Objective To explore the peripheral blood immunoinflammatory indicators that can distinguish bipolar depression from unipolar depression in the early stage of the disease， and build a prediction model to provide evidence for the early identification and early treatment of bipolar depression. Methods Patients who have been hospitalized at least 2 times in Beijing Anding Hospital， Capital Medical University between January 2013 and December 2019 were recruited as research subjects. A total of 487 patients diagnosed with major depressive disorder at the first admission were included， who were divided into unipolar depression group （357 cases） and bipolar depression group （130 cases） according to whether they were diagnosed with bipolar disorder in the subsequent admission records. 102 patients were included in each of the two groups by the propensity score matching method. The demographic data and immunoinflammatory factors of the first hospitalization were compared between the two groups. The selected immunoinflammatory factors include neutrophil count to lymphocyte count ratio （NLR）， monocyte count to lymphocyte count ratio （MLR），platelet count to lymphocyte count ratio （PLR）， C-reactive protein （CRP）， complement 3 （C3）， complement 4 （C4）， immunoglobulin A （IgA）， immunoglobulin G （IgG） and immunoglobulin M （IgM）. Binomial Logistic regression analysis was used to control confounding factors to explore the early predictors for bipolar disorder. Receiver operating characteristic （ROC） curve was used to analyze the early predictive value of the selected immunoinflammatory factors for bipolar disorder. Results There were statistical differences in NLR， MLR， C3 and CRP between unipolar depression group and bipolar depression group （Z=2.004，3.062，2.333，2.233； P ＜ 0.05）. Binomial Logistic regression analysis showed that MLR （OR=1.63， 95%CI=1.206-2.194） and C3 （OR=1.195， 95%CI=1.033-1.383） were early predictors of bipolar depression. ROC analysis showed that the AUC of Logistic regression model for predicting bipolar depression was 0.669， with a sensitivity of 78.4% and a specificity of 53.9%. Conclusions Elevated MLR and C3 levels may be an early predictor of bipolar depression