1142 Obstructive Sleep Apnea, Use of Positive Airway Pressure Therapy, and Dementia Incidence: A Longitudinal Study of Medicare Beneficiaries

Abstract Introduction Although obstructive sleep apnea (OSA) is linked to a higher incidence of dementia, additional data are needed on whether OSA represents a modifiable risk factor for dementia. We assessed the risk of incident dementia among Medicare beneficiaries with and without OSA, and among those with OSA who showed evidence vs. no evidence of adherent positive airway pressure (PAP) use. Methods Data were derived from fee-for-service claims from a random 20% sample of Medicare beneficiaries aged 65y and older who obtained care from 2010-2019 and were free of dementia at baseline. OSA and dementia diagnoses were identified with ICD-9/10 codes. PAP “adherent” users had ≥2 HCPCS codes for PAP supplies separated by ≥30 days and submitted during the first 3 years of the first identifiable claim. Generalized estimating equations were utilized to estimate sex- and race-specific cumulative dementia incidence by OSA status and PAP usage. Models were adjusted for age, race, ethnicity, and area deprivation index. Results Among 14,902,187 beneficiaries, OSA was present in 7% of women and 11% of men. Among beneficiaries with OSA, 35% of women and 39% of men were PAP adherent. By age 85, cumulative dementia incidence was 12.0% in women with OSA versus 9.0% in women without an established OSA diagnosis, and 10.3% versus 7.7% in men with and without an OSA diagnosis. Among beneficiaries with OSA, by age 85y, cumulative dementia incidence in PAP adherent vs. non-adherent use was 10.3% vs. 12.4% for women and 8.8% versus 10.7% for men, suggesting a relative risk reduction of 17.4% and 17.6% respectively. Conclusion OSA is associated with an increased risk of dementia, and exposure to adherent use of PAP is associated with partial mitigation of this risk. These results cannot prove cause-and-effect, but they strengthen existing evidence that OSA is a modifiable risk factor for dementia. Support (if any) This study was supported by funding from the NIH/NIA (R01AG074342, PIs Braley and Dunietz