Towards an Ambient Support System for Continence Management in

Nursing Homes: An Exploratory Study

Hannelore Strauven

1,2 a

, Ine D’Haeseleer

1 b

, Kristof T’Jonck

3 c

, Hans Hallez

3 d

Vero Vanden Abeele

1 e

, Pieter Crombez

and Bart Vanrumste

1,2 f

e-Media Research Lab, KU Leuven, Andreas Vesaliusstreet 13, Leuven, Belgium

imec-Stadius, KU Leuven, Leuven, Belgium

imec-DistriNet, KU Leuven, Bruge, Belgium

Televic Healthcare NV, Izegem, Belgium

Keywords:

Incontinence, Nursing Home, Ambient Monitoring, Exploratory Study.

Abstract:

Time consuming and costly, continence care management has become one of the main care demands in nurs-

ing homes with potential inadequacy negatively impacting residents’ quality of life. While engineering efforts

in this area are increasing, these mainly focus on wearable innovations. To support continence care in nursing

homes in an unobtrusive manner, we developed an ambient sensor system to continuously monitor inconti-

nence events, e.g., saturated incontinence materials or leakages. In an exploratory study in two nursing homes,

we evaluated an early prototype of the sensor system and built annotated data sets. Implemented annotation

devices included a smart sensor mat, a toilet timing predicting device, and manual data entry of continence

care by care personnel. From our analysis of the preliminary study results based on the ﬁrst two residents, we

learned how challenging the ambient monitoring and annotation of incontinence events is. On the basis of the

outcomes, we provide suggestions for further research of ambient sensor systems supporting continence care.

1 INTRODUCTION

In long-term care settings, such as nursing homes,

over 50% of the older adults experience incontinence,

i.e., the involuntary leakage of urine or stool (Offer-

mans et al., 2009). Consequently, continence care

has become one of the main care demands (Wagg

et al., 2017). Over 20% of care time is spent on di-

rect continence care management, e.g., toileting as-

sistance, and changing incontinence materials (Ous-

lander and Kane, 1984). In addition, the average cost

for management is about US$15 per day per resi-

dent (Hu et al., 2003). Continence care is thus time

consuming and costly for nursing homes, moreover,

it also negatively impacts the residents’ quality of

life, e.g., by disturbing their sleep (Ouslander et al.,

https://orcid.org/0000-0002-7233-8137

https://orcid.org/0000-0001-5455-3581

https://orcid.org/0000-0003-4150-4144

https://orcid.org/0000-0003-2623-9055

https://orcid.org/0000-0002-3031-9579

https://orcid.org/0000-0002-9409-935X

1998). In a generally ageing society, the prevalence

of incontinence will only increase, resulting in addi-

tional pressure on the already overburdened staff and

institutional costs (Wilson et al., 2001; Mather and

Bakas, 2002). Previous engineering research focused

on wearables, such as the development of smart in-

continence wear (Fish and Traynor, 2013; Lin et al.,

2017). To the best of our knowledge, no studies in-

volve unobtrusive, ambient sensor technology to sup-

port continence care in nursing homes.

We are interested in the potential of an ambient

sensor system to support continence care in nursing

homes. To evaluate the system, this paper describes

an exploratory ﬁeld study with an early prototype de-

ployed in two nursing homes.

We ﬁrst elaborate on continence care practices,

and how technology approaches can support them.

We then proceed with the implementation of the pro-

totype, together with the applications and devices for

annotation. Following implementation, a ﬁrst evalu-

ation of the system is presented through preliminary

results and user feedback. We close by putting for-

ward considerations for further system development

438

Strauven, H., D’Haeseleer, I., T’Jonck, K., Hallez, H., Abeele, V., Crombez, P. and Vanrumste, B.

Towards an Ambient Support System for Continence Management in Nursing Homes: An Exploratory Study.

DOI: 10.5220/0008963404380446

In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 5: HEALTHINF, pages 438-446

ISBN: 978-989-758-398-8; ISSN: 2184-4305

 2022 by SCITEPRESS – Science and Technology Publications, Lda. All r ights reserved

Table 1: Overview of Ambient Intelligence (AmI) technologies evaluated in nursing homes to support care management.

Authors Care domain Technology Device

Huion et al. Continence Wearable Smart diaper (WetSens)

Traynor et al. Continence Wearable Smart diaper (SIM)

Wai et al. Continence Wearable Smart diaper (iCMS)

Aloulou et al. Activities of daily living Ambient assistive living Unobtrusive sensors & devices

Hori & Nishida Activities of daily living Ambient assistive living Ultrasonic sensors

Rantz et al. Activities of daily living Ambient assistive livings Unobtrusive sensors & devices

Suzuki et al. Activities of daily living Ambient assistive living Infrared sensors

and associated research studies in nursing homes.

1.1 Background

During the day, over 20% of care tasks are directly re-

lated to continence care, increasing to 70% during the

night. For more interdependent older adults, the esti-

mated time reaches nearly one hour per nursing home

resident per day (Ouslander and Kane, 1984; Borrie

and Davidson, 1992). Continence care practices in-

volve periodic manual check-ups of incontinence ma-

terial, toilet visits and continence assessments. Cur-

rent practices lead to unnecessary controls or delayed

interventions, triggering undesired instances such as

disturbed sleep (O’Dell et al., 2008). These are un-

comfortable situations for care personnel and resi-

dents alike.

Efforts have been made to support continence care

management through Ambient Intelligence (AmI)

systems. AmI brings intelligence to physical envi-

ronments and measures them through sensors (Cook

et al., 2009). It has the potential to improve the health-

care domain and mainly consists of two types of tech-

nology (Acampora et al., 2013):

• Body Area Network (BAN): body sensors are at-

tached on clothing or on the body, commonly

known as wearables, or even implanted under the

skin, e.g., to measure body temperature, blood

pressure, or cardiac activity;

• Wireless Mesh Sensor Networks (WMSN): am-

bient sensors are embedded into the environment,

e.g., to measure room temperature, opening doors,

or movement.

Research can be found on the development of smart

incontinence wear, i.e., a BAN with a wearable de-

tecting the saturation level of material via an inte-

grated humidity sensor and alerting care personnel

for required check-ups. Exploratory studies in nurs-

ing home settings evaluated several smart diaper pro-

totypes, identifying and testing technical speciﬁca-

tions (Table 1) (Huion et al., 2019; Traynor et al.,

2014; Wai et al., 2010b). Conducted research to de-

velop an intelligent Continence Management System

(iCMS) for nursing homes via smart incontinence ma-

terials and wetness alert notiﬁcations on a smartphone

for care personnel, faced several technical challenges

(Wai et al., 2008; Wai et al., 2010a; Wai et al., 2010b;

Wai et al., 2011). The researchers concluded that de-

sirable features of an intelligent continence care man-

agement system could range from unobtrusive contin-

uous monitoring to odour-based detection, pointing

towards a recommendation to shift continence care

technology from BANs to WMSNs.

In contrast to the technological solutions to sup-

port continence care, WMSN is already thoroughly

researched and evaluated in the care domain of ac-

tivities of daily living (ADLs) by the name of Ambi-

ent Assistive Living (AAL) (Table 1) (Aloulou et al.,

2013; Hori and Nishida, 2005; Rantz et al., 2010a;

Suzuki et al., 2006). To monitor ADLs, research

utilised ambient sensors and devices, ranging from

low cost sensors (e.g., pressure sensors, motion sen-

sors, bed sensors, and vibration sensors) and devices

of interaction (e.g., camera, speakers, and tablets)

(Aloulou et al., 2013; Rantz et al., 2010a) to ultra-

sonic sensors (Hori and Nishida, 2005) or infrared

sensors (Suzuki et al., 2006).

Monitoring via ambient sensors presents an op-

portunity to develop unobtrusive technology for user

groups like nursing home residents. Although, plenty

of AAL research is conducted in nursing home set-

tings, to date and to the best of our knowledge, no

literature can be found in this area on ambient conti-

nence care monitoring.

1.2 Scope of this Work

This paper focuses on the implementation and evalu-

ation of the prototype of an ambient sensor system to

unmask incontinence events through an exploratory

study. Present nursing home technology and techni-

cally validated devices were used to annotate conti-

nence care events and, afterwards, label them. Our

aim is then to avoid these unpleasant incontinence

events in the future. The ﬁrst results of this study

will enrich further development of the ambient sup-

port system and provide suggestions for future re-

Towards an Ambient Support System for Continence Management in Nursing Homes: An Exploratory Study

439

search studies in nursing home settings.

2 IMPLEMENTATION

The explorative ﬁeld study was carried out in two

nursing homes to build annotated data sets, necessary

for system evaluation and further technical develop-

ment. Data was acquired via four different sources

(Figure 1):

• The sensor system prototype (blue): a developed

non-obtrusive sensor system that was evaluated;

• Manual data entry by care personnel (red): nurse

calls via the nurse call system (green) and conti-

nence care (Report Inco) via the resident care ﬁle

(cyan);

• The Toilet Timing Predicting Device DFree

(pur-

ple): validated non-invasive ultrasound technol-

ogy;

• The smart sensor mat Texible Wisbi

(brown):

validated wet sheet technology.

Figure 1: Implementation overview of the sensor system

prototype, the manual data entry, and technically validated

devices.

2.1 Sensor System Prototype

The aim was to develop a sensor system to identify in-

continence events. To preserve residents’ comfort, we

wanted the system to be unobtrusive and, therefore, to

include ambient sensors.

We developed a sensor system prototype to be at-

tached to a care bed, monitoring and detecting incon-

tinence events (T’Jonck et al., 2019). The prototype

was designed with commercial off-the-shelf compo-

nents. The chosen sensors were breakout boards, i.e.,

dfree.biz/en/

www.texible.at/

click boards, from MikroE

that use the mikroBUS

speciﬁcation: Weather click, Temp&Hum 2 click, Ac-

cel 5 click, and Air quality 5 click. The decision

to integrate the Air quality 5 click sensor was based

on research concluding that this sensor is able to de-

tect concentrations of ammonia (NH3) in the range

of concentrations that are present in urine (Strauven

et al., 2019).

A Raspberry Pi

was used as room gateway

to send sensor data via an Long Term Evolution

(LTE) network to a secure InﬂuxDB

database. Data

was stored in real-time and could be exported from

Grafana

, an open source visualisation application

that supports the direct integration of the InﬂuxDB

database. The Grafana dashboard monitored the sen-

sor modules remotely and in real-time.

The sensor system prototype recorded sensor data

every second. The temperature, humidity and ammo-

nia sensors were ﬁxated in the middle of the bed bar,

located on the side of the care bed (Figure 2). The ac-

celerometer was positioned between the bottom of the

care bed and the mattress. Sensor data was collected

during the entire study.

Figure 2: Picture of the sensor module at the care bed, in-

cluding the Accel 5 click (left), the Weather click and Air

quality click (middle) and Temp&Hum 2 click (right).

2.2 Manual Data Entry

The manual data entry of the care personnel provided

additional insights in the care needs of participants

and was entered through two nursing home technolo-

gies, the nurse call system (NCS) and the resident care

ﬁle (RCF). Care personnel was requested to provide

this data during the entire study. Each annotation was

www.mikroe.com/

www.raspberrypi.org/

www.inﬂuxdata.com/

grafana.com/

HEALTHINF 2020 - 13th International Conference on Health Informatics

440

Table 2: Timeline of the exploratory ﬁeld study in the ﬁrst nursing home for two residents.

Timeline Action

19/07/19 Installation sensor module, NCS application, RCF incontinence report

21/07/19 Start measurement for three weeks with resident 1

29/07/19 Start DFree & Wisbi

12/08/19 End sensor measurement resident 1

14/08/19 Start measurement for three weeks with resident 2

21/08/19 Start DFree & Wisbi

04/09/19 End measurement resident 2

timestamped to link the entry to the sensor data after-

wards.

Residents can call for assistance via the NCS, a com-

munication and management application. If residents

need assistance, they call by pushing an alarm button,

that generates a call at the care personnel’s phone.

For this study, care personnel received smartphones

with the application ‘Call Manager’

, designed for

this purpose, and running on the Televic NCS (Figure

1). When a nurse call was triggered, care personnel

could, via the application, ﬁrst select the reason of the

call from a list and then an appropriate action to assist

the resident from a contextual list.

The RCF is a software application that stores in-

formation about the health and care plan of residents.

Care personnel can obtain all necessary information

about the residents they care for via the RCF. For this

study, additional continence care content was added

to Corilus’ RCF application, ‘Geracc’

, to let care

personnel report more extensively about conducted

continence care tasks (Figure 1). The additional tasks

ranged from the check up of incontinence material to

toilet visits. To facilitate the extra input, a Microsoft

(MS) Surface was installed in participants’ room, al-

lowing care personnel to enter the information right

after taking care.

2.3 Validated Technology

The TripleW Toilet Timing Predicting Device DFree

detects bladder size by using harmless, non-invasive

ultrasound technology. The sensor is placed at the ab-

domen in the area of the pubic bone. The main unit,

i.e., another part of the device connected to the sen-

sor, sends data wirelessly to a secured web portal of

TripleW. Participants were asked to wear DFree for

three to four days.

The smart sensor mat Texible Wisbi is a portable

application and consists of an intelligent mattress pad

that detects the humidity level of the fabric. In ad-

www.televic-healthcare.com/en/solutions/mobile-

alarming

www.corilus.be/en/elderly/careconnect-elderly

dition, Wisbi consists of an occupation detector to

detect presence on the sensor mat. The sensor mat

is attached to an external transmitter that sends the

data through a WLAN network to an application on

a smartphone. The smart sensor mat was placed be-

tween the mattress and sheet of a participant’s care

bed so they lied on the fabric when lying in bed. The

smart sensor mat was used by participants during the

same period as DFree.

2.4 Participants

We recruited ten residents across two Belgian nursing

homes, Sint-Bernardus in Bertem and Biezenheem

in Kortrijk, in collaboration with the nursing home’s

care personnel. Participants’ rooms were provided

with the sensor system prototype for three weeks.

Ethical approval to conduct the research was obtained

from the Social and Societal Ethics Committee of KU

Leuven with protocol number G- 2019 01 1510.

Inclusion criteria for participants were: 65 years

or older, living in one of the two participating nurs-

ing homes, having continence care needs (Katz score

above 1 for Continence (Katz et al., 1963)), and

being able to participate independently, understand-

ing the purpose and involvement and providing con-

sent (Mini-Mental State Examination (MMSE) score

above 18 (Folstein et al., 1975)). An exclusion cri-

terion was when the resident had a separate medical

condition that inﬂuences continence.

3 EVALUATION

Measurements were completed in the ﬁrst nursing

home as the exploratory study was ongoing in the sec-

ond nursing home. Three out of ﬁve selected partici-

pants from the ﬁrst nursing home were excluded: one

removed themselves before the start, one decided to

discontinue during the study, and one was transferred

to another ward. Therefore, this section discusses the

preliminary results from two residents, measured over

a period of three weeks from July to September 2019.

Towards an Ambient Support System for Continence Management in Nursing Homes: An Exploratory Study

441

Figure 3: Time series plots of data from Resident 1: the sensor system prototype with the ammonia (NH3), temperature (T)

and humidity (RH) sensor, and the x-axis of the accelerometer (AccX) (top), the DFree (middle), and the Wisbi wetness and

occupation detector (bottom) over a period of 3.5 days.

Table 2 presents a timeline of all systems and appli-

cations used during the study.

3.1 Data Exploration

Data is explored for resident 1 (Figure 3) and resi-

dent 2 (Figure 4) individually by time series plots over

a period of 3.5 days. Resident 1 was a 92-year-old

woman, used a wheelchair due to a physical disability,

and highly dependent on carers. She was also func-

tionally incontinent (i.e., unable to get to a bathroom

for one or more physical or mental reasons). Resident

2 was 88-year-old woman and an active resident who

independently took care of herself. She experienced

urge incontinence (i.e., unable to postpone the desire

to void).

The results of the sensor system prototype are dis-

played at the top of the ﬁgures, for the ammonia

(NH3), temperature (T) and humidity (RH) sensor,

and the x-axis of the accelerometer (AccX). For res-

ident 2, the accelerometer lost connection during the

study and is not included in the plot. The y-axis of the

NH3 signal is reversed for better understanding, as the

sensor’s response is inversely proportional to the NH3

concentration. As the sensor module was installed at

the care bed, only periods when a resident was in bed

are relevant for data analysis. To illustrate this time

range, the nights are marked grey. Resident 1 went to

bed around 20:00 and woke up around 08:00. Resi-

dent 2 had a different circadian rhythm and went to

bed around 23:00 and woke up around 09:00.

The NH3 signal varied less in time during the

night, compared to the day. This can be explained

by the difference in motion. During the day, residents

were moving around, opening/closing doors and win-

dows, received care, or had visitors. During the night,

the room was quiet, providing less motion and, there-

fore, a more stable signal. For resident 1, we would

expect an increase of the NH3 signal at night when

she was in bed, wearing incontinence material, and a

decrease in the morning when she was lifted out of

bed. For multiple nights, we could identify a gentle

increase of the NH3 signal over time. After the night,

an elevation was observed which might relate to the

change of the night incontinence material in the morn-

ing. For resident 2, we did not expect the same varia-

tion as she went to the toilet at night as well. Hence,

the NH3 signal remained more stable at night.

We noticed that the temperature raised during the

day and lowered during the night, and how the humid-

ity values related inversely to temperature measures.

Both levels and variances were in accordance with our

expectations for the season and environment.

The AccX signal was low during the day and

HEALTHINF 2020 - 13th International Conference on Health Informatics

442

Figure 4: Time series plots of data from Resident 2: the sensor system prototype with the ammonia (NH3), temperature (T)

and humidity (RH) sensor (top), the DFree (middle), and the Wisbi occupation detector (bottom) over a period of 3.5 days.

high during the night, following resident 1’s day/night

regime. Therefore, the accelerometer can be consid-

ered an out-of-bed detector.

The AccX signal remained high after the third night,

when we would expect it to turn low. The reason for

this alteration was uncertain as, for example, it was

possible resident 1 stayed in bed that day, or care per-

sonnel made up the bed and changed the position of

the sensor.

The ﬁgures’ middle sections display the output of

TripleW DFree by the average bladder volume per

minute. The signal was continuously spiking, and

harder to be examined visually in time than foreseen.

In addition, DFree had a short battery lifetime of ca.

24 hours so the device had to be charged periodically,

which explains the gaps in the plots.

The output of Texible Wisbi is visualised at the ﬁg-

ures’ bottom. For resident 2, Wisbi did not detect any

wetness; subsequently only the occupation plot is in-

cluded. When the occupation signal was high, this

largely corresponded with the marked night times, ex-

cept for some disparities. For resident 2, we observed

low occupation periods during the night. The resident

mentioned she went to the toilet at night, so we could

assume these low signal periods coincide with these

instances. To deﬁne the resting night times, the Wisbi

was more accurate as the predeﬁned ranges based on

the habitual times. We could see one wetness peak

for resident 1 at the second morning. The DFree plot

showed a low bladder volume at the same time and

the NH3 signal increased. This means it was possible

that the incontinence material became oversaturated

or leaked. On the other hand, the AccX signal was

low, and in contrast, the occupation detector of Wisbi

high. Since the accelerometer was located behind the

upper part of the care bed and the sensor mat under

the bottom, it was possible the resident was sitting on

the bed. For resident 1, the same unexpected high sig-

nal for the occupation detector was not noticed as for

AccX after the third night. Therefore, it was more

likely that the position of the accelerometer changed,

for example, after making up the bed.

In the window of 3.5 days, one annotation was

made via Geracc, where care personnel assisted resi-

dent 1 at the toilet, and one via Call Manager, where

resident 1 called to be taken off the toilet. We would

expect the bladder volume to decrease before these

annotations, which was observed for the ﬁrst annota-

tion but not visible for the second.

During the measurement period of three weeks,

18 nurse calls were recorded via Call Manager: 16

for resident 1 and two for resident 2. The number

of entries seemed low as resident 1 needed assistance

for each toilet visit and used the NCS to alert care

personnel when ready. For this reason, we would ex-

pect at least two calls per day (i.e., 42 calls for three

weeks). Although the total number of entries was

rather low, Figure 5 gives some insight in the type

of nurse calls. It appeared that resident 1 called more

than resident 2, and mostly for toilet assistance and

Towards an Ambient Support System for Continence Management in Nursing Homes: An Exploratory Study

443

Figure 5: Result of manual data entry by care personnel via

Call Manager for resident 1 and resident 2.

movement. This can be associated with her physi-

cal impairments, whereby she needed more assistance

with ADLs compared to resident 2.

Via Geracc, three continence care records were gener-

ated for resident 1 and none for resident 2. In relation

to the nurse call records, the number of annotations in

Geracc was remarkably low, as there should be at least

a continence care record for each toilet assistance. On

the basis of the three records, we could only identify

that all records entered were toilet visits and that in-

continence material was changed during each toilet

visit, even if the material was not yet saturated.

3.2 User feedback

When we ﬁnished the study in the ﬁrst nursing home,

we obtained user feedback. Residents who partici-

pated were asked how they experienced their partici-

pation in a one on one conversation with a researcher.

Care personnel of the ward was asked for their feed-

back in group by two researchers during their daily

ward meeting.

Both residents were unfamiliar with the devices

and applications installed in their room for this study.

It was an adjustment of their environment which felt

uncomfortable. However, after a few days, they

adapted to the changes. One resident enjoyed to be

involved and notiﬁed care personnel when a device

was not working properly (i.e., when the lights were

not blinking). Wearing the sensor of DFree was per-

ceived tolerable, at least for a few (ca. four) days. The

level of tolerance depended mostly on the mobility of

the resident. For example, resident 2 walked around

during the day and had to think about the clothes she

could wear when wearing DFree, so others would not

see the device. The use of Wisbi went largely unno-

ticed by the residents.

Care personnel found it difﬁcult to recommend

residents as suitable participants. In a nursing home,

only a limited number of residents have a MMSE

score above 18 and are able to understand the purpose

of and involvement in a research study.

The sensor system attached to the care bed might be

designed more robust or cased for care personnel to

ease making the bed or cleaning underneath. Largely,

care personnel had little experience with smart elec-

tronic devices, e.g., smartphones. This made anno-

tating challenging for them as they were learning to

operate a new device as well as a new application.

DFree was perceived as easy to use, once they found

the proper sensor location on the abdomen of the res-

ident. They preferred a longer battery lifetime as they

had to remove and reload the device daily. At last,

they experienced Wisbi as straightforward to use.

4 DISCUSSION AND

CONCLUSION

As mentioned before (see Table 1), technical inno-

vations to support continence care in nursing homes

are predominantly wearable systems. After conduct-

ing several trials, including exploratory studies, re-

searchers (Wai et al., 2010b) stressed the importance

of an unobtrusive system and odour-based detection.

When looking at other care domains, e.g., ADLs, us-

ing AmI to design unobtrusive technology is already

further explored. On the basis of the need to support

continence care by ambient monitoring, a ﬁrst proto-

type was developed.

Through the implementation and evaluation of the

prototype via an exploratory study, we evaluated the

preliminary results from two participants of the ﬁrst

nursing home. They provided us with sensor data

from the prototype, together with data from devices

for annotation purpose to obtain a reliable ground

truth. In this paper, we analysed and compared data

from all devices in a window of 3.5 days. Patterns

were found that illustrated the alteration of the en-

vironment among day and night, and indicated that

the accelerometer can be used for out-of-bed (or oc-

cupation) detection. The NH3 signal was highly in-

ﬂuenced by an environment in motion. During the

night, the signal was smoother as the environment was

more steady. The visual identiﬁcation of incontinence

events appears to be challenging. To lower the inﬂu-

ence of motion, we suggest to locate the NH3 sensor

at a position on the care bed where less motion oc-

curs, e.g., under the sheets, and to integrate the sensor

system in a robust case.

Finding and providing an appropriate annotation

method in exploratory studies to label ground truth

is known to be challenging, but also incredibly valu-

able for further qualitative, in-depth data analysis (Al-

HEALTHINF 2020 - 13th International Conference on Health Informatics

444

shammari et al., 2017).

Once initiated, a trial could not be repeated, as

nursing home residents are frail elderly and would not

have the resilience for a second try. Every day with-

out annotation can be seen as an expensive loss (Hein

et al., 2017). Common practice for annotating AmI

studies is to videotape the recording session and to

label the data based on the video footage afterwards

(Plotz et al., 2012). Although this technique provides

an accurate labelling, it is ethically inappropriate to

videotape in real-life situations due to privacy reasons

(Aicha et al., 2017). Our sensor setup was in nursing

home residents’ rooms for three weeks and would in-

terfere with the privacy of the residents as well as their

relatives and nursing home staff, who all enter the

room during the study. In addition, it is uncertain that

incontinence events occurring under a sheet, would be

noticed on video footage. For this reason, we opted

for a combination of manual data entry and techni-

cally validated devices (DFree and Wisbi) to enable

ground truth labelling. The obtained data of Wisbi

was straightforward, however, the data from DFree

turned out to be harder to interpret than expected.

Manual data entry by care personnel, e.g., via log

sheets or questionnaires, was similarly part of our pro-

tocol as previously mentioned by AmI research (Table

1) (Aloulou et al., 2013; Rantz et al., 2010b; Suzuki

et al., 2006). However, in our study, we encoun-

tered multiple difﬁculties to let care personnel anno-

tate. Care personnel was inexperienced with the use

of smartphones, which caused insecurity in using the

device and generated issues directly related with the

smartphone, instead of the application use. These is-

sues eventually resulted in frustrations and a reduced

usage of the devices. As a result, the number of an-

notations were limited. To obtain an enriched set of

manual data entries for exploratory studies in real-life

settings, such as nursing homes, we suggest to extend

the information about the use of the annotation appli-

cations with an explanation of the used device or en-

abling annotation through devices that have already

been used by the annotators prior to the study.

We will continue our research with an in-depth

analysis of all acquired data and ground truth la-

belling to deconstruct this complex data and identify

incontinence events, e.g., the saturation level of the

incontinence material or leakages. Further research

of an ambient monitoring system supporting conti-

nence care should consider the altering environment

and monitor at places with the lowest impact. In ad-

dition, care personnel should be assisted more thor-

oughly and consistently in the annotation process.

ACKNOWLEDGEMENTS

The imec.icon project DISCRETE runs from Octo-

ber 1, 2018 to September 30, 2020 and joins forces

of commercial partners Televic Healthcare, Corilus,

Distrac Group and the knowhow of Zorg Kortrijk and

WZC Sint-Bernardus, with the scientiﬁc expertise of

researchers from van imec-KU Leuven-DISTRINET,

imec-KU Leuven-STADIUS and KU Leuven-HCI.

The project is funded by Flanders Innovation & En-

trepreneurship. We would like to thank the nursing

homes’ care personnel and residents for their partici-

pation during the exploratory ﬁeld study, and our col-

league, Katta Spiel, for their feedback that greatly im-

proved the paper.

REFERENCES

Acampora, G., Cook, D. J., Rashidi, P., and Vasilakos, A. V.

(2013). A survey on ambient intelligence in health-

care. Proceedings of the IEEE, 101(12):2470–2494.

Aicha, A. N., Englebienne, G., and Kr

ose, B. (2017). Un-

supervised visit detection in smart homes. Pervasive

and Mobile Computing, 34:157–167.

Aloulou, H., Mokhtari, M., Tiberghien, T., Biswas, J., Phua,

C., Kenneth Lin, J. H., and Yap, P. (2013). Deploy-

ment of assistive living technology in a nursing home

environment: methods and lessons learned. BMC

Medical Informatics and Decision Making, 13(1):42.

Alshammari, N., Alshammari, T., Sedky, M., Champion, J.,

and Bauer, C. (2017). Openshs: Open smart home

simulator. Sensors, 17(5):1003.

Borrie, M. J. and Davidson, H. A. (1992). Incontinence

in institutions: costs and contributing factors. CMAJ:

Canadian Medical Association Journal, 147(3):322–

328.

Cook, D. J., Augusto, J. C., and Jakkula, V. R. (2009).

Ambient intelligence: Technologies, applications, and

opportunities. Pervasive and Mobile Computing,

5(4):277–298.

Fish, P. and Traynor, V. (2013). Sensor technology: a smart

way to manage continence.

Folstein, M. F., Folstein, S. E., and McHugh, P. R. (1975).

“Mini-mental state”: A practical method for grading

the cognitive state of patients for the clinician. Journal

of Psychiatric Research, 12(3):189–198.

Hein, A., Kr

uger, F., Bader, S., Eschholz, P., and Kirste,

T. (2017). Challenges of Collecting Empirical Sensor

Data from People with Dementia in a Field Study.

Hori, T. and Nishida, Y. (2005). Ultrasonic Sensors for

the Elderly and Caregivers in a Nursing Home. pages

110–115.

Hu, T.-W., Wagner, T. H., Bentkover, J. D., LeBlanc, K.,

Piancentini, A., Stewart, W. F., Corey, R., Zhou, S. Z.,

and Hunt, T. L. (2003). Estimated economic costs

of overactive bladder in the United States. Urology,

61(6):1123–1128.

Towards an Ambient Support System for Continence Management in Nursing Homes: An Exploratory Study

445

Huion, A., Decalf, V., Kumps, C., Witte, N. D., and Ev-

eraert, K. (2019). Smart diapers for nursing home

residents with dementia: a pilot study. Acta Clinica

Belgica, 74(4):258–262.

Katz, S., Ford, A. B., Moskowitz, R. W., Jackson, B. A.,

and Jaffe, M. W. (1963). Studies of Illness in the

Aged: The Index of ADL: A Standardized Measure

of Biological and Psychosocial Function. JAMA,

185(12):914–919.

Lin, S. H., Kajiyama, K., and Wu, T. (2017). Smart Di-

aper: How It Works. In Proceedings of the 2017

ACM International Joint Conference on Pervasive and

Ubiquitous Computing and Proceedings of the 2017

ACM International Symposium on Wearable Comput-

ers, UbiComp ’17, pages 129–132, New York, NY,

USA. ACM. event-place: Maui, Hawaii.

Mather, K. F. and Bakas, T. (2002). Nursing assistants’

perceptions of their ability to provide continence care.

Geriatric Nursing, 23(2):76–81.

O’Dell, Katharine, K., Jacelon, C., and Morse, A. N.

(2008). ’I’d Rather Just Go On as I Am’ - Pelvic Floor

Care Preferences Of Frail, Elderly Women in Residen-

tial Care. Urologic Nursing, 28(1):36–47.

Offermans, M. P. W., Moulin, M. F. M. T. D., Hamers, J.

P. H., Dassen, T., and Halfens, R. J. G. (2009). Preva-

lence of urinary incontinence and associated risk fac-

tors in nursing home residents: A systematic review.

Neurourology and Urodynamics, 28(4):288–294.

Ouslander, J. G., Buxton, W. G., Al-Samarrai, N. R., Cruise,

P. A., Alessi, C., and Schnelle, J. F. (1998). Nighttime

Urinary Incontinence and Sleep Disruption Among

Nursing Home Residents. Journal of the American

Geriatrics Society, 46(4):463–466.

Ouslander, J. G. and Kane, R. L. (1984). The Costs of Uri-

nary Incontinence in Nursing Homes. Medical Care,

22(1):69–79.

Plotz, T., Chen, C., Hammerla, N. Y., and Abowd,

G. D. (2012). Automatic Synchronization of Wear-

able Sensors and Video-Cameras for Ground Truth

Annotation–A Practical Approach. pages 100–103.

IEEE.

Rantz, M., Skubic, M., Alexander, G., Aud, M., Wake-

ﬁeld, B., Galambos, C., Koopman, R., and Miller, S.

(2010a). Improving Nurse Care Coordination With

Technology. Computers, informatics, nursing : CIN,

28:325–32.

Rantz, M., Skubic, M., Alexander, G., Popescu, M., A Aud,

M., J Wakeﬁeld, B., Koopman, R., and J Miller, S.

(2010b). Developing a Comprehensive Electronic

Health Record to Enhance Nursing Care Coordina-

tion, Use of Technology, and Research. Journal of

gerontological nursing, 36:13–7.

Strauven, H., Hallez, H., Vanden Abeele, V., and Vanrum-

ste, b. (2019). The Design and Development of an

Ambient Sensor System to Detect Incontinence.

Suzuki, R., Otake, S., Izutsu, T., Yoshida, M., and Iwaya,

T. (2006). Monitoring Daily Living Activities of

Elderly People in a Nursing Home Using an In-

frared Motion-Detection System. Telemedicine and

e-Health, 12(2):146–155.

T’Jonck, K., Strauven, H., Vankeirsbilck, J., Hallez, H., and

Boydens, J. (2019). Design and implementation of

an unobtrusive sensor system to support incontinence

care of elderly in nursing homes. Sozopol, Bulgaria.

Traynor, V., Yu, P., Hailey, D., and Fleming, R. (2014). An

exploration of the effects of introducing a telemoni-

toring system for continence assessment in a nursing

home. Journal of Clinical Nursing, 23:3069–3076.

Wagg, A., Kung Chen, L., Johnson II, T., Kirschner-

Hermanns, R., Kuchel, G., Markland, A., Orme, S.,

Ostaszkiewicz, J., Szonyi, G., and Wyman, J. (2017).

Incontinence in frail older persons. Incontinence : 6th

International Consultation on Incontinence, Tokyo,

September 2016, pages 1309–1442.

Wai, A. A. P., Foo, S. F., Jayachandran, M., Biswas, J., Nu-

gent, C., Mulvenna, M., Zhang, D., Craig, D., Pass-

more, P., and Lee, J.-E. (2010a). Towards develop-

ing effective Continence Management through wet-

ness alert diaper: Experiences, lessons learned, chal-

lenges and future directions. pages 1–8. IEEE.

Wai, A. A. P., Foo, S. F. V., Jayachandran, M., Biswas, J.,

Nugent, C., Mulvenna, M., Zhang, D., Craig, D., Pass-

more, P., and Lee, J.-E. (2010b). Technical develop-

ment and clinical evaluation of intelligent continence

management system at nursing home. pages 345–352.

IEEE.

Wai, A. A. P., Fook, F. S., Jayachandran, M., Song, Z.,

Biswas, J., Nugent, C., Mulvenna, M., Lee, J.-E., and

Yap, L. K. P. (2008). Smart wireless continence man-

agement system for elderly with dementia. pages 33–

34. IEEE.

Wai, A. P., Foo, S., Biswas, J., Donnelly, M., Parente, G.,

Nugent, C., and Yap, P. (2011). Smart phone reminder

system for managing incontinence at nursing home.

pages 254–259. IEEE.

Wilson, L., Brown, J. S., Shin, G. P., Luc, K.-O., and Subak,

L. L. (2001). Annual direct cost of urinary inconti-

nence. Obstetrics & Gynecology, 98(3):398–406.

HEALTHINF 2020 - 13th International Conference on Health Informatics

446