Severity assessment in animals is an ongoing field of research. In particular, the question of objectifiable and meaningful parameters of score-sheets, as well as their best combination, arise. This retrospective analysis investigates the suitability of a score-sheet for assessing severity and seeks to optimise it for predicting survival in 89 male Sprague Dawley rats (Rattus norvegicus), during an experiment evaluating the influence of liver cirrhosis by bile duct ligation (BDL) on vascular healing. The following five parameters were compared for their predictive power: (i) overall score; (ii) relative weight loss; (iii) general condition score; (iv) spontaneous behaviour score; and (v) the observer’s assessment whether pain might be present. Suitable cut-off values of these individual parameters and the combination of multiple parameters were investigated. A total of ten rats (11.2%; 10/89) died or had to be sacrificed at an early stage due to pre-defined humane endpoints. Neither the overall score nor any individual parameter yielded satisfactory results for predicting survival. Using retrospectively calculated cut-off values and combining the overall score with the observer’s assessment of whether the animal required analgesia (dipyrone) for pain relief resulted in an improved prediction of survival on the second post-operative day. This study demonstrates that combining score parameters was more suitable than using single ones and that experienced human judgement of animals can be useful in addition to objective parameters in the assessment of severity. By optimising the score-sheet and better understanding the burden of the model on rats, this study contributes to animal welfare.