Activity pattern monitoring: how it learns a daily rhythm, and why offline is not a quiet room

What is a daily rhythm?

A daily rhythm is the shape of a normal day, written as numbers. Most systems watch four: wake time, the first movement of the morning; settle time, the last movement before the bedroom goes still; active minutes, minutes of movement by room; and night trips, how often someone got up at night.

Averages have a problem. One week in hospital drags the average and makes the next normal week look strange. A median takes the middle value and ignores the odd week, so the system can ask one question: is today far outside that middle? WalledCare's cloud uses a median and a median-based spread over a 28-day window for its per-room baseline (manufacturer specification, September 2026; not published on the product site).

How long does it take to learn a routine?

Research systems used long histories. Shahid and colleagues modelled nine people living alone with 80 to 355 days each; in about 64 days their detector sent 421 notifications to eight apartments, 373 of them reassurance (JMIR Aging, 2022).

envoyatHome says its Insight reports surface a behaviour that is "changing over hours, days, or weeks" (envoyatHome, vendor claim). The same page describes the product as sold to caregivers of older adults in the United States. We found no Canadian availability and no published learning window on its public site (checked 6 September 2026).

WalledCare raises no rhythm alert until at least 14 days have been observed in its 28-day window, a default setting (manufacturer specification, September 2026; not published on the product site). Two weeks is a floor, not a promise. A rhythm with a weekly outing takes longer.

In week one, expect frequent messages, and expect most to say "normal." Timing data cannot explain why: a system knows the bathroom was busy at 3 a.m., but not about the flu.

The only Canadian trial here is the Donelle trial below; the routine-learning research is European, so those numbers describe those homes.

What can a changed rhythm mean?

It can track real health change. The Hayes study above covered more than 108,000 person-hours, but used an array of research sensors and called its own result a demonstration of feasibility. A single room sensor is not that tool. It cannot predict dementia. A camera-free presence sensor like WalledCare does not detect falls, does not measure breathing, and is not a medical device. A changed rhythm is a reason to phone, visit, or ask a doctor.

Why offline is not the same as no movement

A room can be in three states: empty; occupied but still; and not reporting, where the sensor has lost power or Wi-Fi. The last two look alike in an app. A heartbeat separates them. WalledCare's firmware says "I am here and the room is empty" every 60 seconds. Its console shows a room offline after 3 minutes with no ping; an offline alert is sent after 10 minutes with no data. Quiet-household alerts are separate, firing after 30 minutes of no daytime movement and 120 at night (manufacturer specification, September 2026; not published on the product site).

A stillness alert does not tell you someone has fallen. It tells you a room has been occupied and still past the threshold, which can be a nap or a long read. It is not fall detection (see fall detection without a wearable).

Connectivity is not a given: 72% of Canadians 75 and older used the internet in 2022, leaving more than one in four without it (Statistics Canada, Canadian Internet Use Survey, 2023). If an app treats "no data" as a calm day, nobody will know.

Does rhythm monitoring keep people at home?

The Canadian evidence is weak. A multiprovincial randomized trial tested passive remote monitoring in home care with motion sensors, cameras and door alarms (Ontario 174 patient-caregiver pairs, 60 monitored; Nova Scotia 139 pairs, 45 monitored) (Donelle et al., JMIR Aging, 2025). Measuring moves to a higher level of care within one year, it found a hazard ratio of 0.7 in Ontario (95% CI 0.3–1.4, P=.30) and 1.1 in Nova Scotia (95% CI 0.3–3.7, P=.90). The trial was small, so "no significant effect" is the right reading, not "proven not to work."

What to ask any vendor

• How many days before pattern alerts start?
• Is a lost connection a separate alert, and after how many minutes?
• Does it store video or audio? Is the data kept in Canada?
• Who is called when data stops?

PIPEDA defines personal information as "any factual or subjective information, recorded or not, about an identifiable individual", and applies to private-sector organizations handling it "in the course of a commercial activity" (OPC, PIPEDA in brief). On that definition, room-level presence data tied to a named person is personal information. Quebec, Alberta and British Columbia have private-sector privacy laws that may apply instead, and several provinces, Ontario among them, have health privacy laws declared substantially similar to PIPEDA (OPC, Provincial and territorial privacy laws). General information, not legal advice.