Rapid shallow breathing index (RSBI, respiratory frequency [f] divided by tidal volume [Vt]) has been used to prognosticate liberation from mechanical ventilation (LMV). We hypothesize that dynamic changes in RSBI predict failed LMV better than isolated RSBI measurements. We conducted a retrospective study of patients who were mechanically ventilated (MV) for longer than 72 hours. Failed LMV was defined as need for reinstitution of MV within 48 hours post-LMV. Ventilatory frequency (f) and Vt (liters) were serially recorded. The instantaneous RSBI (i-RSBI) was defined as f/Vt. Dynamic f/Vt ratio (d-RSBI) was defined as the ratio between two consecutive i-RSBI (f/Vt) measurements ([f(2)/Vt(2)]/[f(1)/Vt(1)]). RSBI Product (RSB-P) was defined as (i-RSBI × d-RSBI). Data from 32 patients were analyzed (Acute Physiology and Chronic Health Evaluation II 13.4, male 69%, mean age 57 years). Mean length of stay was 19.5 days (11.5 ventilator; 14.1 intensive care unit days). For LMV failures, mean time to reinstitution of invasive MV was 20.8 hours. All patients had pre-LMV i-RSBI less than 100. Failed LMVs had higher i-RSBI values (68.9, n = 18) than successful LMVs (44.2, n = 23, P < 0.01). Failures had higher d-RSBI (1.48) than successful LMVs (1.05, P < 0.04). The RSB-P was higher for failed LMVs (118) than for successful LMVs (48.8, P < 0.01) with failures having larger proportion of pre-LMV d-RSBI values greater than 1.5 (39.0 vs 10.7%, P < 0.03). Pre-LMV RSB-P may offer early prediction of failed LMV in patients on MV for longer than 72 hours despite normal pre-LMV i-RSBI. Divergence between RSB-P for successful and failed LMVs occurred earlier than i-RSBI divergence with a greater proportion of pre-LMV d-RSBI greater than 1.5 among failures.