P4 "The socioeconomic profile of alcohol-attributable mortality in South Africa"

Dipl.-Psych. Charlotte Probst

supervisor: Prof. Hans-Ulrich Wittchen, Prof. Jürgen Rehm

submitted: 8. November 2017

Introduction: Globally, illness and life expectancy follow a social gradient that puts people of lower socioeconomic status (SES) at higher risk of dying prematurely. Research from high income countries showed a particularly wide socioeconomic gap in alcohol-attributable mortality and found that alcohol use contributes to the socioeconomic differences. However, as of now little evidence is available from low and middle income countries. South Africa is an upper middle income country in which income is highly unequally distributed. Furthermore, alcohol use is one of the most important risk factors for premature mortality in South Africa. With 11 liters of pure alcohol, the per capita consumption in South Africa was five liters above the average in the African region in 2010.

Aims and objectives: The overall aim of this dissertation project was to generate a socioeconomic profile of cause-specific alcohol-attributable mortality among South Africa’s general adult population (defined as 15 years of age or older), taking age and sex into account. The objectives of the five component studies were to investigate coverage of alcohol, i.e., the proportion of the “true” alcohol per capita that got reported in South African nationally representative surveys (study I); to estimate the relative risk (RR) of human immunodeficiency virus/acquired immune deficiency syndrome (HIV/AIDS) mortality for people of low SES compared to people of high SES (study II); to identify potential interactive effects between alcohol use and SES on the HIV-risk (study III); to calculate alcohol-attributable fractions (AAF) of HIV/AIDS mortality and respective mortality rates, stratified by SES, age and sex (study IV); and finally to calculate AAFs and respective mortality rates from all major causes of death attributable to alcohol consumption, stratified by SES, age and sex (study V).

Design: A comparative risk assessment of alcohol-attributable mortality in the adult general population was performed for subgroups defined by SES, age and sex. The calculation of AAFs required measures of exposure to alcohol by subgroup and the cause-specific risk associated with each level of exposure. Mortality statistics in each subgroup were required for the calculation of alcohol-attributable death counts and rates. A consistent measure of SES was used in all calculations, in order to increase comparability across the computational steps.

Data sources: Cross-sectional data from in total six nationally representative surveys for the years 2003 (N=8,115), 2005 (N=16,110), 2008 (N=13,055), 2012 (N=27,070 and N=18,688), and 2015 (N=22,741) were used. Longitudinal data on mortality came from a cohort study based out of KwaZulu-Natal (years 2000 to 2014; N=87,029; 757,404 person-years; 23,002 deaths). Aggregate data on “true” alcohol per capita consumption as well as population and mortality statistics by cause of death, age, and sex were retrieved from public sources. Cause-specific risk functions associated with alcohol use were taken from a recent overview paper.

Statistical analyses: Coverage was investigated based on five cross-sectional data sets. Socioeconomic differences in HIV/AIDS mortality were investigated based on a systematic literature search. RRs were pooled for measures of SES separately using random-effects meta-analyses. Additionally, Cox regressions were calculated based on the cohort data in order to ensure a coherent operationalization of SES across computational steps. The resulting cause-specific hazard ratios were used to split mortality statistics by SES. Interactive effects between SES and alcohol use on HIV-risk were analyzed using multinomial regression models based on survey data.

Comparative risk assessments were performed in study IV, looking at HIV/AIDS mortality specifically (two SES groups), and in study V for all major alcohol-attributable causes of death (three SES groups). The prevalence of current alcohol use and irregular heavy drinking occasions in all subgroups was calculated using survey data. A standard triangulation technique was used to correct for underestimation of population-level alcohol use based on survey data. Alcohol-attributable mortality rates were age-standardized using the World Health Organization reference population. Monte Carlo simulation techniques were used to calculate uncertainty intervals (UI).

Results: The survey data covered between 11.8% (95% UI 9.3-16.2%; 2005) and 19.4% (95% UI 14.9- 24.2%; 2003) of the total alcohol used per capita (study I). The meta-analysis on socioeconomic differences in HIV/AIDS mortality in South Africa included 10 studies comprising over 175,000 participants and 6,700 deaths (study II). For income (RR 1.6, 95% confidence interval (CI) 1.2-2.1), asset score (RR 1.6, 95% CI 1.1-2.4) and employment status (RR 1.5, 95% CI 1.2-1.9), persons of low SES were at increased risk of dying from HIV/AIDS. The RR of 1.1 for education was not statistically significant (95% CI 0.7-1.7). The analyses of interactive effects showed that current drinkers of low SES had an elevated HIV-risk in all three investigated surveys (study III). The relative risk ratio (RRR) raged from 1.9 (95% CI 1.3-3.0; 2012) to 3.5 (95% CI 2.0-6.1; 2008).

The comparative risk assessment of HIV/AIDS mortality found higher AAFs in adults of low compared to high SES (study IV). The age-standardized mortality rate attributable to alcohol in South Africa in 2012 was 31 (95% UI 22-41) and 230 (95% UI 109-352) per 100,000 adults in males of high and low SES, respectively. In women, the respective rates were 11 (95% UI 6-16) and 76 (95% UI 31-145). The comparative risk assessment including all major causes of death attributable to alcohol use showed that in 2015 about 62,300 (95% UI 27,000-103,000) adults died from alcohol-attributable causes of death in South Africa in 2015 (study V). Age-standardized, alcohol-attributable mortality rates per 100,000 adults were highest for the low (727, 95% UI 354-1,208) followed by the middle (377, 95% UI 165-687) and the high SES group (163, 95% UI 71-289). The socioeconomic differences in AAFs were highest for mortality from infectious diseases and among men aged 35 to 54 years.

Conclusions: To the best of my knowledge this dissertation project was the first to quantify alcohol- attributable mortality in South Africa for different SES groups. Alcohol use contributed to vast socioeconomic differences in mortality, confirming findings from high income countries. Elevated AAFs for people of low and middle SES arose from higher levels of consumption among current drinkers and not from the prevalence of current alcohol use per se. For HIV/AIDS interactive effects were identified that contributed to elevated AAFs in people of low SES. As income inequalities and socioeconomic inequalities in alcohol-attributable mortality are increasing in many countries, the investigation of such interactive effects for a broader spectrum of outcomes and settings will be of great importance for future research. The final socioeconomic profile of alcohol-attributable mortality allowed for the identification of HIV/AIDS as a potential driver of socioeconomic differences and of middle-aged men as a high-risk group. Targeting preventive measures and interventions accordingly promises to yield the highest reductions in alcohol-attributable mortality and related socioeconomic inequalities in South Africa.