OPTIMAL AUTO-REMINDER-CALLING ALGORITHM FOR

SELF-REPORTING TYPE SAFETY MONITORING SYSTEM

BY USING TELEPHONE FOR ELDERLY PEOPLE

Jun Sasaki

, Takuhide Kikuchi

, Masanori Takagi

, Keizo Yamada

Michiru Tanaka

and Akiko Ogawa

Faculty of Software and Information Science, Iwate Prefectural University, Takizawa, Iwate, Japan 020-0193

Iwateshiga Inc., 31-81-D, Hannokizawayama, Takizawa, Iwate, Japan 020-0173

Faculty of Social Welfare, Iwate Prefectural University, Takizawa, Iwate, Japan 020-0193

Keywords: Information system, Elderly people, Safety monitoring, Auto-calling.

Abstract: An increasingly aging society is a significant problem in advanced countries. Safety monitoring is required

for elderly people especially in a rural area. This paper describes a self-reporting type safety monitoring

system by using telephone and the field experimental results in Iwate prefecture and Aomori Prefecture in

Tohoku area of Japan. In this system, a user (an elderly person) makes a daily wellness call to the Iwate

Prefecture Council of Social Welfare (IPCSW). The IPCSW staff calls the user to check his/her wellness if

no wellness-report is received from the user. However, the system introduced a new work load to the

IPCSW staff when many users forgot to make the call. Our project is to reduce the staff’s phone work load

by an automatic wellness-report reminder call. In this paper, the daily reporting time of long-term using

users is analyzed and a new algorithm to determine an optimal auto-calling scheduler for each user based on

the analyzed results is proposed.

1 INTRODUCTION

Social isolation of elderly people has become a

significant problem in Japan as well as in other

developed countries (Japanese Cabinet Office,

2011). In addition, the “solitary death” rate among

senior citizens is increasing year by year. In

December 2009, Iwate Prefectural University and

the Iwate Prefectural Council of Social Welfare

(IPCSW) developed and introduced a self-reporting

type safety monitoring system by using telephone

for elderly living alone to minimize the occurrence

of “solitary death”. A problem exists in that, in an

automated scheduling system, elderly people

sometimes forget to report their status to the system.

Therefore, an IPCSW staff has to establish over

telephone the status of those forgetful users.

Unfortunately, confirmation itself becomes a burden

in the case of many users. By contrast, this system

has a preset auto-calling function for these daily

confirmations. However, if confirmation is

performed via this auto-calling function, the self-

sending motivation for a user decreases. This

research proposes an algorithm for optimally-

scheduling automated calls in a self-reporting type

safety monitoring system for elderly people.

2 SELF-REPORTING TYPE

SAFETY MONITORING

SYSTEM

Figure 1 shows the self-reporting type safety

monitoring system that we developed and introduced

to Iwate prefecture and Aomori prefecture in

Tohoku area of Japan. In the system, elderly people

(users, clients) can send a status report by pushing a

selected phone key, such as “1= Fine”, “2=Not so

fine”, “3=Bad” and “4= I want to talk”, in response

to a voice command. Subsequently, an IPCSW staff

can confirm the user’s status on a Web page. The

Web server can send the information by e-mail to

pre-registered neighbors and family members living

away from the users. Nearby support members

(supporters) can also send status information of an

362

Sasaki J., Kikuchi T., Takagi M., Yamada K., Tanaka M. and Ogawa A..

OPTIMAL AUTO-REMINDER-CALLING ALGORITHM FOR SELF-REPORTING TYPE SAFETY MONITORING SYSTEM BY USING TELEPHONE

FOR ELDERLY PEOPLE.

DOI: 10.5220/0003762203620365

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2012), pages 362-365

ISBN: 978-989-8425-88-1

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

elderly user by cell-phone, e-mail or a website

information board. The IPCSW staff can also send

daily messages for the entertainment of elderly users.

In essence, the main feature of the system is that

elderly users can send status information themselves,

and various people such as an IPCSW staff,

neighbors, nearby supporters, and families can

monitor and share that status information. Crucial to

the implementation of this system is using the

mobility and capability of the users to report health

and status information. This particular information

cannot be obtained using other systems such as in

sensor-type monitoring and emergency calling.

These systems have problems of lack of privacy and

frequent false alarms. Furthermore, in our system,

since users only require a telephone connection, the

initial set-up cost is not an issue and it is very easy

to operate (

Report of Iwate Prefecture Council of Social

Welfare, 2009).

Figure 1: Self-reporting type safety monitoring system.

The number of users has been increasing and it

becomes about 400, recently. In June 2011, the

number of users, supporters and managers is over

700, 500 and 150, respectively and the total number

of system users is over 1,300.

Some elderly users occasionally forget to send

status information for various reasons such as being

very busy or through some medical condition such

as progressive dementia. In this case, an IPCSW

manager can check the user’s status by phoning such

users at a specified time. However, if there is large

number of users, the task involved becomes

burdensome. According to our experience in field

tests on the system in the Kawai area of Miyako city

in the Iwate Prefecture of Japan, the average no-call

rate for 30 to 40 users was 12.3 %. This result means

that 3 or 4 users on average per day would not report

status information, and these users needed to have

their status established by direct calling. If the

number of system users was to increase to 1,000,

which would be a typical expected number in the

future, an IPCSW officer would have to make over

100 telephone calls per day, placing a heavy

scheduling load on them.

The system we have developed has an automatic

calling function that schedules daily specified times

for users. The method for determining the specified

time was not studied and this function is not used

now. If the specified time was too early, users’

motivation in sending status reports would decrease,

and if too late, the delay would be of concern to

relatives. In this paper, we have analyzed the daily

scheduling data for each user, and studied an

algorithm that optimizes auto-calling times in these

user-interface monitoring systems.

Conventional sensor-type monitoring systems are

used to notify any abnormal activity in an elderly

user by detecting the difference from normal

behavior patterns (Shigeki Aoki, 2002, Yoshimitu

Sinagawa, 2005, 2006). Those studies have yielded

algorithms for determining normal/abnormal

patterns using daily-acquired sensor data. Our

algorithm decides the optimal auto-scheduling time

for each user and is basically different from those

studies for sensor-type systems.

3 ANALYSYS OF STORED DATA

3.1 Variability of Daily Reporting Time

Our algorithm objective aims to automatically

determine a scheduling times for calls from the

system to each elderly user so as not to decrease the

motivation of a user from reporting in but also not to

be too late in the day. To develop the algorithm, we

analyzed actual reporting time data obtained from

our experimental system.

Carrying out the experiments, we found that

there were various user types on the daily reporting

time. Figure 2 shows a distribution graph on the

daily reporting time of a typical-user type. The

average reporting time of this user type is 7:50 am.

We can see that the user type usually sends the

safety information between 7:30 am and 8:00 am.

On the other hand, there is random-user type,

which is difficult to know a usual reporting time

because the time distribution area is very wide and

there seems to be no regulation. We think the system

should wait for a long time to make an automatic

reminder call for the random-user type.

In contrast, there is an accurate-user type, which

is reporting the wellness report everyday at almost

OPTIMAL AUTO-REMINDER-CALLING ALGORITHM FOR SELF-REPORTING TYPE SAFETY MONITORING

SYSTEM BY USING TELEPHONE FOR ELDERLY PEOPLE

363

the same time for example just 7:50 am. If the

accurate-user type would send the information at the

different time from usual time, the possibility to be

abnormal condition seems to be high. The research

goal is to find the abnormal condition and make a

reminder call to the user as sooner than other type

users from system. So, we think there is an optimal

calling time depending on various type users.

Figure 2: Distribution of self-reporting time of a typical-

user type.

We believe that the data should be obtained from

long term users (real system users) and therefore we

selected data for 67 elderly people from the current

440 users, as these selected users habitually reported

in over more than a 360 day period.

Table 1 shows the degree of variability which

indicates the difference in daily report times for each

user in our study group. In Table 1, IQR stands for

the “Interquartile Range”, which is defined as

IQR= Q

(1)

where Q

is the 25 percentile and Q

is the 75

percentile of daily reporting times for a user.

Table 1: Variability of daily reporting time.

IQR Number of Users

1 hour ~ 22

45 minutes ~ 1 hour 5

30 minutes ~ 45 minutes 8

15 minutes ~ 30 minutes 10

7 minutes ~ 15 minutes 10

0 minutes ~ 7 minutes 12

With regard to the IQR, 22 out of 67 users had a

variability of more than 1 hour, while 45 (= 67 - 22)

users were within 1 hour. We found that about half

the number (n=32) of the long-term users (n=67)

tend to send their daily status at similar times within

30 minutes. If a user in this group sends the status

information more than 30 minutes from the average

reporting time, he/she might forget to report. If a

user in the “1 hour-IQR” group (n=22) did not report

status information, then it is difficult to identify

whether they might have forgotten to report because

the possibility of reporting after 1 hour for such

users is comparatively high.

3.2 Trend in Reporting for Users

Next, we analyzed the trend in reporting times for

the long-term active users (n=67). Figure 3 shows

the trend in the IRQ difference between two

consecutive days for all users in all systems over a

period of 360 days. As a result, we found that the

difference in IQR has large variability over an initial

period of 36 days, after which the variability

becomes small. After 36 days, the variability for

95 % of the users becomes less than ±30 minutes.

Figure 3: IQR difference between consecutive two days.

3.3 The Proposed Algorithm

Based on the above analysis, we propose the

following algorithm to determine the optimal calling

time:

1. Record the report times for daily calls for each

elderly user.

2. If the number of days, D

, that the system has

been using is over a certain threshold (for example,

36 days within 5%, as determined from Figure 3),

calculate the average report time, T

and IQR (= Q

3. Determine a certain threshold time, T

, for a

report time. The threshold time (T

) is calculated as

follows,

= Q

– k*IQR, (2)

HEALTHINF 2012 - International Conference on Health Informatics

364

where, “k” is a coefficient to determine fitting value

for the actual field work. We suppose “k” will be

between 0 and 1. In actual field work, there is a

limit time for example 5:00 pm to make a reminder

call. So, we set the limit time T

, and the threshold

time (T

) must not be exceed the limit time (T

Namely,

< T

(3)

4. If an elderly user does not call after a threshold

time, T

, the system make a reminder call to the

user automatically.

5. If the threshold time, T

th ,

in the actual field work

is too large or too small, the coefficient, k (usually, 0

< k < 1), can be changed to an appropriate value.

6. Repeat steps 15 until the system is not required.

With this algorithm, the system can provide auto-

calling for non-reporting users with an individual

auto-calling time for variable-reporting users. For

example, the system would provide auto-calling for

a user with small variability range after a relatively

short threshold time but auto-calling for a user with

large variability range after a relatively long

threshold time.

According to the proposed algorithm, we expect

to realize effective confirmation without

compromising user motivation to report and to

reduce the daily work load of IPCSW staff in

establishing the information by telephoning.

4 SUMMARY

In this paper, we analyzed the accumulated

experimental data of reporting times for elderly

people in the current self-reporting type monitoring

system by using telephone, and we proposed a new

algorithm that determines the auto-calling time in

the system. Before we apply this algorithm to the

current serf-reporting type safety monitoring system,

we will validate the algorithm by simulation using

actual data.

We plan also to evaluate the effectiveness of our

proposal by interviewing elderly users and IPCSW

staff.

ACKNOWLEDGEMENTS

This research has been achieved with many

contributors. We would like to say thank to Japan

Science and Technology (JST) Agency for

subsidizing our project. And we also appreciate to

staff of Iwate and Aomori Prefectural Social Welfare

Centres and relative Social Welfare Centres and

Iwate Prefectural University for developing the

prototype systems and cooperation on our

experiment.

REFERENCES

Japanese Cabinet Office, “FY 2011 White Paper on

Ageing Society Edition”, 2011.

Report of Iwate Prefecture Council of Social Welfare,

“Survey Report on monitoring the elderly”, 2009.

Shigeki Aoki, Masaki Onishi, Atushi Kojima, Yasuhiro

Sugawara, “Recognition of a Solitude Senior’s

Behavioral Pattern Using Infrared Detector”, IEICE

technical report. WIT, 101 Welfare Information

Technology (703), 43-48, 2002-03-11.

Yoshimitu Sinagawa, Toshio Kishimoto, Shigeru Ohta,

“Detection of an Unusual Day for the Elderly Living

Alone Using the Classification of Movement Pattern”,

Kawasaki Medical Welfare Journal 15 (1), 175-181,

2005.

Yoshimitu Sinagawa, Toshio Kishimoto, Shigeru Ohta,

“Development of an Algorithm for Automatic

Emergency Calls Using Non-Response Intervals of

Infrared Sensors”, Kawasaki Medical Welfare Journal

15 (2), 553-563, 2006.

OPTIMAL AUTO-REMINDER-CALLING ALGORITHM FOR SELF-REPORTING TYPE SAFETY MONITORING

SYSTEM BY USING TELEPHONE FOR ELDERLY PEOPLE

365