Intelligent Assistive Technology and Systems Lab - click to go to homepage
IATSL develops assistive technology that is adaptive, flexible, and intelligent, enabling users to participate fully in their daily lives. Learn more about our research

Visit us:

Room 438

500 University Ave.

Toronto, Canada

P 416.946.8573

F 416.946.8570


Send us mail:

160 - 500 University Ave.

Toronto, ON, M5G 1V7



email us!


Follow IATSL on Twitter


Frailty Toolkit: Home-based Frailty Assessment and Prediction Using Internet of Things and Artificial Intelligence

Keywords: Frailty, Assessment, Sensor, Internet of Things, Machine Learning, Smart Home, Predictive Modeling

Overview of Research

Population ageing is accelerating rapidly worldwide, from 461 million people older than 65 years in 2004 to an estimated 2 billion people by 2050, which has profound implications for the planning and delivery of health and social care. The most problematic expression of population ageing is the clinical condition of frailty. Frailty is a condition in which an individual is in a vulnerable state at increased risk of adverse health outcomes and/or dying when exposed to a stressor. An average of 10.7% and 41.6% community-dwelling older adults are frail and pre-frail, respectively. These people have a substantially increased risk of fall, disability, hospitalization or mortality. However, current clinical frailty assessments have limitations such as time-consuming, subjective, and therefore can lead to delayed and inaccurate assessment. The rapid development of technology such as sensors and artificial intelligence in recent years enables monitoring frailty-related criteria in order to assess frailty early and accurately.

This research aims to design, develop and validate a technology-based frailty toolkit that can be used in home settings to assess and predict frailty. The research questions are:

  1. Which frailty operational criteria and technology are the best fit for assessing frailty in home settings based on clinical evidence and technical feasibility?
  2. How do sensor data of the frailty toolkit compare with clinically validated data with respect to accuracy?
  3. Which predictive algorithms for predicting frailty are the most appropriate to use with respect to accuracy, sensitivity, and specificity?

We plan to use sensors to collect data for measuring frailty-related criteria and use artificial intelligence to analyze the data in order to assess for frailty. The assessment will be validated by comparing with the clinical gold reference standard, the Friedís Frailty Index (FFI). FFI focuses on physical frailty which is clinically characterized by slow walking speed, weight loss, weak muscle strength, self-report exhaustion, and low activity levels. To guide for frailty criteria and technology selection for this research, we have conducted a systematic review on frailty and technology. The review synthesized existing technologies for assessing frailty, evaluated the clinical evidence and feasibility of the technologies and suggested new potential technologies.

The frailty toolkit system includes a wearable activity tracker to measure walking speed and activity levels, a bathroom scale to track weight loss, a grip ball to measure upper extremity muscle strength and a voice-based smart speaker to make conversations with older adults in order to collect exhaustion information. The data from the heterogeneous sensors will be real-time streamed to a cloud-based streaming analytics software service via the internet of things platform. The streaming analytics service uses multiple machine learning classifiers to classify frailty statuses based on the data. A web portal will be built for data visualization. The system architecture of the proposed frailty toolkit is shown in Figure 1.

System Architecture of the Frailty Toolkit

Figure 1. System Architecture of the Frailty Toolkit

Research Team

Chao Bian, University of Toronto

Bing Ye, University of Toronto

Kathy McGilton, University of Toronto

Babak Taati, Toronto Rehabilitation Institute

Alex Mihailidis, University of Toronto