This paper develops a model to assess the number of nurses needed to ensure both healthy patients and caregivers. We propose a model with random arrivals and exits of patients who may be of a single type (or several), and calculate the average care time they can receive. We show that the mean care time does not depend only on the mean number of patients in the unit. Actually, the probability distribution of the new patients per time step plays a central role. In the Poissonian case, we obtain totally explicit results, prove that the mean care time converges to a constant and give numerical examples. We also propose an analysis of the impact of working conditions on the average time that can be devoted to a patient. Four scenarios are proposed with numerical applications. Our analysis provides insights into current discussions on the introduction of caregiver ratios in hospitals to improve both the quality of care and caregivers’ working conditions.