MODELLING HAZARD OF BECOMING ALCOHOLIC USING PARAMETRIC AND NON-PARAMETRIC METHODS

Hazard function shows the dependence of chance of onset of alcoholism on time. A cross-sectional study applying multistage stratified sampling method was conducted using questionnaire administered in seven out of 47 counties to current alcohol consumers and persons in rehabilitation centres which captured latent variables: peer influence, family attention, personality, economic status, environmental setting, socio-cultural, drinking habits and patterns. Parametric hazard function was determined using Birnbaum-Saunders (B-S) model. Person-period dataset and discrete-time proportional hazards model were used to determine non-parametric hazard rates. Parametric hazard function based was an upside-down curve, while failure rate was an increasing curve. Non-parametric hazard of alcoholism is an upside-down curve with bumps and valleys. The non-parametric failure rate of alcoholism based on all factors was an increasing upward curve. Probability of becoming alcoholic increased from 0.31% when drinking once per week to 57% when drinking seven times a week. Based on years since onset of drinking it rose from 0.0059% in year one to 70% after 30 years. Thus, probability of becoming alcoholic increased with frequency of drinking per week and with time since onset of alcohol taking. Females had higher hazard of becoming alcoholic than males and was affected by factors differently. B-S distribution was confirmed suitable to model time from onset to diagnosis by upside-down parametric hazard function. Hazard of alcoholism is highly affected by gender, personality and peer influence. Anti-alcoholism programs to focus in reducing the frequency of drinking.