• horse;
  • colic;
  • survival analysis;
  • penalised Cox regression;
  • random effects modelling


Colic surgery is a frequently performed operation with high postoperative mortality. This study was undertaken to identify variables associated with decreased postoperative survival. We used data from 321 horse years of postoperative survival time to model the probability of survival following recovery from colic surgery. Continuous variables were modelled using a 6 variable, penalised Cox regression model. This demonstrated approximately linear relationships between survival and the following variables: increase in packed cell volume (PCV), intestinal resection length, time to surgery (interval between onset of colic and surgery) and duration of surgery. No significant decrease in survival was demonstrated with increasing age of the patient or with heart rate. The only categorical variable to be significantly associated with decreased survival was epiploic foramen entrapment. The final, fixed effects Cox proportional hazards model of postoperative survival included the variables epiploic foramen entrapment, PCV, resection length and duration of surgery, each variable adjusted for the nonlinear relationship with time to surgery. Residual variation in postoperative survival attributable to professional personnel (referring veterinary surgeon, anaesthetist and surgeon) was explored by fitting each as a random effects term in the model. Little of the residual variation could be attributed to any category of personnel. Model diagnostics indicated little influence by individual outliers on model parameters and little evidence of subjects poorly predicted by the final model. The study highlights factors influencing the long-term survival of horses recovering from colic surgery and proposes a model that can be used to inform prognosis.