Passive monitoring that respects dignity: what to measure and what to leave alone

There is a version of technology in elderly care that nobody wants: a camera in the living room, an app that pesters, a wearable that has to be charged and remembered, and a family member watching a dashboard from two hundred miles away. It is intrusive, it is resented, and it usually ends with the device in a drawer.

There is a better version, and the difference is not the hardware. It is what you choose to measure. The most effective monitoring in a home does not watch a person at all. It watches the ordinary traces that daily life leaves behind, and it only speaks up when those traces stop appearing.

Activity patterns, not observation

Consider what you actually want to know about a parent living alone. Almost never “what are they doing right now?” Nearly always “is today roughly like other days?”

That question can be answered without watching anything. A door contact on the fridge tells you someone has eaten. A metering plug on the kettle tells you the morning has started. A motion sensor in the hallway tells you movement between rooms is happening at the usual times. None of these records a person; they record that the house is being lived in.

The signal you are looking for is deviation from a pattern, not any single event. The kettle not being used by eleven in the morning means something only because it is normally used by eight. This is why these systems need a few weeks of observation before they are useful, and why a system installed and immediately configured with alert thresholds will generate nonsense for a month.

The measurements that carry the most information

The kitchen

The single richest source. A metering smart plug on the kettle or coffee machine, and a contact sensor on the fridge door, between them tell you about waking, eating and drinking — three of the things that change first when someone becomes unwell. They are also completely unobtrusive: nobody notices a plug.

Movement between rooms

Two or three motion sensors — hallway, living room, bathroom approach — establish that someone is moving around the home. You are not trying to track a person; you are establishing that transitions are happening. Long gaps between transitions during waking hours are the useful signal.

The bathroom

Handle with care, because this is the most private space in the house. A motion sensor outside the door, or a door contact, gives you visits without any sensing inside. That is almost always sufficient, and it avoids a conversation that is genuinely difficult to have.

Temperature

Underrated and important. Older people are more vulnerable to both cold and heat, and are also more likely to under-heat a home for cost reasons and not mention it. A simple temperature sensor in the living room and bedroom, with a low threshold alert, catches something that has real health consequences and that nobody will volunteer.

Doors to outside

An external door opening at three in the morning matters. So does an external door that has not opened for four days, which may indicate someone has stopped going out. Both are worth knowing; neither requires observing the person.

What we would not fit

Being clear about the limits is what makes the rest acceptable.

Cameras in living spaces. Very occasionally justified, always after an explicit conversation, and never as a default. A camera changes how a person behaves in their own home, and that cost is usually higher than the information is worth. If the concern is falls, use radar rather than video.

Microphones. Same reasoning, more so.

Anything requiring daily interaction. A system that needs someone to press a button, confirm a prompt or charge a device every night has introduced a new task into the life of the person it was meant to help. It will lapse, and its lapsing will be misread as a problem.

Bedroom cameras or bathroom sensing. There is almost never a case that justifies these in a private home.

Getting the alerting right

This is where most home-built systems fail, and the failure is always in the same direction: too many alerts, then all alerts ignored.

Use long windows

Almost nothing in this domain needs a response within minutes. “No movement detected between 08:00 and 11:00” is a useful alert. “No movement for twenty minutes” is noise, and will fire every time someone reads a book.

Distinguish informational from actionable

A daily summary — woke around seven, kettle used three times, went out at eleven, home by two — is reassuring and asks nothing of anybody. Reserve push notifications for genuine deviations: no morning activity by late morning, an external door open in the middle of the night, a living room below fifteen degrees.

Escalate to a person, not a screen

An alert should end with somebody phoning. The system's job is to prompt a human check-in, and the check-in is what actually helps. A dashboard that nobody is contractually obliged to look at is not a safety net.

Expect to tune it for a month

Everyone's routine is idiosyncratic. Someone who sleeps until ten, or who spends whole afternoons in the garden with no indoor motion at all, will break naive thresholds immediately. Plan on a month of adjustment, and involve the person in it — they can explain the pattern far faster than you can infer it.

Reliability matters more here than anywhere else

If a lighting automation fails, someone flips a switch. If reassurance monitoring fails silently, the family believes everything is fine because no alerts arrived — which is exactly what the system reports when it is broken.

Three defences. Choose a hub that processes automations locally, so a broadband fault does not disable everything. Put the hub and router on a small uninterruptible power supply, because they draw very little and a power cut otherwise takes the whole system down. And most importantly, monitor the monitor: an alert if a sensor has not reported in twenty-four hours, and an alert if the hub itself goes quiet. Silence must never be interpreted as good news.

A modest starting kit

You can achieve most of the value with very little. This is roughly where we would begin.

1. A metering plug on the kettle and a contact sensor on the fridge. Between them, the best single indicator of a normal morning.
2. Two motion sensors, hallway and living room, for movement between rooms.
3. Contact sensors on external doors, for night-time and going-out patterns.
4. Two temperature sensors, living room and bedroom, with a low-temperature alert.
5. A local-processing hub on a UPS, with sensor-silence alerts configured on day one.

Then log for a month before setting a single deviation alert. The data will tell you what normal looks like, and normal is different for everybody.

The conversation comes first

None of this should be installed without the knowledge and agreement of the person living there. That is partly an ethical matter and partly a practical one: covert monitoring is discovered, and being discovered destroys trust in a way that is very hard to repair.

The framing that tends to work is honest and specific: this is so we do not have to phone every morning to check, and so you are not obliged to reassure anybody. Presented that way, activity monitoring is a route to less intrusion rather than more — and that is a genuine argument, not a sales pitch. We look at how to have that conversation properly in consent, dignity and who sees the data.