Objective To investigate perioperative attention network in children with emergence delirium (ED) under sevoflurane anesthesia and adverse behavioral changes within postoperative two weeks. Methods A total of 110 children, aged 4 to 12 years, who underwent tonsil/adenoidectomy under sevoflurane anesthesia were selected. According to the occurrence of ED, they were divided into two groups: a delirium group (n=42) and a non‑delirium group (n=68). Anesthesia induction was performed in both groups using 6%−8% sevoflurane through inhalation and sufentanil at 0.4 μg/kg, sulfenammonium cisatracurium at 0.2 mg/kg, and dexamethasone at 0.1 mg/kg through intravenous injection, followed by maintenance of anesthesia using 2%−3% sevoflurane through inhalation, and remifentanil at 0.1−0.5 μg·kg−1·min−1 through intravenous infusion. Their general information [sex ratio, age, body mass index, the modified Yale Preoperative Anxiety Scale (mYPAS) score before surgery, and mask acceptance score], time to extubation, anesthesia duration, operation duration, the length of post‑anesthesia care unit (PACU) stay, the Faces, Legs, Activity, Cry and Consolability (FLACC) scores (the highest scores), and the use of rescue analgesics. Their attention network efficiency (including sub‑networks such as alerting, orienting, and executive control), correctness rates, and mean reaction time (RT) were recorded before surgery and on postoperative day 1. Furthermore, the Post Hospitalization Behavior Questionnaire for Ambulatory Surgery (PHBQ‑AS) scores were evaluated through telephone follow‑up at postoperative two weeks to evaluate behavioral changes in the children. Results There were no statistical differences in general information between the two groups (all P>0.05). No statistical differences were found in time to extubation, anesthesia duration, operation duration, the length of PACU stay, the FLACC score (the highest scores), or the use of rescue analgesics (all P>0.05). However, compared with those before surgery, the efficiency of executive control network was significantly weakened in the delirium group on postoperative day 1 (P<0.05). Compared with the non‑delirium group, the executive control network efficiency in the delirium group was weakened on postoperative day 1 (P<0.05). There were no statistical differences in the alerting and orienting network efficiency, correctness rates and mean RT between the two groups (all P>0.05). The PHBQ‑AS score in the delirium group was higher than that in the non‑delirium group at postoperative two weeks (P<0.05). Furthermore, no statistical differences were found in adverse behavioral changes (the percentages of children with improved, worsened, or unchanged behaviors) at postoperative two weeks between the two groups (all P>0.05). Conclusions The efficiency of executive control network in children with ED does not return to baseline levels on postoperative day 1, with increased PHBQ‑AS scores at postoperative two weeks.