This paper investigates Machine Learning methods for predicting vitamin D levels in patients using biochemical parameters, age, and sex. Vitamin D is essential for proper physiological functions, and its deficiency is associated with many diseases. However, vitamin D testing is often expensive and not always accessible, highlighting the need for alternative predictive approaches. By applying and comparing different Machine Learning models, including Linear Regression, Random Forest, Support Vector Regression (SVR), and Huber Regression, this study evaluates their effectiveness in estimating vitamin D status without systematic blood testing. The results show that SVR achieved the best performance, offering an optimal balance between accuracy and robustness. These findings demonstrate the potential of Machine Learning to provide cost-effective, reliable tools for supporting medical decision-making and improving healthcare resource management.