SELECTION AND SELF‐SELECTION: HOW TO DETERMINE THE REAL IMPACT OF ALCOHOL ON HEALTH‐CARE UTILIZATION AND COSTS?

The paper by Anzai et al. (2005) on alcohol consumption and the use of health services finds a U-shaped relationship between the level of consumption and in-patient health care utilization (and costs), and an inverse relationship with frequency of out-patient care (and costs). In arriving at this conclusion, the paper avoids many of the shortcomings of the literature, i.e. small sample size, self-selection of relatively well-off social strata, drawing causal conclusions from cross-sectional data, retrospective reporting and self-report on utilization. However, there are still the following questions remaining which the study could address more adequately. The group of life-time abstainers may be characterized by other traits and behaviours responsible for the left side of the U-shaped curve/inverse relationship. There is some indication for this kind of explanation (Bondy & Rehm 1998; Cryer et al. 2001), especially in samples where life-time abstention is a behaviour shared by only a relatively small minority. While this explanation cannot be excluded, its omnibus character and unspecificity leaves a bitter taste. In the paper by Anzai and colleagues, a relatively large group of people with high levels of disease were excluded: 3361 people with either stroke, myocardial infarction, liver disease and cancer at baseline, and 1886 ex-drinkers. This corresponds to a sample size reduction of about 25%, mainly of people who drive the overall health-care costs in the sample. The reasons for specifically excluding these diseases are not clear. Alcohol is related to more than 60 International Classification of Diseases (ICD) codes, many not excluded from the analysis, but, on the other hand, is not related to all cancers (Rehm et al. 2003). However, combined with the relatively short follow-up time these exclusions present a problem, as there were probably not sufficient new events of these diseases within the follow-up period. Part of the difference between the high morbidity costs associated with alcohol in the traditional indirect analyses (e.g. Single et al. 1998) and the results of Anzai et al. (2005) can be explained by the fact that high-cost and alcohol-related diseases were excluded from these latter analyses. With respect to the second group systematically excluded, the ex-drinkers, I see no reason why they could not have been part of the statistical analyses, except for the trend calculations. In the current sample and procedures is the treatment of alcohol use disorders included, both in terms of the in-patient and out-patient services and, if so, what role does it play in respect to costs? In what way is alcohol consumption itself related to shorter in-patient stays and less frequent out-patient visits? There are some indications that people with relatively high levels of alcohol and tobacco use may have shorter stays in hospitals because they cannot exhibit these behaviours there (Single et al. 1996). Is there any information on patterns of drinking, especially in relation to irregular heavy drinking occasions, which have been linked to injury and some chronic disease (Rehm et al. 2003)? These open questions should not take away from the strengths of the study and the analyses. The Anzai et al. study is certainly one of the best-controlled and thus most informative studies on the topic. However, for the remaining questions, it would be extremely valuable to have additional analyses on the other characteristics of life-time abstainers, and similar types of analyses with a longer follow-up period. Also, some sensitivity analyses including the people with various diseases at baseline, plus the estimated alcohol relationship, would help in determining the real costs or potential savings incurred by alcohol.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.