Prevention and early detection of metabolic and reproductive disorders can significantly reduce veterinary costs and loss of performance. The objective of the study was to predict these disorders in dairy cows using the models that include milk characteristics. A total of 1783 Holstein cows from three herds were involved in the study. Data from monthly milk performance tests (milk yield, milk composition including milk fatty acids) were recorded from September 2023 to December 2024. In addition, the health of the cows was monitored and diagnosed metabolic and reproductive disorders were documented. In total, metabolic disorders were detected in 33 primiparous (PP) and 77 multiparous (MP) cows, while reproductive problems were detected in 93 PP and 253 MP cows. Binary logistic regression models were developed to predict the incidence of disorders in PP and MP cows using milk yield and milk composition data as explanatory variables. Receiver operating characteristic (ROC) curves, which graphically represent the relationship between true positive and false positive rates, and area under the curve (AUC), which quantifies the accuracy of the model, were used to assess the predictive ability of the models. The Index of Union method, which minimizes the absolute differences between the AUC and the diagnostic measures (sensitivity and specificity), was used to determine the optimal cut-off values. For the prediction of metabolic disorders in PP cows, the resulting model included milk fat, C18:0, and C18:1 (AUC 0.81) and correctly identified 72.7% of affected animals when the optimal cut-off value was applied. In MP cows, a model based on milk protein, C16:0, C18:1 and C18:0 resulted in an AUC of 0.72 and correctly predicted 64.3% of metabolic disorders. With regard to reproductive disorders in PP cows, a model with milk fat, C18:0 and C18:1 (AUC 0.85) was able to detect 78.5% of the affected animals. Milk composition, particularly fatty acid content, appears to be a suitable indicator for predicting metabolic and reproductive disorders in Holstein dairy cows, and binary logistic regression has proven to be an effective method for developing prediction models.