Health risk behaviour and persistent and incident depression among middle-aged and older adults in South Africa

The aim of this study was to assess the association between five health risk behaviours and persistent and incident depressive symptoms in a longitudinal study in South Africa. The sample included 5059 men and women (⩾40 years) in 2014/2015, and 4176 in 2018/2019 of the ‘Health and Ageing in Africa: A Longitudinal Study of an INDEPTH Community in South Africa’. Depressive symptoms were assessed with the Centre for Epidemiologic Studies Depression scale. Multivariable logistic regression analysis was conducted to assess the associations between five health risk behaviours and persistent and incident depressive symptoms. Results indicate that 5.0% of participants had depressive symptoms at both wave 1 and 2 (persistent depressive symptoms), and 27.9% had incident depressive symptoms in wave 2. Higher education and moderate baseline physical activity were negatively associated and those with cardiovascular disease were positively associated with persistent depressive symptoms. Middle wealth index was negative, and being HIV positive and baseline tobacco use were positively associated with incident depressive symptoms. In conclusion, of five health risk behaviours assessed (inadequate fruit/vegetable intake, alcohol dependence, tobacco use, physical activity, and sedentary behaviour), only moderate physical activity was protective against persistent depressive symptoms, and tobacco use was associated with incident depressive symptoms.


Participants
The first wave (W1) or the baseline included a sample of 5059 participants (⩾40 years) and at the second wave (W2) 4176 of the Wave 1 participants (at W2: 595 had died, 254 declined participation, 34 were not traced, and the response rate was 94%).

Instruments
Outcome variable. At baseline, DS were assessed with the 'Centre for Epidemiological Studies-Depression Scale eight-item scale (CES-D 8)', with 'a cut-off of three or more symptoms signifying DS' (Radloff, 1977) (Cronbach's alpha 0.66). Using a cutoff point of ⩾3, the CES-D 8 had a 'sensitivity of 71 percent and a specificity of 79 percent using the CIDI-SF diagnosis as the true caseness' (Steffick, 2000).
At follow-up, DS were assessed with the CES-D 20-item measure (Radloff, 1977) (Cronbach's alpha 0.79). The total scores for the 20 items range from 0 to 60, with higher scores indicating a higher DS. The cut-off for major DS was 20 or more, which has been observed to yield 'Sensitivity = 0.83, Specificity = 0.78, diagnostic odds ratio = 16.64' (Vilagut et al., 2016).
Health risk behaviour at baseline. Current tobacco use included current smokeless tobacco use and/ or current smoking (Gómez-Olivé et al., 2018).
Alcohol dependence was measured using the Cut, Annoyed, Guilty, and Eye (CAGE) scale (Ewing, 1984); Cronbach's alpha was 0.82 in this study.
Fruit and vegetable intake was sourced from questions: (1) 'How many servings of fruit/vegetables do you eat on a typical day? (on any one day)' (Gómez-Olivé et al., 2018).
Sedentary behaviour levels (<4 hr, 4 to <8 hr, and ⩾8 hr/day) were sourced from one item on the 'time usually spend sitting or reclining on a typical day?' (hours/minutes) from the GPAQ (Armstrong & Bull, 2006;van der Ploeg et al., 2012).

Covariates
Sociodemographic information included country of birth, age, sex, educational level, marital status, and an asset-based household wealth quintile.
Functional disability was obtained from the six items Activities of Daily Living (ADL) scale, including 'bathing, dressing, eating, getting in/out of bed using the toilet, and walking' (Katz et al., 1963); Cronbach's alpha for the ADL scale was 0.86 in this study.
Stroke, angina, heart attack, and heart failure (cardiovascular disease) and HIV status were assessed by self-reported health care provider diagnosis (Gómez-Olivé et al., 2018).

Procedure
We analysed longitudinal data from two consecutive waves of the 'Health and Ageing in Africa: A Longitudinal Study of an INDEPTH Community in South Africa (HAALSI)'. Full details of the methodology have been shown previously (Gómez-Olivé et al., 2018). Briefly, participants were randomly sampled from the 'Agincourt health and socio-demographic surveillance system (AHDSS)' in South Africa. The first wave (W1) or the baseline survey was conducted between November 2014 to November 2015, with a response rate of 85.9% (Gómez-Olivé et al., 2018), and the second wave (W2) was undertaken between October 2018 and November 2019, with a response rate of 94% . Prior to the survey, the study was introduced to the community in the study area. During the survey, local trained field workers introduced the study and obtained informed consent in the local language to all identified potential participants by visiting them at their homes. Adult household members (⩾40 years) who agreed to participate in the study were assessed with an interview-administered structured questionnaire (translated into xiTsonga) using computer-assisted personal interview (CAPI) and physical measurements (Gómez-Olivé et al., 2018).

Data analysis
First, the proportion of participants with persistent and incident DS was calculated and described. Logistic regression analyses were used to estimate (1) longitudinal persistent DS and (2) incident DS. Main predictors included health risk behaviours, controlled for sociodemographic factors, the presence of cardiovascular disease, HIV infection, functional limitations, and BMI. Odds ratio (OR) and 95% confidence intervals (95% CI) show the results from the logistic regressions. The values p ⩽ 0.05 were considered statistically significant. Longitudinal analyses included 'inverse probability weights that accounted for the probabilities of mortality and attrition over the followup' (Kobayashi, Morris, et al., 2021). All analyses were conducted using Stata SE 15.0 (College Station, TX, USA).

Sample characteristics by persistent and incident DS
Overall, 204 adults of 4929 participants who had DS in Wave 1 (5.0%) screened positive for DS at both Waves 1 and 2 (persistent DS), and overall, 965 adults of 3468 participants without DS in Wave 1 (27.9%) had incident DS in Wave 2. Table 1 provides the sample characteristics by persistent and incident DS.

Associations with persistent and incident DS
Participants with higher education had lower odds of persistent DS in comparison with those with lower education, and those with a cardiovascular disease had higher odds of persistent DS than those without cardiovascular disease. Regarding health risk behaviours, participants with baseline moderate physical activity had lower odds of persistent DS. Baseline alcohol dependence, tobacco use, inadequate fruit and vegetable intake, and sedentary behaviour were not significantly associated with persistent DS.
Participants with middle wealth index had lower odds of incident DS in comparison with those with lower wealth index, and those with being HIV positive had higher odds of incident DS than those being HIV negative. Regarding health risk behaviours, participants with baseline tobacco use had higher odds of incident DS. Baseline alcohol dependence, physical activity, inadequate fruit and vegetable intake, and sedentary behaviour were not significantly associated with incident DS (Table 2).

Discussion
In this first longitudinal study in Africa, we found that from five health risk behaviours assessed (alcohol dependence, tobacco use, inadequate fruit and vegetable intake, physical activity, and sedentary behaviour), only moderate physical activity was protective against persistent DS, and tobacco use was associated with incident DS. In line with a four-country prospective study (Cabello et al., 2017), we found that tobacco use was associated with incidence DS. Some research (Tully et al., 2010) suggest the existence of some shared genetic influences between tobacco use and depression, and other suggestions include that tobacco use triggers specific inflammation markers linking tobacco use with depression (Berk & Jacka, 2012). It is also possible that tobacco users at baseline (Wave 1) may have experienced DS prior to Wave 1 which may have increased the likelihood of engaging in tobacco use (Strine et al., 2008). More research is needed to show if the cessation of tobacco use can ameliorate DS, as found in high-income countries (Taylor et al., 2014), in South Africa (Cabello et al., 2017). Furthermore, our survey did not find that depressed tobacco users at Wave 1 had a higher likelihood to have persistent DS in Wave 2. Similar results were also found in the four-country prospective study (Cabello et al., 2017). More research may provide information if persistent DS increases tobacco use (van Gool et al., 2007). However, contrary to systematic review evidence (Dishman et al., 2021;Matison et al., 2021;Schuch et al., 2018), in our study, alcohol dependence, inadequate fruit and vegetable intake, physical activity, and sedentary behaviour were not associated with incident DS. In line with a meta-analysis of prospective studies (Huang et al., 2020), we found that sedentary behaviours associations were not significantly associated with incident DS as evaluated by screening, such as the CES-D. Therefore, we cannot confirm the hypothesis that health risk behaviours, apart from tobacco use, are used as a form of 'self-medication of DS' (Cabello et al., 2017). However, it is possible that participants with incidence DS in Wave 2 may have had subthreshold DS at Wave 1, which is also associated with considerable impairment in psychosocial and work functioning (Haller et al., 2014). Furthermore, depressed participants with moderate physical activity were less likely to also be depressed in Wave 2 in this study. This result is in agreement with some previous research (Boschloo et al., 2014;Cabello et al., 2017). However, lower physical activity was not related to incent DS in this study. Some findings seem to provide evidence that exercise interventions in people with depression may reduce depression symptoms and prevent relapses (Trivedi et al., 2006). The study found that having a cardiovascular disease was associated with persistent DS, and being HIV positive was associated with incident DS. The relationship between DS and cardiovascular disease may be bidirectional, explaining the link between cardiovascular disease and persistent DS (Hare et al., 2014). Incident depression in people with HIV may be explained by HIV-induced brain injury or antiretroviral medication (American Psychiatric Association, 2012).
One study limitation could be that at Wave 1, the CES-D 8 and at Wave 2, the CES-D 20 was used to evaluate DS. In additional analysis converting the CES-D 20 (by selecting the CES-D 8 items and converting the response format from four choices (1) 'rarely/none of the time, (2) some of the time, (3) occasionally or a moderate amount of time, and (4) most or all of the time' to 1 or 2 = yes and 3 or 4 = no) into the CES-D 8 at Wave 2 (Steffick, 2000), we found similar results as in our original analysis. However, the study used only screening measures and not structured psychiatric evaluation to establish DS caseness. Most measures used in this article were by self-report, which may have led to recall bias and social desirability bias, such as in the case of tobacco and alcohol use and physical activity.

Conclusion
From five health risk behaviours assessed (alcohol dependence, tobacco use, inadequate fruit and vegetable intake, physical activity, and sedentary behaviour), only moderate physical activity was protective against persistent DS, and tobacco use was associated with incident DS. Health behaviour interventions to reduce DS seem not warranted.

Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: 'HAALSI (Health and Aging in Africa: A Longitudinal Study of an INDEPTH Community in South Africa) is sponsored by the National Institute on Aging (grant no. 1P01AG041710-01A1) and is conducted by the Harvard Center for Population and Development Studies in partnership with Witwatersrand University. The Agincourt HDSS was supported by the Wellcome Trust, UK (grant nos. 058893/Z/99/A, 069683/Z/02/Z, 085477/Z/08/Z, and 085477/B/08/Z), the University of the Witwatersrand and South African Medical Research Council'. Karl Peltzer https://orcid.org/0000-0002-5980-0876