Prediction of Vitamin D Deficiency by Simple Patient Characteristics

Background: Vitamin D status is currently diagnosed by measuring serum 25-hydroxyvitamin D [25(OH)D].

Objective: This study aimed to develop a risk profile that can be used to easily identify older individuals at high risk of vitamin D deficiency.

Design: This study was performed within the Longitudinal Aging Study Amsterdam, an ongoing cohort study in a representative sample of the Dutch older population (n = 1509 for the development sample and n = 1100 for the validation sample). Prediction models for serum 25(OH)D concentrations <50 and <30 nmol/L were developed by using backward logistic regression. Risk scores were calculated by dividing the individual regression coefficients by the regression coefficient with the lowest β to create simple scores.

Results: Serum 25(OH)D concentrations <50 and <30 nmol/L were present in 46.2% and 17.5% of participants, respectively. The model for the prediction of concentrations <50 nmol/L consisted of 13 easily assessable predictors, whereas the model for concentrations <30 nmol/L contained 10 predictors. The resulting areas under the curve (AUCs) were 0.78 and 0.80, respectively. The AUC in the external validation data set was 0.71 for the <50-nmol/L model. At a cutoff of 58 in total risk score (range: 8-97), the model predicted concentrations <50 nmol/L with a sensitivity of 61% and a specificity of 82%, whereas these values were 61% and 84%, respectively, at a cutoff of 110 in the total risk score (range: 6-204) in the model for concentrations <30 nmol/L.

Conclusions: Two total risk scores, including 13 or 10 predictors that can easily be assessed, were developed and are able to predict serum 25(OH)D concentrations <50 and <30 nmol/L accurately. These risk scores may be useful in clinical practice to identify persons at risk of vitamin D deficiency.