Vitamin D is commonly supplemented to dairy cows as vitamin D3 to support calcium homeostasis and in times of low sunlight exposure. Vitamin D has beneficial immunomodulatory and anti-inflammatory properties. Serum 25-hydroxyvitamin D [25(OH)D] concentrations fluctuated during lactation, with the lowest concentrations measured in healthy cows within 7 d of calving. However, it is unknown if serum 25(OH)D concentrations measured during the previous lactation are associated with transition diseases or health risk factors in dairy cattle. We collected serum samples from 279 dairy cattle from 5 commercial dairy herds in Michigan at dry-off, close-up, and 2-10 d in milk (DIM). Vitamin D concentrations were determined by measuring serum 25(OH)D by radioimmunoassay. Total serum calcium was measured by colorimetric methods. Body condition scores (BCS) were assigned at the time of blood collection. Clinical disease incidence was monitored until 30 d postparturition. Separate bivariable logistic regression analyses were used to determine if serum 25(OH)D at dry-off, close-up, and 2-10 DIM was associated with various clinical diseases including mastitis, lameness, and uterine disorders (classified as metritis, retained placenta, or both) and increased urine ketone concentrations at P < 0.05. Among all significant bivariable analyses, multivariable logistic regression analyses were built to adjust for potential confounding variables including parity, BCS, season, and calcium. Receiver operator characteristic (ROC) curve analyses were used to determine optimal concentrations of serum 25(OH)D. We found that higher serum 25(OH)D concentrations at dry-off and close-up predicted increased urine ketone concentrations in early lactation, even after adjustment for confounders. Alternatively, we found that lower serum 25(OH)D at 2-10 DIM was associated with uterine diseases. Optimal concentrations for serum 25(OH)D at dry-off and close-up for lower risk of increased urine ketone concentrations were below 103.4 and 91.1 ng/mL, respectively. The optimal concentration for serum 25(OH)D at 2-10 DIM for uterine diseases was above 71.4 ng/mL. These results indicate that serum 25(OH)D at dry-off and close-up may be a novel predictive biomarker for increased urine ketone concentrations during early lactation. Increased urine ketone concentrations are not necessarily harmful or diagnostic for ketosis but do indicate development of negative energy balance, metabolic stress, and increased risk of early lactation diseases. Predicting that dairy cattle are at increased risk of disease facilitates implementation of intervention strategies that may lower disease incidence. Future studies should confirm these findings and determine the utility of serum 25(OH)D concentrations as a predictive biomarker for clinical and subclinical ketosis.