To describe etiology, short-term outcomes and prognostic accuracy of serial PELD scores in PALF. Retrospective analysis of children aged =16 yr, admitted with PALF under the QLTS, Brisbane, Australia, between 1991 and 2011. PELD-MELD scores were ascertained at three time points (i) admission (ii), meeting PALF criteria, and (iii) peak value. Fifty-four children met criteria for PALF, median age 17 months (1 day-15.6 yr) and median weight 10.2 kg (1.9-57 kg). Etiology was known in 69%: 26% metabolic, 15% infective, 13% drug-induced, 6% autoimmune, and 9% hemophagocytic lymphohistiocytosis. Age <3 months and weight <4.7 kg predicted poor survival in non-transplanted children. Significant independent predictors of poor outcome (death or LT) were peak bilirubin > 220 孯L and peak INR > 4. Serial PELD-MELD scores were higher in the 17 (32%) transplant recipients (mean: [i] 26.8, [ii] 31.8, [iii] 42.6); highest in the 12 (22%) non-transplanted non-survivors (mean: [i] 31.6, [ii] 37.2, [iii] 45.7) compared with the 25 (46%) transplant-free survivors (mean: [i] 25.3, [ii] 26.0, [iii] 30.3). PELD-MELD thresholds of =27 and =42 at (ii) meeting PALF criteria and (iii) peak predicted poor outcome (p < 0.001). High peak bilirubin and peak INR predict poor outcome and serial PELD-MELD is superior to single admission PELD-MELD score for predicting poor outcome.