The Larson Miller (LM) parameter method is a commonly used prediction method for endurance strength experiments, which considers the high-temperature creep of the metal materials as a thermal activation process, expresses the steady-state creep rate using the Arrhenius equation, and extrapolates the long-term creep life through short-term accelerated creep tests. However, the LM parameter method usually overestimates the extrapolation results of creep life, and the larger is the creep life extrapolation ratio, the more is the overestimation. To improve the accuracy of the LM parameter method for extrapolating creep life, a long-term creep life extrapolation method based on an improved LM parameter method is proposed by introducing a modified parameter Cf into the LM parameter equation to consider the microstructure degradation and damage mechanism of the metal materials under high temperature with a long time. Then, the established extrapolation method was validated by using high-temperature creep life data of 316 series stainless steel materials in ASME and RCC-M specifications, respectively. The results show that it is obvious for the overestimation of creep life when the extrapolation ratios are about 1:3.3 and 1:10 by using the LM parameter method, and almost all data points were above the 30% error dispersion line. The extrapolated data points obtained by the improved LM parameter method can basically fall within the 30% error dispersion line, which depicts that the present method can significantly improve the accuracy of the extrapolated long-term creep life and it has high reliability and wide applicability to solve the overestimating problem of the extrapolated creep life caused by the changes in material constant C due to the long-term microstructural damage in high-temperature environments. In addition, the present method can provide strong support for predicting the long-term creep life of the metal materials in engineering.