Medicare claims data are used to model hospital utilization patterns for the 4,660 survivors and 2,867 decedents of the Longitudinal Study on Aging (LSOA). When the volume of hospital utilization was collapsed into four categories based on the mean annual number of hospital episodes and consistency was defined as a maximum absolute deviation from that mean of 1.5 or less, 42.6 percent of the LSOA respondents were found not to have any hospital admissions, and another 24.7 percent were found to be consistently low users. Only 4.8 percent were consistently high users, with an additional 6.8 percent being inconsistently high users. Multiple regression identified prior physician and hospital utilization, as well as poor perceived health, as the most important predictors of the volume and consistency of hospital utilization among survivors, as well as of the volume of hospital utilization among decedents. Among decedents, consistency was primarily related to geographic region and prior physician utilization.