Indirect Estimation of 25-Hydroxyvitamin D Reference Intervals Using Data Mining

Background: Reference intervals for 25-hydroxyvitamin D are significantly influenced by seasonal variations and population-specific factors. This study aimed to establish local, population-specific Vitamin D reference intervals using the refineR algorithm and to compare these results with the conventional non-parametric method using a hospital database. Methods: A total of 127,220 laboratory results from 2022 to 2025 were retrospectively analyzed. Data were filtered based on adult age, morning fasting specimens, outpatient status, and normal parathyroid hormone levels (14-72 ng/L). Following filtration, Wilcoxon and Kruskal-Wallis tests were employed to evaluate differences between potential subgroups. Consequently, 24,036 eligible results were stratified by age (18-40, 41-65, and >65 years) and sex (female and male). Reference intervals were calculated using the refineR algorithm and the non-parametric percentile method. Results: The distribution of vitamin D levels was found to be right-skewed [Median: 12.92 ng/mL (4.21-149)]. The upper reference limits obtained via refineR were consistently lower than those derived from the non-parametric method. Both methods showed a linear increase in median values and upper reference limits with advancing age. Females in the 18-40 age group exhibited the lowest vitamin D profile. The summer concentrations were significantly higher than those of other seasons (p <0.05). Conclusions: The refineR algorithm managed outliers and pathological results more effectively than the non-parametric method. Our findings highlight the clinical necessity for age- and sex-specific subgroups, rather than relying on the manufacturer's single reference range. Implementing these population-specific intervals can enhance diagnostic accuracy and prevent misclassification, facilitating earlier identification of vitamin D-related metabolic and skeletal disorders.