AimAlthough the relationships between duration of untreated psychosis (DUP) and outcomes are often assumed to be linear, few studies have explored the functional form of these relationships. The aim of this study is to demonstrate the potential of recent advances in curve fitting approaches (splines) to explore the form of the relationship between DUP and global assessment of functioning (GAF).MethodsCurve fitting approaches were used in models to predict change in GAF at long-term follow-up using DUP for a sample of 83 individuals with schizophrenia.ResultsThe form of the relationship between DUP and GAF was non-linear. Accounting for non-linearity increased the percentage of variance in GAF explained by the model, resulting in better prediction and understanding of the relationship.ConclusionThe relationship between DUP and outcomes may be complex and model fit may be improved by accounting for the form of the relationship. This should be routinely assessed and new statistical approaches for non-linear relationships exploited, if appropriate.