Long-term care products cover the risk of permanent loss of autonomy. In the event of a loss of autonomy, the insurer must pay an annuity until the death or recovery of the insured. As recovery is very unlikely, long-term care product risk modellers in many countries assume that the only cause of ending annuity payments is death. Therefore, estimating the mortality of disabled insured individuals is crucial for insurers, as it significantly impacts the pricing and reserving of long-term care products. Multiple pathologies can lead to the loss of autonomy of an individual. Experience data show that the pathology responsible for the long-term care needs of an individual has a significant impact on its mortality. Therefore, accounting for the pathology is important, especially for reserving. However, the small number of data observations does not enable insurers to estimate a single mortality table for each pathology. In this paper, we present two clustering approaches to create groups of pathologies with similar mortality rates. This allows us to aggregate the data of different pathologies and reduce the number of different mortality tables without losing too much specific information on each pathology. The first method relies on GLM trees, while the second method is a generalized K-means approach. We then show that accounting for pathologies using the clusters obtained from the proposed methods improves the predictive performance of mortality. Finally, estimating different mortality rates according to pathology allows insurers to improve reserving and to study the impact of an increase or decrease in the incidence of a specific pathology on the mortality of disabled insured individuals.