Steroid‐refractory (SR) acute graft‐versus‐host disease (aGVHD) is one of the leading causes of early mortality after allogeneic hematopoietic stem cell transplantation (allo‐HSCT). We investigated the efficacy, safety, prognostic factors, and optimal therapeutic protocol for SR‐aGVHD patients treated with basiliximab in a real‐world setting. Nine hundred and forty SR‐aGVHD patients were recruited from 36 hospitals in China, and 3683 doses of basiliximab were administered. Basiliximab was used as monotherapy (n = 642) or in combination with other second‐line treatments (n = 298). The cumulative incidence of overall response rate (ORR) at day 28 after basiliximab treatment was 79.4% (95% confidence interval CI 76.5%–82.3%). The probabilities of nonrelapse mortality and overall survival at 3 years after basiliximab treatment were 26.8% (95% CI 24.0%–29.6%) and 64.3% (95% CI 61.2%–67.4%), respectively. A 1:1 propensity score matching was performed to compare the efficacy and safety between the monotherapy and combined therapy groups. Combined therapy did not increase the ORR; conversely, it increased the infection rates compared with monotherapy. The multivariate analysis showed that combined therapy, grade III–IV aGVHD, and high‐risk refined Minnesota aGVHD risk score before basiliximab treatment were independently associated with the therapeutic response. Hence, we created a prognostic scoring system that could predict the risk of having a decreased likelihood of response after basiliximab treatment. Machine learning was used to develop a protocol that maximized the efficacy of basiliximab while maintaining acceptable levels of infection risk. Thus, real‐world data suggest that basiliximab is safe and effective for treating SR‐aGVHD.