Well-known biases associated with hedge fund data have created difficulties in analysing the factors behind the failure of hedge funds. This study was interested in quantifying the factors influencing hedge fund failure and to construct a framework for forecasting the time varying probability of survival for individual hedge funds. Based on a robust filter for the identification of failure times, a Cox proporational hazards model was used to determine the effects of historical performance and fund characteristics on true hedge fund failure. Covariates included fund size, return measures, leverage , strategy, liquidity, minimum investment, fee structure and domicile. By adopting an evaluation procedure developed from the theory of signal detection, it was found that the specification of the Cox proportional hazards model incorporating fixed factors had predictive skills for forecasting the occurrence of failure in individual hedge funds. Importantly, the model is robust against variations in covariate definition, evaluation timing, data filter as well as the thresholds used to identify failure times.