To investigate the efficacy of secondary antifungal prophylaxis (SAP) in patients with hematological diseases, all medical records within two consecutive years were retrospectively reviewed. In all, 57 patients with a history of invasive fungal infections received 149 cycles of further therapy for their underlying hematological diseases. Logistic regression and recursive partitioning were used to discriminate high risk patients for failure of SAP. After a median follow-up period of 120 (12-1,080) days, 11 cases (7.4/100 cycles) experienced failure of SAP, including 5 cases during allo-SCT and 6 cases during chemotherapy, corresponding to cumulative incidence at 24.5%. Multivariate logistic regression revealed two risk factors for failure of SAP: use of high dose corticosteroid (OR 13.5, 95% CI 3.09-58.6, P = 0.0005) and duration of neutropenia ≥ 14 days (OR 7.47, 95% CI 1.69-32.9, P = 0.008). Recursive partitioning found that patients with these two risk factors were in high risk, with SAP failure rate as high as 50.0%. In conclusion, use of high dose corticosteroid and duration of neutropenia ≥ 14 days were risk factors of SAP failure. Patients with the two risk factors concurrently were in high risk and needed special concern.