Prognostic factors for response and survival in higher-risk myelodysplastic syndrome patients treated with azacitidine (AZA) remain largely unknown. Two hundred eighty-two consecutive high or intermediate-2 risk myelodysplastic syndrome patients received AZA in a compassionate, patient-named program. Diagnosis was RA/RARS/RCMD in 4%, RAEB-1 in 20%, RAEB-2 in 54%, and RAEB-t (AML with 21%-30% marrow blasts) in 22%. Cytogenetic risk was good in 31%, intermediate in 17%, and poor in 47%. Patients received AZA for a median of 6 cycles (1-52). Previous low-dose cytosine arabinoside treatment (P = .009), bone marrow blasts > 15% (P = .004), and abnormal karyotype (P = .03) independently predicted lower response rates. Complex karyotype predicted shorter responses (P = .0003). Performance status ≥ 2, intermediate- and poor-risk cytogenetics, presence of circulating blasts, and red blood cell transfusion dependency ≥ 4 units/8 weeks (all P < 10⁻⁴) independently predicted poorer overall survival (OS). A prognostic score based on those factors discriminated 3 risk groups with median OS not reached, 15.0 and 6.1 months, respectively (P < 10⁻⁴). This prognostic score was validated in an independent set of patients receiving AZA in the AZA-001 trial (P = .003). Achievement of hematological improvement in patients who did not obtain complete or partial remission was associated with improved OS (P < 10⁻⁴). In conclusion, routine tests can identify subgroups of patients with distinct prognosis with AZA treatment.