Historically, time preferences are modelled by assuming constant discounting, which implies a constant level of impatience. The prevailing empirical finding, however, is decreasing impatience (DI), meaning that levels of impatience decrease over time. Theoretically, such changes in impatience are crucial to understand behavior and self-control problems. Very few methods exist to measure DI without being restricted to or confounded by certain assumptions about the discounting function or utility curve. One such measure is the recently introduced DI-index, which has been applied to both monetary and health outcomes. The DI-index quantifies the deviation from constant impatience and is flexible enough to capture both increasing and decreasing impatience. In this study, we apply the DI-index to measure impatience for health outcomes in a reference-dependent framework. That is, we measure impatience for both health gains and health losses compared to a reference- point, in individual and societal settings, using a within-subjects design (n = 98). We allowed for both positive and negative discounting, since negative discounting has been observed for losses (i.e. preferring to incur losses earlier rather than later) in earlier work. To capture changes in time inconsistency when subjects show negative discounting (i.e. patience), we modify the DI-index to a decreasing (im)patience (DIP)-index, which can be applied without loss of generality. As in earlier work, we observe large heterogeneity in time consistency; i.e., a mix of decreasing, increasing and constant (im)patience. Across all DIPindices elicited, increasing impatience was the modal preference for those satisfying impatience, and decreasing patience for those satisfying patience. No systematic differences were observed between health gains and losses or between societal and individual outcomes. This suggests that for health outcomes both patient and impatient individuals assign more importance to time differences delayed further in the future.