75% of Memory Clinic Patients Have Undiagnosed Sleep Apnea

Sleep disturbances are increasingly recognized as modifiable risk factors for dementia, yet they remain largely undetected in the clinical settings where at-risk patients are most concentrated — memory clinics. This study from the Healthy Brain Ageing Clinic in Sydney, Australia, represents one of the largest and most comprehensive examinations of sleep disorders in a memory clinic population, using a multi-modal approach combining self-report questionnaires, wrist actigraphy, and full polysomnography (PSG) over a 15-year period from 2009 to 2024.

The sample comprised 1,234 adults aged 50 and older (mean age 67.2 years, 46% male) presenting with cognitive concerns. Participants were classified as having subjective cognitive impairment (SCI), mild cognitive impairment (MCI — amnestic or non-amnestic), or early dementia using standardized consensus criteria. Sleep was assessed across multiple dimensions: OSA via PSG and prior clinical history; sleep duration and circadian phase via both self-report and actigraphy; and insomnia, sleep quality, and RBD symptoms via validated questionnaires including the Insomnia Severity Index, Pittsburgh Sleep Quality Index, and REM Sleep Behavior Disorder Screening Questionnaire.

The prevalence findings were striking. Polysomnography revealed that 75.3% of participants met criteria for OSA, yet only 12.7% had a prior diagnosis — meaning the vast majority were undetected. Insomnia symptoms affected 12.0% of participants, 54.3% reported poor sleep quality, and 14.2% endorsed symptoms consistent with RBD. Critically, self-report and actigraphy measures of sleep duration showed poor concordance. Self-reported short sleep (≤6 hours) was reported by 30.5% of participants, compared to only 8.5% by actigraphy. Similarly, self-reported long sleep (≥9 hours) was endorsed by 10.2%, versus 5.1% by actigraphy. This systematic discrepancy suggests patients in this population significantly misperceive their own sleep duration, which has major implications for clinical reliance on self-report alone.

Regarding cognitive associations, OSA was significantly linked to impaired global cognition (MMSE) and verbal memory (RAVLT delayed recall, p<0.05). Prolonged sleep duration was the most robust predictor of cognitive dysfunction, associated with deficits in global cognition, processing speed (TMT-A), verbal memory (RAVLT), and executive function (TMT-B), all at p<0.05. Multinomial logistic regression further showed that long sleep duration significantly predicted a higher likelihood of being classified as amnestic MCI (aMCI) compared to SCI. Unexpectedly, poor sleep quality was associated with better memory performance, a counterintuitive finding the authors suggest may reflect a reporting bias or compensatory hyperarousal in this population.

The clinical implications are substantial. Given that pathological changes underlying dementia begin 10–20 years before symptom onset, and that memory clinic attendees are already in a high-risk window, undetected sleep disorders represent a missed opportunity for intervention. The authors argue that routine, objective sleep screening — including PSG or at minimum validated actigraphy — should be integrated into memory clinic workflows. Relying on self-report alone is insufficient and potentially misleading. Treating OSA and addressing sleep duration abnormalities may offer a meaningful, modifiable pathway to slow cognitive decline in this vulnerable population.