Determinants of Serum Vitamin D level; A Data Mining Approach

Abstract

Introduction: Serum vitamin D levels are related to a wide spectrum of factors including low sunlight exposure, high oxidative stress, low physical activity and sleep disorders. In this paper we are going to investigate the most crucial parameters associated with serum vitamin D levels in survey of ultraviolet intake by nutritional approach (SUVINA) study with a data mining approach.Material and Methods: Data including demographic, anthropometric, clinical and laboratory information were extracted from the SUVINA dataset comprising 289 subjects who were enrolled into our study. The XGBoost algorithm was used to define the most important features related to vitamin D level in our population.Results: Applying XGBoost modeling for vitamin D level showed that the presented scheme can determine the most important determinants of serum vitamin D level with an accuracy of 91%. Pro-oxidant anti-oxidant balance (PAB), body fat percentage, physical activity level (PAL), age, restless leg syndrome (RLS), and dietary inflammatory index (DII) density were the most important variables correlated with vitamin D deficiency.Conclusion: Using XGBoost and with an accuracy of more than 90%, we showed that the six most important risk factors for vitamin D deficiency are PAB, PAL, age, body fat percentage, RLS and DII density, respectively.

Keywords

• vitamin d
• oxidative stress
• machine learning
• data mining algorithm