Circadian dysrhythmias occur commonly in critically ill patients reflecting variable effects of underlying illness, ICU environment, and treatments. We retrospectively analyzed the relationship between clinical outcomes and 24-h urinary 6-sulfatoxymelatonin (aMT6s) excretion profiles in 37 critically ill patients with shock and/or respiratory failure. Nonlinear regression was used to fit a 24-h cosine curve to each patient's aMT6s profile, with rhythmicity determined by the zero-amplitude test. From these curves we determined acrophase, amplitude, phase, and night/day ratio. After assessing unadjusted relationships, we identified the optimal multivariate models for hospital survival and for discharge to home (vs. death or transfer to another facility). Normalized aMT6s rhythm amplitude was greater (p = 0.005) in patients discharged home than in those who were not, while both groups exhibited a phase delay. Patients with rhythmic aMT6s excretion were more likely to survive (OR 5.25) and be discharged home (OR 8.89; p < 0.05 for both) than patients with arrhythmic profiles, associations that persisted in multivariate modelling. In critically ill patients with shock and/or respiratory failure, arrhythmic and/or low amplitude 24-h aMT6s rhythms were associated with worse clinical outcomes, suggesting a role for the melatonin-based rhythm as a novel biomarker of critical illness severity.