Independent living sensor review: a general check on day-to-day patterns, 1 August to 6 September 2026

Weekly averages of detected activity across the five complete weeks were about 167, 118, 113, 111 and 101 episodes per day — a marked step down after the first week, followed by a smaller continuing drift rather than a steady decline. Across the whole period, the first and second halves averaged about 138 and about 109 episodes per day, and movement between monitored areas fell in the same direction, from about 97 to about 72 transitions per day. Three days, 5 to 7 August, recorded daily totals (228, 200 and 197) above the screening range calculated from this period's own average and spread, so the headline comparison is sensitive to whether that first week is included: from 8 to 21 August the average was about 120 episodes per day against about 106 per day from 24 August to 6 September, a difference of roughly 12% and much smaller than any comparison that includes the first week. Periods of three hours or more with no detected activity at all were no more frequent later in the period (2 of the last 18 days) than earlier (6 of the first 18 days).

Two independent measures — how often activity was detected and how often the individual moved between monitored areas — point the same way, which makes the direction reasonably dependable, although the available evidence cannot establish whether a difference of this size lies beyond ordinary week-to-week variation. The shape matters more than the headline percentage: most of the difference occurred between the first week and the second, and the weeks since have been comparatively similar to each other. Because longer undetected stretches did not become more common, the later pattern looks like fewer and shorter detections spread through the day rather than long unexplained absences from the monitored rooms. Indoor temperatures also fell by roughly two degrees Celsius across the monitored rooms over the same weeks; this is noted as co-occurring context only, as the data cannot establish whether the two are connected. The most useful next step is to check what was happening in early August and whether the quieter pattern since matches the individual's own account.

Bathroom-area detected episodes fell from about 40 per day in the first half of the period to about 26 per day in the second half, and detected bathroom-area minutes fell from about 147 to about 89 minutes per day. The reduction remains when the busy first week is excluded: about 34 episodes per day from 8 to 21 August against about 23 per day from 24 August to 6 September. Splitting by time of day across two matched 18-day windows, bathroom-area episodes beginning between 06:00 and 22:00 fell from 607 to 347, while those beginning between 22:00 and 06:00 were essentially unchanged (215 then 224), so the reduction is a daytime one. The bathroom device itself continued reporting on all 18 days of both windows, with light readings slightly higher in the later window (3,191 then 3,706) and temperature reporting maintained, while its positive movement detections fell from 1,899 to 1,165 and its presence readings from 845 to 612.

This is the most specific pattern in the data: counts, duration and the raw sensor detections all move together, the change is confined to daytime hours, and the device continued reporting other readings throughout, which makes a sensor fault or dropout an unlikely explanation. Detected bathroom-area activity is not the same as personal care or use of the bathroom, and fewer detections cannot establish reduced washing, changed continence needs or any change in the individual's ability to manage them — they may be entering the room less often, or moving less within it. It is nonetheless the observation most worth checking directly with the individual, because a sustained daytime change in this area is something sensor data can flag but only a conversation or assessment can explain.

Across 36 analysed nights, bathroom-area activity between 22:00 and 06:00 averaged about six episodes per night, ranging from 2 to 12, with at least three episodes on 35 of the 36 nights. Comparisons at different window lengths disagree in direction, so no clear change in this overnight frequency was identified over the period, in contrast to the daytime reduction. The longest nightly period without any detected activity averaged about two hours (median 104 minutes, ranging from 54 to 234 minutes) and likewise showed no clear sustained change. The first detected activity of the morning moved earlier by roughly 35 to 40 minutes on average — about 05:04 across the first 12 days against about 04:25 across the last 12 — a direction seen at every window length tested, although individual days still ranged from 04:00 to 07:26.

Overnight bathroom-area activity is a settled feature of this dataset rather than something that developed during the period. A quiet period is only an interval with no detected activity: it is not evidence of sleep, waking or sleep quality, and the frequency of overnight detections must not be read as disturbed sleep. What the evidence does offer is a consistent, checkable pattern — several detected trips to the bathroom area on almost every night, and a first detection of the day that is now typically before half past four — which can be compared with the individual's own account and with any continence or medication information held elsewhere. First detected activity is not proof of when the individual got up.

Kitchen-area detected episodes fell from about 52 per day in the first half of the period to about 41 per day in the second half, a direction seen at every window length tested. Using matched fortnights that exclude the busy first week, detected kitchen minutes fell from about 175 to about 147 minutes per day, around 16% lower. Appliance-power telemetry is recorded for the kitchen, but usage sessions have not been derived because the required approved policies are not in place, so no appliance-use evidence can be produced and none should be inferred.

Detected kitchen-area activity measures movement in that part of the home; it is not a measure of cooking, eating or drinking, and the reduction cannot establish anything about nutrition or the ability to prepare food. The most useful point here is the evidence gap: appliance power data is the one signal that could have offered an independent check on kitchen routine, and it is unavailable in this run. The absence of derived appliance sessions must not be read as an absence of appliance use. Until that cross-check exists, any question about meals is better explored directly with the individual than inferred from movement detections.

Bedroom-area activity showed no clear change: comparisons at different window lengths disagree in direction for both detected episodes and detected minutes, with some shorter comparisons higher and some longer comparisons lower. Because bedroom-area activity held broadly steady while bathroom-area activity fell, the balance between rooms shifted: across matched 18-day windows the bathroom's share of all detected room activity fell from about 29% to about 24%, the bedroom's share rose from about 34% to about 39%, and the kitchen's share stayed close to 38%.

This is useful contrary evidence: the reduction seen elsewhere is not a whole-home drop in detected activity, and nothing suggests the bedroom became less active. The rise in the bedroom's share is arithmetic — it follows from a steady bedroom level alongside a falling bathroom level — and should not be read as the individual spending more time in the bedroom, or as the bedroom being their main area. Read with the earlier findings, the picture is of a change concentrated in particular areas and times of day rather than a general withdrawal.

Activity is sensed in three areas only — bathroom, bedroom and kitchen. The living room has ambient temperature and battery reporting but no movement or presence sensing, and no door or exit sensor exists anywhere in the dataset, so property exits and time spent away from home could not be measured or checked. Continuous sensor availability was not confirmed for any of the 38 overnight windows that intersect the assessment, and like-for-like sensor coverage between the earlier and later comparison periods has not been independently verified.

These gaps bear directly on how the earlier findings should be read. Time spent in the living room, in any unmonitored space, or outside the home produces no detections at all, so a fall in detected activity may partly reflect where activity happened rather than how much occurred — and with no exit sensor there is no way to separate the two. Equally, because overall sensor continuity is unconfirmed, a quieter later period cannot be entirely distinguished from reduced sensor availability, even though the bathroom device was shown to keep reporting throughout. None of this removes the observed differences, but it does mean the findings should be treated as questions to explore with the individual and other records, not as a complete account of their day.