Objective To evaluate the effect of preoperative cognitive function training on postoperative delirium (POD) in frail elderly patients. Methods A total of 88 frail elderly patients, aged 65−75 years, with body mass index (BMI) of 20−30 kg/m2 and American Society of Anesthesiologists (ASA) grade Ⅱ or Ⅲ, who underwent elective laparoscopic radical gastrointestinal tumor surgery under general anesthesia were selected. According to the random number table method, they were divided into two groups: an intervention group (n=42) and a control group (n=46). The intervention group was subject to cognitive function training 5 d before surgery, 60 min per time, twice per day, while the control group was not given cognitive function training before surgery. All the patients underwent conventional preoperative surgical and anesthetic education. The concentrations of serum interleukin‑6 (IL‑6), tumor necrosis factor‑α (TNF‑α) and S100 calcium‑binding protein β (S100β) were measured on pre‑operative day 1 and on post‑operative day 1. The incidence of POD was assessed on post‑operative days 1, 3 d and 7 d using the Confusion Assessment Method (CAM). Their Montreal Cognitive Assessment (MoCA) scores, Self‑rating Anxiety Scale (SAS) scores and Self‑rating Depression Scale (SDS) scores were recorded 1 d before cognitive function training and 1 d before surgery. The incidence of postoperative nausea and vomiting (PONV), the incidence of pulmonary infection and the length of postoperative hospitalization stay were recorded. The postoperative Visual Analogue Scale (VAS)score, the rate of rescue analgesia, and the number of effective and ineffective patient‑controlled intravenous analgesia (PCIA) compressions were recorded. Results Compared with the control group, after cognitive function training, the intervention group showed increases in MoCA scores (P<0.05), and decreases in SAS and SDS scores (all P<0.05). Compared with those 1 d before cognitive function training, the intervention group presented increases in MoCA scores 1 d before surgery (P<0.05), and decreases in SAS and SDS scores (all P<0.05). Compared with the control group, the intervention group showed significant reduction in the levels of serum IL‑6, TNF‑α, and S100β 1 d after surgery (all P<0.05). Compared with those 1 d before surgery, the levels of serum IL‑6, TNF‑α, and S100β in the two groups significantly increased 1 d after surgery (all P<0.05). Compared with the control group, the intervention group showed significant decreases the incidence of POD on post‑operative days 1 and 3 and the total incidence of POD (all P<0.05), with shortened length of postoperative hospitalization stay (P<0.05). There were no statistical differences in the incidence of POD 7 d after surgery, PONV incidence, the rate of pulmonary infection, VAS score, the rate of rescue analgesia, and the number of effective and ineffective PCIA compressions between the two groups (all P>0.05). Conclusions Preoperative cognitive function training can relieve the postoperative inflammatory response in frail elderly patients, decrease the incidence of POD in patients, and shorten their length of hospitalization stay.