Background/Objectives: While the descriptive trajectory of COVID-19 is well-documented, there is a methodological gap in quantifying the precise magnitude of risk reduction across multi-year pandemic phases in Brazilian subnational units. This study aimed to fill this gap by applying Generalized Linear Models (GLMs) to quantify the temporal transition of epidemiological risks (Incidence, Mortality, and Case Fatality) in Ceará (2020–2023), using the first year of the pandemic as a statistical baseline. Methods: Ecological time-series study was conducted using official surveillance data. We employed GLMs with Poisson distribution to calculate Rate Ratios (RRs) and 95% Confidence Intervals, allowing for a robust comparative risk modeling between 2020 (reference) and subsequent years (2021–2023). Results: Modeling revealed a significant epidemiological dissociation between transmission and severity. While the risk of incidence remained high through 2022 (RR = 1.42), the mortality risk showed an earlier and more drastic decline, with a 68% reduction as early as 2022 (RR = 0.32) and 99% in 2023 (RR = 0.01). The Case Fatality Rate (CFR) risk decreased consistently from 2021 onwards, reaching its lowest point in 2023 (RR = 0.09; 91% reduction). Conclusions: Between 2020 and 2023, Ceará transitioned to reduced COVID-19 severity. Despite ecological design and data limitations, these findings underscore the importance of resilient health systems and equitable immunization.