A Statistical Analysis for the Evaluation of the Use of Wearable and

Wireless Sensors for Fall Risk Reduction

Giovanna Sannino, Ivanoe De Falco and Giuseppe De Pietro

Institute on High-Performance Computing and Networking (ICAR), National Research Council of Italy (CNR), Naples, Italy

Keywords:

Falling Risk, Physical Activity, Body Mass Index, Statistical Analysis, Correlation, Wearable Sensors, Mobile

Devices.

Abstract:

The aim of this study is to investigate the correlation between, on the one hand, personal and life-style indi-

cators and, on the other hand, the risk of falling. As indicators we consider here for each subject age, body

mass index, and information about physical activity habits, while a subject’s risk of falling is estimated by

the Mini-BES test score. Three different groups of subjects are taken into account, namely healthy, suffering

from metabolic diseases and suffering from cardiovascular diseases. Firstly, we aim at ﬁnding explicit linear

correlations for any pair of parameters. Secondly, we wish to pay attention to whether or not these correlations

change as the health state of the subjects does. The ﬁnal goal is to move the ﬁrst steps towards the design of

a system composed by wearable sensors, a mobile device, and an app that would be able to help people in

improving their life-style so as to decrease their falling risk.

1 INTRODUCTION

Falls have been shown to result in increased morbidity

and are considered the cause of the yearly loss of more

than 17 million years worldwide that are spent in dis-

ability (World Health Organization . Ageing and Life

Course Unit, 2008). Many studies have been dedi-

cated to fall detection, as e.g. (Sannino et al., 2015).

Identifying individuals with a high fall risk is often a

signiﬁcant part of prevention programs. The assess-

ment of the risk of falling is a major and effective

prevention tool that allows identifying intrinsic and

extrinsic risk factors. These latter help determine the

most suitable interventions, thus reducing, or in some

cases even eliminating, falls.

The goal of this study is threefold.

Firstly, we aim to carry out a statistical analysis

to inquire into the existence of clear correlations be-

tween, on the one hand, some of the most widely

considered body parameters, as age and Body Mass

Index, and physical activity tests, and, on the other

hand, the risk of falling, represented through the score

of the Mini-Balance Evaluation Systems (Mini-BES)

test (Franchignoni et al., 2010).

Secondly, we wish to diversify our statistical anal-

ysis, so as investigate whether or not these correla-

tions change when healthy or unhealthy subjects are

considered. We wish to take into account here two

different wide classes of diseases. The ﬁrst class con-

tains metabolic problems such as hypo- and hyper-

thyroidism, hypo- and hyper-glycemia, and so on.

Approximately 34% of the worlds adult population

has the cluster of risk factors that is metabolic syn-

drome (Mozumdar and Liguori, 2011). The second

class, instead, makes reference to diabetes, hypo- and

hyper-tension, vascular and heart-related problems.

Cardiovascular diseases (CVD) are responsible for

30% of all deaths (17.5 million) (World Health Or-

ganization and others, 2005).

Thirdly, we wish to move the ﬁrst steps towards

the opening of a path to the use of wearable sensors

and mobile devices for the on-line monitoring and the

real-time evaluation of a subject’s falling risk through

the consideration of the above found relationships.

This latter goal would make fall risk assessment

much easier, because subjects would not need to un-

dergo the classical Mini-BES test, rather they could

estimate it at home in their everyday life by simply us-

ing a small set of wearable sensors. Namely, a sensor

could estimate the Body Mass Index (BMI), whereas

a second could keep track of the subject’s physical ac-

tivity. Based on the measured data, an app on the sub-

ject’s mobile device could act as a kind of an advisor,

by providing them with a view of their general health

state, and with useful suggestions as well. Moreover,

subjects with a potentially moderate-to-high falling

Sannino G., De Falco I. and De Pietro G.

A Statistical Analysis for the Evaluation of the Use of Wearable and Wireless Sensors for Fall Risk Reduction.

DOI: 10.5220/0006295805080516

risk assessment would be adviced to meet a doctor so

as to possibly undergo a real test. This approach could

lead to easily performing a kind of a ’mass screening’

with reference to the risk of falling. To fulﬁll the two

ﬁrst above statistical goals, in this paper we will make

use of a set of personal and life-style information con-

tained in a real-world database making reference to to

the risk of falling. Then, we will describe the body

system we propose, based on some wearable senors,

a mobile device, and an app.

This paper is organized as follows. Section 2 re-

ports on the related work on ﬁnding correlations be-

tween personal parameters and falling risk. Informa-

tion about the database is provided in Section 3. The

statistical analysis is shown and discussed in Section

4. In Section 5 some considerations are given on the

use of the results of the statistical analysis for the de-

sign of a monitoring system based on wearable sen-

sors. Finally, our conclusions and future works follow

in Section 6.

2 RELATED WORKS

One of the ﬁrst papers trying to ﬁnd correlations be-

tween personal parameters and falling risk was (Gard-

ner et al., 2000). In it, the objective was to assess

the effectiveness of exercise programs in preventing

falls (and/or lowering the risk of falls and fall related

injuries) in older people. Their conclusion was that

exercise is effective in lowering falls risk in selected

groups and should form part of falls prevention pro-

grams.

In (Hue et al., 2007) the aim of the study was to

determine the contribution of body weight to predict

balance stability. Their experiments suggest that body

weight may be an important risk factor for falling.

In (Faulkner et al., 2009) the authors examined

potential independent effects of lifestyle on fall risk.

Not smoking and going outdoors frequently or in-

frequently were independently associated with more

falls, indicating lifestyle-related behavioral and envi-

ronmental risk factors are important causes of falls in

older women.

Although not directly tied to fall risk, the paper

(Shekharappa et al., 2011) dealt with similar ideas, in

fact the aim was to ﬁnd a correlation between body

mass index and cardiovascular parameters in obese

and non-obese in different age groups. The results

showed a statistically signiﬁcant increase in heart rate,

systolic blood pressure and diastolic blood pressure in

obese subjects when compared to non-obese in all age

group. Moreover, there was a positive correlation be-

tween body mass index and heart rate, systolic blood

pressure, diastolic blood pressure, mean blood pres-

sure and pulse blood pressure.

The relationship between Body Mass Index and

stability has been investigated in (Ku et al., 2012).

Namely, the aims of that study was to examine the

impact of BMI and gender on static postural control.

Their conclusion was that BMI do have an impact on

postural control during both bipedic stance and uni-

pedic stance.

The effect of the type, level and amount of phys-

ical activity in falls and fall-related injuries was ex-

amined in (Pereira et al., 2014). Their conclusions

were that being active, especially sufﬁciently active,

reduces fall-related injuries by decreasing falls and by

safeguarding against severe injuries when falls occur.

A study was conducted in (Shahudin et al., 2016)

to investigate the effects of age on physical activ-

ity level, strength and balance towards fall risk index

(FRI) among women, as well as identifying the main

contributing factors towards FRI test performance.

That study suggested that women aged 20–73 years

were found to associate their FRIs mostly with age,

followed by strength, balance, and lastly, physical ac-

tivity.

3 THE DATABASE

To carry out our investigation, we have taken

advantage of the Human Balance Evaluation

database, collected at the Biomechanics and

Motor Control Laboratory (BMCLab) of the

Federal University of ABC, Sao Paulo, Brazil

(http://demotu.org/datasets/balance/), and freely

available in PhysioNet (Goldberger et al., 2000).

This database was collected while performing sta-

bilography tests over a set of subjects. Each of those

subjects had to perform standing tasks under four dif-

ferent conditions: by keeping their eyes opened or

closed, and while standing on a rigid surface or on an

unstable one. Each condition was tested three times,

with the order of the conditions being randomized

among subjects. A total of 1930 trials performed by

164 different subjects are given in this database. Each

1 minute recording is sampled at 100Hz and low pass

ﬁltered at 10Hz.

Moreover, and most importantly to us, the fol-

lowing qualitative tests were employed on each

subject, and the replies/outcomes recorded in the

database: Short Falls Efﬁcacy Scale International

(FES-I) (Kempen et al., 2008) (seven questions plus

the score), the Short version of the International Phys-

ical Activity Questionnaire (IPAQ) (Craig et al., 2003)

(eight questions plus the score), Trail Making Test

(four pieces of information), Mini Balance Evalua-

tion Systems (Mini-BES) Tests (Franchignoni et al.,

2010) (fourteen values plus the score). Furthermore,

the subjects were also interviewed about some of their

socio-cultural, demographic, and health information,

including their age, medications, and illnesses.

Consequently, each database item contains 63 at-

tributes. The database, apart from the raw data record-

ings, also includes a BDSinfo ﬁle that contains meta-

data describing the conditions of the stabilography tri-

als, the information from the anamnesis, and the re-

sults of the qualitative evaluations. Because, as stated

above, a subject has 12 ﬁles for the force platform

data, there are 12 rows for each subject in this ﬁle.

In these 12 rows, the only column that has rows with

different values is the column identifying the trial (the

ﬁle name). The content of all the other columns are

simply repeated over the 12 rows. As result, the

BDSdata ﬁle has the header plus 1930 rows and 64

columns. The complete list of the attributes can be

found in (Santos and Duarte, 2016).

Starting from this database, we have conducted an

analysis phase by creating a new database composed

by 6 items for each of the 164 subjects. The parame-

ters taken into account in our study are:

• x

: age group

• x

: Body Mass Index (BMI)

• x

: IPAQ 1: minutes per week of vigorous physi-

cal activity according to the short IPAQ question-

naire

• x

: IPAQ 2: minutes per week of moderate physi-

cal activity according to the short IPAQ question-

naire

• x

: IPAQ 3: minutes per week of low physical

activity according to the short IPAQ questionnaire

• x

: the total score of the Mini-BES test

In the short IPAQ questionnaire used to create

the Human Balance Evaluation database, the vigorous

physical activities are deﬁned as: heavy lifting, dig-

ging, aerobics, or fast bicycling. The moderate ones,

instead are considered as: carrying light loads, bicy-

cling at a regular pace, or doubles tennis. Finally, the

low physical activities include: walking at work and

at home, walking to travel from place to place, and

any other walking that is done solely for recreation,

sport, exercise or leisure.

As concerns the value for IPAQ

1 for a sub-

ject in our database, this is computed starting from

the subject’s answers to short IPAQ questions 1a

(days per week of high-level physical activity) and 1b

(hours per day of high-level physical activity through:

IPAQ 1 = IPAQ 1a · IPAQ 1b. The same mechanism

holds true for the computation of IPAQ 2 and IPAQ 3.

The value of the score for the Mini-BES test is

computed through the answers of the subject to 14

questions, each of which can be assigned a value

equal to 0, 1, or 2, the higher the better. Therefore,

the value of the Mini-BES test score can range within

0 and 28, where a higher value means that the subject

has a lower falling risk.

Moreover, we have divided the subjects in the

database into three groups:

• healthy: they are the subjects with no disease at

all. This has resulted in a number of 56 individu-

als;

• metabolic diseases: this group contains all the in-

dividuals who declared problems related to hyper

- or hypo-thyroidism, hyper- or hypo-glicemia,

and so on. This group contains 32 subjects;

• cardiovascular diseases: this group is composed

by all the individuals with hyper- or hypo-tension,

cardiovascular problems, or diabetes. There are

41 people in this group.

It should be pointed out that we excluded from the

groups 48 subjects who were not healthy, yet they

suffered from diseases other than those reported in

the above two groups. As examples, some of them

suffered from melanoma, breast cancer, hepatitis,

Parkinson, arthrosis, asthma, dermatitis, rhinitis, gas-

tritis, kidney stones, sickle cell anemia,tendinitis, and

so on. Moreover, there are 13 people in the database

who suffer from both endocrinological and cardio-

vascular diseases. These have been assigned to both

groups.

4 STATISTICAL ANALYSIS

For each of the three groups of subjects described in

the previous section we have performed a correlation

analysis among the chosen database parameters. By

doing so, we have been able to obtain the correlation

value for each pair of parameters. Let’s recall here

that a correlation value between two parameters is in

the range [-1.0, 1.0], where positive values represent

direct correlations and negative values inverse corre-

lations, and the higher the absolute value the stronger

the correlation.

Moreover, for each of these pairs we have created

a ﬁgure, in which we report the raw data, and have

computed and drawn the best line for the linear re-

gression that best ﬁts the data, and have also reported

the related R-squared (R

) value. R-squared is a sta-

tistical measure of how close the data are to the ﬁt-

Table 1: Correlation values between pairs of parameters for healthy subjects.

age BMI IPAQ 1 IPAQ 2 IPAQ 3 Mini-BES

age 1.00

BMI 0.45 1.00

IPAQ 1 0.56 0.29 1.00

IPAQ 2 0.30 -0.08 0.20 1.00

IPAQ 3 -0.05 -0.02 -0.06 0.21 1.00

Mini-BES -0.30 -0.24 -0.14 -0.06 0.05 1.00

ted regression line. A value of 0 for R

indicates that

the model explains none of the variability of the re-

sponse data around its mean, whereas an R

of 1 in-

dicates that the regression line perfectly ﬁts the data.

These regression lines and their R

values are very

useful because from them fruitful information can be

obtained. In the next three subsections all these ﬁnd-

ings are shown for the three groups, respectively.

4.1 Healthy Subjects

Table 1 reports the correlation values for all of the

considered parameters.

In it the generic cell (i, j) contains the correlation

value between the pair of parameters i and j. Very

high values (≥ 0.50) and very low ones (≤ −0.50) are

shown in dark grey. Moderate values, lying in [-0.49,

0.30] and [0.30, 0.49], are highlighted in light grey.

All the pairs in which the Mini-BES test score appears

have been considered for further analysis. For each

of them the corresponding ﬁgures are shown, which

contain information about the linear regression too.

A ﬁrst remark that can be made concerns the pairs

of parameters for which the correlation values are

high or moderate, i.e. those for which the values in

the tables are shown in dark grey or light grey, re-

spectively.

As concerns the healthy subjects, the only strong

correlation is between the age and the IPAQ 1, i.e. the

vigorous activity, and it a positive correlation. Basi-

cally, this says that as long as healthy people get old,

they go on exercising vigorously. Also quite high is

the moderate direct correlation between age and BMI,

meaning that the older a healthy subject, the more

obese she/he is. Moreover, a moderate direct cor-

relation is shown between age and IPAQ 2, similar

to that between age and IPAQ 1, but with reference

to moderate physical activities. Furthermore, a mod-

erate inverse correlation exists also between age and

Mini-BES. This suggests that for healthy subjects the

higher the age the lower the value of the Mini-BES

test score, hence the more probable the subject will

be prone to falls.

Fig. 1 shows that this group is mainly composed

by young adults. In fact the average age is 31.32 years

y"="$0,051x"+"25,043"

R²"="0,08925"

10"

15"

20"

25"

30"

0,00" 20,00" 40,00" 60,00" 80,00" 100,00"

miniBES"Test"value"

Age"

Age"$"miniBestTEST"correlaFon"

in"Healthy"Subjects"

Best_T"

Linear"(Best_T)"

Figure 1: Analysis of parameters age and Mini-BES Test

score for healthy subjects.

!"#"$%&'(()*"+"',&(-'"

./"#"%&%0,-1"

(%"

(0"

'%"

'0"

-%"

%" 0" (%" (0" '%" '0" -%"

2343567"89:;"<=>?9"

5@A"<=>?9"

5@A"$"234359:;8678"BCDD9>=EC4"

34"F9=>;G!"7?HI9B;:"

59:;J8"

K349=D"L59:;J8M"

Figure 2: Analysis of parameters BMI and Mini-BES Test

score for healthy subjects.

y"="$0,0018x"+"23,786"

R²"="0,01825"

10"

15"

20"

25"

30"

0" 200" 400" 600" 800" 1000"

miniBES"Test"value"

minutes"of"acDvity"in"a"week"

IPAQ1"$"miniBESTest"CorrelaDon"

in"Healthy"Subjets"

Best_T"

Linear"(Best_T)"

Figure 3: Analysis of parameters IPAQ 1 and Mini-BES

Test score for healthy subjects.

±14.90. Moreover, the average value for the Mini-

BES test score is 23.44 ± 2.54, which is quite a high

value suggesting that healthy people have a scarce fall

risk.

A closer examination of Figures 1, 2, 3, 4, and 5

provides more precise information about the relation-

Table 2: Correlation values between pairs of parameters for metabolic subjects.

age BMI IPAQ 1 IPAQ 2 IPAQ 3 Mini-BES

age 1.00

BMI 0.52 1.00

IPAQ 1 0.04 -0.17 1.00

IPAQ 2 0.13 0.08 0.09 1.00

IPAQ 3 0.08 0.26 -0.01 0.03 1.00

Mini-BES -0.78 -0.54 -0.06 -0.03 -0.25 1.00

!"#"$%&%%%'(")"*+&'+,"

-."#"%&%%+/0"

/%"

/'"

*%"

*'"

+%"

%" '%%" /%%%" /'%%"

1232456"789:";<=>8"

123>:89"?@"<ABC2:!"23"<"D88E"

FGHI*"$"1232456789:"J?KK8=<B?3"

23"L8<=:M!"6>NO8A:9"

489:P7"

Q238<K"R489:P7S"

Figure 4: Analysis of parameters IPAQ 2 and Mini-BES

Test score for healthy subjects.

!"#"$%$$&'(")"&*%*+&"

,-"#"$%$$&+*"

'$"

'."

&$"

&."

*$"

$" .$" '$$" '.$" &$$" &.$" *$$"

/010234"5678"9:;<6"

/01<867"=>":?@A08!"01":"B66C"

DEFG*"H"/010234"5678"I=JJ6;:@=1"

01"K6:;8L!"4<MN6?87"

2678O5"

P016:J"Q2678O5R"

Figure 5: Analysis of parameters IPAQ 3 and Mini-BES

Test score for healthy subjects.

ship between the Mini-BES test score and each of the

other parameters considered in this study. Namely,

the precise form of their relationship, under a linear

hypothesis, is shown.

The slope of the line drawn in each ﬁgure provides

intuitive visual understanding of the relationship: a

down-bound line means an inverse linear relationship,

an up-bound one a direct dependence, and the more

inclined the line the higher the amount of this rela-

tionship. Consequently, lines that are almost horizon-

tal imply a substantial independence between the two

parameters. As an example, this is the case shown in

Fig. 5.

4.2 Subjects with Metabolic Diseases

Table 2 reports the correlation values for all of the

considered parameters.

As it has been for the healthy group, here too the

generic cell (i, j) contains the correlation value be-

tween the pair of parameters i and j. The same con-

vention used for that group applies also in this case to

highlight some speciﬁc cells in the table. Similarly to

the previous case, here too all the pairs in which the

Mini-BES test score appears have been considered for

further analysis. For each of them the corresponding

ﬁgures are shown, which contain information about

the linear regression too.

As far as the metabolic patients are taken into

account, three parameter pairs have high correlation

values, namely age-BMI, age-Mini-BES, and BMI-

MiniBES. The ﬁrst is a direct correlation, meaning

that as the age increases so does BMI, as it is quite

frequent in humans, be they healthy or suffering from

some disease. Of higher interest for our purposes

are the other two correlations. Age and Mini-BES

test score are strongly and inversely correlated, which

means that as the age of these diseased subjects in-

creases the Mini-BES test score decreases, so older

subjects suffering from metabolic diseases are more

prone to falls. Moreover, also BMI and Mini-BES test

score are strongly and inversely correlated, meaning

that the more obese a metabolic subject, the higher

probability she/he has of falling.

In this case, as Fig. 6 reveals, the age of this group

is quite higher than that for healthy subjects. In fact,

the average is 62.80 years ±17.91. The average value

for the Mini-BES test score for these subjects, instead,

is 19.67 ± 4.10, i.e. about four points worse that that

for healthy people.

By looking at Figures 6, 7, 8, 9, and 10, it can

be visually understood that for people suffering from

metabolic diseases changes in IPAQ

2 and IPAQ 3 al-

most do not affect the Mini-BES test score, while the

opposite is true for the age, BMI, and IPAQ 1. In par-

ticular, the R

value for the correlation between age

and Mini-BES test score is equal to 0.60908, so we

are conﬁdent that the regression line well ﬁts the data.

4.3 Subjects with Cardiovascular

Diseases

Table 3 reports the correlation values for all of the

considered parameters.

Also for this group, the generic cell (i, j) in the

table reports the correlation value between the pair

of parameters i and j. The cells in this table have

Table 3: Correlation values between pairs of parameters for cardiovascular subjects.

age BMI IPAQ 1 IPAQ 2 IPAQ 3 Mini-BES

age 1.00

BMI -0.08 1.00

IPAQ 1 -0.10 -0.29 1.00

IPAQ 2 -0.34 -0.18 0.15 1.00

IPAQ 3 -0.08 0.17 -0.03 0.03 1.00

Mini-BES -0.39 0.01 -0.07 0.17 0.22 1.00

y"="$0,1787x"+"30,899"

R²"="0,60908"

10"

15"

20"

25"

30"

0" 20" 40" 60" 80" 100"

miniBES"Test"value"

Age"

Age"$"miniBestTEST"correlaGon"

in"PaGents"with"Metabolic"diseases"

Best_T"

Linear"(Best_T)"

Figure 6: Analysis of parameters age and Mini-BES Test

score for metabolic subjects.

y"="$0,6008x"+"34,465"

R²"="0,29433"

10"

15"

20"

25"

30"

0" 5" 10" 15" 20" 25" 30" 35"

miniBES"Test"value"

BMI"value"

BMI"$"miniBestTEST"correlaFon"

in"PaFents"with"Metabolic"diseases"

Best_T"

Linear"(Best_T)"

Figure 7: Analysis of parameters BMI and Mini-BES Test

score for metabolic subjects.

y"="$0,0006x"+"19,852"

R²"="0,00324"

10"

15"

20"

25"

30"

0" 500" 1000" 1500" 2000" 2500"

miniBES"Test"Value"

minutes"of"acDvity"in"a"week"

IPAQ1$"miniBestTEST"correlaDon"

in"PaDents"with"Metabolic"diseases"

Best_T"

Linear"(Best_T)"

Figure 8: Analysis of parameters IPAQ 1 and Mini-BES

Test score for metabolic subjects.

been highlighted by using the same convention as

done for the two previous groups. Similarly to the

two above described cases, also for this group all the

pairs in which the Mini-BES test score appears have

been considered for further analysis. For each of them

the corresponding ﬁgures are shown, which contain

information about the linear regression too.

Finally, when the cardiovascular subjects are con-

sidered, no correlation can be deﬁned as strong, the

!"#"$%&%%%'(")"*+&,-."

/0"#"%&%%%+1"

*%"

'%"

2%"

%" 3%%" *%%%" *3%%" '%%%" '3%%" 2%%%"

4565789":;<=">?@A;"

456A=;<"BC"?DEF5=!"56"?"G;;H"

IJKL'$"45657;<=:89:"DBMM;@?EB6"

56"J?E;6=<"G5=N"O;=?PB@5D"Q5<;?<;<"

7;<=R:"

S56;?M"T7;<=R:U"

Figure 9: Analysis of parameters IPAQ 2 and Mini-BES

Test score for metabolic subjects.

y"="$0,0041x"+"20,189"

R²"="0,06131"

10"

15"

20"

25"

30"

0" 500" 1000" 1500"

miniBES"Test"Value"

minutes"of"acDvity"in"a"week"

IPAQ3$"miniBestTEST"correlaDon"

in"PaDents"with"Metabolic"diseases"

Best_T"

Linear"(Best_T)"

Figure 10: Analysis of parameters IPAQ 3 and Mini-BES

Test score for metabolic subjects.

highest one being a moderate inverse correlation be-

tween age and Mini-BES test score. This is quite sim-

ilar to that already seen for the metabolic subjects, al-

though with a lower tie between the two parameters.

Also, age and IPAQ 2 show a moderate inverse corre-

lation, that is the opposite as that for healthy subjects:

healthy people tend to exercise when they get older,

whereas cardiovascular ones tend to not work out.

Also for the cardiovascular subjects, as it was for

the metabolic ones, the average age is quite higher

than that for the healthy subjects. In fact, as shown in

Fig. 11, it is equal to 72.27 ±6.61, which is higher

than that of metabolic people too. As for the average

score of the Mini-BES test for this group, it results to

be equal to 18.02 ± 3.81, i.e. even worse that that of

the metabolic subjects.

For this group the Figures 11, 12, 13, 14, and 15

do not show any particularly strong correlation, nor

do they report any sufﬁciently high value for R

, apart

from, possibly, the case of age and Mini-Best test

score.

In conclusion, the main result from this statisti-

y"="$0,2227x"+"34,116"

R²"="0,14905"

10"

20"

30"

0" 20" 40" 60" 80" 100"

miniBES"Test"value"

Age"

Age"$"miniBestTEST"correlaGon""

in"Cardiac"PaGents"

Best_T"

Linear"(Best_T)"

Figure 11: Analysis of parameters age and Mini-BES Test

score for cardiovascular subjects.

!"#"$%$$&'(")"*&%+,*"

-."#",%/01$2"

*$"

*2"

/$"

/2"

,$"

$" *$" /$" ,$" '$"

3454607"89:;"<=>?9"

6@A"<=>?9"

6@A"1"345469:;8078"BCDD9>=EC5"

45"F=DG4=B"H=E95;:"

69:;I8"

J459=D"K69:;I8L"

Figure 12: Analysis of parameters BMI and Mini-BES Test

score for cardiovascular subjects.

!"#"$%&%%%'(")"*+&*,,"

-."#"%&%%//0"

*%"

1%"

0%"

%" *%%%" 1%%%" 0%%%" '%%%" /%%%"

2343567"89:;"<=>?9"

234?;9:"@A"=BC<3;!"D9E"F99G"

HIJK*"$"234359:;8678"B@EE9>=C@4""

34"L=EM3=B"I=C94;:"

59:;N8"

O349=E"P59:;N8Q"

Figure 13: Analysis of parameters IPAQ 1 and Mini-BES

Test score for cardiovascular subjects.

cal analysis is that, when a subject suffers from a

metabolic disease, she/he has a probability of falling

that is higher than that of an equally aged cardiovas-

cular subject, and much higher than that of a healthy

peer.

5 USE OF THE STATISTICAL

ANALYSIS RESULTS IN A

MONITORING SENSOR-BASED

SYSTEM

The statistical analysis made in this preliminary study

has shown that, even though moderate, a correlation

exists between the risk of falling (the mini-BES test

score) and the personal and lifestyle indicators.

These results mean that it is imaginable to real-

ize a monitoring system in order to give speciﬁc rec-

!"#"$%$$&&'"("&)%*&+"

,-"#"$%$.$/)"

&$"

0$"

.$"

$" *$$" &$$$" &*$$" 0$$$" 0*$$" .$$$"

1232456"789:";<=>8"

123>:89"?@"<AB;2:!"C8D"E88F"

GHIJ0"K"1232489:7567"A?DD8=<B?3""

23"L<DM2<A"H<B83:9"

489:N7"

O238<D"P489:N7Q"

Figure 14: Analysis of parameters IPAQ 2 and Mini-BES

Test score for cardiovascular subjects.

!"#"$%$$&'(")"*+%,,-"

./"#"$%$&$'-"

*$"

'$"

-$"

$" *$$" '$$" -$$" ,$$" &$$" 0$$"

1232456"789:";<=>8"

123>:89"?@"<AB;2:!"C8D"E88F"

GHIJ-"K"1232489:7567"A?DD8=<B?3""

23"L<DM2<A"H<B83:9"

489:N7"

O238<D"P489:N7Q"

Figure 15: Analysis of parameters IPAQ 3 and Mini-BES

Test score for cardiovascular subjects.

ommendations about the diet and the amount and the

type of physical activity so as to improve the subject

well-being with respect to the risk of falling.

Nowadays, mobile devices, such as smartphones

or tablets, wearable devices, such as smartwatches or

bands, and wireless healthcare devices, such as smart

digital scales, are widely used and it is demonstrated

that they are valid tools to monitor body and life-

styles parameters. For these reasons, thanks to these

devices, it is possible to collect in real time data about,

for example, the weight of a subject or the type and

the amount of activity performed during a day or a

week.

This collected information could be analyzed in

real time on a mobile device in order to give a prompt

feedback to the subject about her/his risk of falling

and in order to guide the subject to develop new habits

to reduce the estimated risk of falling.

As an example, for the subjects with metabolic

disease the statistical analysis has shown that there is

a strong and inverse correlation between the BMI and

the Mini-BES test score, so a wearable sensor-based

monitoring system could give speciﬁc recommenda-

tions in order to not only reduce the obesity, but also

to reduce the risk of falling knowing that more obese

a metabolic subject, the higher probability she/he has

of falling.

Within our laboratory, several mobile applica-

tions have been implemented aimed to monitor dif-

ferent kinds of healthcare parameters, as for example

(Forastiere et al., 2016; Sannino and De Pietro, 2014).

The correlation results found in this study for healthy

subjects and cardiac subjects will be easily added into

them respectively.

Unfortunately, there is no unique app with the pos-

sibility to have a speciﬁc knowledge base for each

subject in order to suggest different recommendations

for the three different groups of subjects.

For this reason, a new mobile health application

is under development in order to take into account

the different results obtained for the different classes

of people examined, e.g. healthy subjects, subjects

with metabolic diseases or subjects with cardiac dis-

eases. The app will be able to monitor body indica-

tors, physiological data, and physical activity infor-

mation by using wearable sensors, be they or not com-

pliant to the Continua Alliance guidelines (Carroll

et al., 2007). Of course, some of these sensors will be

used for long periods, as e.g. those for physical activ-

ity monitoring, therefore they are affected by the typ-

ical problems related to battery charge. Other types

of sensors, instead, will be employed more rarely, as

for example smart digital scales that are typically used

just once in a day.

6 CONCLUSIONS AND FUTURE

WORK

In this paper the correlation between personal and

life-style indicators and the risk of falling has been

investigated.

As indicators we have considered here for each

subject age, body mass index, and information about

physical activity habits, while a subject’s risk of

falling has been estimated by the Mini-BES test

score. Three different groups of subjects have been

taken into account, namely healthy, suffering from

metabolic diseases and suffering from cardiovascular

diseases.

Firstly, explicit linear correlations have been

found for any pair of parameters. Secondly, atten-

tion has been paid to whether or not these correlations

change as the health state of the subjects does.

Finally, some ﬁrst steps have been moved towards

a system, composed by wearable sensors, a mobile

device, and an app, that would be able to help people

in improving their life-style so as to decrease falling

risk.

In the near future we aim at implementing the

sensor-based system.

Moreover, due to the fact that the data set from

Physionet used in this paper looks highly clustered

with little outliers, e.g. most healthy patients are

around 20 years old, we plan to start a cooperation

phase with the University of Naples ”Federico II” in

which they will provide us with some volunteers with

different ages in order to better balance the database.

Within this cooperation we will supply the system

to the volunteers, so as to test its effectiveness and

usefulness.

ACKNOWLEDGEMENTS

This work has been supported by the project

“eHealthNet: Ecosistema software per la Sanit

a Elet-

tronica” (PON03PE 00128 1) ﬁnanced within the

P.O.N. “Research and Competitiveness” call of the

Italian Ministry for University and Research.

REFERENCES

Carroll, R., Cnossen, R., Schnell, M., and Simons, D.

(2007). Continua: An interoperable personal health-

care ecosystem. IEEE Pervasive Computing, 6(4):90–

94.

Craig, C. L., Marshall, A. L., Sjstrm, M., Bauman, A. E.,

Booth, M. L., Ainsworth, B. E., Pratt, M., Ekelund,

U., Yngve, A., Sallis, J. F., and Oja, P. (2003). In-

ternational physical activity questionnaire: 12-country

reliability and validity. Medicine and science in sports

and exercise, 35(8):13811395.

Faulkner, K. A., Cauley, J. A., Studenski, S. A., Landsittel,

D. P., Cummings, S. R., Ensrud, K. E., Donaldson, M.,

Nevitt, M., of Osteoporotic Fractures Research Group,

S., et al. (2009). Lifestyle predicts falls independent

of physical risk factors. Osteoporosis international,

20(12):2025–2034.

Forastiere, M., De Pietro, G., and Sannino, G. (2016). An

mhealth application for a personalized monitoring of

ones own wellness: Design and development. In Inno-

vation in Medicine and Healthcare 2016, pages 269–

278. Springer.

Franchignoni, F., Horak, F., Godi, M., Nardone, A., and

Giordano, A. (2010). Using psychometric techniques

to improve the balance evaluation systems test: the

mini-bestest. Journal of Rehabilitation Medicine,

42(4):323–331.

Gardner, M. M., Robertson, M. C., and Campbell, A. J.

(2000). Exercise in preventing falls and fall related

injuries in older people: a review of randomised con-

trolled trials. British journal of sports medicine,

34(1):7–17.

Goldberger, A. L., Amaral, L. A., Glass, L., Hausdorff,

J. M., Ivanov, P. C., Mark, R. G., Mietus, J. E., Moody,

G. B., Peng, C.-K., and Stanley, H. E. (2000). Phys-

iobank, physiotoolkit, and physionet components of

a new research resource for complex physiologic sig-

nals. Circulation, 101(23):e215–e220.

Hue, O., Simoneau, M., Marcotte, J., Berrigan, F., Dor

J., Marceau, P., Marceau, S., Tremblay, A., and Teas-

dale, N. (2007). Body weight is a strong predictor of

postural stability. Gait & posture, 26(1):32–38.

Kempen, G. I., Yardley, L., Van Haastregt, J. C., Zijlstra,

G. R., Beyer, N., Hauer, K., and Todd, C. (2008). The

short fes-i: a shortened version of the falls efﬁcacy

scale-international to assess fear of falling. Age and

ageing, 37(1):45–50.

Ku, P., Osman, N. A., Yusof, A., and Abas, W. W. (2012).

Biomechanical evaluation of the relationship between

postural control and body mass index. Journal of

biomechanics, 45(9):1638–1642.

Mozumdar, A. and Liguori, G. (2011). Persistent increase

of prevalence of metabolic syndrome among us adults:

Nhanes iii to nhanes 1999–2006. Diabetes care,

34(1):216–219.

Pereira, C. L., Baptista, F., and Infante, P. (2014). Role

of physical activity in the occurrence of falls and fall-

related injuries in community-dwelling adults over 50

years old. Disability and rehabilitation, 36(2):117–

124.

Sannino, G., De Falco, I., and De Pietro, G. (2015). A

supervised approach to automatically extract a set of

rules to support fall detection in an mhealth system.

Applied Soft Computing, 34:205–216.

Sannino, G. and De Pietro, G. (2014). A mobile system for

real-time context-aware monitoring of patients health

and fainting. International journal of data mining and

bioinformatics, 10(4):407–423.

Santos, D. A. and Duarte, M. (2016). A public data set of

human balance evaluations. PeerJ, 4:e2648.

Shahudin, N. N., Yusof, S. M., Razak, F. A., Sariman,

M. H., Azam, M. Z. M., and Norman, W. M. N. W.

(2016). Effects of age on physical activity level,

strength and balance towards fall risk index among

women aged 20–73 years. In Proceedings of the 2nd

International Colloquium on Sports Science, Exercise,

Engineering and Technology 2015 (ICoSSEET 2015),

pages 25–34. Springer.

Shekharappa, K., Smilee, J. S., Mallikarjuna, P. T., Veda-

vathi, K. J., and Jayarajan, M. P. (2011). Correlation

between body mass index and cardiovascular parame-

ters in obese and non obese in different age groups.

International Journal of Biological & Medical Re-

search, 2(2):551–555.

World Health Organization . Ageing and Life Course Unit

(2008). WHO global report on falls prevention in

older age. World Health Organization.

World Health Organization and others (2005). Preventing

chronic diseases: a vital investment: Who global re-

port.