This paper assesses the extent that systematic mortality improvement varies with individual risk characteristics. To do this, a Lee-Carter model is used to assess if mortality improvement varies for groups of individuals with similar risk characteristics along with an individual mortality model that allows for heterogeneity with time trends to assess systematic risk. Data from the U.S. Health and Retirement Study (HRS) is used since this provides longitudinal, individual level data. Our results are highly relevant to life insurers, pension funds and regulators assessing the future impact of improvement trends in mortality on their premiums and liabilities. Mortality trends differ across individuals reflecting the different risk factors and particularly the prevalence of different diseases such as high blood pressure, cancer and heart problems. Models that are based on aggregate population level trends and differing only by gender and age are not adequate in quantifying mortality trends and risks.