A New, Obesity-specific Healthy Eating Index (OS-HEI)

Because of the strong relationship between health status and diet quality, many indices have been developed to measure diet quality. Although most of these indices are intended to show a relationship with health outcomes, in all indices this association has been proved to be moderate, especially in the case of obesity. Moreover, all suggested indices are based on a high level of subjectivity in relation to choosing the components to build up the index and the cut-off points for each component. Even in the case of less subjective indices, such as the Healthy Eating Index (HEI), it depends on a review of scientific literature on nutrition and health effects done by a distinguished panel of scientists and the HEI components and their weighting are objectively based on this review. Both issues have reinforced the need to develop a diseaseand obesity-specific diet quality index, which is the main objective of this paper. To reduce subjectivity, we propose the use of a new, non-parametric approach – the multivariate adaptive regression splines (MARS) – to develop our new, Obesity-specific Healthy Eating Index (OS-HEI), which is a data-driven, non-parametric tool and allows for interaction between the different items. The data used comes from the 2007–08 and 2009–10 National Health and Nutrition Examination Surveys (NHANESs). While the data from 2007– 08 has been used to develop the new index, the data from 2009–10 has been used to validate the results. The traditional HEI-2010 index has been used as a benchmark. Results indicate that the OS-HEI notably outperforms the HEI-2010 in predicting obesity prevalence.