open access

https://hdl.handle.net/2115/68300

Optimal information networks : Application for data-driven integrated health in populations

英語

2018-02-02

American Association for the Advancement of Science

Science Advances

4(2)

e1701088

Development of composite indicators for integrated health in populations typically relies on a priori assumptions rather than model-free, data-driven evidence. Traditional variable selection processes tend not to consider relatedness and redundancy among variables, instead considering only individual correlations. In addition, a unified method for assessing integrated health statuses of populations is lacking, making systematic comparison among populations impossible. We propose the use of maximum entropy networks (MENets) that use transfer entropy to assess interrelatedness among selected variables considered for inclusion in a composite indicator. We also define optimal information networks (OINs) that are scale-invariant MENets, which use the information in constructed networks for optimal decision-making. Health outcome data from multiple cities in the United States are applied to this method to create a systemic health indicator, representing integrated health in a city.

学術雑誌論文

Version of Record

Creative Commons Attribution-NonCommercial 4.0 International

2375-2548