Postoperative cognitive dysfunction (POCD) is a common postoperative complication observed in elderly patients. However, the diagnosis of POCD is not very satisfactory as no specific biomarkers have been classified. It is necessary to identify new diagnostic markers to better understand the pathogenesis of POCD. We performed liquid chromatography with a time-of-flight mass spectrometer- (LC/Q-TOF-MS-) based metabolomics study to investigate POCD. A total of 40 metabolites were differentially expressed between POCD and non-POCD patients. In this study, we investigated whether phosphatidylserine (PS) (17:2/0:0), with an area under the curve value of 0.966, was a potential sensitive and specific biomarker for the diagnosis and prognosis of POCD. Pathway analysis showed that fatty acid metabolism, lipid metabolism, and carnitine metabolism were significantly altered in POCD. Network analysis indicated that nitric oxide signaling, PI3K-AKT signaling, mTOR signaling, and mitochondrial dysfunction were related to the pathogenesis of POCD. This study showed that metabolic profiling was meaningful when studying the diagnosis and pathogenesis of POCD.