SafeBeing leverages Machine Learning and Big Data analytics to deliver alerts, reminders, notifications, actionable insights and predictive analytics in real-time. Using these tools, SafeBeing not only provides clinicians with a 24/7 profile of patients’ activities, but also detects variations in activities of daily living that can be a cause for concern, triggering alerts to clinicians and family members as necessary. Abnormalities that are a cause for concern can be personalized to each patient, and alerts and notifications can be adjusted accordingly to avoid overloading clinicians with non-critical data.
Applicable Scenarios:
Use Case 1: Monitoring of Seniors at Home and in Elder Care Facilities
Seniors are in need of monitoring now more than ever. To date, seniors have relied on healthcare professionals, family caregivers, and senior day care centers to oversee their health and wellbeing. Yet, both at home and at care facilities, seniors are left unprotected because of the isolation and social distancing protocols that are in place now due to COVID-19. In practice, families are keeping their distance from their loved elders to avoid the risk of exposing them to the disease, healthcare professionals are unable to provide the same level of care, and seniors are staying home, missing both medical appointments and other day-to-day interactions that provided them with a layer of protection. These seniors can benefit from a simple way of monitoring their health and wellbeing. SafeBeing can remotely connect both caregivers and family members with seniors, identifying falls, emergencies, risks for hospitalization, and overall decline.
In senior communities, SafeBeing creates a safer environment by remotely monitoring patients in their rooms as well as preventing patient roaming to minimize patient-to-patient and patient-to-provider transmission of the illness.
Use Case 2: Monitoring of COVID-19 Patients in Quarantine at Home for Signs of Decline
People who were identified as carriers of COVID-19 and need to stay quarantined at home are currently completely unmonitored. Their health and wellbeing can be monitored with SafeBeing and those at-risk for decline can be identified earlier and at-scale. SafeBeing’s insights can be integrated with data from thermometers, blood pressure cuffs, and oxygen saturation monitors, as well as electronic health records to further increase their predictive power.
Use Case 3: Monitoring at Home of High-Risk Patients After Hospital Discharge to Increase Hospital Capacity
Other patients with non-COVID-19 related conditions are being sent home to free healthcare resources to deal with the emergency COVID-19 response. However, these patients still need to be monitored. SafeBeing can monitor these patients at scale with no installations or complicated setup, and it can identify those that need provider attention. Reductions in hospital readmissions through earlier detection of clinical decline has already demonstrated in the elder care population and will help maintain hospital capacity for COVID-19 patients.
In all three scenarios, SafeBeing creates a way for high-risk patients and seniors to be monitored and cared for as needed without constraining the healthcare system or putting them at risk of exposure to the virus. SafeBeing will enhance provider productivity by triaging at-risk patients based on symptoms and signs of decline and provide an automated layer of oversight that manual labor resources currently do not have due to the COVID-19 response.
SafeBeing Components:
The SafeBeing solution has the following components:
Caregiver Dashboard - A web interface that presents alerts, reminders, notification, insights and predictive analytics to the healthcare professional
User app - It can be used by the patient on his mobile phone
Caretaker app - It can be used by caregivers as a "mobile dashboard" or by family members of the patient
Smartband - A lightweight, water-proof band that is being worn by the patient on their dominant hand
Insights and Predictive Analytics:
The existing version of SafeBeing is ready and deplorable within 48 hours' notice. It supports the following specific insights:
Decline in sleep quality or quantity
Decline in activity over time
Wandering beyond designated safe zones
Fall detection and risk for falling
Risk for pressure ulcer through movement analytics
Risk for developing urinary tract infection
High daytime somnolence
Low fluid intake
Risk of dehydration
Emergency button
Compliance Management (low battery alerts, missing data/no connectivity alerts, reminders to wear the band)
To better assist providers with patient monitoring needs, the SafeBeing functionality is being enhanced to include better communication, heart rate monitoring, oxygen saturation, medication management and intake detection, a HIPAA-compliant chat feature, and symptoms surveys. With SafeBeing, patients can have the monitoring they need at home, relieving the healthcare system and connecting caregivers and family while maintaining isolation protocols. In elder care applications, the platform has already demonstrated reduced rehospitalizations, earlier identification of symptomatic and at-risk patients, viable in-place monitoring and greater productivity per healthcare provider.
The next version of the platform (Q3 2020) will have additional functionality including heart rate, oxygen saturation, Wi-Fi connectivity, HIPAA-compliant messaging, symptom surveys, and more.
Example of a Gesture Detection- Drinking and the Science Behind It
Drinking is a type of hand-to-mouth gesture. Such events are being detected in real-time based on accelerometer and gyroscope signals. It is important to monitor drinking episodes and volumes for detecting a risk for dehydration, as well as for monitoring medication adherence and essential activity in general.
Drinking events possess the following characteristics:
Social activity – Drinking is a social activity. It is performed during conversation and social events. Consequently, the gesture is not very compact in time and does have many permutations (different ways to drink)
Drinking gesture- the drinking gesture is composed of three distinct regions (see Figure 2):
Hand to mouth – in this region the hand with the glass is elevated towards the mouth. This region ends when the glass reaches the lips. The graph demonstrates acceleration and deceleration specific patterns in the accelerometer and a peak in the gyroscope signal.
Sip – in this region the person rolls the glass to enable fluids to flow from the glass to the mouth. There is no acceleration or deceleration in the accelerometer signal. The accelerometer signal variance is very low. There is also a significant angular velocity over the X-axis due to the rolling of the glass.
Hand from mouth – in this region the hand is lowered from the mouth. The sensor activity is similar to the hand-to-mouth region but reversed
It has high occurrences – a single drinking event may contain an average of six hand-to-mouth gestures (sips). A very hot beverage (e.g. tea) can have up to 30 sips.
In order to build the model, the programmers used a sophisticated boosting algorithm coupled with meticulously designed features. These features are formulated using deep domain knowledge of drinking gestures and hand-to-mouth gestures.