The Impact of Monitoring Device: Evidence from the RCT in Japanese Long-Term Care Nursing Facilities

The intervention is the installation and activation of "Nemuri SCAN," an under-mattress monitoring device manufactured by Paramount Bed Co., Ltd. The device uses pressure sensors to continuously detect residents' sleep/wake states, bed-exit events, and physiological proxies including respiration rate and heart rate. Data are transmitted to a central monitoring system accessible to night-shift care staff via tablets or desktop displays. Treatment clusters receive device installation under all resident mattresses, along with standardized staff training on system operation and alert response protocols. Care workers can monitor resident status in real-time without physical room visits, enabling more efficient triage of nocturnal care needs. Control clusters continue business-as-usual monitoring practices (periodic physical room checks) during the study period. Control clusters receive device access after the main intervention period ends, per the program agreement with Kanagawa Prefecture.

2025-12-21

2026-01-30

We organize primary outcomes into five conceptually distinct outcome families, with one primary endpoint per family:

1. Care worker stress: SRS-18 (Stress Response Scale-18) total score (0–72), a validated psychological stress measure.
2. Care worker physical activity: Daily steps during work shifts measured via wearable devices.

3. Resident sleep efficiency: Sleep efficiency (%) = 100 × (total sleep time / time in bed), derived from device telemetry.

4. Resident safety: Falls per 1,000 resident-nights (cluster-level rate in the post-intervention window).

5. Resident well-being: WHO-5 Well-Being Index rescaled to 0–100 (proxy-completed by caregivers).

1. Time-use during night shifts: Share of time across activity categories (e.g., direct care, monitoring, documentation) recorded via time-study survey.

2. Care worker activity (distance and calories): Distance traveled and calories expended during work shifts measured via wearable devices.

3. Night-visit counts and purposes: Number and reasons for nocturnal room visits recorded by care workers.

4. Text-based psychological burden index: Embedding-based measure of caregiver psychological burden derived from daily work logs using natural language processing (BERT-family model).

5. Device-derived sleep metrics for residents: Wake after sleep onset (WASO, minutes), nocturnal awakenings (count), and out-of-bed minutes from device telemetry.

6. Physiological proxies: Respiration rate and heart rate from the monitoring device.

7. Near-miss incidents ("hiyari-hatto"): Count of near-miss events from routine incident reports.

8. Falls and accident counts: Number of falls and accidents from facility records.

9. Hospitalizations: Count of hospital admissions during the study period.
Pressure ulcers: Incidence of new pressure ulcers.

10. Use of physical restraints: Frequency of physical restraint use.

11. Resident vitality (Vitality Index): Sum score (0–10) across five domains (waking pattern, communication, feeding, toileting, rehabilitation/activity), proxy-recorded by caregivers.

This study employs a matched-pairs cluster randomized controlled trial (CRCT). The unit of randomization is the cluster, defined as either an independent floor within a facility or an entire facility. Pairing: Clusters are paired based on governing corporation, service type, facility scale, resident case-mix, and night-shift staffing structure using a two-stage matching algorithm. Randomization: Within each matched pair, one cluster is randomly assigned to treatment and one to control (1:1 allocation) using a reproducible random seed. Sample: Approximately 20 facilities in Kanagawa Prefecture, yielding ~40 clusters (~20 pairs), ~600 care workers, and ~640 residents. Timeline: Baseline data collection (5-day window) occurs prior to device installation. The intervention runs from December 2025 through February 2026, with follow-up data collection during the intervention period. Control condition: Control clusters continue business-as-usual monitoring and receive device access after the intervention period ends.

Not available

Randomization will be done in office by a computer.

Cluster (facility floor or entire facility). Within each matched pair, one cluster is randomized to treatment and one to control.

Yes

~40 clusters (facility floors or facilities)

~600 care workers; ~640 residents

~20 clusters treatment, ~20 clusters control (1:1 within each of ~20 matched pairs)