Overnight Supervision of Alzheimer’s Disease Patients

in Nursing Homes

System Development and Field Trial

Laura Montanini

, Laura Raffaeli

, Adelmo De Santis

, Antonio Del Campo

, Carlos Chiatti

Giorgio Rascioni

, Ennio Gambi

and Susanna Spinsante

Information Engineering Department, Universit

a Politecnica delle Marche, Ancona, Italy

Italian National Research Centre on Ageing (INRCA), Ancona, Italy

ArieLAB S.r.l., Ancona, Italy

Keywords:

Environmental Sensors, Unobtrusive Monitoring, Activity Detection, Alarm Notiﬁcation, Behavioural

Analysis.

Abstract:

The number of patients affected by Alzheimer’s disease among the population is currently growing, while

the availability of resources for their assistance is decreasing. A solution for this problem is provided by

the use of Ambient Assisted Living technologies, with the objectives to prolong the independent living of

patients at home, to relieve assistance burden on caregivers, and to improve care effectiveness in nursing

homes. This paper describes an integrated system designed to support the work of nurses during the night, to

ensure comfort and safety of Alzheimer’s disease patients in nursing homes. The project started from a similar

solution designed for home use, suitably re-engineered for adoption in nursing homes. The system has been

designed according to nurses’ requirements and expectations, both by revising some existing functionalities,

and by developing new components. The results gained from an experimental trial are also presented and

discussed.

1 INTRODUCTION

Staff working in care and nursing homes typically ex-

periences a high workload, due to the need of carry-

ing out a lot of tasks in a relatively short period of

time. Usually, this happens because of budgetary re-

strictions on the amount of personnel recruited, with

respect to the number of patients cared after. In

fact, since 2010, due to the global economic cri-

sis, growth in public health spending came almost

to a halt across the OECD (Organisation for Eco-

nomic Cooperation and Development), with even re-

ductions in many countries. Since then, the spending

growth has been very slow, often in line with over-

all economic growth (Organization for Economic Co-

operation and Development, 2015). Despite the cur-

rent trend of moving long-term care out of institu-

tions into patients’ home premises, the role of nurs-

ing homes remains relevant, especially for those pa-

tients affected by chronic diseases, like dementia or

Alzheimer’s disease (AD), who cannot be assisted at

home. Information and Communication Technolo-

gies (ICT) should be exploited to improve the work-

ing conditions of the care staff, and to improve the

quality of care. Experiences showed that the impact

of technology on underlying clinical work processes

should be carefully evaluated and analysed. Possible

blocks in the execution of routine procedures due to

the adoption of technology tend to distract staff from

care issues, and can result in new errors. Typically,

in reaction to this condition, nurses develop problem-

solving behaviours that involve bypassing new tech-

nology, or adapting work process so as to minimize

disruption in operational procedures (Bowens et al.,

2010; Lowry et al., 2015; Huston, 2013).

Several ICT-based solutions have been proposed

to facilitate home-caring of people affected by de-

mentia or AD during the night hours. In fact, night-

time activity is a common occurrence in persons with

dementia, which increases the risk for injury and unat-

tended home exits, and impairs the sleep patterns of

caregivers (Lee et al., 2014; Kim et al., 2014). Tech-

nology has been applied to develop tools that alert

caregivers of suspicious nighttime activity, to help

Montanini, L., Raffaeli, L., Santis, A., Campo, A., Chiatti, C., Rascioni, G., Gambi, E. and Spinsante, S.

Overnight Supervision of Alzheimer’s Disease Patients in Nursing Homes - System Development and Field Trial.

In Proceedings of the International Conference on Information and Communication Technologies for Ageing Well and e-Health (ICT4AWE 2016), pages 15-25

ISBN: 978-989-758-180-9

prevent injuries and unattended exits (Occhiuzzi et al.,

2014; Mao et al., 2015; Vuong et al., 2013). Night-

time attendance of patients affected by dementia or

AD may be difﬁcult to manage also in nursing homes,

especially because the number of nurses available is

reduced, with respect to daily hours. As a conse-

quence, it is of interest to evaluate the applicability

of technology for night monitoring of AD patients in

nursing homes, in order to assess the impact of tech-

nology on nurses’ work ﬂows, and on the quality of

assistance provided to patients.

This paper describes an integrated system for the

monitoring of AD patients, realized by evolving and

updating an already existing product named UpTech

(Chiatti et al., 2013). The UpTech project focused

on AD patients and their family caregivers; it was

carried out as a multi-component randomized clini-

cal trial (RCT), integrating previous evidence on the

effectiveness of AD care strategies, in a comprehen-

sive design, to reduce the burden of family caregivers

of AD patients, and to maintain AD patients at home.

Indeed, often the relatives who take care of AD pa-

tients are subjected to high levels of stress, that could

also contribute to the onset of physical problems. The

positive outcomes of the UpTech experimental phase

(Pombo et al., 2015), providing the use of technolog-

ical devices as alternative or complementary form of

support, have suggested its application in a different

scenario, represented by the nursing homes. The aim

of the UpTech RSA project is to support and help as-

sistance of AD patients in nursing homes, during the

night hours, by means of a set of sensors located in

patient’s room, and suitable software applications to

detect dangerous events and raise alerts for the nurses.

When dealing with monitoring of people, this con-

dition is often seen as violating the privacy of the

user. Therefore, in order to satisfy the requirement

of providing an unobtrusive monitoring, only simple

environmental sensors have been employed in the Up-

Tech RSA solution, that are less intrusive and more

acceptable than other options, like wearable devices,

or video cameras. Wireless sensors have been chosen

and used: on one hand, this enables a simple installa-

tion, on the other hand, power consumption is a crit-

ical aspect, which has to be evaluated at the design

stage.

The paper is organized as follows: the context of

application of the proposed technology is discussed

in Section 2, whereas Section 3 is focused on design

and deployment issues. The ﬁeld trial implementa-

tion is presented in Section 4, and the results gath-

ered from the practical use of the technology in a

real nursing home are discussed in Section 5, show-

ing how the data collected from sensors may be trans-

lated into useful information for understanding the

patients’ needs and requirements. Finally, Section 6

concludes the paper and suggests possible future de-

velopments.

2 CONTEXT

Dementia is becoming increasingly prevalent world-

wide and is today considered as one of the

most burdensome disease for the western societies.

Alzheimer’s Disease is the most common form of de-

generative dementia. Generally, the onset of the ill-

ness occurs in the pre-senile age, however it could be

even earlier. A person with dementia can live 20 years

or more after diagnosis, during which he/she experi-

ence a gradual change of the functional and clinical

proﬁle. As consequence of the disease, a progres-

sive loss of cognitive capacity is occurring, eventually

leading to disability and to a severe deterioration of

quality of life. During the so-called “dementia jour-

ney”, the disease affects not only the patients but also

their informal (e.g. families) and formal (e.g. care

staff) caregivers, on whom the bulk of the care burden

falls (Chiatti et al., 2015).

Up-to-date, there is no cure for dementia thus

the attention to the symptomatic non-pharmacological

treatment for the patients and their caregivers has be-

come increasingly relevant, especially as the litera-

ture shows that these can be more effective that most

of available drugs (Spijker et al., 2008). Although

home remains the preferred place for care delivery,

a substantial number of patients need to access (per-

manently or temporarily) to residential care facilities,

when home care is no longer feasible. In the resi-

dential context, infrastructure and stafﬁng levels are

not always adequate to manage residents with demen-

tia. Residential care services are indeed labour inten-

sive and the quality of care here depends largely on

the stafﬁng level and characteristics (Kahanp

a et al.,

2016; Milte et al., 2016). As the ongoing ﬁnancial

crisis is reducing the budget available for residential

care services, a detrimental effect on personnel stan-

dards might occur. This concrete risk of staff short-

comings might, in turn, lead to a substantial propor-

tion of avoidable hospitalisations, use of emergency

departments, icreased carers’ burden and stress, and

inappropriate use of chemical and physical restraints

(e.g. antipsychotics).

The literature suggests that education, training and

support of available staff, supervision, improvement

of job satisfaction could be effective measures to in-

crease quality of care in this care setting (Institute of

Medicine, 1986). In addition, technologies and other

ICT4AWE 2016 - 2nd International Conference on Information and Communication Technologies for Ageing Well and e-Health

environmental factors have been identiﬁed as the most

promising measures to improve working conditions

in the residential care setting, to reduce the care bur-

den and to improve the overall quality of care (Freed-

man, 2005; Ancker et al., 2015). The potentials of

new technologies have been tested to reduce the need

for constant monitoring of dementia patients, increase

their safety and wellbeing within the residential set-

ting. So far, however, few solutions have manage to

survive to the prototyping phase and have been con-

cretely exploited in the market.

3 DESIGN AND DEPLOYMENT

The system described in this paper represents an evo-

lution of a project named UpTech, aimed at improv-

ing the quality of life of both AD patients living at

home and their family caregivers. This project in-

volved nurses and social workers, who periodically

went to the patients’ houses, and the installation of

technological kits. Each kit consisted of a network of

wireless sensors installed in the house, for the moni-

toring of the patient. Data were processed by a central

control unit and, in case of danger, a notiﬁcation was

sent to the caregiver. The new system, called UpTech

RSA, targets the nursing home environment and has

been devised primarily for the overnight monitoring

of patients, when there is a lack of personnel in the

building. Moreover, the main differences between the

two systems concern:

• number of users: in the nursing home, multiple

patients are monitored at the same time. Thus, the

central control unit is able to manage data coming

from more than one set of sensors;

• sensors: different types of sensors are employed,

due to the diversity of the physical environment;

• system architecture: the whole network can be

seen as a set of sub-networks, one for each room;

• alarm management: the monitoring system is an

aid for the nurses, the notiﬁcations are not sent to

the remote caregiver as in the previous system.

The project development stage conducted in the

Laboratory was aimed ﬁrst at the improvement of the

previous UpTech kit, secondly at the design and im-

plementation of the modules required for the new sys-

tem. In particular, the radio transceivers ﬁrmware was

re-designed, to implement an efﬁcient data acquisi-

tion and transmission procedure. At the same time,

particular attention was paid to the energy consump-

tion exhibited by the transmission nodes, by taking

into account the values of power absorption in the dif-

ferent operation phases, and implementing all the pos-

sible strategies for its reduction. As for the new com-

ponents, the following modules have been designed:

the structure of the database used to store the collected

information, and the applications necessary to imple-

ment the decision algorithms, in charge of making ac-

tions depending on particular values of the acquired

data.

The system requirements have been identiﬁed by

collecting nurses’ requests, thus the developed func-

tionalities are related to the usual daily care proce-

dures. Speciﬁcally, the set of sensors installed in each

room enables the following functionalities:

• door opening detection;

• window opening detection;

• “French-window” opening detection;

• presence in bed detection;

• presence in the bathroom detection.

The door opening detection is achieved using a mag-

netic sensor, wireless connected by Sub-GHz tech-

nology at a frequency of 868 MHz to a gateway, by

means of a properly designed electronic equipment.

Figure 1: Magnetic sensors for windows opening detection.

Figure 2: Self-calibrating mat sensor, for under-the-

mattress positioning.

Similarly, the detection of windows opening is

made through the same technology (see Figure 1).

The user’s presence in the bathroom is detected by a

self-powered Passive Infrared Sensor (PIR), which is

connected to the radio transmitter module. For ease

of installation, and to avoid damage to the ﬁxtures

of the building, these sensors have been placed on

top of the entry doors of the bathrooms. A mat sen-

sor has been adopted to detect the user’s presence in

Overnight Supervision of Alzheimer’s Disease Patients in Nursing Homes - System Development and Field Trial

bed; it is available in two versions, with and without

self-calibration. The sensor without self-calibration is

placed over the mattress, under the sheets, while the

other one is placed under the mattress (Figure 2), and

therefore it appears more comfortable for the patients

and for the daily operations of bed maintenance. The

gateway represents a central node that forwards data

to a PC located at the nurses’ station. Then, the ap-

plication running on the PC ﬁlters the incoming in-

formation. Data related to events are saved in a local

database (DB), while those referred to the operating

status of the sensors are veriﬁed in order to monitor

the correct operation of the technology kits.

The electronic boards transmit an event to the cen-

tral server every time there is a status change, that is,

for example, activation/deactivation of the PIR sensor,

or opening/closing the door. Accordingly, the data

stored in the database contain the sensor information

(id, gateway address, name and type), the date and

time when the notiﬁed event occurred, and the status

of the sensor represented in binary format as follows:

• activation: state = 1;

• deactivation: state = 0.

In addition, the server assigns a unique id to each

DB row in order to implement a robust mechanism

for transmitting information to the mobile interface.

This allows the mobile device to identify one or more

missing events, and to request them back from the

server. In fact, a mobile Android application has been

developed, running on a tablet or smartphone, and so

easily portable. This allows the nursing staff to re-

ceive event notiﬁcations even when they are outside

the nurses’ station and cannot access the ﬁxed desk-

top interface. Events data, properly processed, are

displayed through not only mobile, but also desktop

interfaces (Figures 3 and 4). In the ﬁrst case, the

user can see a scrollable list of events identiﬁed by

the name of the sensor that generated it and the room

Figure 3: Mobile interface running on a smartphone.

Figure 4: Screenshot of the Desktop Interface, two sections

version.

name, as shown in Figure 3. Each event is tagged with

a colored circle: depending on the associated level of

alert, the circle may be green, yellow or red. In the

latter case, two versions are available:

• a two sections version: the interface is divided

into two parts. On the right there is a scrollable

list of the events acquired by the sensors, while

on the left the status of the sensors in each room

is shown. There is a top bar which becomes

coloured and ﬂashing when an event occurs;

• a multi-user version: the main screen shows all

the rooms monitored. When an event occurs in

one room, the corresponding frame becomes col-

ored. By clicking on the box, it is possible to see

the details of sensors state.

Given the wireless transmission mode of the sen-

sor nodes and their battery supply, the monitoring

of the sensors state itself becomes very important.

Therefore, a procedure for the periodic sending of

alive messages has been implemented in the sensors.

They are constantly monitored by the central process-

ing system, that generates alarm messages in the case

of failure. Despite its importance, this procedure is

extremely critical, because sending alive messages

too frequently causes an increase in the batteries con-

sumption. Otherwise, the transmission of the alive

message at a lower frequency can give rise to long

time intervals in which the sensor is not active, but

the system is not informed about the failure. When

an alive message does not reach the local server at the

expected time, the latter notiﬁes a malfunction of the

sensor node to the nurse, who can promptly ﬁnd out

the problem and act accordingly.

4 FIELD TRIAL

4.1 Experimental Set-up

The system described in Section 3 is already avail-

ICT4AWE 2016 - 2nd International Conference on Information and Communication Technologies for Ageing Well and e-Health

able as a prototype. Following the initial develop-

ment phase in the Laboratory, aimed to better adapt

the technology to the emerged operational require-

ments, the prototype has been installed in the nursing

home “Villa Cozza” in Macerata (Italy). In this phase,

the supervision of two rooms (tagged as room 2 and

room 3) has been implemented, while the ﬁnal ver-

sion of the system will be able to dynamically accept

a plurality of rooms, depending on the operating re-

quirements. Each room is equipped with a sensors kit

consisting of three magnetic sensors (one applied onto

the window, one onto the French window, and one

onto the room front door), a PIR sensor in the bath-

room, and a force sensor placed in the bed, as shown

in Figure 5. A single gateway device has been used to

manage wireless communications with the sensors in

the two rooms.

Figure 5: Floor plan of the two rooms equipped with the

UpTech RSA sensors in the nursing home “Villa Cozza”,

Macerata (Italy).

In room 2 two female patients are housed, only

one suffering from Alzheimer’s disease. Her bed has

been equipped with a force sensor. The other one is

not autonomous and can move only by a wheelchair;

consequently, the events generated by the different

sensors can be originated only by the movement of the

ﬁrst patient. In room 3, instead, a single female pa-

tient is housed, also suffering from Alzheimer’s dis-

ease, but in this case she can not move autonomously.

As the system represents a support tool for improv-

ing the safety of patients, it can be well-compared to

an alarm system. Moreover, the type of sensors em-

ployed do not collect personal data of the two patients

involved. According to the national laws, in this case

the ethical approval is not required.

A critical issue encountered during the installa-

tion phase has been to enable the communication be-

tween the gateway, positioned in the corridor in front

of the two rooms, and the central server, located in

the nurses’ station on the upper ﬂoor. Such a problem

arises because the building where the nursing home

is located is not equipped with a communication in-

frastructure (e.g. a Local Area Network): there are

no network cables, or WiFi coverage. Moreover, the

nursing home is hosted in a historic building and, as

often happens in such cases, the walls are thick and

made of concrete, thus making wireless communica-

tions very difﬁcult. Both a Power Line Communica-

tion (PLC) and a mixed wireless infrastructure (WiFi

and Hiperlan) have been experimented, ﬁnally select-

ing the wireless solution as the supporting commu-

nication architecture. In order to overcome obsta-

cles like metal doors and thick walls, that limit signal

propagation, multiple Access Points (APs) and links

have been setup.

4.2 Evaluation Survey

Some weeks after the installation a survey for the

evaluation of the system has been conducted over 18

nurses. Although some of them are not very familiar

with the technology, the results are highly positive. In

Table 1, some of the most signiﬁcant questions and

results are listed. The 100% of respondents believes

the kit is easy to use and recommends it for the mon-

itoring of AD patients in nursing homes during the

night. All the nurses state the system has not been

a source of stress for them. In fact, its introduction

does not generate further work for the staff. They just

carried on the usual activities, but with an additional

monitoring tool. Only the 6% of nurses believes that

it was stressful for patients. Indeed, operators have re-

ceived some sporadic grumbles due to the discomfort

produced by one of the bed sensors. As mentioned

in Section 2, the bed force sensor without calibration

must be placed between mattress and sheet: this may

annoy the patient during sleep time, due to a differ-

ence in thickness. This leads us to conclude that the

sensor with calibration is preferable, as it has to be

placed under the mattress, and will be consequently

used in the subsequent installations. Apart from that,

patients have noticed any change.

Moreover, the nurses stated that, during the trial

period, there have been some dangerous episodes de-

tected by the kit, such as the opening of a window

during the night, and a patient’s fall out of the room.

In both cases the system detected the alarming situa-

tion and the staff was been able to intervene promptly.

Despite the positive opinions, some problems were

found, in particular due to the occurrence of false

alarms. They were caused primarily by failures in the

communication link, resulting in multiple sending of

alarm events.

Still considering nurses’ opinions, some ideas for

improving the system were identiﬁed. First, false

alarms must be avoided, as they can generate a feel-

Overnight Supervision of Alzheimer’s Disease Patients in Nursing Homes - System Development and Field Trial

Table 1: The opinion of the nurses about the experimental deployment of the UpTech RSA technology at the nursing home

“Villa Cozza”.

Question Yes No

Is the kit easy to use? 100% 0%

Do you think that the patients monitored have suffered a

stress?

6% 94%

Do you think that the kit has been a source of stress for

nurses?

0% 100%

Would you recommend the use of this kit in nursing

homes?

100% 0%

Question Positive Medium Negative

Overall opinion on the technological kit 89% 11% 0%

Question Yes Quite a lot No

Do you think that the kit can improve the assistance pro-

vided in nursing homes?

61% 39% 0%

ing of distrust by operators against the entire system.

Secondly, customizing different alarms for each user

would be preferable, since each patient has different

behavioural and health conditions. Finally, imple-

menting an even more friendly user interface would

encourage the adoption of the system by nurses unfa-

miliar with technology.

5 DATA ANALYSIS

5.1 Context Characterization

In addition to the real-time monitoring of patients,

it is possible to perform several types of analysis

on the data collected by UpTech RSA sensors over

time, such as obtaining information on the patient’s

habits and, as a consequence, detecting any changes

or unusual behaviours. In the following, some sample

graphs are shown, representing selected daily activi-

ties of the monitored patients, obtained thanks to the

events detected by the sensors. The analysis refers to

data collected from May to June, 2015, by the sensors

located in both the monitored rooms.

First of all, in order to give signiﬁcance to the

analysed data, some information about the patients

and the daily activities conducted in the Alzheimer’s

ward are necessary. Table 2 represents a sort of daily

diary. Patients remain within the ward during the day:

they can stay together in the common areas, where

they also have lunch and dinner, and can go in/out of

the rooms whenever they want. The entry doors of

the rooms are generally closed during the night. They

are opened by the shift nurse who performs two in-

spection rounds per night, in order to verify that the

Table 2: Diary of daily activities.

Time Activity

7:30 Rooms cleaning

7:00 - 10:00 Patients get out of beds

Morning Patients stay in the common ar-

eas, can go in/out of the rooms

11:30-12:30 Lunch in the dining room

Afternoon Patients stay in the common ar-

eas, some of them have a rest

17:30-18:30 Dinner in the dining room

19:00-21:00 Patients go to bed

22:00 First nurses’ check round

3:00 Second nurses’ check round

patients are sleeping and do not need assistance.

In room 2 there are two patients: only one is mon-

itored through a bed sensor, because she suffers from

AD and often wakes up in the night and goes out of

the room. The other patient moves by wheelchair and

is not able to get off the bed on her own. The AD pa-

tient in room 3 has bed rails, so she can not get out of

the bed autonomously during the night.

Although the system is able to monitor the pa-

tients throughout the entire day, the interesting events

are those occurring during the night. In that pe-

riod, in fact, the user is left alone for most of the

time and thus the data acquired are more signiﬁcant.

The graphical visualization of the analysis output pro-

vided in the following sub-section has the ability to

help the reader in recognizing and understanding a

large amount of data, and in easily identifying anoma-

lies and behavioural patterns that would not be obvi-

ous otherwise.

ICT4AWE 2016 - 2nd International Conference on Information and Communication Technologies for Ageing Well and e-Health

5.2 Data Representation and Analysis

The raw data collected by the sensors installed in the

rooms are often difﬁcult to interpret. Therefore, in or-

der to carry out the data analysis, ﬁrst of all it is nec-

essary to ﬁnd a representation allowing to understand

them immediately. Lotﬁ et al. (Lotﬁ et al., 2012) af-

ﬁrm that, among the various representation methods

presented in the literature, the start-time/duration is

the most effective one for large data sets. The data

acquired from each sensor can be seen as a binary

signal, in which the value “1” is the activation and

the value “0” is the deactivation. Representing in-

formation according to a start-time/duration method

means converting the binary signal into two separate

sequences of real numbers corresponding to the start-

time and duration of each activity, respectively. Fig-

ure 6 shows the start-time/duration graphs of the ac-

tivity detected by the bed sensor, i.e. presumably

sleeping, for each room. Each point on the graph indi-

cates a “sleep” and is characterized by a start-time (on

the abscissa) and a duration (on the ordinate). All ac-

tivities lasting less than 10 minutes have been ignored

because they could indicate sensor activations and de-

activations due to involuntary movements of the sub-

ject while asleep.

Looking at the charts is easy to notice the triangu-

lar shape assumed by the set of points. This result was

expected because life in the nursing home is sched-

uled by the daily diary and, thus, the sleeping activ-

ities are bounded by speciﬁc and almost ﬁxed time

constraints. Therefore, it seems plausible that patients

never go to bed before 6:30 PM, and the sleep dura-

tion is inversely proportional to the start-time. The

sparse distribution of points in the triangular-shape di-

agrams indicates that the monitored subject wakes up

several times during the night. In Figure 6 (b) a group

of points is located between 12:30 PM and 13:30 PM:

this suggests that sometimes the patient has a rest after

lunch. On the other hand, looking at Figure 6 (a), the

presence of two outliers (highlighted by red circles)

becomes immediately evident.

Analytically, a ﬁrst detection of outliers is per-

formed using clustering techniques. In the present

case the K-means algorithm is applied (Nazerfard

et al., 2010), which allows condensing the data. Dif-

ferent techniques can be used to separate normal data

and outliers (Chandola et al., 2009). In this case, a

variation of the threshold ﬁltering method have been

chosen: it consists in both comparing a speciﬁc fea-

ture of the points with a threshold and excluding the

outliers. Speciﬁcally, for each cluster identiﬁed, and

for each point in the cluster, the considered feature is

the euclidean distance between one point and the oth-

ers belonging to the same cluster. Such distances are

then compared against a threshold empirically cho-

sen: all points whose distance exceeds the threshold

are considered outliers. Moreover, to improve the

clustering effect, another iteration of the algorithm

is performed, by excluding the abnormalities found

from the dataset. Clustering is employed as a pre-

processing method, and it can be considered as the

basic level of data analysis. It does not provide a

deﬁnitive result, in fact its application to the dataset

has the only aim to help understanding data by means

of a graphical representation.

Another information that can be extrapolated by

combining the data obtained from the bed sensor with

those detected by other sensors, is the identiﬁcation

of the action carried out after the user came out of

bed. This will enable the possibility to calculate the

occurrences of predeﬁned patterns of activities, in-

stead of single ones. Such an analysis allows to iden-

tify potentially dangerous situations with respect to

behaviours commonly exhibited by the subject, and

not considered as alarms. Each point on the graphs

in Figure 7 indicates a “sleep” and is characterized

by a start-time (on the abscissa) and an end-time (on

the ordinate). As for the start-time/duration, the start-

time/end-time representation requires the conversion

of the binary signal in two separate sequences of real

numbers which in this case correspond to the start-

time and end-time of the activity. The type of activity

shown is still the sleeping, but, according to the ac-

tion carried out subsequently, the shape and colour of

the marker changes. In fact, the graphs show, for each

room, the actions executed within 4 minutes after the

patient got out of bed (end-time), i.e.:

• door opening (marked as a green circle);

• window opening (marked as a black square);

• presence in the bathroom (marked as a blue cross);

• no other activity (marked as a red cross).

This kind of representation has been chosen to em-

phasize, especially in Fig. 7 (b), that some of the

actions are performed only when the patient gets up

at certain times. For example, the patient in room 3

enters the bathroom within 4 minutes after waking up

only in the morning, i.e. only when nurses remove the

rail from the bed. The other activations and deactiva-

tions occurring during the night could indicate that the

subject has moved or was seated up on the bed, while

the openings of the door or window are probably due

to the presence of the medical staff.

Conversely, looking at Figure 7 (a), the observer

can notice the patient very often goes to the bathroom

or opens the door immediately after getting up. This

agrees with the reports of the nurses concerning the

Overnight Supervision of Alzheimer’s Disease Patients in Nursing Homes - System Development and Field Trial

(a) Room 2

(b) Room 3

Figure 6: Start-time/duration graphs of the “sleeping” activity detected from May to June 2015, respectively in (a) room 2,

and (b) room 3.

ICT4AWE 2016 - 2nd International Conference on Information and Communication Technologies for Ageing Well and e-Health

(a) Room 2

(b) Room 3

Figure 7: Start-time/end-time graphs representing the activities performed after waking up by the patients housed respectively

in (a) room 2, and (b) room 3, from May to June 2015.

fact that the monitored elder is very lively, and often

gets up during the night.

In Table 3, the percent occurrence rates of each

activity described above are given, limited to the night

hours.

The analysis described so far is just the very ﬁrst

step to identify the user’s behavioural patterns and ab-

normal situations. Until now, we focused on the rep-

Overnight Supervision of Alzheimer’s Disease Patients in Nursing Homes - System Development and Field Trial

Table 3: Hit rate of the getting up action followed re-

spectively by the action of entering the bathroom, open-

ing/closing the door, or opening/closing the window, in the

time slot between 09:00 PM and 06:00 AM.

Event detected after

awakening

Room2 Room3

Presence in the bathroom 76% 15%

Door opening 14% 2%

Window opening 0% 0.2%

None 10% 82.8%

resentation and visualization of data, extracting some

preliminary information on the habits of two mon-

itored patients. Nevertheless, there is still a long

way to go. Although the detection of outliers can

be very useful in this context, however, it is neces-

sary to set up a predictive system able to identify in

advance any anomalous situation to help the nursing

home staff making the necessary arrangements. As

already hinted, one of the aspects emerged during dis-

cussion with nurses is the need of alarm personaliza-

tion. In fact, a situation may be potentially dangerous

for a user, while it may be harmless for another one.

This strongly depends on motor and cognitive skills of

each patient. Although this can be done manually by

nurses via graphical user interfaces, a signiﬁcant con-

tribution comes from the analysis of patients’ habits.

One of the future developments is to extend the be-

havioural analysis in the long term, aimed at recog-

nizing unusual, and, therefore, potentially dangerous

situations and notifying them to the staff, in a com-

pletely automatic way.

6 CONCLUSION AND FUTURE

WORK

The issue addressed in this paper is the need to of-

fer assistance to a growing number of AD patients, by

providing solutions that can be applied both at their

homes, and in nursing homes. Different types of pa-

tients have their own speciﬁc requirements: this case

deals with AD, but it could be possible to adapt the

current solution to patients with different pathologies,

by changing the sensors selected and offering proper

services. The results obtained from the ﬁrst exper-

imental installation of the monitoring system high-

lighted the effectiveness of the proposed solution to

support the nurses during the night supervision of pa-

tients. The effectiveness could grow even more by

extending the pool of monitored patients. Monitoring

all the patients leads to increase the convenience for

the nurses and, above all, the degree of safety of the

patients. For example, by monitoring the presence to

bed at night time for all the patients, it could be pos-

sible to detect any accidental fall or necessity to help.

Other advantages brought by the UpTech RSA pro-

posal are the introduction of innovative technologies

in the nursing home facility, the efﬁcient use of hu-

man and technical resources, and the quality of care

improvement.

ACKNOWLEDGEMENTS

This work was partially funded through an Innova-

tion Voucher issued in the framework of the WIDER

project (Green Growing of SMEs: Innovation and De-

velopment in the Energy Sector in the Med Area), and

partially by the AAL 20141041 project “Home4Dem”

(HOMEbased ICT solutions FOR the independent liv-

ing of people with DEMentia and their caregivers),

co-founded by the Active and Assisted Living Joint

Platform.

REFERENCES

Ancker, J. S., Witteman, H. O., Hafeez, B., Provencher,

T., de Graaf, M. V., and Wei, E. (2015). The invisi-

ble work of personal health information management

among people with multiple chronic conditions: Qual-

itative interview study among patients and providers.

J Med Internet Res, 17(6):e137.

Bowens, F. M., Frye, P. A., and Jones, W. A. (2010).

Health information technology: integration of clini-

cal workﬂow into meaningful use of electronic health

records. Perspectives in health information manage-

ment / AHIMA, American Health Information Man-

agement Association, 7:1d.

Chandola, V., Banerjee, A., and Kumar, V. (2009).

Anomaly detection: A survey. ACM Comput. Surv.,

41(3):15:1–15:58.

Chiatti, C., Masera, F., Rimland, J., Cherubini, A.,

Scarpino, O., Spazzafumo, L., and Lattanzio, F.

(2013). The up-tech project, an intervention to sup-

port caregivers of alzheimer’s disease patients in italy:

study protocol for a randomized controlled trial. Tri-

als, 14(1).

Chiatti, C., Rimland, J. M., Bonfranceschi, F., Masera, F.,

Bustacchini, S., and Cassetta, L. (2015). The up-

tech project, an intervention to support caregivers of

alzheimer’s disease patients in italy: preliminary ﬁnd-

ings on recruitment and caregiving burden in the base-

line population. Aging & mental health, 19(6):517–

525.

Freedman, V. A. (2005). Barriers to implementing tech-

nology in residential long-term care settings. Polisher

Research Institute.

Huston, C. (2013). The impact of emerging technology on

ICT4AWE 2016 - 2nd International Conference on Information and Communication Technologies for Ageing Well and e-Health

nursing care: Warp speed ahead. OJIN: The Online

Journal of Issues in Nursing, 18(2).

Institute of Medicine (1986). Improving the Quality of Care

in Nursing Homes. The National Academies Press,

Washington, DC.

Kahanp

a, A., Noro, A., Finne-Soveri, H., Lehto, J., and

Per

a, M.-L. (2016). Perceived and observed quality

of long-term care for residents - does functional abil-

ity account? International Journal of Older People

Nursing.

Kim, S.-S., Oh, K. M., and Richards, K. (2014). Sleep dis-

turbance, nocturnal agitation behaviors, and medical

comorbidity in older adults with dementia: Relation-

ship to reported caregiver burden. Research in Geron-

tological Nursing, 7(5):206–214.

Lee, D., Heo, S. H., Yoon, S.-S., Chang, D.-I., Lee, S.,

Rhee, H.-Y., Ku, B. D., and Park, K.-C. (2014). Sleep

disturbances and predictive factors in caregivers of pa-

tients with mild cognitive impairment and dementia. J

Clin Neurol, 10(4):304.

Lotﬁ, A., Langensiepen, C., Mahmoud, S., and Akhlagh-

inia, M. (2012). Smart homes for the elderly demen-

tia sufferers: identiﬁcation and prediction of abnormal

behaviour. Journal of Ambient Intelligence and Hu-

manized Computing, 3(3):205–218.

Lowry, S. Z., Ramaiah, M., Patterson, E. S., Brick, D., Gib-

bons, M. C., and Paul, L. A. (2015). Integrating elec-

tronic health records into clinical workﬂow: An ap-

plication of human factors modeling methods to spe-

cialty care in ‘obstetrics and gynecology’ and ‘oph-

thalmology’. Technical report, National Institute of

Standards and Technology.

Mao, H.-F., Chang, L.-H., Yao, G., Chen, W.-Y., and

Huang, W.-N. W. (2015). Indicators of perceived use-

ful dementia care assistive technology: Caregivers’

perspectives. Geriatrics & Gerontology International,

15(8):1049–1057.

Milte, R., Shulver, W., Killington, M., Bradley, C., Rat-

cliffe, J., and Crotty, M. (2016). Quality in residen-

tial care from the perspective of people living with de-

mentia: The importance of personhood. Archives of

Gerontology and Geriatrics, 63:9–17.

Nazerfard, E., Rashidi, P., and Cook, D. (2010). Discover-

ing temporal features and relations of activity patterns.

In Data Mining Workshops (ICDMW), 2010 IEEE In-

ternational Conference on, pages 1069–1075.

Occhiuzzi, C., Vallese, C., Amendola, S., Manzari, S.,

and Marrocco, G. (2014). Night-care: A passive

{RFID} system for remote monitoring and control

of overnight living environment. Procedia Computer

Science, 32:190 – 197. The 5th International Confer-

ence on Ambient Systems, Networks and Technolo-

gies (ANT-2014), the 4th International Conference on

Sustainable Energy Information Technology (SEIT-

2014).

Organization for Economic Co-operation and Development

(2015). Focus on health spending: Oecd health statis-

tics 2015.

Pombo, N., Spinsante, S., Chiatti, C., Olivetti, P., Gambi,

E., and Garcia, N. (2015). Assistive technologies for

homecare: Outcomes from trial experiences. In ICT

Innovations 2015 Proceedings, Workshop ELEMENT

2015.

Spijker, A., Vernooij-Dassen, M., Vasse, E., Adang, E.,

Wollersheim, H., Grol, R., and Verhey, F. (2008).

Effectiveness of nonpharmacological interventions in

delaying the institutionalization of patients with de-

mentia: A meta-analysis. Journal of the American

Geriatrics Society, 56(6):1116–1128.

Vuong, N. K., Goh, S. G. A., Chan, S., and Lau, C. T.

(2013). A mobile-health application to detect wander-

ing patterns of elderly people in home environment.

In 2013 35th Annual International Conference of the

IEEE Engineering in Medicine and Biology Society

(EMBC), pages 6748–6751. Institute of Electrical &

Electronics Engineers (IEEE).

Overnight Supervision of Alzheimer’s Disease Patients in Nursing Homes - System Development and Field Trial