What eight weeks of home sleep research asked of older adults with memory difficulties

Since the previous research note · Earlier expectation not tested

Track continuity

Since the previous research note: What researchers can learn from better and worse sleep in older adults

Previous expectation: Reproducible and clinically informative links between reported sleep quality and cognition could help clinicians decide when sleep complaints warrant fuller assessment.

What this study adds: RESTED demonstrated sustained home collection of multiple sleep and cognitive measures in a selected cohort that included people with mild cognitive impairment or mild dementia. Its feasibility findings did not test the earlier expectation about clinically useful questionnaire-defined relationships.

---

What changed in this study

Sleep research often depends on measurements taken repeatedly over time. A single clinic night can show detailed information, but it may also interrupt a person’s usual routine. Questionnaires are easier to use at scale, yet sleep reports can be affected by memory problems and may not capture sleep staging or other fine-grained features.

The RESTED study tested whether older adults with mild cognitive impairment or mild dementia could manage a demanding set of sleep and cognitive measurements from home. Published on May 28, 2025, it was a prospective observational feasibility study. Its central question was practical: could people take part, keep using the equipment and provide enough usable data over eight weeks?

Among 41 eligible participants, 40 completed the main monitoring period. The study therefore showed that sustained remote data collection was possible in this selected group. It did not test whether changing sleep prevents dementia, improves cognition or produces a clinical diagnostic tool.

Why repeated home measurements matter

Sleep varies from night to night. Illness, stress and ordinary changes in routine can all affect what happens during sleep. Repeated home measurements can help researchers see that variation instead of treating one night in an unfamiliar setting as representative.

The alternative is not simply a choice between a questionnaire and a laboratory. Self-reported sleep is convenient and inexpensive, but it can be difficult to interpret when participants have memory impairment or limited awareness of their symptoms. Polysomnography, commonly used for detailed objective sleep measurement, typically requires specialist setup and analysis in a controlled clinic environment. That makes it harder to repeat frequently and may alter the way someone sleeps.

RESTED explored a middle path: several forms of measurement collected in the participant’s own home, across many ordinary nights. The study was about whether this arrangement could work reliably enough for future mechanistic or intervention research.

A study built around ordinary nights

Participants wore an Axivity AX3 wrist activity monitor and completed digital sleep diaries for 56 days. The wrist device provided actigraphy data, while the diaries captured participants’ own reports about their sleep. They also completed three web-based cognitive tasks twice a week: choice reaction time, forward digit span and self-ordered search.

The schedule became more intensive for one week. During those seven nights, participants used a Dreem 2 wireless sleep EEG headband to record sleep at home. They also completed cognitive tasks daily, took part in four verbal memory assessments with a researcher by videoconference and provided saliva samples for measurements related to melatonin timing and the cortisol awakening response.

The equipment was supplied as part of a study kit that included charging equipment, an overnight pulse oximeter, saliva collection materials and printed instructions. Participants who requested one could receive a tablet. The arrangement was remote in where much of the data collection happened, but it was not unsupported or entirely virtual.

Success depended on what counted as complete

The headline retention result was strong: 40 of the 41 eligible participants completed all eight weeks of home monitoring. The one withdrawal happened after baseline assessments and training but before home data collection began. The 40 completers also reached the six-month follow-up.

The amount and quality of data varied by method. During the intensive week, participants recorded an average of 6.4 of the requested seven EEG nights, equivalent to 91.4% completeness. Seventy-three percent provided all seven nights. The average automated record-quality measure was 84.4%. Completeness and signal quality are related, but they are not the same: a recording can be started and still contain less usable signal than researchers would prefer.

Actigraphy covered a longer period. Participants provided an average of 52.8 analyzable nights out of 56, or 94.3% completeness, and 60% supplied all 56 analyzable nights. Everyone recorded at least 14 days. Some missed days because of forgetting, short breaks or minor skin irritation, while battery failure affected two recordings.

The online cognitive tasks showed why a high average does not mean every participant completed every scheduled session. Average completion was 79.8% of allocated tasks, yet only six of the 40 participants completed all three tasks on all 21 occasions. Depending on the task, 18% to 23% of participants supplied complete data across every scheduled session.

These figures describe different schedules and different definitions of success. Completing most assigned tasks can provide a useful longitudinal record even when a participant misses some sessions. At the same time, a future trial would need to decide which measurements are essential, how much missing data is acceptable and how much support is realistic.

Collecting a sample was not the same as timing it

The saliva measurements illustrated another layer of difficulty. Most melatonin samples were collected: 272 of 280 requested samples, or 97.1%. For samples that could be analyzed, completeness was 92.5%.

Timing was harder to preserve. Of the 236 melatonin samples with participant-reported times, 50.8% were recorded within 15 minutes of the scheduled time. The study could therefore show that samples had generally been collected while also revealing that collection did not always happen at the intended moment.

Cortisol sampling was also mostly completed, with 97.5% of the requested samples collected and 91.7% completeness for analyzable samples. Among the timings that could be compared with EEG-estimated awakening, some samples appeared to have been recorded before the estimated wake time and were removed because the timing was suspected to be incorrect.

The digital sleep diaries created a similar, less technical issue. Several free-text answers did not follow the requested format. Researchers could manually interpret many of them, but some entries remained ambiguous, especially reports about how long it took to fall asleep or how long nighttime awakenings lasted. Home data collection can reduce the burden of clinic visits while increasing the need for careful data checking.

Home research still needed a human support system

Participants received in-person and remote training and written instructions, and were offered a refresher session before the intensive week. The research team offered reminders and responded to questions by email, phone, video call or messaging. Every participant had at least one in-person visit during the study, including visits for equipment or data-related tasks.

Relatives and partners also played a role for some participants. Thirty percent of the 40 analyzed participants reported receiving outside help with study tasks. Assistance included reminders, setting up devices and troubleshooting technical problems. It was not required for everyone: many participants produced good-quality data with little or no help from someone outside the research team.

This matters for how the result should be understood. High retention did not show that older adults with cognitive impairment can complete complex research without assistance. It showed that participation was possible within a package of training, instructions, researcher contact, occasional visits and, for some people, help from relatives.

The study also exposed a dependency that is easy to overlook. Existing devices and digital services can be quicker to deploy and easier to use than newly built systems, but researchers may still depend on manufacturers or providers for technical support and continuity. Changes made by providers caused problems for some study tasks and contributed to data loss. A remote study needs not only willing participants but also stable tools and dependable infrastructure.

A willing cohort is not the whole population

RESTED recruited community-dwelling adults aged 50 or older who had internet access at home. Participants with mild cognitive impairment or mild dementia were included in groups related to probable Alzheimer’s disease, Lewy body disease or Parkinson disease, alongside cognitively healthy controls. Advanced dementia, acute or terminal illness and some unrelated conditions were excluded.

Recruitment took place over 17 months. Of 129 people identified as potentially eligible before screening, 44 consented, a consent rate of 34%. Three were later withdrawn after screening, leaving 41 eligible participants. This funnel is part of the feasibility story: the people who completed the study had already passed several practical and clinical filters.

The cohort was small and socially narrow. All 40 analyzed participants identified as White British, most were male, and adults over 80 were underrepresented. The researchers also noted that women and minoritized ethnic groups were underrepresented, as is common in dementia research. Most participants who answered a technology questionnaire reported frequent smartphone or smart-device use, while only a minority had previously used a wearable to monitor sleep.

People who were less confident with technology, less interested in digital monitoring or unable to use the internet at home may have been less likely to enroll. The study’s success therefore demonstrates sustained participation among a selected group, not broad accessibility across everyone living with memory or thinking difficulties. It also does not establish feasibility for advanced dementia or for communities that were inadequately represented.

The findings are best read as a practical proof of delivery. Researchers were able to collect repeated sleep and cognitive measurements over time, but the work still required choices about participant selection, training, data validation and support. Nor did the study test the earlier question of whether questionnaire-defined sleep and cognition relationships could guide clinical assessment; its contribution was the delivery of repeated measurements, not a clinical association or treatment result.

What home monitoring could open up for dementia research

The study points toward future trials in which older adults contribute repeated sleep and cognitive measurements while staying closer to their usual routines. That could make it possible to evaluate sleep-focused interventions with fewer burdensome overnight research visits, provided the measurements remain reliable and the support system is sustainable.

The potential benefit would extend beyond researchers. Participants might take part without repeatedly traveling to a specialist center, relatives could have flexible ways to help when needed, and research teams could observe changes across many nights rather than relying on a small number of clinic assessments. Those benefits remain conditional on tools that work consistently, study designs that accommodate different levels of ability and recruitment that reaches people beyond this selected cohort.

RESTED’s practical achievement was not proving that better sleep changes dementia outcomes. It was showing that, with structured support, some older adults with mild cognitive impairment or mild dementia could sustain a demanding program of home-based sleep and thinking assessments. That capability gives future studies a workable starting point for asking the next clinical questions.

Key Sources

• Primary source: Longitudinal Remote Sleep and Cognitive Research in Older Adults With Mild Cognitive Impairment and Dementia: Prospective Feasibility Cohort Study.
  DOI or Journal Page · Peer-Reviewed Paper · 2025-05-28