Many studies have pointed out that patients with vitamin D deficiency have a longer stay in the intensive care unit and a poor prognosis. Previous multi- center prospective observational study in Taiwan reveals that the prevalence of vitamin D deficiency in critically ill patients in northern Taiwan is 59%, and the prevalence of severe vitamin D deficiency is 18%. Several prediction models of vitamin D deficiency had been built for the general population but not patients admitted in intensive care units.

This multi-center retrospective study aims to develop and validate a score-based prediction model for severe vitamin D deficiency in critically ill patients. Investigators will review the data of previous multicenter, prospective, observational study. For temporal validation, the data will be divided into a derivation cohort (first 80% of the data set based on chronology) and a validation cohort (the remaining data set). The development and validation of the models will be carried out following the recommendations established in the Transparency Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) initiative.

Inclusion Criteria:

• Patients admitted to intensive care units

Exclusion Criteria:

• aged < 20 years
• admitted to intensive care unit longer than 28 days
• body max index < 18 kg/m2
• receive vitamin D supplement > 3000 IU/day
• previously admitted to intensive care unit within 3 months
• have parathyroid disease, rickets, or severe liver cirrhosis [Child C]

Critically ill patients