Can Demographic and Health Surveys (2007–2024) Capture Alcohol Use Trends in Zambia?
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Alcohol is a leading modifiable risk factor for non-communicable disease. Zambia’s National Alcohol Policy and the World Health Organization’s target of a 10% relative reduction in the harmful use of alcohol both require that the trend be monitored. Alcohol items appear in four rounds of the Zambia Demographic and Health Survey (ZDHS), and those rounds are widely treated as a trend series, although whether they are one has never been tested.
Methods
Secondary analysis of ZDHS 2007, 2013–14, 2018 and 2024 (women aged 15–49, men aged 15–59). A direct current-use item is available in three rounds, with three different instruments for men and two for women. Four identification strategies were applied in ascending order of assumption: nesting bounds, which exploit the fact that a seven-day window falls within a thirty-day window and that within undated current status; restriction to the fieldwork months common to both rounds; a lifetime-use analogue available in 2024; and an instrument-constant partner-report series available in all four rounds, validated by linking each woman to her co-resident husband. Estimation throughout was design-based.
Results
The conventional series suggests a fall in current drinking among men from 42.0% (95% CI 39.9–44.1) in 2007 to 28.2% (95% CI 27.0–29.3) in 2024, and among women from 11.1% (95% CI 9.9–12.4) to 8.8% (95% CI 8.0–9.6). Neither change is sign-identified. Placed on a common thirty-day basis with matched fieldwork months, the 2013–14 to 2024 change lies between -7.6 and +1.2 percentage points for men and between -0.2 and +3.7 for women. The instrument-constant proxy fell from 53.7% in 2007 to 37.7% in 2018, then plateaued at 37.0% in 2024; a constant-decline model is rejected (Q = 15.9, 2 df, p = 0.0003). Sensitivity of the proxy against husbands’ own reports fell from 86.0% to 66.8%. The 2024 cross-section is unaffected and is reported in full.
Conclusions
These data do not establish the apparent national decline in alcohol use. Differences in reported prevalence across ZDHS rounds substantially reflect instrument change, reference-period shift, fieldwork seasonality and decay in proxy reporting. On present evidence Zambia cannot monitor its alcohol commitments from national survey data. Trend monitoring would require a consistent alcohol module restored to the questionnaire, an occasion-based heavy-drinking item, and the reporting of fieldwork month.
