Why do sleep disturbance prevalence estimates vary so widely across healthy older adults, MCI, and Alzheimer's disease?

The same person can count as a sleep case or not, depending on the ruler

"Sleep disturbance" is not one thing. A study may ask people whether they sleep badly, or it may record brain waves, eye movements, and breathing overnight in a lab. These capture different information, and they do not always agree. A systematic review of 71 studies found that objective measures gave more consistent results than subjective ones [1]. That matters for prevalence: if two studies use different instruments, or different cut-offs on the same instrument, they can report very different shares of people with a "sleep disturbance" even when the underlying sleep is similar.

The review also found that the specific sleep features most affected in MCI and Alzheimer's disease were REM sleep (the dreaming stage), sleep efficiency (how much of the time in bed is actually spent asleep), sleep latency (how long it takes to fall asleep), and total sleep duration [1]. A study that defines disturbance by one of these features will produce a different count than one that relies on a general complaint about sleep quality.

Sleep problems get more common and more severe as cognition declines

Across the cognitive spectrum, the direction is consistent: people with MCI and Alzheimer's disease show worse sleep than cognitively healthy older adults, on both subjective and objective measures [1]. The review concluded that the frequency and severity of sleep disturbance appear to follow the evolution of cognitive impairment [1]. In practical terms, a sample drawn from a memory clinic will usually yield higher rates than a sample drawn from the general community, because the clinic sample is further along that trajectory.

A large imaging study using Alzheimer's Disease Neuroimaging Initiative data found the same gradient: people with MCI and Alzheimer's disease had more sleep disturbance than normal controls across brain regions examined [2]. That study did not report prevalence estimates, but it supports the idea that the groups differ in how much sleep disturbance is present, not just in whether it is present.

Who gets recruited changes the rate as much as the diagnosis does

Population setting matters. A study of older adults in rural China examined sleep disturbance in people with and without MCI and reported prevalence estimates for both groups [3]. Rates from a rural community sample are not directly comparable to rates from a volunteer research cohort or a specialist clinic, because the people, their environments, and their willingness to report problems all differ. This is one reason two studies of "MCI" can publish quite different numbers.

Within MCI and Alzheimer's disease, biological factors may also sort people into higher- and lower-sleep-problem subgroups. An integrative review of seven studies found that carrying the APOE ε4 gene — a known risk factor for MCI and Alzheimer's disease — was associated with poorer sleep quality, including shorter total sleep time, reduced REM sleep, lower sleep efficiency, longer sleep latency, and more time awake after falling asleep [4]. The same review noted that the combined relationship between sleep disturbance and APOE ε4 on cognitive decline was inconsistent across studies [4]. If a sample happens to contain more ε4 carriers, its sleep-problem rate could shift, though the evidence here is an association, not a demonstrated cause.

Reading a prevalence number: ask what was measured and in whom

When you see a headline figure for sleep disturbance in older adults, the useful questions are: Was sleep measured by questionnaire or by overnight recording? Which sleep feature counted as disturbed? Were participants recruited from the community, a clinic, or a research registry? And what stage of cognitive decline were they in? A cross-sectional study of older adults with MCI and cognitively healthy controls compared prevalence and correlates of sleep disturbance between the two groups [5], illustrating that even within one study design, the comparison group you choose shapes the number you report.

The honest summary is that the wide range is not mainly a contradiction. It reflects real differences in sleep across the cognitive spectrum, combined with real differences in how sleep is defined and who is studied. The most robust statement the evidence supports is directional: sleep disturbance is more frequent and more severe in MCI and Alzheimer's disease than in healthy aging [1][2], and objective measures tend to be more consistent than subjective ones [1]. Exact percentages should be read with their method and their population attached.

Sources used in this answer

Related research questions