ethiopia, there are large gaps in treatment for mental health problems in general and for depression in particular. 16 However, renewed efforts are being made to improve the detection and treatment of depression in primary health-care settings through WHO's Mental Health Gap Action Programme (mhGAP). 17 Greater understanding of the effect of untreated depression on the management of diseases important for public health, such as tuberculosis, is vital and would help ensure holistic care.
The aim of this study was to examine the impact of comorbid depression on treatment outcomes in people with tuberculosis in Ethiopia and on their health-related quality of life and level of disability. and 12 were health centres. Facilities were eligible for inclusion if they had staff trained in mhGAP, including the detection and treatment of depression. We recruited study participants within 1 month of starting antituberculosis treatment and who: (i) were aged 18 years or older; (ii) had no plans to move out of the study area; (iii) were well enough to be interviewed, as judged by the interviewer or prospective participant; (iv) had not been an inpatient for more than 5 days in the previous month; and (vi) had not been diagnosed with multidrug-resistant tuberculosis. Between 23 December 2014 and 4 February 2015, we consecutively invited people who fulfilled these criteria to participate in the study by health professionals running tuberculosis clinics at the study centres. Trained nurse research assistants provided those willing to participate with detailed information and obtained written informed consent or witnessed a thumb print at data collection. In Ethiopia, all people with newly diagnosed tuberculosis are treated using the directly observed treatment, short course (DOTS) approach: a combination of rifampicin, ethambutol, isoniazid and pyrazinamide is administered for the first 2 months and, subsequently, a combination of rifampicin and isoniazid is given for an additional 4 months. 18

Study variables
The primary outcome variable was treatment default: a patient who defaulted was defined by the Ethiopian Federal Ministry of Health as one "who has been on treatment for at least four weeks and whose treatment was interrupted for eight or more consecutive weeks". 18 The timing of treatment default was taken to be midway between the last successful attempt to contact the person and the first unsuccessful attempt. Other treatment variables were treatment success, treatment failure and death due to any cause. Treatment was defined as successful if either the patient was cured (i.e. the sputum smear or culture became negative during, or in the last month of, treatment) or the treatment course was completed. Treatment was defined as having failed if the sputum smear or culture was positive 5 months or later after the start of treatment or a multidrug-resistant strain was present, irrespective of sputum smear or culture findings. Data on these variables were obtained from each centre's tuberculosis register, which did not contain information after the time of referral on patients who were transferred to another area. The study protocol has been published elsewhere. 19 The secondary outcome variables were quality of life and disability, which were assessed on three occasions over 6 months: (i) at baseline, at the start of the intensive treatment phase; (ii) at 2 months, after completion of the intensive treatment phase; and (iii) at 6 months, after completion of all tuberculosis treatment. Quality of life was assessed from responses to the question, "How would you rate your healthrelated quality of life?" and scored from zero for worst imaginable to 10 for best  : 3 to 14), with a high score indicating better perceived social support. 32 d Tuberculosis stigma was assessed at 2 months in 592 participants using a 10-item scale (range: 10 to 50), on which a high score indicated a high level of stigma.

Research
Depression and tuberculosis treatment in Ethiopia Fentie Ambaw et al.
imaginable. 20 Such single-item methods have been used successfully in population surveys, clinical settings and clinical interviews and found to be valid in indicating vulnerability to death due to all causes. 21,22 No validated, tuberculosisspecific, quality-of-life instrument is available. Disability was assessed using the interviewer-administered version of the 12-item WHO Disability Assessment Schedule, version 2.0. 23 This tool has been shown to be useful for assessing disability in primary care patients with depression and is able to capture changes over time. 24,25 Moreover, it has been validated in Ethiopia and showed convergent validity with other predictors of impaired functioning in people with depression. 26 At the three assessments, health professionals asked respondents if they were being treated for any mental illness, including depression.
Our exposure variable was probable depression which was identified using the nine-item version of the Patient Health Questionnaire and defined conservatively as a score of 10 or above. 27,28 The nine-item version has been validated in two different treatment settings in Ethiopia. 28,29 In the baseline assessment in our study, this version of the questionnaire was found to have construct validity and acceptable internal consistency, with an α of 0.81 and a mean inter-item correlation coefficient of 0.33. 6 Participants who responded positively to the questionnaire item on suicidal ideation were referred for evaluation and treatment to health workers who had received training in mental health care as part of WHO's mhGAP. 30 We took into account a range of possible confounding variables such as age, sex, educational level, household income, marital status, religion, ethnicity and place of residence (i.e. urban versus rural). Data on these variables were obtained at baseline using a structured questionnaire. The duration of tuberculosis symptoms before diagnosis was self-reported by participants at baseline and information on the type of tuberculosis infection (i.e. pulmonary or extrapulmonary) was obtained from tuberculosis registers. The presence of any diagnosed comorbid chronic illnesses was also reported by participants themselves and whether or not they had an HIV infection, was recorded in tuberculosis registers. Use of substances, such as alcohol, tobacco and khat, was assessed using WHO's Alcohol, Smoking and Substance Involvement Screening Test, version 3.1. 31 Each participant's perceived level of social support was assessed at baseline using the three-item Oslo-3 scale, which ranges from 3 to 14, with a high score indicating better perceived social support. 32 This scale has previously been reported to work well in tuberculosis patients in Ethiopia. 33 In addition, we assessed the stigma of tuberculosis at 2 months by adapting a 10-item tuberculosis stigma scale 34 translated into Amharic; 19 a high score indicated a high level of stigma. We also assessed participants' perceptions about tuberculosis at 2 months: perceived tuberculosis severity was categorized as mild, moderate or severe; tuberculosis treatment was perceived as not helpful; somewhat helpful; or very helpful; and perceived barriers to tuberculosis treatment were identified from a yes or no answer to the question: "Are there barriers to taking your medications as prescribed?"

Data analysis
Study variables and the participants' characteristics are presented using descriptive statistics. We estimated the association between probable depression at baseline and treatment default, treatment success and death using multivariate Poisson regression analysis with a robust variance estimator and present the results as risk ratios. 35 The follow-up time was included as a weighting variable in the analysis of these outcomes. We did not perform multivariate analysis for treatment failure because there were only six cases. We assessed differences in quality-of-life and disability scores between participants with and without probable depression at baseline and at 6 months using the independent samples t test and the Mann-Whitney U test. To examine the change in health-related quality-of-life and disability scores between tuberculosis diagnosis and the end of antituberculosis treatment, we used a multilevel, mixed-effects, generalized linear model to fit data from the three measurement times (i.e. baseline, 2 months and 6 months), with the three measurement times nested within individuals and individuals nested within each of the 14 primary care centres. The analysis was performed using Stata   36 We calculated the study sample size using Stata version 12.0 for a power of 80%, a confidence level of 95% and an estimated prevalence of treatment default among tuberculosis patients without depression of 2.5%. 37 In addition, the sample size had to be sufficient to detect a 5.0 percentage point increase in the prevalence of treatment default among people with comorbid depression when the ratio of nonexposure to exposure to depression was 2:1. With these parameters, the required sample size was 639, which was increased to 703 to include a 10% contingency for potential losses to follow up. The study was approved by the Institutional Review Board of the College of Health Sciences of Addis Ababa University (number 027/14/Psy).

Results
In total, 657 people were recruited. However, as 9 people were subsequently found to have been misdiagnosed with tuberculosis, the study analysis included data on only 648. Participants' ages ranged from 18 to 85 years and 53.7% (348/648) were male ( Table 1). The median time between starting tuberculosis treatment and the first assessment was 0 days (interquartile range, IQR: 2). Face-to-face follow up assessments were conducted with 91.4% of those with tuberculosis (592/648) at 2 months and 82.1% (532/648) at 6 months. The median time from the start of antituberculosis treatment to the second assessment was 56 days (IQR: 1) and the median time to the third assessment was 160 days (IQR: 3). Data on treatment outcomes at 6 months were available for 88.7% (575/648) of participants. Overall, 9.6% (62/648) were transferred out of study sites (Fig. 1). At baseline, 53.9% (349/648) scored 10 or higher on the nine-item Patient Health Questionnaire and were classified as having probable depression ( Table 2) here was no significant difference at baseline between those who completed the study and those who were transferred out in the frequency of At 6 months, the treatment default rate was significantly higher among participants with probable depression at baseline than among those without: 3.9% (12/309) versus 0.8% (2/266), respectively (P < 0.05). Similarly, the proportion who had died was significantly higher among those with probable depression: 7.8% (24/309) versus 1.9% (5/266) in those without (P < 0.01). In addition, the treatment success rate was significantly lower in those with probable depression: 87.1% (269/309) versus 96.6% (257/266) in those without (P < 0.001; Table 3). On multivariate analysis, treatment default by 6 months was independently associated with probable depression (adjusted risk ratio, aRR: 9.09; 95% confidence interval, CI: 6.72 to 12.30), as was death (aRR: 2.99; 95% CI: 1.54 to 5.78). However, there was no significant association with treatment success (aRR: 0.95; 95% CI: 0.91 to 1.00), though the upper confidence bound was borderline for significance (Table 4).
Probable depression at baseline was also associated with quality of life (Fig. 2) and disability (Fig. 3). The mean quality-of-life score at baseline was lower among those with probable depression than among those without (4.7 versus 5.7, respectively; P < 0.001) and the median disability score was higher (30.0 versus 18.0, respectively; P < 0.001). These differences remained significant at 6 months, when the mean quality-of-life score among those with and without probable depression was 5.0 and 6.0, respectively, (P < 0.001) and the median disability score was 22.0 and 14.0, respectively, (P < 0.001; Table 3). On multivariate analysis, quality of life at 6 months was significantly and negatively associated with probable depression (β = −0.07; 95% CI: −0.07 to −0.06) and disability was positively associated (β = 0.83; 95% CI: 0.67 to 0.99). There was no significant change over time in either mean quality-of-life score (β = −0.02; 95% CI: −0.13 to 0.09) or median disability score (β = −0.07; 95% CI: −0.33 to 0.22; Table 5). Data on factors other than depression that were associated with tuberculosis treatment outcomes are shown in Table 4 and data on factors associated with quality of life and disability are shown in Table 5.

Discussion
Our study provides evidence that people with tuberculosis in Ethiopia who had probable depression at the start of treatment were significantly more likely to default on treatment or die. Moreover, their chance of successful treatment was lower. Previous studies have also reported that depression compromises adherence to essential scheduled health care. 11 With tuberculosis, treatment default leads to transmission of the infection to others, thereby raising the odds of further defaults, 38 and increases the risk of multidrug-resistant disease. 39 In agreement with our observations, systematic reviews and large populationbased studies in both high-and low-income settings have found that mortality is increased in people with depression and that the association is maintained across patient groups. [40][41][42][43] However, depression is a more serious concern for people with tuberculosis in Ethiopia because comorbid depression has been found in the majority. 6 The mechanism by which depression increases mortality is likely to be complex. Although 113 of our 648 study participants reported suicidal ideation, we were not able to confirm its contribution to the mortality observed. One systematic review found that suicide contributed to less than 1.0% of deaths in medical samples like ours. 41 Moreover, in our study those who refused to be tested, were transferred to another area or whose diagnosis changed were excluded. b Quality of life was assessed using the question, "How would you rate your health-related quality of life?" and scored from 0 (worst) to 10 (best). c Quality-of-life and disability were assessed in 648 participants at baseline and in 532 at 6 months. d Disability was assessed using the interviewer-administered version of the 12-item World Health Organization Disability Assessment Schedule, version 2.0 (score range: 0 to 60). 23 Depression and tuberculosis treatment in Ethiopia Fentie Ambaw et al.

Research
Depression and tuberculosis treatment in Ethiopia Fentie Ambaw et al.
area, the commonest cause of death in people with severe mental illness is infectious disease. 44 In people with tuberculosis, depression may increase mortality through decreased self-care, including failure to take medications as prescribed, 11 and through disability leading to poverty and substandard living conditions. One possible biological mechanism is depression-associated immune suppression. 45 We also found that probable depression in people with tuberculosis was associated with poorer quality of life and greater disability, both at the start and after completion of antituberculosis treatment. In previous studies, neither quality of life nor the degree of disability returned to levels normal for the population by the end of tuberculosis treatment. 8,46 Possible explanations are underlying depression, the quality of tuberculosis care falling short of international standards and the socioeconomic consequences of the illness and its associated stigma. 47,48 One implication of these findings is that evaluating disability only during episodes of tuberculosis is likely to underestimate the disease burden as continuing disability after clinically successful treatment would be ignored. 49 The treatment default rate we observed was markedly lower in females, rural residents and people living with HIV. Treatment adherence may have been stronger in people with an HIV infection, because of the counselling and additional support given to them in the Ethiopian health-care system, particularly in rural communities. 50 Treatment default was also associated with higher khat use, substance use has previously been found to reduce the tuberculosis treatment success rate. 51 In Ethiopia, the implementation of integrated care for people with mental, neurological and substance use disorders does not include khat use disorder as a target condition, because little is known about its adverse consequences and, because practical interventions are lacking. The level of disability was higher in females, in participants who perceived their episode of tuberculosis as severe, in those with pulmonary rather extrapulmonary tuberculosis and in those with a high level of stigma, indicating that both physical and psychosocial factors may contribute to the development of disability. 52  Although we found that comorbid depressive symptoms in people with tuberculosis are often associated with poorer outcomes, even with successful treatment, generally health-care providers in Ethiopia do not assess depression in these people or provide evidencebased treatment. Consequently, unnoticed comorbid depressive symptoms may hamper efforts to end tuberculosis. National tuberculosis treatment guidelines may need to address depressive symptoms directly and health professionals should be trained to detect and treat depression in the context of the disease.
Our study has several of limitations. In our sample size calculation, we assumed that the tuberculosis treatment default rate in people without depression was 2.5%, which was based on a national report. We found a rate of 0.8%. However, we were still able to obtain estimates even with this sample size and do not believe it critically affected our findings. Second, in the low-income setting of our study, participants could have had undiagnosed, comorbid physical illnesses. Third, as we did not know whether or not participants were treated for depression outside the study centres, we may have overestimated the frequency of untreated depression. Nonetheless, as few people with depression receive treatment in Ethiopia, it is unlikely that misclassification of untreated depression seriously affected our results. Fourth, poverty may not have been fully captured by our sociodemographic variables and may have been a confounding factor. Fifth, as quality of life was assessed using a single question, no detailed information on different dimensions of quality of life was available. Sixth, we had no information on whether participants transferred out of the study area differed significantly from others in treatment outcomes or final quality-of-life or disability scores. However, there were no differences at baseline. Finally, our conclusions cannot be extended to tuberculosis patients who are hospitalized, are being retreated, or have multidrug-resistant disease.
Nevertheless, consecutive patients were recruited and our study sample was reasonably representative, because several sites were included, all eligible people were invited to participate and data were collected over a long enough period to take seasonal variations into account. 53 Use of the DOTS approach

Time from start of tuberculosis treatment (months)
Without probable depression at baseline With probable depression at baseline baseline 2 6 Note: Disability was assessed using the interviewer-administered version of the 12-item World Health Organization Disability Assessment Schedule, version 2.0 (score range: 0-60). 23 Research Depression and tuberculosis treatment in Ethiopia Fentie Ambaw et al.
Depression and tuberculosis treatment in Ethiopia Fentie Ambaw et al.