An interpretable survey-based model of Khat consumption intensity in Sana’a, Yemen

Objectives: This study identifies the sociodemographic, economic, health-related, and social factors associated with khat consumption intensity among consumers in Sana’a, Yemen, and develops an interpretable model for classifying consumption into four ordered levels: light, moderate, regular, and heavy. Method: The study used a stratified random sample of 452 adult residents of the Capital Municipality of Sana’a, including 286 khat consumers. Participants were divided into two strata according to khat-use status: consumers and non-consumers. Data were collected through a structured questionnaire administered online and in paper form. Annual consumption intensity was derived from self-reported frequency and quantity of use, classified into four ordered categories, and modeled using ordinal logistic regression after screening candidate predictors. Results: The model demonstrated order-aware predictive performance, with tolerance accuracy within ±1 class of 92% and a quadratic weighted kappa of 0.67. Balanced accuracy reached 0.548, exceeding the 0.25 chance baseline for four classes. Higher consumption was associated with next-day withdrawal-like symptoms, particularly sleep-paralysis-like episodes, family and social exposure to khat, self-employment, and afternoon, night-time, or group-based consumption. Lower consumption was associated with older age, absence of next-day symptoms, preference for a non-consuming partner, price sensitivity, and washing khat before use. Conclusion: High-intensity khat consumption in Sana’a appears to reflect the combined influence of dependence-like symptoms, social normalization, and economic responsiveness. The findings support health-awareness campaigns, family- and workplace-based prevention, youth recreational alternatives, and pricing-related policy interventions. Because the sample was limited to Sana’a, generalization to other regions of Yemen should be made cautiously.