Objective: To understand what impact hospital inpatient occupancy levels have on patient throughput by analysing one hospital's occupancy levels and the rate of patient discharge. Methods: A four-stage model was fit to hospital admission and separation data and used to analyse the per-capita separation rate according to the patient load and the impact of hospital over-census actions. Results. Per-capita separation rates are significantly higher on days when the hospital declares an over-census due to emergency department crowding. Per-capita separation rates are also higher or lower on days with 8-10% higher or lower patient loads, respectively, but the response is not nearly as strong as the response to an over-census declaration, and is limited to patients with an elapsed stay of 10 days or more. Within the medical division there is an increase in per-capita separation rates on over-census days, but no significant difference in per-capita release rates for different patient loads. Within the surgical division there is no significant difference in per-capita separation rates on over-census days compared with other days, but the patient load does make a significant difference. Conclusion: Staff do discharge a greater proportion of long-stay patients when the hospital is experiencing high demand and a lower proportion when occupancy is low, but the reasons driving those changes remains unclear.