Most dairy cows experience a certain degree of negative energy balance (NEB) after calving, which may result in various health problems. The aim of the study was to predict the incidence of NEB in Holstein cows using milk components collected during milk performance tests conducted in the period from day 5 to day 35 of lactation. A total of 732 cows from 5 herds were used in the experiment. Depending on the blood concentration of non-esterified fatty acids (threshold 0.6 mmol/l) they were considered healthy (n = 621) or in NEB (n = 111). A binary logistic regression model was developed to predict NEB incidence from milk yield and milk components including milk fatty acids as explanatory variables. The results of the regression analysis applied to the data from individual herds and from all herds combined showed that milk fat content and milk C18:1 and C18:0 concentrations were the main significant factors (P < 0.05) in the diagnosis of NEB. Receiver operating characteristic (ROC) curves were then constructed to evaluate the performance of the model and to visualize its use as a diagnostic test. The area under curve (AUC) was used as a measure of the diagnostic accuracy of the model outcome, with an AUC of 1 meaning perfect diagnosis and an AUC of 0.5 indicating no discriminative ability. When milk fat content and milk C18:1 and C18:0 concentrations were included in the model applied to the dataset containing the cows from all herds, the resulting AUC value was 0.82, indicating quite high diagnostic ability. The optimal prediction cut-point was determined by Index of Union method. The resulting sensitivity and specificity values were 0.73 and 0.73, respectively. This means that the test detected 73% of real positive NEB cases, while the false positive rate was 27%, which makes it reasonably useful for on-farm diagnosis of energy-imbalanced cows. It was concluded that milk fat, C18:1 and C18:0 could be used as biomarkers potentially applicable for screening the incidence of NEB in early lactation Holstein cows. However, the results need to be validated by further research and large-scale studies.