A Case for Using Sensor Technology to Monitor Disruptive Behavior of Persons With Dementia

A Case for
Using Sensor Technology to
Monitor Disruptive Behaviors
of Persons with Dementia
Maria Yefimova & Diana Lynn Woods
UCLA School of Nursing
November 3, 2012

2012 AAAI Fall Symposium: AI for Gerontechnology 1

Overview
 Behavioral Symptoms of Dementia
 Definition
 Measurement Issues
 Sensor Networks
 Sensor Selection
 Data Processing Algorithms
 Meaningful Output
 Ethics of Technology
 Future Research 2

Population of Interest
 35.6 million people worldwide (Wimo & Prince, 2010)
 Affected by dementia - Alzheimer’s disease
most common
 Progressive memory loss
 Need help with daily activities
 “Disruptive behaviors” (BSD)
 Cost of care $604 billion, more than 1% GDP
(Wimo & Prince, 2010)

3

Problem of Interest
 “Behavioral Symptoms of Dementia” (BSD)
 Challenging to care providers
 Consume time and effort
 Unsafe, resulting in accidents
 Lead to institutionalization
 Increase cost of care

4

Defining BSD
 Restless body movements (tapping, banging,
picking clothes)
 Vocalization (muttering, repetitive questions,
screaming)
 Sleep disturbances (night time awakening)
 Pacing, wandering
 Resisting daily care (showering, dressing)

5

Goal of Research on BSD
 Inform clinical practice on prevention and
management through
 Identifying high risk individuals
 Predicting escalation of behavior
 Timing is crucial for interventions
 Developing individualized plans

6

Challenges and Opportunities
 Heterogeneous population
 Within-individual variability of behavior and
response
 Data is costly and difficult to analyze
 BSD manifestation
 Behaviors vary in intensity, cluster together
 Non-linear patterns of escalation/de-escalation
Can technology provide new tools to study BSD?
7

Faces of Dementia
Alzheimer’s Disease International - http://www.alz.co.uk/
8

Deploying Sensor Networks
 Multimodal sensors to capture all aspects of
behavior
 E.g. agitation recognition rate - 59% ultrasonic alone,
73% with pressure, up to 94% with sensor fusion
(Biswas, Jayachandran & Thang, 2006)

Sensors Behavior
Motion Restlessness
Radar Tapping/Banging
GPS tracking Wandering
Acoustic Vocalization
Pressure (bed) Sleep disturbances
Video Daily Activities 9

Technical Considerations
 Wearable
 Specific to individual, mobile
 Size and placement must be tolerated
 Environmental
 Unobtrusive and more acceptable
 Static position, tied to specific environment
 Wired versus Wireless
 Power consumption and battery life, cable
management
10

Data Processing
Algorithms chosen based on assumptions about
behavior
 Well-defined sequences of events (eg. ADLs)
 Decision trees with classifiers (Maurer et al., 2007),
temporal logic models (Rugnone et al., 2007)
 Sporadic and variable (eg. BSD)
 Fuzzy logic (Fook, 2007), Bayesian models (Biswas,
Jayachandran & Thang, 2006)

Goal: activity recognition and prediction
11

Algorithm Considerations
 Obtaining “baseline” to recognize deviations
 Learning data can’t be modeled in lab setting
 Not generalizable across individuals (Algase et al., 2010)
 Inferring meaning
 Timing and micro-context (Biswas et al., 2010)
 Information about surrounding environment

12

Meaningful Output
 Output readable by end user (researchers,
clinicians, caregivers)
 Visual representation of data
 Clinically relevant time intervals
 Validation against “ground truth”
 Direct observation by experienced clinicians
 Caution in using existing scales and measures

13

Ethical Issues
 Monitoring may be perceived as
“intrusive surveillance” (Price 2007)
 Moderated by perception of technology
 From older adult and caregiver/family
 Consent with the cognitively impaired
 Proxy consent
Preserve privacy and safety of information

14

Future Direction
 Currently limited to small samples
 Feasibility studies in labs, with non-elders
 Translating to “real world” setting
 Potential to use in evaluating interventions
 Financial and human capital considerations
 Cost-benefit analysis

15

Conclusion
 Collaboration between clinicians and
developers
 Learning a new, common “language”
 Technology as a research tool
 To understand the phenomenon
 To evaluate interventions
Goal: improving health of vulnerable older
adults and reducing healthcare costs
16

Questions? Suggestions?
m.yefimova@ucla.edu

Supported by
John A. Hartford Foundation’s
National Hartford Centers of Gerontological
Nursing Excellence
Award Program 17

A Case for Using Sensor Technology to Monitor Disruptive Behavior of Persons With Dementia

More Related Content

Similar to A Case for Using Sensor Technology to Monitor Disruptive Behavior of Persons With Dementia (20)

Recently uploaded (20)

A Case for Using Sensor Technology to Monitor Disruptive Behavior of Persons With Dementia

Editor's Notes