The prognostic nutritional index (PNI) has been applied in acute myocardial infarction (AMI) recently.However, the application of PNI in AMI needs verification. This was a prospective cohort study. Patients diagnosed with AMI were enrolled. PNI was calculated as (serum albumin (SA in g/L)) + (5 × total lymphocyte count (TLC) × 109/L). Modified PNI (mPNI) was analyzed by logistic regression analysis to reset the proportion of SA and TLC. The primary outcome was all-cause death. A total of 598 patients were enrolled; 73 patients died during follow-up. The coefficient of SA and TLC in the mPNI formula was approximately 2:1. The area under the receiver operating characteristic curve of SA, TLC, PNI, mPNI and GRACE in predicting death for patients with AMI was 0.718, 0.540, 0.636, 0.721 and 0.825, respectively. Net reclassification improvement (NRI) between PNI and mPNI was 0.230 (p < 0.001). Integrated discrimination improvement (IDI) was 0.042 (p = 0.001). Decision curve analysis revealed that mPNI had better prognostic value for patients with AMI than PNI; however, it was not superior to SA. Thus, PNI may not a reliable prognostic predictor of AMI; after resetting the formula, the value of PNI in predicting prognosis of AMI is almost entirely due to SA.