Survival Outcomes and Machine Learning-Based Prediction of 12-Month Mortality in Glioblastoma Before and During the COVID-19 Pandemic: A SEER Population-Based Study

Background and Objectives: The COVID-19 pandemic disrupted cancer diagnosis and treatment pathways worldwide. Glioblastoma is an aggressive primary brain malignancy requiring timely multimodal care. This study evaluated survival outcomes among glioblastoma patients diagnosed before and during the COVID-19 pandemic and prepared a dataset for machine learning-based prediction of 12-month mortality. Materials and Methods: Patients aged ≥20 years diagnosed with glioblastoma between 2018 and 2021 were identified from the SEER database using ICD-O-3 histology codes 9440/3, 9441/3, and 9442/3. Patients were categorized as pre-COVID period (2018–2019) or COVID period (2020–2021). OS and CSS were evaluated using Kaplan–Meier curves, log-rank tests, and Cox regression models. Machine learning models predicted 12-month all-cause mortality using registry variables. Results: The final cohort included 9914 patients; 4819 were diagnosed pre-COVID and 5095 during COVID. Median OS was 11 months pre-COVID and 10 months during COVID; 12-month OS was 44.3% and 41.2%, respectively. Median CSS was 11 months in both periods; 12-month CSS was 46.9% and 44.1%, respectively. COVID-period diagnosis was modestly associated with poorer OS (adjusted HR 1.050, 95% CI 1.006–1.095, p = 0.025) and CSS (adjusted HR 1.048, 95% CI 1.003–1.095, p = 0.035). Machine learning models showed moderate discrimination for 12-month mortality prediction. Conclusions: Glioblastoma patients diagnosed during the COVID period had modestly poorer OS and CSS in conventional survival analyses; however, competing-risk analysis did not show a significant association with cancer-specific death. Registry-based machine learning models provided moderate 12-month mortality prediction, supporting their potential utility for population-level prognostic assessment.