Sex differences in alcohol and tobacco use in Ugandan adults: A log-linear analysis of interaction patterns

Alcohol and tobacco use remain major contributors to non-communicable disease burden worldwide. Understanding how these behaviors intersect across sexes is crucial for effective prevention strategies. This study investigated the interrelationships between sex, alcohol use, and tobacco use among Ugandan adults using log-linear modeling. Data were drawn from the 2014 Uganda WHO STEPwise Approach to Non-Communicable Disease Risk Factor Surveillance (STEPS) survey, a nationally representative cross-sectional study including 3,987 adults aged 18-69 years. "Current use" was defined as self-reported consumption of alcohol or tobacco within the past 30 days. Frequencies and proportions were stratified by sex, and a series of hierarchical log-linear models were fitted to assess independence and interaction effects. Model fit was evaluated using the Likelihood-Ratio Chi-square (G2) and Dissimilarity Index (D) statistics. Alcohol use was more prevalent among females (60.0%, 95% CI: 56.8-63.1%) than males (38.9%, 95% CI: 36.2-41.7%), while tobacco use was slightly higher among males (9.05%, 95% CI: 7.4-10.9%) than females (7.95%, 95% CI: 6.4-9.8%). Combined alcohol and tobacco use was relatively uncommon but marginally higher in females (4.89%, 95% CI: 3.7-6.4%) than males (3.20%, 95% CI: 2.3-4.5%). The Homogeneous Association model provided the best fit (G2 = 0.93, p = 0.334, D = 0.5201), indicating that two-way interactions adequately captured the observed data. The Alcohol × Tobacco interaction was non-significant (p = 0.50), suggesting that alcohol and tobacco behaviors occurred independently once sex differences were accounted for-implying distinct behavioral or social drivers for each substance. This analysis revealed sex-specific patterns in substance use, with higher alcohol consumption among women and greater tobacco use among men. The absence of a significant alcohol-tobacco interaction suggests these behaviors operate within different social or environmental contexts rather than as co-occurring habits. These findings underscore the importance of sex-sensitive and behavior-specific prevention strategies and highlight the utility of log-linear modeling for exploring complex behavioral associations in public health data.