After several years of implementation, the original Welfare Quality scoring model for dairy cows appears to be highly sensitive to the number and cleanliness of drinkers and not enough to the prevalence of diseases, and as a consequence may not fit the opinion of some animal welfare experts. The present paper aims to improve the Welfare Quality calculations for the criteria ‘Absence of prolonged thirst’ and ‘Absence of disease’ in dairy cows, so that the results are more sensitive to input data and better fit experts’ opinion. First, we modified the calculation of ‘Absence of prolonged thirst’ by linearizing the calculation for drinkers’ availability to avoid threshold effects. Second, we modified the calculation of ‘Absence of disease’ by applying a Choquet integral on the three lowest spline-based scores for each health disorder to limit compensation between health disorders. Third, we performed a global sensitivity analysis of the original and the alternative scoring models. Fourth, we compared the results obtained with the original and the alternative models with 8 experts’ opinions on two subsets composed of 44 and 60 farms, respectively, inspected using the Welfare Quality protocol and on which experts gave their opinion on the overall level of animal welfare. Results show that the alternative model significantly reduced the ‘threshold effects’ related to the number of drinkers and the compensation between health disorders. On the first subset, the alternative model fits the experts’ opinion slightly better than the original model (P = 0.061). On the second subset, the models performed equally. In conclusion, the proposed refinements for calculating scores are validated since they significantly reduced ‘threshold effects’ and the influence of measures related to drinkers. It also reduced the compensation between health disorders by considering only the three lowest scores and thus increasing the influence of measures related to health disorders, and slightly improve at overall score level the accordance with experts’ opinion.