=Paper=
{{Paper
|id=Vol-2699/paper31
|storemode=property
|title=Frailty Detection using Presence in a Room
|pdfUrl=https://ceur-ws.org/Vol-2699/paper31.pdf
|volume=Vol-2699
|authors=Ramesh Balaji,Evelyn Tan Sio Keow,Srinivasa Raghavan Venkatachari,Hwee Pink Tan
|dblpUrl=https://dblp.org/rec/conf/cikm/BalajiKVT20
}}
==Frailty Detection using Presence in a Room==
https://ceur-ws.org/Vol-2699/paper31.pdf
Frailty Detection using Presence in a Room
Ramesh Balajia, Evelyn Tan Sio Keowb, Srinivasa Raghavan Venkatacharic, and Hwee Pink
Tan d
a
Tata Consultancy Services, Research & innovation, Chennai, Tamilnadu ,India
b
SMU-TCS iCity Lab, Singapore Management University, Singapore, Singapore
c
Tata Consultancy Services, Research & innovation, Chennai, Tamilnadu ,India
d
SMU-TCS iCity Lab, Singapore Management University, Singapore, Singapore
Abstract
The elderly population is steadily
increasing in most countries due to
a decline in birth and mortality rate.
The population of senior citizens
living independently has also
become significant. This has led to
an active research focus in geriatric
wellness.
One of the common effects of
ageing is Frailty which is seen in
the Elderly population. This
research is focused on detecting
activities specific to frailty in
typical natural home environment
of the geriatric population living
alone. To this end, we evaluated
only Passive Infra-Red (PIR)-based
motion and magnetic door contact
sensors for detection of frailty
through a unique combination of
custom and statistical techniques to
achieve reasonable accuracy. This
is further validated using ground
truth obtained through surveys to
understand the level of frailty of
selected elderly people.
.
Keywords
Elderly, Frailty, Passive Infrared Sensors, Algorithm, Median, Unobtrusive, Supervised
Learning
Proceedings of the CIKM 2020 Workshops, October 19-20, 2020,
Galway, Ireland
EMAIL: ramesh.balaji@tcs.com (A. 1); evetan25@gmail.com (A.
2); venkatachari.raghavan@tcs.com (A. 3); hptan@smu.edu.sg
(A. 4)
ORCID: 0000-0003-0344-0186 (A. 1); 0000-0001-6533-7622 (A.
2); 0000-0002-2907-0446 (A. 3) ; 0000-0002-8279-1429 (A. 4)
©️ 2020 Copyright for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0).
CEUR Workshop Proceedings (CEUR-WS.org)
1. Introduction flexibility in analyzing the sensor data and
validating with multiple survey that include
baseline and follow-up and interacting with
In many countries, there is a significant rise
Elderly and caregivers. This is not something
in the number of senior citizens living
that will be available to us if we gone with
independently. As a result, there is an active
public datasets.
research focus in the wellness and care of the
geriatric population. In this research, we focus
on frailty, which is defined as a “clinical
syndrome in which three or more of the
following criteria were present: unintentional
weight loss (10 lbs in past year), self-reported
exhaustion, weakness (grip strength), slow
walking speed, and low physical activity [2]”.
In the following sections, we describe our
approach in understanding the symptoms of
frailty with the right processing techniques so
that the right intervention can be applied to the
individual.
2. Data Description
Through the Smart Homes and Intelligent
Figure 1: Sensor-based monitoring system and
Neighbors to Enable Seniors (SHINESeniors)
[1] initiative, over 90 homes of elderly living sensor data
alone in Singapore have been instrumented
with non-intrusive and privacy-preserving 3. Sensor based Frailty Detection
sensors. Each home is equipped with at least 4
PIR motion sensors, one in each room. Each In an earlier related work [3], the authors
PIR sensor detects motion by sensing the examined the possibility of using the in-home
change in temperature between the room and a sensor-based monitoring system to detect
body temperature. In addition, a door contact frailty in 46 participants of the SHINESeniors
sensor is mounted at the main entrance of the project. Baseline and follow-up surveys were
dwelling. Finally, each elderly is also given a conducted in March 2016 and March 2017, and
help button on a lanyard, which they can sensor data 30 days prior to the survey date
activate in case they need help. All sensors were used for frailty detection.
communicate with backend cloud servers
through a home-based gateway transmitting Motivated by promising results from the
data over a 3G data network. Figure 1 shows the above study, this follow-up study proposes a
layout of the sensor-based monitoring system. Sensor-based Frailty algorithm based on data
(both sensor and survey) obtained in a later time
This study utilizes sensor data collected over period (i.e., November 2017 to March 2019).
two time periods: (i) between November 2017 We first describe the algorithmic approach
and July 2018 (Baseline) and (ii) between (illustrated in Figure 2) to understand Frailty in
August 2018 and March 2019 (Follow-up). In the elderly based on sensor data. Specifically,
addition, it utilizes survey data obtained from we aim to detect the symptoms of Frailty based
two surveys conducted on the elderly: baseline on the following daily living patterns of the
in March 2018 and follow-up in December elderly.
2018. The survey data includes information
about their demographics, mental health, • Movement within the home (i.e.,
physical health and psychosocial well-being, number of transitions from room to
and serve as ground truth validation for the room)
sensor-based frailty algorithm. The objective of
selecting this specific dataset is due to the
• Time spent in bedroom during the p dbd Percentage of days with
increased bedroom duration in
daytime (i.e., between 0600 to 1800) daytime
• Number of outings (i.e., number of p ddc Percentage of days with
reduced door contact events
door contact events)
pf Frailty percentage computed
from sensor feature dataset
The intuition for using only the above three MED(D) Function to return median
average
patterns will certainly help in determining the
COUNT(D) Function to return the size of
effective movement of the elderly within and dataset D
outside the home. Plus, given only the QUERY(D, Function to query the dataset D
availability PIR and Door sensor data we can “x ”) and return subset that satisfies
condition x
only able to determine the above three
movement patterns effectively.
The notations used in the Sensor-based Step 0 – Extract Features
Frailty algorithm is listed in Table 1. Each step
of the algorithm is described in the following. The Featured Sensor Dataset basically takes
the raw sensor input and create as many features
as possible so that it makes the dataset very
detailed and clear. The clarity in coming up
with featured dataset will itself reveal the data
patterns in more simplistic way and we can
have options to mine the dataset.
Table 2
Raw Sensor Dataset
Figure 2: Sensor-based Frailty Detection
Algorithm
The original raw dataset of sensor
approximately comes with Sensor ID, Sensor
Table 1
datetime and sensor location (refer Table-2) ,
Notations used in Sensor-based Frailty
then an algorithm on top of this raw sensor data
Algorithm
Notation Meaning
which will first create a Featured Sensor
Dsf Sensor Feature Dataset dataset in time series.
Ddrc Subset of feature dataset
corresponding to daily room change The featured dataset will include but not
Ddbd Subset of feature dataset limited to Observationtimestamp, elderlyid,
corresponding to daily bedroom
duration in daytime Date, Hour, Minute, Second, Day, Week,
Dddc Subset of feature dataset Month, Year, Weeday/Weekend,
corresponding to daily door contact FromLocation, Tolocation, Timespent,
events RoomChangeIndicator, “Daytime (Y/N)”,
idrc Indices of sensor feature dataset
entries corresponding to daily room “Timeperiod” i.e. whether it is morning 6.00 am
changes to 9.00 pm or Afternoon 3.00 pm to 6.00 pm
idbd Indices of sensor feature dataset etc.
entries corresponding to daily
bedroom duration in daytime
iddc Indices of sensor feature dataset Every row in raw sensor data will convert
entries corresponding to daily door from the above data structure. For example, If
contact events the raw sensor data has 1000 rows for 3 months
p drc Percentage of days with
reduced room change as an example, this will have 3 months of
enriched features. Few key fields of a sample
Featured Dataset is given in Table 3
duration in the daytime. By extracting a subset
The objective of this step is to extract useful comprising days for which the daily duration
features from time-stamped motion sensor and is above the median, we can determine the
door contact sensor readings that are relevant to percentage of days for which the elderly
frailty based on the raw sensor data. Table 3 experienced increased daily bedroom duration
provides a sample sensor feature dataset that as an indication of frailty. This is illustrated as
illustrates the movement of elderly1 from follows:
location to location, the duration in each
location, as well as the time period for which idbd = QUERY(Dsf ,“RoomChangeInd = N,
the transition/sojourn happens. FromLocation = bedroom,
ToLocation=bedroom,
Table 3 TimePeriod=["M6to9","M9to12","A12to3","E
Sample of Sensor Feature Dataset (Elderly1) 3to6"] & group by Date”)
Ddbd = Dsf{ idbd }
iidbd = QUERY(Ddbd ,“Ddbd > MED(Ddbd )”)
pidbd = (|Ddbd { iidbd } | / | Ddbd |) * 100
Step 1c – Calculate Movement outside
Home
Based on the ToLocation feature, we can
extract the daily total door event (i.e., daily
outings) over the duration of the dataset. By
Step 1a – Calculate movement between extracting a subset comprising days for which
rooms the daily total is below the median, we can
determine the percentage of days for which the
Based on the RoomChangeInd feature, we elderly experienced reduced daily door counts
can extract the daily total room changes over (i.e., reduced outings) as an indication of frailty.
the duration of the dataset. By extracting a This is illustrated as follows:
subset comprising days for which the daily total
is below the median, we can determine the iddc = QUERY(“ToLocation = Door & group
percentage of days for which the elderly by Date”)
experienced reduced daily room changes as an Dddc = Dsf{iddc}
indication of frailty. This is illustrated as irddc = QUERY(Dddc ,“Dddc < MED(Dddc)”)
follows: prddc = (|Dddc { irddc } | / | Dddc |) * 100
idrc = QUERY(Dsf, “RoomChangeInd = Y & Step 2 – Compute Frailty
group by Date”) This is the final step in fundamentally
Ddrc = Dsf{idrc} determining frailty of the elderly. Here, we
irdrc= QUERY(Ddrc ,“Ddrc < MED(Ddrc)”) assign weightage to each of the previous steps
prdrc = (|Ddrc { irdrc } | / | Ddrc |) * 100 to give more specific preference to certain
processes based on the domain understanding
The introduction of percentage in all the of the environment as follows:
steps (1a, 1b and 1c) helped us in determination • Step 1a - Movement between
of quantitative change in terms of Movement
Rooms - 50%
between Rooms (Step 1a) , Calculate Bedroom
dwell time (Step 1b) and Movement outside • Step 1b – Stay in Bedroom during
Home (Step 1c). Daytime – 30%
• Step 1c – Movement outside Home
Step 1b – Calculate Bedroom dwell time – 20%
during Daytime
Accordingly, we compute the Frailty
Based on the RoomChangeInd,
Percentage as follows:
FromLocation, ToLocation and TimePeriod
features, we can extract the daily bedroom
pf = prdrc*0.5 + pidbd *0.3 + prdc*0.2
Table 4
Some key advantages of the proposed Frailty percentage / indices
algorithm are:
• The algorithm uses only sensor data
which is considered unobtrusive rather
using other methods like wearable
devices.
• The algorithm does not require any
individual health information as it uses
only sensor data,
• The algorithm does not need a labelled
dataset unlike supervised machine
learning,
• The algorithm understands frailty
through statistical and custom formula
which can be easily fine-tuned for
optimization,
• Finally, the algorithm bases its findings 4.1 Frailty Classification
from a featured dataset that is derived
from the raw sensor data. According to an epidemiology study in
Singapore [4], the prevalence of pre-frailty and
For ground truth validation of the sensor- frailty among the elderly aged 65 years and
based frailty detection algorithm, we used up to older in Singapore stands at approximately
38 deficits to compute the frailty index in this 6.2% and 37% respectively. Accordingly, we
study as a higher number of deficits used has use the 45th percentile split at each time period,
been shown to yield better results. both for the frailty percentage and frailty index,
to divide the elderly into two groups: robust and
frail. The corresponding 45th percentile values
4. Results are provided in Table 4. Accordingly, the frailty
classification is given in Table 5.
In this section, we present some results
obtained for the proposed sensor-based frailty Table 5
detection algorithm for the period (i) November Frailty classification
2017 and July 2018 (Baseline) and (ii) between
August 2018 and March 2019 (Follow-up).
Among the SHINESeniors participants, 65
residents participated in the baseline survey,
and 47 participated in the follow-up survey. As
the profile of the residents who participated in
the surveys are different, we selected 39
residents with similar demographics. Out of
these 39 participants, those with insufficient
sensor data for the 2 periods, or who did not
participate in both surveys, were removed,
resulting in 11 residents (elderlies) for this
study.
From Table 5, it is observed that the Sensor-
based Frailty classification matches the Survey-
based Frailty classification for 13 out of 22
instances, achieving an accuracy of 59%.
To provide further insights into the results,
we perform case studies on two participants
(Elderly 3 & Elderly 1) by investigating their
frailty percentage and component scores, frailty
index, and other anecdotal information
obtained from the study. The latter includes
history of help request activation, fall history,
employment information etc. Specifically, we
plot Ddrc, Ddbd and Dddc for these elderlies over
the baseline duration in Figure 3 & 4
respectively. Figure 4: Select activity patterns of Elderly 1
4.2 Frailty Change (Baseline vs
Follow-up)
According to the Sensor-based Frailty
Detection algorithm, the frailty percentage
increased for 7 elderlies (became more frail)
and decreased for 4 elderlies (became less frail).
Figure 3: Select activity patterns of Elderly 3 Based on the survey, the frailty index increased
for 7 elderlies (became more frail), decreased
From Figure 4, we can observe a reduction for 3 elderlies (became less frail) and remained
in daily room movement door events plus unchanged for 1 elderly. The frailty change
increase in bedroom stay during daytime, which according to the survey is plotted in Figure 5.
are indicative signs of frailty. In fact, Elderly 3
experienced a fall and had just returned home
from nursing care in early December 2017.
Moreover, Elderly 3 requested for help through
the help button several times due to reduced
movement during this period. These
observations corroborate with Elderly 3 being
classified as frail based on the frailty percentage
(sensor data) as well as the frailty index (survey
data).
To further substantiate the help button
correlation with frailty, we examined another Figure 5. Frailty Change in Elderly (Baseline to
elderly, Elderly 1, who also activated the help Follow-up Survey)
button a few times seeking help. Based on the
sensor data analysis of Elderly 1 in Figure 4, we Among the 7 elderlies detected to be more
also observe a reduction in daily room frail based on sensor data, 5 of them were also
movement stay and door events, along with few deemed as more frail according to the survey.
increments on change in bedroom duration. Among the 4 elderlies detected to be less frail
Again, these observations corroborate with based on sensor data, 2 of them were also
Elderly 1 being classified as frail. deemed as less frail based on the survey. This
corresponds to a frailty change detection
accuracy of 63%. This suggests that sensor-
based frailty detection may be more accurate in
detecting frailty change over time as compared
to frailty classification.
5. Conclusion Human Aspects of IT for the Aged Population.
Applications in Health, Assistance, and
Entertainment. ITAP 2018. Lecture Notes in
In this paper, we demonstrated the
Computer Science, vol 10927. Springer, Cham.
feasibility of using in-home unobtrusive
sensors to autonomously detect frail or pre-frail
[4] Merchant RA, Chen MZ, Tan LWL, Lim
elderly as well as detect frailty change in the
elderly over time for community dwelling MY, Ho HK, van Dam RM. Singapore Healthy
Older People Everyday (HOPE) Study:
elderly living alone. Based on motion sensors
Prevalence of Frailty and Associated Factors in
installed in each room of the apartment and one
Older Adults. J Am Med Dir Assoc.
door contact sensor on the main door, we
2017;18(8):734.e9-734.e14.
proposed a frailty detection algorithm based on
doi:10.1016/j.jamda.2017.04.020.
movement between rooms, duration in
bedroom in the daytime, as well as door event
[5] Rockwood K, Mitnitski A. Frailty in
counts. These behavioral aspects of the elderly
relation to the accumulation of deficits. The
have been shown in a previous study to be
Journals of Gerontology Series A: Biological
useful in discriminating frail / pre-frail elderly
from robust elderly. This sensor-based frailty Sciences and Medical Sciences, 2007. 62(7): p.
722-727. doi:10.1093/gerona/62.7.722
algorithm is validated by frailty indices
computed based on deficit accumulation [5]
from surveys.
Using surveys conducted in March 2018
(baseline) and December 2018 (follow-up) and
sensor data from November 2017 to July 2018
(baseline) and August 2018 to March 2019
(follow-up) from 11 elderlies, we achieved
frailty classification accuracy and frailty
change detection accuracy of 59% and 63%
respectively. The 11 elderlies were split into
two groups: frail and non-frail using a 45th
percentile split, based on actual prevalence of
frailty from an epidemiology study conducted
in Singapore.
6. References
[1] SHINESENIORS OVERVIEW
URL: https://icity.smu.edu.sg/node/996
[2] Linda P. Fried, Catherine M. Tangen,
Jeremy Walston, Anne B. Newman, Calvin
Hirsch, John Gottdiener, Teresa Seeman,
Russell Tracy, Willem J. Kop, Gregory Burke,
Mary Ann McBurnie, Frailty in Older Adults:
Evidence for a Phenotype, The Journals of
Gerontology: Series A, Volume 56, Issue 3, 1
March 2001, Pages M146–
M157, https://doi.org/10.1093/gerona/56.3.M1
46
[3] Goonawardene N., Tan HP., Tan L.B.
(2018) Unobtrusive Detection of Frailty in
Older Adults. In: Zhou J., Salvendy G. (eds)