To perform therapeutic drug monitoring of total and free plasma valproic acid (VPA) concentrations in clinical samples and to analyze the related factors. The total VPA concentration in plasma was determined by ultrahigh-performance liquid chromatography with precolumn derivatization with α-bromoacetophenone, and the free VPA concentration was determined by liquid chromatography-tandem mass spectrometry after the plasma was treated by hollow fiber centrifugal ultrafiltration. Regression analysis was performed to examine the associations between free plasma VPA, total plasma VPA, and the plasma protein binding rate. The impact of individual situations, outpatient or inpatient factors, and drug combinations on VPA concentrations were examined. Of the 569 clinical samples, 268 were inpatients and 301 were outpatients, and the total VPA concentration in 138 cases (24.2%) was lower than the effective treatment concentration range; the total and free VPA concentrations in outpatient samples were 11.0% and 26.1% higher than those of inpatients, respectively. There was no linear relationship between the free and total VPA concentrations. The relationship equation between the plasma protein binding rate and free VPA concentrations was as follows: Y = 0.0255X2 - 1.1357X + 97.429 (r = 0.8011). The total and free VPA concentrations were significantly decreased after the coadministration of phenobarbital (83.7% and 64.3% of the control group, P < 0.05) or carbapenem antibiotics (32.0% and 32.7% of the control group, P < 0.05). The total VPA concentrations in patients with epilepsy at our hospital was lower than the effective treatment concentration range, which was inadequate for epilepsy control; the total VPA concentrations of outpatients were higher than those of inpatients; as phenobarbital affects VPA metabolism, therapeutic drug monitoring is recommended. Carbapenem antibiotic coadministration with VPA should be avoided because carbapenem antibiotics can lead to the failure of VPA antiepileptic treatment.