A prediction tool for vitamin D deficiency in New Zealand adults.

Abstract

This study aims to develop a model for predicting vitamin D deficiency in New Zealand adults using easily accessible clinical characteristics. Data were derived from the Vitamin D Assessment (ViDA) study dataset. Included participants in the main analysis were aged 50-84 years and resided in Auckland, New Zealand. The dataset was split into a discovery dataset in which the prediction model was developed (n = 2036) and a validation dataset in which it was tested (n = 2037). The prediction model was developed using clinical characteristics in a logistic regression analysis with deseasonalised serum 25OHD (DS-25OHD) as the dependent variable. DS-25OHD < 40 nmol/L was found in 8.2% of European participants, 18.8% of Māori participants, 23.1% of Pacific participants and 52.2% of South Asian participants. Predictors for DS-25OHD < 40 nmol/L in the European sub-cohort included increasing age, female sex, higher body mass index, current smoking, no alcohol intake, lower self-reported general health status, lower physical activity hours, lower outdoor hours and no use of vitamin D-containing supplementation. The area under the curve in the discovery dataset was 0.73, and in the validation dataset was 0.71. Of those with a prediction score ≥ 10 (total risk score range 0-21.5), the sensitivity and specificity for predicting vitamin D deficiency was 0.90 and 0.41, respectively. Non-European ethnicity is an important risk factor for vitamin D deficiency. Our vitamin D deficiency prediction model performed well and demonstrates its potential as a tool that can be integrated into clinical practice for the prediction of vitamin D deficiency.