Summary of Methods & Results
Wenig utilized data from a population-based multi-region survey, the German Interview and Examination Survey for Children and Adolescents, which measured anthropometric, sociodemographic and healthcare utilization data. In comparison to the MEPS analyses from the USA, a major benefit to this analysis is the increased precision of actual anthropometric data compared with self-report data, which suffer from concerns about validity of parentally or self-reported anthropometric data. The author chooses different cut-offs for overweight and obesity in childhood from those described by the American Medical Association Expert Panel and the International Obesity Task Force. While Toschke et al. have found that risk factor identification does not vary by choice of this or the German cut-offs used by Wenig, the implications of different cut-offs choice on incremental healthcare utilization or costs has not been examined carefully.
As other studies have done, Wenig controlled for socioeconomic status, place of residence and age group as part of the multivariable regression analysis to identify incremental costs associated with elevated BMI group. Focus of this analysis on individuals aged 3–17 years old was appropriate because of ongoing difficulty with categorization of BMI group in younger children. Participant/parent recall was used to quantify visits by type, and then prices as established by a national costing guideline were used to convert visits into costs. Analysis of direct costs was also limited in that rehabilitation and medication costs were not measured directly, and the German Interview and Examination Survey for Children and Adolescents does not provide data on indirect costs that do not relate to treatment for an obesity-related condition but do result from obesity (e.g., reduced economic productivity from lost school/work days).
Analysis of healthcare expenditure and utilization data is challenging, especially in children because zero outcomes are frequent, and the frequency of distribution of these outcomes is positively skewed. Two-part models are almost always the preferred choice, with the first part of the model predicting expenditure/non-expenditure or use/non-use, and the second part, estimated independently of the first, predicting level of use or cost.[14,15] These two-part regression analyses account for the impact that obesity has on the presence/absence of expenditures/utilization and further weight the results obtained from a linear regression using only those observations with non-zero expenditures/utilization. Wenig takes an alternative approach, using a one-part model and setting zero costs to €5. The justification used to support this decision is the low (4.4%) frequency of zero costs (although this frequency is not described by subcategory), and log transformation is used to account for the positive skew in the distribution. Tests are not described to assess whether a persistent skew remains after this transformation, or to compare the model fit for alternative statistical modeling to the gamma, log-link model that was chosen. The study uses complex survey design weights and assesses whether there is a different result without use of such weighting. Bonferroni adjustments for multiple simultaneous comparisons are not described.
The author finds that overweight and obese children have higher physician costs (~€62/year for obese patients and €27/year for overweight patients) and, on further logistic regression, higher likelihood of being 'high utilizers' of healthcare services. Total costs do not show a clear relationship with severity of obesity, although positive and significantly higher total costs are noted among overweight children.
Expert Rev Pharmacoeconomics Outcomes Res. 2011;11(1):47-50. © 2011 Expert Reviews Ltd.
Cite this: Quantifying the Economic Consequences of Childhood Obesity and Potential Benefits of Interventions - Medscape - Feb 01, 2011.