Abstract:Excessive daytime sleepiness (EDS) is linked to adverse clinical outcomes. This study evaluated changes in a validated tool to assess EDS, the Epworth Sleepiness Scale (ESS) and mortality risk. This retrospective cohort study included Veterans receiving sleep-related services in the Department of Veterans Affairs (VA) from October 4, 1999 to August 18, 2018, with two qualifying ESS measures. ESS values were extracted from patient notes using a validated natural language processing (NLP) pipeline (96% accuracy). ESS scores were categorised as Normal (0–10) or Abnormal (11–24). Patients were grouped based on ESS changes: Normal-Normal, Normal-Abnormal, Abnormal-Abnormal and Abnormal-Normal. Cox proportional hazards models adjusted for time, age, sex, race and comorbid conditions assessed the risk of 5-year all-cause mortality. Among 17,967 qualifying Veterans (mean age: 56.3 (SD 13.5) years), 11.75% died within 5 years of the second ESS measure. At baseline, 9342 (52.0%) had EDS, for whom 2232 (12.4%) improved to normal by the second exam (Abnormal-Normal). The Normal-Abnormal group had a 25% higher adjusted all-cause mortality risk within 5 years (aHR: 1.25, 95% CI: 1.09, 1.44) compared to the Normal-Normal group, with progressively increasing risk after age 55. In contrast, neither persistent abnormal sleepiness (Abnormal-Abnormal) nor improvement from abnormal to normal (Abnormal-Normal) was associated with significantly different mortality risk compared to the Normal-Normal group. ESS can efficiently identify EDS, which may serve as a clinical marker for 5-year all-cause mortality risk, particularly among Veterans seeking VHA sleep services aged 55 and older.