Creatinine is a key biomarker for diagnosing and monitoring kidney disease, so rapid and sensitive testing is very important. Raman spectroscopy is particularly suitable for quantitatively detecting the creatinine in the human environment because it is sensitive to subtle changes in the concentration of the analyte. In this work an effective strategy is provided to promote the activity of surface-enhanced Raman scattering spectroscopy by enhancing the photon-induced charge transfer efficiency at low temperature. The nano-gold icosahedron (Au ₂₀) is obtained by the seed-growing method, which is used as an active substrate for SERS. The ultra-low temperature (98 K) SERS detection technology is used to realize the rapid and sensitive detection of the dye molecule crystal violet (CV) and creatinine in normal saline. The experimental results show that at room temperature of 296 K, the detection limit of Au ₂₀substrate for CV molecules is as low as 10 ^–12mol/L, and the signals are uniform; at a low temperature of 98 K, the detection limit of CV molecules can reach 10 ^–14mol/L, which is two orders of magnitude lower than that at 296 K. As a result, the adopted cryogenic temperature can effectively weaken the lattice thermal vibration and reduce the release of phonons, then suppress phonon-assisted non-radiative recombination. So, it will increase the number of photo-induced electrons to participate in the photo-induced charge transfer efficiency. Finally, we perform the label-free detection of creatinine in saline by using an Au ₂₀substrate. The results show that the detection limit of the SERS substrate for creatinine is 10 ^–6mol/L at 296 K, and the linear correlation coefficient of the 1619 cm ^–1peak is 0.9839. At a low temperature of 98 K, the detection limit of creatinine concentration is as low as 10 ^–8mol/L, and the linear correlation coefficient of the 1619 cm ^–1peak becomes 0.9973. It can be seen that low temperature may further improve the detection limit of creatinine concentration and the linearity of characteristic peak. In summary, the current work provides a new idea for accurately detecting the creatinine concentration in the field of biomedicine.