OBJECTIVES: This study examines the dynamics of out-of-pocket health care spending by looking at the persistence of such spending among Medicare beneficiaries. It provides insight into how long high health care out-of-pocket spending persists, its association with certain health conditions and health shocks, and identify the determinants of the likelihood that an individual will be a persistent high health care spender. METHODS: This study uses the Health and Retirement Study (HRS) from 2001 to 2010. Descriptive statistics show concentration and persistence of out-of-pocket health care spending. Logistic model is used to study the likelihood of persistently being in the top percentiles of out-of-pocket health care spending among elderly adults who were initially in the top percentiles and to identify determinants of such probabilities.  RESULTS: High out-of-pocket spending persisted for about six years and then remained fairly stable afterwards. Having a certain health condition such as diabetes, strokes or emotional problems increased the probability of persistence by up to 9 percentage points. Compared to elderly adults who were covered by only Medicare, people having other government health insurance such as Medicaid, CHAMPUS, VA, or other military programs were about 19 percentage points less likely to be a persistently high out-of-pocket spender. By contrast, people who were covered by Medicare and any supplementary health insurance were more likely to be a persistently high spender.  CONCLUSIONS: Non-trivial number of elderly adults remained in the same upper percentiles of the distribution for multiple years. The crucial factors that triggered high out-of-pocket spending on health care for several years were having certain chronic conditions or experiencing the onset of new health condition, even after controlling individual characteristics. Additionally, having an existing health insurance that supplements Medicare coverage or the acquisition of a new supplementary health insurance had a significant impact on the probability of persistence.