Development of an alcohol intervention model for predicting healthcare costs, life years, quality-adjusted life years and using for economic evaluation

TL;DR: An alcohol intervention model that predicts life years, quality adjusted life years (QALYs), and healthcare costs classified by the Alcohol Use Disorder Identification Test (AUDIT) screening tool and other various risk factors related to alcohol consumption was developed and transferred to the Thai setting.

Abstract: Objectives To develop an alcohol intervention model that predicts life years (LYs), quality adjusted life years (QALYs), and healthcare costs classified by the Alcohol Use Disorder Identification Test (AUDIT) screening tool and other various risk factors related to alcohol consumption. Furthermore, the developed model was transferred to the Thai setting. Methods Eight Scottish Health Surveys from 1995-2012 were linked to Scottish morbidity records and death records for the period 1981 to the end of 2013. Parametric survival analysis was used to estimate the hazard risks of first alcohol-related and non-alcohol related hospitalisations and deaths. For men and women, multivariate data analyses were applied separately for each gender in modelling the utility score, risks of subsequent hospitalisation and annual healthcare costs within the follow-up period. Risk profiles were used for the covariates of the models as follows: age, socio-economic status, health condition, alcohol drinking (i.e. AUDIT and binge drinking), smoking, body mass index, and physical activity. According to the under-reporting bias of alcohol consumption among the survey population, this study adjusted the reported alcohol consumption using alcohol sales data. Multiple imputation approach was applied to deal with missing data. A health-state transition model with annual cycle length was developed to predict LYs, QALYs, lifetime costs, and cost-effectiveness. Probabilistic sensitivity analysis was also performed to deal with parameter uncertainty. Moreover, a methodological transferability protocol of the Thai study was detailed. Results The sample size of the cohort was 46,230. The developed model showed the association between drinking and alcohol-related and non-alcohol related hospitalisations and deaths which were calculated as LYs and QALYs. Other risk factors were also taken into account that would likely affect the outcomes of interest. The modelling showed that an increasing AUDIT score and the number of cigarettes per day were associated with an increased risk of first alcohol-attributable hospitalisation. Predicted outcomes for a male aged 30 year with high-risk drinking levels (AUDIT >7) were worse than males with low risk drinking (AUDIT ≤7), with approximately 5 LY gained and 7 QALY gained. The same results for females were obtained for high-risk drinking (AUDIT >4) compared to low-risk drinking (AUDIT ≤4), with approximately 10 LY gained and 12 QALY gained. Furthermore, an economic evaluation was performed to compare the no-intervention situation with a hypothetical health promotion intervention - which aimed to stop drinking (measured by the AUDIT) and smoking (measured by the number of cigarettes per day) behaviours. To compare the costs and benefits of the hypothetical intervention and no intervention over the lifetime period, a within-trial analysis combined with the developed model was able to capture both short- and longer-term consequences (i.e. LYs, QALYs, and healthcare costs) of the intervention. Finally, the model was able to compare cost-effectiveness ratio between risk behaviours without the new intervention and the modified risk behaviours when the new intervention is implemented. Conclusions The study highlights the potential and importance of developing health economic models utilising data from routine national health surveys linked to national hospitalisation and death records. The developed framework can be used for further economic evaluation of alcohol interventions and other health behaviour change interventions. The framework can further be transferred to other country settings.