Objective criteria for predicting survival of chronic grass sickness cases are currently lacking. To determine whether the rate and/or magnitude of bodyweight change during hospitalisation of chronic grass sickness cases can provide an objective predictor of survival to discharge from hospital. Clinicians' recorded indication(s) for euthanasia were also reviewed. Single centre retrospective observational study. Case records of all horses admitted for management of chronic grass sickness to The Dick Vet Equine Hospital between 1998 and 2013 were analysed. Case background, survival to hospital discharge, indication(s) for euthanasia, disease duration at admission and bodyweight changes during the hospitalisation period were analysed, and data for survivors and nonsurvivors compared. Percentage weight change was calculated for 7 day intervals up to 28 days (0-7, 7-14, 14-21, 21-28 days) and for entire periods from the first weight recorded (0-7, 0-14, 0-21, 0-28 days). These results were used to estimate survival probability conditional on weight change. The study sample comprised 213 horses, with 114 survivors (53.5%) and 99 (46.5%) nonsurvivors. Compared with nonsurvivors, survivors had significantly lower median maximum bodyweight loss as a percentage of first weight (survivors 5.9%, interquartile range 1.8-13.5; nonsurvivors 12.7%, 6.4-17.3). Throughout all time periods analysed, survivors had significantly lower median bodyweight loss than nonsurvivors, but no specific time period was more predictive of survival. Highest percentages of total bodyweight loss for individual horses were comparable for survivors (36%) and nonsurvivors (37%). Survival prediction curves reporting percentage survival rates for all time periods analysed provided data to aid prediction of chronic grass sickness survival. Overall, nonsurvivors had greater bodyweight loss than survivors. Rapidity and magnitude of bodyweight loss were equally predictive of outcome. Percentage survival prediction curves provide objective data to aid discussion of prognosis, but greater predictive specificity with associated sensitivity is required for clinical decision making in individual cases.