The study analyzed the dynamics of platelet counts in 280 patients with sepsis admitted to the intensive care unit in 2022–2024 using group trajectory modeling (GBTM) during the first three days. Three trajectories were identified: a persistent low (71.79%), a decline in the high (20.00%), and an increase in rebound (8.21%). Patients with persistently low levels had the highest 28-day mortality (71.63%) versus 55.36% and 13.04% in the other groups (p < 0.05). After adjustment, Cox regression showed that the group with a high level decrease had a 42% lower risk of death (HR = 0.58, 95% CI: 0.36–0.92, p = 0.02) and the group with an increase 93% lower (HR = 0.07, 95% CI: 0.02–0.26, p < 0.001). Mediation analysis demonstrated that the effect of trajectories on mortality is partially mediated by changes in SOFA scores and log-transformed APTT. These findings support better risk stratification and personalized sepsis care.