Poverty and Substance Use in South African Tuberculosis Patients Goedele M. Louwagie, DMed, MMed; Edwin Wouters, PhD; Olalekan A. Ayo-Yusuf, BDS, MPH, PhD Objectives: To determine whether social support and depressive symptoms mediate the relationship between poverty and substance use in tuberculosis (TB) patients in South Africa. Methods: We performed structural equation modeling with the latent constructs socioeconomic status (SES), social support and “substance use” (tobacco smoking, alcohol problem, illicit drug use) in 1005 male TB patients. Results: Higher SES directly (standardized coefficient= -0.27) and indirectly reduced substance use. Indirectly SES provided increased social

S

outh Africa has one of the highest tuberculosis (TB) burdens in the world, an epidemic fuelled by the human immunodeficiency virus (HIV) pandemic.1,2 HIV-infected individuals are at an elevated risk of TB2 and an estimated 60% of TB patients in South Africa are HIV-positive.1 Furthermore, TB treatment outcomes are poor, because treatment is often interrupted or fails, and/or patients die.1 Among the underlying causes of adverse outcomes include limited healthcare, clinical and socio-economic factors, and substance use.3-7 Alcohol and smoking have been established to be risk factors for acquiring TB disease, and continued use of alcohol and tobacco products once a person has contracted TB lowers the chances of successful treatment.4,7-11 It has been shown that smoking and alcohol-related problems affect poor and disadvantaged groups to a disproportionate degree.12-16 Such disparities between social groups-

Goedele MC. Louwagie, Senior lecturer, School of Health Systems and Public Health, University of Pretoria, Pretoria. Edwin Wouters, Assistant professor, Department of Sociology and Research Centre for Longitudinal and Life Course Studies, University of Antwerp, Antwerp and Centre for Health Systems Research and Development, University of the Free State, Bloemfontein. Olalekan A. Ayo-Yusuf, Director, School of Oral Health Sciences, University of Limpopo MEDUNSA campus and Extraordinary professor, School of Health Systems and Public Health, University of Pretoria, Pretoria. Correspondence Dr Louwagie; [email protected]

Am J Health Behav.™ 2014;38(4):501-509

support (coefficient=0.37), which was associated with reduced substance use (coefficient=-0.15). Higher SES diminished depressive symptoms. Model fit was acceptable. Separate models for tobacco smoking, illicit drug use or alcohol problem produced similar findings. Conclusions: Poverty alleviation and social support-based interventions may benefit male TB patients with substance use. Key words: tuberculosis, alcohol, tobacco, substance Am J Health Behav. 2014;38(4):501-509 DOI: http://dx.doi.org/10.5993/AJHB.38.4.3

may be due to a lack of access by poor and disadvantaged groups to basic resources (including access to healthcare and education) and exposure to harmful social and biological environments,17,18 but links among socio-economic status (SES), health behaviors, and ultimately, disease are only partially explained by these factors. Hence, several authors have studied the mediating role of intraand inter-personal reserve capacity and cognitiveemotional factors, in particular for non-communicable diseases.17,18 However, as far as we could ascertain, the exact pathways through which SES influences “substance use” (defined as the construct of tobacco smoking and/or having an alcohol problem and/or illicit drug use for the purpose of this paper) in patients with infectious diseases such as TB have not yet been elucidated. In particular, the role of modifiable factors such as social support and depressive symptoms is unknown. Greater understanding of the direct and indirect effects of socio-economic and psycho-social factors on substance use in TB patients would be helpful in developing remediating strategies because psycho-social factors may be more amenable to change than socio-economic status. Besides substance use, other mental health problems, such as anxiety and depressive symptoms, are common in TB and HIV/TB co-infected patients.19-22 This may be a result of neurological disease manifestations and drug side-effects, but, more importantly, this may be related to the psy-

501

Poverty and Substance Use in South African Tuberculosis Patients chological effects of dealing with a life-threatening disease.9 Depression and/or anxiety (measured separately or with a scale combining measures such as the Kessler psychological distress scale)23 may in turn increase alcohol use, illicit drug use, or tobacco consumption, closing the vicious circle of illness and mental health.9,24 In considering possible pathways among SES, depressive symptoms and/or anxiety, social support, and substance use, it is necessary to review evidence from studies that use conceptual models which include different elements in the chain, and studies that report on individual links involving only some of the elements. As has been mentioned above, poor people are at increased risk of smoking, and the adverse consequences of drinking and drug use.12-16 There is also some evidence of a link between poverty and psychological distress in both TB and nonTB patients,18,20,21,25,26 and between psychological distress and smoking, alcohol, and drug use.24,2628 Individuals suffering from anxiety or depression may smoke or drink more to deal with negative emotions.18 SES also may affect people’s reserve capacity, including social support.18,25,26,29 Positive social support may diminish substance use either directly or indirectly, by reducing depressive symptoms and/or anxiety.18,27,30,31 Conversely, negative social support which reinforces continued smoking, alcohol use, or drug use may be an even more important determinant of these behaviors than the effect of positive social support in diminishing such behaviors.27,32,33 Complete theoretical models encompassing SES, reserve capacity, negative or positive emotions, intermediate endpoints (health behavior and biological changes), and ultimately, illness, have been proposed by several authors.17,18 Gallo and Mathews18 reviewed the evidence available on the role of negative emotions in the link between socio-economic status and health. These authors propose a general framework which suggests that low SES increases exposure to stressful life experiences and simultaneously shapes a person’s reserve capacity (tangible, interpersonal, and intrapersonal). Both stress and low reserve capacity exacerbate a person’s vulnerability to negative emotions and cognitions. These elements together influence health behaviors adversely, and induce hormonal and metabolic changes, ultimately leading to increased morbidity and mortality.18 Similar theoretical models have been applied in several real-life populations, using structural equation modeling. Sheehan25 has demonstrated that the effect of economic stress on substance use and low birth weight was mediated through low social support in a large cohort of women in the United States. Likewise, higher SES was a predictor of higher smoking cessation rates both directly and indirectly (through increased social support, decreased negative affect, and higher self-efficacy) in adult smokers in the same country.29 Mulia et al26 studied the effect of poverty on psychological distress and problem drinking over time in a co-

502

hort of poor American women, finding that stressful events and neighborhood disorder significantly increased the risk for problem drinking, largely by increasing psychological distress. Their study could not prove that more social support buffered the effect of social stressors on problem drinking.26 To our knowledge, psychosocial pathways from SES to substance use have not been determined in TB patients. Therefore, the aim of this study was to ascertain the role played by social support and depressive symptoms in mediating the relationship between poverty and substance use in TB patients with high HIV co-infection rates in South Africa. In our initial theoretical model, we hypothesized that SES would influence substance use both directly and indirectly via its effect on social support and depressive symptoms. We also hypothesized that social support would buffer the effect of depressive symptoms on substance use.18 METHODS Participants For this analysis we used baseline data collected for a smoking cessation study. The study was conducted at the 6 largest public service TB clinics in a large urban township in Tshwane Metropolitan Municipality in South Africa. The population in this township is predominantly black Africans. All the adult patients seeking TB treatment at these clinics were consecutively enrolled from September 2011 to April 2013. Patients who were already on TB treatment for over a month, patients too ill to be interviewed, and those unable to communicate in the local language or in English were excluded from the study. Details of the study have been published elsewhere.34 As current smoking, alcohol problems, and illicit drug use were uncommon in female participants, data analysis for this paper was limited to the male participants. Furthermore, it is also plausible that the predictors and mediators of substance use may differ for men and women. Measurements Trained fieldworkers administered semi-structured questionnaires to eligible patients in English or in the most commonly spoken local language. To ensure accuracy of the translation into the local language, questionnaires were back-translated into English. Participants were asked questions about tobacco, alcohol, and illicit drug use, their socio-economic situation, social support, and recent depressive symptoms. Socio-economic measures included education (3 ordinal categories), household income (4 ordinal categories), a 6-item asset index (fixed telephone, mobile phone, television, radio, refrigerator and computer; interval-level variable), the number of rooms in the dwelling, and how often the participant had gone to bed hungry in the past month (never, 1-7 days, >7 days). TB and HIV-related information was obtained from the individual stan-

Louwagie et al

Table 1 Socio-economic and Psycho-social Characteristics of Male TB Patients (N = 1005a) All Participants n/N (%) Mean (SD)

Variable

Median (IQR)

Smoking Characteristics Non-smoker

378/1005 (62.4)

Current smoker

627/1005 (37.6)

Alcohol Problem (CAGE score ≥2 )

b

Score ZAR 6000

34/989 (3.4)

Hunger in Past Month 0 days

886/1001 (88.5)

1-7 days

104/1001 (10.4)

>7 days

11/1001 (1.1)

Number of Rooms in Dwellinge (N = 1001)

4.18 (1.95)

4 (3-5)

Emotional/informational support (range 8-40) (N = 991)

33.0 (7.05)

33 (30-40)

Tangible support (range 4-20) (N = 994)

16.6 (3.83)

17 (15-20)

Affectionate support (range 3-15) (N = 991)

11.7 (3.64)

12 (9-15)

Positive social interaction (range 3-15) (N = 993)

10.5 (3.85)

12 (9-14)

Additional item (range 1-5) (N = 985)

3.8 (1.37)

4 (3-5)

Social Support

f

Depressive Symptoms in Past 2 Weeksg No

619/931 (66.5)

Yes

312/931 (33.5)

Note. a Excluding 2 people with unknown current smoking status b CAGE = Cut-Down, Annoyed, Guilt Eye-opener; Cronbach alpha for Cage Score = 0.89 c HIV status unknown for 72 participants d ZAR= South African Rand, ZAR 8.7 ≈ 1 US dollar e Excluding kitchen/bathroom/shed or outdoor rooms unless a household member lives in such a room f Cronbach alpha=0.95 for emotional/informational support, 0.95 for tangible support, 0.95 for positive social interaction and 0.95 for affectionate support g 57 participants were unsure and information was missing for 18 participants

Am J Health Behav.™ 2014;38(4):501-509

DOI:

http://dx.doi.org/10.5993/AJHB.38.4.3

503

Poverty and Substance Use in South African Tuberculosis Patients

Table 2 Sample Correlation Matrix with Variances on the Diagonal Emotional Support

Tangible Support

Affetionate Support

Education

Asset Score

Household Income

Depressive Symptoms

Current Tobacco Smoking

Emotional Support

49.677

Tangible Support

0.618*

14.672

Affectionate Support

0.481*

0.446*

13.252

Education

0.151*

0.121*

0.166*

0.541

Asset Score

0.122*

0.143*

0.123*

0.310*

0.833

Household Income

0.170*

0.139*

0.166*

0.273*

0.269*

0.603

Depressive Symptoms

-0.030

-0.035

0.105*

-0.146*

-0.181*

0.045

Current Tobacco Smoking

-0.155*

-0.176*

-0.112*

-0.204*

-0.115*

-0.179*

0.017

Alcohol Problem

-0.203*

-0.191*

-0.059

-0.139*

-0.115*

-0.056

0.127*

0.613*

Illicit Drug Use

-0.118*

-0.071

-0.034

-0.175*

-0.054

-0.228*

0.177*

0.631*

Alcohol Problem

Illicit Drug Use

0.415*

* = Significant correlation (p < .05)

dardized patients’ TB records. Questions regarding tobacco use were adapted from the Global Adult Tobacco Survey questionnaire, an expert-developed questionnaire used for tobacco prevalence surveys across the world.35,36 Current smoking was defined as having smoked tobacco in the past month (yes/no); daily and occasional smokers were grouped together as occasional smoking was uncommon. The CAGE (CutDown, Annoyed, Guilt, Eye-opener) questionnaire was used to identify participants with an alcohol problem. The CAGE questionnaire is a brief validated screening tool frequently used in clinical settings, including in South Africa.37,38 Answers to the 4 items were added up and a new binary variable created whereby a score of 2 or more indicates a likely alcohol problem.39 Illicit drug use was measured by enquiring into the use of a range of commonly used illicit substances: marijuana/cannabis, cocaine/crack, methamphetamine and others (the latter category was added to make provision for uncommonly used non-listed substances, if any). In the actual questionnaire respondents could choose any answer out of 7 categories (from 0 days per month substance use up to all 30 days) for each of the 4 listed drugs. Answers to these questions were later categorized as a yes/no for any type of illicit drug use to collapse categories with few observations. Social support was measured using the Rand Medical Outcomes Survey Social Support (MOSSS) questionnaire, a reliable, validated, and widely used measure with 4 categories of functional social support: emotional/informational, tangible, affectionate, and positive social interaction.40 The 19 items were rated on a Likert-type scale, with responses ranging from “none of the time” to “all of the time”. The MOS-SS has also been applied in a variety of South African patient populations.41-43

504

The one-item question: “Have you been down or felt depressed most of the day nearly every day for the past 2 weeks? Yes/no” was used to screen for possible depression. The validity of a one-item question as a rapid screening tool has been demonstrated in several populations.44,45 The study was piloted at all 6 clinics over a period of one month. Statistical Analysis Data were double-entered in Microsoft Excel by experienced data capturers, and were exported into STATA version 12 for data comparison, cleaning, and initial analysis.46 Cronbach alpha internal consistency coefficients were calculated for the scales. Descriptive summary statistics consisted of percentages for categorical variables and means with standard deviations or medians with interquartile ranges, as appropriate. For the structural equation modeling, the MPlus Software Statistical package was used.47 Because the model contained both continuous and categorical variables, coefficients for categorical variables were estimated with probit regression, whereas those for continuous variables were estimated with linear regression. Due to the non-normality of the scores and the inclusion of categorical variables, free parameters were estimated using the weighted least squares with robust standard errors method. Regression coefficients were standardized with the appropriate standardization method. For the measurement component of the model, we built latent variables for the constructs social support, SES, and substance use. The model was tested for model fit using the following indices: the chi-square goodness-of-fit test,48 the Root Mean Square Error of Approximation (RMSEA) (a value of 0.05 or less indicates a good fit),49 the Bentler Comparative Fit index (CFI),50 the Tucker-Lewis Index (TLI) (both the CFI

Louwagie et al

Figure 1 Final Model for Substance Use

Note. a Socio-economic status b Coefficient not significant; Numbers in bold next to arrows are standardised regression coefficients; Italic numbers are R-squared multiple correlations values (% variability explained by model). Model fit indices: Chi-square = 79.121 [df 30, p < .0001]; RMSEA = 0.030; CFI = 0.956; TLI = 0.935; WRMR = 0.944

and TLI should have values above 0.90 for model acceptance), and the Weighted Root Mean Square Residual (WRMR) (a value of less than 0.90 indicates a good fit).51 RESULTS Characteristics of Participants A total of 1005 men were included in this analysis. Excluded from the analysis were all women (N = 919), 2 participants who refused to answer to tobacco smoking related questions, and those excluded a priori from the larger study based on the above mentioned exclusion criteria (428 patients of which 247 were children). The mean age of study participants was 40.8 years and 85% of those with a test result were HIV-positive (HIV status missing in 72 cases). Participants commonly smoked (37.6%) or had an alcohol problem (27.2%), but self-reported illicit drug use was comparatively rare (6.8%) and comprised mainly cannabis (60 out of 67 illicit drug users). Few respondents reported using methamphetamine (n = 9), cocaine (n = 6), or other unlisted illicit drugs (n = 6) (results not presented in the table). Over three-fourths of the participants had not completed high school, or had only pri-

Am J Health Behav.™ 2014;38(4):501-509

mary education or none. Just over three-fourths of participants (76.6%) lived in a household with a combined income of ZAR 2500 (approximately 300 US dollars) or less. One-third (33.5%) of the participants reported recent depressive symptoms. Self-reported social support was high for all 5 dimensions of the MOS-SS (Table 1). Correlations among Predictors The correlations among the observed variables (Pearson product moment, tetrachoric, polychoric, biserial or polyserial as appropriate) are displayed in Table 2. Variables of the same latent construct were highly correlated, as expected. Measurement Model We constructed a latent factor “substance use” for tobacco smoking, a likely alcohol problem and illicit drug use. All 3 indicator variables loaded well on the latent construct. For the MOS-SS, we excluded the “additional item” because of a low factor loading (factor loading

Poverty and substance use in South African tuberculosis patients.

To determine whether social support and depressive symptoms mediate the relationship between poverty and substance use in tuberculosis (TB) patients i...
997KB Sizes 0 Downloads 3 Views