Original article 65
Towards objective evaluation of balance in the elderly:
validity and reliability of a measurement instrument
applied to the Tinetti test
Lorenzo Panellaa, Carmine Tinellib, Angelo Buizzac, Remo Lombardic
and Roberto Gandolfic
The aim of the present study was the validation of an
instrument for evaluating balance, applied to the Tinetti test.
Trunk inclination was measured by inclinometers during
the Tinetti test in 163 healthy participants scoring 28/28 in
the Tinetti scale (controls: 92 women, 71 men; age 19–85
years), and 111 residents in old people’s homes, able to
autonomously perform the test, but scoring less than
28/28 (test group: 78 women, 33 men; age 55–96 years).
Trunk inclination was quantified by 20 parameters, whose
standardized values were summed and provided an overall
performance index (PTOT). PTOT reliability was evaluated
by Cronbach’s a, and its validity by item scale correlation,
discriminant validity and concurrent validity. Influence of
age and sex was assessed by a logistic regression model.
Repeatable and consistent measurements were obtained
(Cronbach’s a = 0.88). Parameter distribution was
significantly different in controls and patients (P < 0.001).
Optimal PTOT threshold for discriminating between normal
and abnormal performance (153.9/200) corresponded to
sensitivity of 88.3%, specificity of 84.7% and area under the
receiver operating characteristics curve of 0.93. PTOT
correlated with the Tinetti scale score, its partial, balance-
related score and Barthel’s Index, but not with the Mini
Mental State score. PTOT correlated with age and level of
performance but not with sex; correlation with age did not
prevent the possibility of discriminating between different
levels of performance and between normal and abnormal
performance. The instrument provided objective
discrimination between different performance levels, in
particular, between normal and altered performance.
Die Studie dient der Validierung eines Instruments zur
Gleichgewichtsevaluierung unter Anwendung des Tinetti-
Tests. Die Rumpfneigung wurde mittels Neigungsmessern
bei 163 gesunden Studienteilnehmern beim Tinetti-Test
gemessen, deren Wertung auf der Tinetti-Skala bei 28/28
lag (Kontrollgruppen: 92 Frauen, 71 Männer im Alter von
19–85 Jahren), sowie bei 111 Bewohnern von Senioren-
heimen, die den Test selbständig durchführen konnten,
aber eine Wertung von weniger als 28/28 erzielten
(Testgruppe: 78 Frauen, 33 Männer im Alter von 55–96
Jahren). Die Rumpfneigung wurde anhand von 20 Para-
metern quantifiziert, deren standardisierte Werte addiert
und zu einem Leistungsindex (PTOT) zusammengefasst
wurden. Die Zuverlässigkeit von PTOT wurde durch
Cronbachs Alpha und seine Gültigkeit nach Item-Skala-
0342-5282

c 2008 Wolters Kluwer Health | Lippincott Williams & Wilkins
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
korrelation und diskriminanter und konkurrenter Validität
evaluiert. Der Einfluss von Alter und Geschlecht wurde
anhand eines logistischen Regressionsmodells erfasst.
Es wurden wiederholbare und konsistente Messwerte
erworben (Cronbachs a = 0,88). Die Parameterverteilung
unterschied sich bei den Kontroll- und Patientengruppen
stark (P < 0,001). Die optimale PTOT-Schwelle für die
Unterscheidung zwischen normaler und anomaler Leistung
(153,9/200) entsprach einer Empfindlichkeit von 88,3%,
einer Spezifität von 84,7% und einer ROC-Kurve (receiver
operating characteristics curve) von 0,93. PTOT korrelierte
mit dem Tinetti-Skala-Score, seinem partiellen, balance-
bezogenen Score und dem Barthel-Index, aber nicht mit
dem Ergebnis des Mini-Mental-Status-Tests. PTOT korre-
lierte mit dem Alter und dem Leistungsgrad, nicht aber mit
dem Geschlecht; die Korrelation mit dem Alter schloss die
Möglichkeit einer Diskriminierung zwischen unterschiedli-
chen Leistungsstufen und einer normalen und anomalen
Leistung nicht aus. Daraus lässt sich schließen, dass
das Instrument eine objektive Abgrenzung zwischen
unterschiedlichen Leistungsebenen und insbesondere
zwischen einer normalen und geänderten Leistung bot.
El objetivo de este estudio es validar un instrumento
utilizado para evaluar el equilibrio postural, aplicado
a la prueba de Tinetti. La inclinación del tronco se midió
mediante inclinómetros durante la realización de la prueba
de Tinetti a 163 participantes sanos cuyos puntajes en la
escala de Tinetti fueron de 28/28 (controles: 92 mujeres,
71 varones; edad: 19 a 85 años), y a 111 residentes de
hogares de ancianos con capacidad para realizar la prueba
de manera independiente pero cuyos puntajes fueron
menores que 28/28 (grupo experimental: 78 mujeres, 33
varones; edad: 55 a 96 años). La inclinación del tronco se
cuantificó mediante 20 parámetros, cuyos valores estan-
darizados se sumaron para determinar el ı́ndice general de
ejecución (PTOT). La fiabilidad del PTOT se evaluó
mediante el alfa de Cronbach; y su validez, mediante una
escala de correlación y mediante la determinación de la
validez discriminante y de la validez concurrente. La
influencia de la edad y del sexo se determinaron mediante
un modelo de regresión logı́stica. Se obtuvieron criterios
de valoración repetibles y consistentes (a de Cron-
bach = 0,88). La distribución paramétrica fue significativa-
mente diferente en los participantes del grupo control y en
los pacientes (P < 0,001). El umbral óptimo del PTOT para
 66 International Journal of Rehabilitation Research 2008, Vol 31 No 1
diferenciar entre una ejecución normal y una ejecución
anormal (153,9/200) correspondió a una sensibilidad del
88,3%, a una especificidad del 84,7% y a un área por
debajo de la curva de las caracterı́sticas operativas del
receptor de 0,93. El PTOT se correlacionó con los puntajes
de la escala de Tinetti, con su puntaje parcial para el
equilibrio postural, y con el Índice de Barthel, pero no con
el puntaje del Examen Mı́nimo del Estado Mental. El PTOT
se correlacionó con la edad y el grado de ejecución, pero
no con el sexo. La correlación con la edad no impidió
discriminar entre los distintos grados de ejecución ni entre
las ejecuciones normales y las anormales. Se puede
concluir que este instrumento permitió una discriminación
objetiva entre los distintos grados de ejecución, en
particular entre las ejecuciones normales y las anormales.
Cette étude avait pour objet de valider un instrument
d’évaluation de l’équilibre, appliqué au test de Tinetti.
L’inclinaison du tronc a été mesurée par deux inclinomètres
durant le test de Tinetti chez 163 sujets en bonne santé qui
ont obtenu un score de 28/28 sur l’échelle de Tinetti
(contrôles: 92 femmes, 71 hommes; âge: 19–85 ans), et
111 résidents en maison de retraite, capables d’effectuer le
test de manière autonome, mais obtenant des scores
inférieurs à 28/28 (groupe de test: 78 femmes, 33
hommes; âge: 55–96 ans). L’inclinaison du tronc a été
quantifiée selon 20 paramètres, dont les valeurs standar-
disées ont été recoupées et reportées sur un indice de
performance global (PTOT). La fiabilité de l’indice PTOT a
été évaluée par a de Cronbach, et sa validité par corrélation
de l’échelle des items, validité discriminante et validité
concourante. L’influence de l’âge et du sexe a été évaluée
par un modèle de régression logistique. Des mesures
répétables et homogènes ont été obtenues (a de
Introduction
Although it is known that many aspects of balance are
associated with an individual’s risk of falling, a standar-
dized and valid screening instrument to identify people
at risk is still unavailable and there is no evidence to
support including screening for falls in periodic health
examinations of old people (Elford, 1994; Raı̂che et al.,
2000). One of the major disadvantages of most of the
current tests for balance disorders is that they rely on an
examiner’s subjective judgment (Gill et al., 2001). More-
over, vague descriptors of balance, such as ‘poor’, ‘fair’ and
‘good’, do not have consistent operational definitions
and their repeatability has not been tested (Camicioli
et al., 1997).
The Tinetti balance scale (Tinetti, 1986) is one of the
most widely used tests for evaluating balance and its
features possibly associated with falls (Tinetti, 1986;
Camicioli et al., 1997; Raı̂che et al., 2000; Gill et al., 2001).
It is considered to have good interrater reliability and
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Cronbach = 0,88). La distribution des paramètres différait
considérablement entre le groupe de contrôle et les
patients (P < 0,001). Le seuil optimal de l’indice PTOT
permettant de discriminer entre des résultats normaux
et anormaux (153,9/200) correspondait à une sensibilité
de 88,3%, une spécificité de 84,7%, et à une zone se situant
sous la courbe des caractéristiques opératoires du
receveur de 0,93. Le PTOT affichait une corrélation avec le
score sur l’échelle de Tinetti, son score partiel orienté à
l’évaluation de l’equilibre et avec l’indice de Barthel, mais
pas avec le score Mini Mental State. Le PTOT affichait une
corrélation avec l’âge et le niveau de performance, mais
pas avec le sexe; la corrélation avec l’âge n’empêchait pas
la possibilité de discriminer entre les différents niveaux de
performance et entre les résultats normaux et anormaux.
On peut en conclure que l’instrument constitue un outil de
discrimination objectif entre les différents niveaux de
performance, en particulier, entre des résultats normaux
et altérés. International Journal of Rehabilitation Research
31:65–72
c 2008 Wolters Kluwer Health | Lippincott
Williams & Wilkins.
International Journal of Rehabilitation Research 2008, 31:65–72
Keywords: balance, objective measurement, receiver operating
characteristics, Tinetti balance scale, validation
a
Rehabilitation Unit, Clinical Institute ‘Humanitas’, Rozzano, Milan, bBiometry Unit,
Polyclinic Hospital ‘San Matteo’ and cDepartment of Computers and Systems,
University of Pavia, Pavia, Italy
Correspondence to Dr Lorenzo Panella, MD, UO di Riabilitazione, Istituto Clinico
Humanitas, Via Manzoni, 56 I, 20089 Rozzano MI, Italy
Tel: + 39 02 822446716/601; fax: + 39 02 82244693;
e-mail: lorenzo.panella@humanitas.it
Received 7 November 2006 Accepted 5 February 2007
concurrent validity (Topper et al., 1993; Cipriany-Dacko
et al., 1997). The Tinetti scale, however, also suffers from
typical limitations of any qualitative evaluation scale.
Indeed, (i) its results depend on subjective judgments, so
that it cannot provide internationally standardized
measurements; (ii) it uses a scale of integers which
limits resolution; and (iii) it is subject to the ceiling
effect, which does not allow us to identify individuals
with only a few balance problems (Raı̂che et al., 2000).
Only instrumental measurement may warrant objective
measurement, and only objective measurements ‘provide
a scientific basis for communication between profes-
sionals, documentation of treatment efficacy, and scien-
tific credibility’ within the medical community, whereas
there is large evidence for ‘the lack of reliability of
clinicians’ unaided measurement capabilities’ (Hinderer
and Hinderer, 1998). Moreover, instrumental measure-
ments may ensure fine resolution as well as avoid (floor
and) ceiling effects.

On the basis of these considerations, we built a prototype
measurement system for evaluating balance, the Pavia
Instrumented Tinetti Test (PITT). It measures the
participant’s movement during the execution of the
items of the Tinetti balance scale (Tinetti, 1986), and
computes performance indicators from these measure-
ments. It has been designed and realized following the
suggestions of the Joint Commission on Accreditation
of Health Care Organisation about tests and instruments
for functional evaluation, that should be practical, easy to
use, insensitive to outside influences, inexpensive and
designed for efficient administration (Hinderer and
Hinderer, 1998).
This instrument has already been described elsewhere
(Lombardi et al., 2001; Panella et al., 2002), in this study
we wanted to evaluate its sensitivity and specificity in
discriminating between different levels of performance,
and, in particular, between normal and altered perfor-
mance, and to verify whether it suffers from ceiling
effect.
We used the Tinetti balance scale as our criterion
standard because of its diffusion. In this study, however,
we focused only on the manoeuvres of the Tinetti scale
related to balance evaluation.
Materials and methods
Participants
Healthy participants (control)
A total of 163 healthy participants (92 women and 71
men), almost uniformly distributed in the age range 19–
85 years (mean 48.3 ± 18.3 years), underwent the PITT.
They were considered healthy if they did not complain of
any pathology of the lower limbs and/or central or
peripheral balance control system, and had scored 28/28
in the Tinetti test.
Although all patients were over 55 years old (see below),
younger people were also considered in the control
sample for wider validation of the instrument and better
definition of normal performance.
Patients
The test group comprised 111 patients (78 women and 33
men) in the age range 55–96 years (mean 78.6 ± 9 years),
resident in two old people’s homes in Lombardy (Italy).
Independently of the pathology, inclusion criteria were
that the patient scored less than 14/14 in the subset
of Tinetti test manoeuvres considered by the PITT
(see below), but was able to understand the relevant
instructions and to carry out the PITT autonomously.
The rationale for this choice was that the study was
aimed at assessing the ability of PITT to objectively
measure different levels of performance and to discrimi-
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Validation of instrumented Tinetti test Panella et al. 67
nate between normal and altered performance, and not
its ability to identify a specific pathology or equilibrium
disease.
All the patients underwent the PITT and 40 were also
assessed by Barthel’s Index (BI), the Mini Mental State
(MMS) evaluation and the whole Tinetti scale, to verify
possible correlations between the results of the PITT and
any deficit in autonomy (BI) and/or cognitive disorders
(MMS), whereas Tinetti’s test was the criterion standard.
All the scales and the PITT were administered by
specifically trained physiotherapists.
The authors claim that human experimentation was
compliant with the ethical rules of the Declaration of
Helsinki, and approved by the ethical committees of the
concerned institutions.
Balance test
The test protocol consisted of repeating that subset of
the whole set of manoeuvres of the Tinetti scale which is
specifically related to balance (Tinetti, 1986). This
subset is the one usually adopted for screening purposes
in Italian geriatric units.
The following manoeuvres were considered:
(1) Sitting balance for about 10 s.
(2) Arising from the seat (sit-to-stand), without using
the hands to help the movement.
(3) Standing balance with open eyes for about 20 s; the
first 4 s, just after completing the sit-to-stand
movement, were considered as ‘immediate
standing’, the subsequent period was considered as
‘prolonged standing’.
(4) Standing balance with closed eyes, for about 15 s.
(5) Turning balance during a 3601 on-the-spot turn
(clockwise or anticlockwise as preferred).
(6) Sitting down (stand-to-sit) on the seat.
Hardware and software
The participant’s movements during the test were
quantified by measuring his/her trunk inclination by
means of two inclinometers (Midori Precision PMP-
S30TX, Tokyo, Japan), placed at the level of the sternum.
Their main nominal characteristics were: measurement
range ± 901, linearity range ± 301 with linearity of
± 1% of full scale (301), resolution better than 0.031,
weight 35 g. They were placed on a polycarbonate
support, with their planes of maximal sensitivity perpen-
dicular to each other. When placed on the partici-
pant’s chest these planes were almost parallel to the
participant’s sagittal and frontal planes and measured
antero-posterior and medio-lateral trunk inclinations,
respectively. Estimated overall accuracy and repeatability
 68 International Journal of Rehabilitation Research 2008, Vol 31 No 1

of the measurement system in usual test conditions, with parameter and Range the min–max excursion of the
reasonably careful application on the participant’s chest, values measured for this parameter.
were better than 2% and 7.5%, respectively (Lombardi
et al., 2001). In the latter case, the following transformation was used:
yi ¼ 10ðMax  xi Þ=Range
Signals from the inclinometers were A/D converted at 50
samples/s using 12-bit words, and acquired by a PC. where Max is the maximal value measured for that
parameter and the other variables are defined as above. In
this way, after standardization for each parameter a higher
Signal processing consisted of computing 20 parameters
score indicated a better performance.
(Table 1) quantifying: (i) excursion of trunk tilt during
either seated or upright stance, (ii) duration, amplitude
The sum of all the standardized values was used as the
and velocity of antero-posterior inclination of the trunk
global performance score (PTOT).
during sit-to-stand and stand-to-sit movements, (iii)
number of steps and time necessary to complete the
The reliability (internal consistency or homogeneity of
on-the-spot rotation (Lombardi et al., 2001; Panella et al.,
items) was evaluated by Cronbach’s a.
2002).
The validity of the PTOT was analyzed by
Statistical analysis
To have a uniform scale for all the 20 considered
(1) Item scale correlation, that is, correlation between
parameters and to summarize them in one figure, their
single items and the PTOT. Correlation coefficients
values were standardized to the interval [0, 10] by linear
(r) lower than 0.25 were considered to indicate a poor
transformation. The transformation differed according to
correlation, between 0.25 and 0.40 a reasonable
whether the parameter reflected better performance by
correlation, between 0.40 and 0.75 a good correlation
higher or by lower values. In the former case, the
and above 0.75 an excellent correlation (Colton, 1974).
following transformation was used:
(2) Discriminant validity, or the capacity to discriminate
yi ¼ 10ðxi  MinÞ=Range between participants with different characteristics,
that is, in our study, the fact of being healthy or not.
where yi is the standardized parameter value, xi its The discrimination power between the normal and
original value, Min the minimal value measured for this the patient was quantified by the threshold PTOT

Table 1 Descriptive statistics of the 20 standardized parameters in the healthy participants and in the patients
Item Healthy participants (N = 163) Patients (N = 111)

Mean SD CV (%) Min Max Mean SD CV (%) Min Max

1 8.2 0.8 10.2 0.0 10.0 8.1 1.0 12.7 4.2 8.7
2 4.2 8.7 6.0 3.7 10.0 9.0 1.1 13.1 0.0 9.6
3 9.8 0.4 4.0 6.1 10.0 9.3 1.5 17.5 2.8 10.0
4 9.3 0.9 10.1 0.0 10.0 8.6 1.1 13.1 3.0 10.0
5 9.5 0.7 7.0 3.6 10.0 8.5 1.4 16.6 3.4 9.9
6 4.4 2.1 46.8 0.0 10.0 1.2 2.2 103.0 0.1 9.9
7 5.0 2.1 42.4 0.0 10.0 1.8 1.6 74.6 0.1 10.0
8 8.3 1.3 15.1 1.2 10.0 8.4 1.5 19.4 0.0 9.3
9 9.7 0.8 8.1 0.0 10.0 9.7 0.4 4.7 7.5 9.9
10 8.2 0.4 5.3 7.0 10.0 7.7 1.4 19.5 0.0 8.6
11 9.3 0.3 3.1 7.9 10.0 8.9 1.5 17.5 0.0 9.7
12 9.3 0.8 8.4 0.0 10.0 9.2 0.6 6.2 7.0 9.6
13 9.1 0.8 9.2 0.0 10.0 8.8 1.0 11.9 2.3 9.5
14 9.4 0.4 4.6 7.8 10.0 8.3 1.5 19.1 0.2 9.6
15 9.4 0.4 4.3 7.4 10.0 8.7 1.6 19.9 0.0 10.0
16 9.0 0.8 9.3 2.9 10.0 7.9 1.9 24.8 0.0 10.0
17 8.3 1.4 17.0 1.0 9.8 7.6 1.7 22.9 1.7 10.0
18 8.8 0.9 10.5 4.1 10.0 7.4 1.8 24.7 0.8 10.0
19 4.3 2.1 49.0 0.4 10.0 1.5 1.1 64.0 0.0 5.3
20 3.5 2.1 59.3 0.4 9.8 1.7 1.6 77.6 0.0 10.0
PTOT 161.7 9.2 5.7 128.5 181.4 139.3 13.4 9.7 107.6 165.9

1, SD of lateral trunk sway when sitting; 2, SD of frontal trunk sway when sitting; 3, time for trunk inclination during sit-to-stand; 4, time for trunk straightening during sit-to-
stand; 5, overall time taken for sit-to-stand manoeuvre; 6, mean velocity of trunk inclination during sit-to-stand; 7, mean velocity of trunk straightening during sit-to-stand; 8,
SD of lateral trunk sway during immediate standing; 9, SD of frontal trunk sway during immediate standing; 10, SD of lateral trunk sway during prolonged standing, eyes
open; 11, SD of frontal trunk sway during prolonged standing, eyes open; 12, SD of lateral trunk sway during prolonged standing, eyes closed; 13, SD of frontal trunk
sway during prolonged standing, eyes closed; 14, time taken to complete a 3601 turn on the spot; 15, number of steps to complete a 3601 turn on the spot; 16, time for
trunk inclination during stand-to-sit; 17, time for trunk straightening during stand-to-sit; 18, overall time taken for stand-to-sit manoeuvre; 19, mean velocity of trunk
inclination during stand-to-sit; 20, mean velocity of trunk straightening during stand-to-sit.
CV, coefficient of variation; mean, mean value; min, minimum value; max, maximum value; SD, standard deviation.

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

corresponding to the highest sensitivity and
specificity of the test, and by the area under the
receiver operating characteristics (ROC) curve
(Metz, 1978). To assess the possible influence of
age and sex on PTOT, a logistic regression model was
used, with age, sex and PTOT as independent
variables. The results of such model were expressed
as odds ratio (OR) with its 95% confidence interval
(95% CI).
(3) The concurrent validity was evaluated by correlating
the PTOT with the criterion standards, that is, the
global score of the Tinetti test and its partial score
relating to balance.
(4) The correlations between the PTOT and either the
BI or the MMS were evaluated by using the
Spearman’s rank correlation coefficients.
Analysis of variance test and post-hoc least significant
differences t-test were used to compare group means.
All tests were two-sided. A P < 0.05 was considered as
statistically significant.
Analyses were performed with Statistica for Windows
(StatSoft Inc., 2004, Tulsa, Oklahoma, USA) and
MedCalc software (Schoonjans et al., 1995).
Results
The descriptive statistics of the 20 parameters provided
by the PITT have been published elsewhere (Panella
Fig. 1
100
Healthy: N = 163; Mean = 161.7; StdDv = 9.2; Max = 181.4; Min = 128.5
Patients: N = 111; Mean = 138.5; StdDv = 13.4; Max = 165.9; Min = 107.6
80
60
No of obs
40
20
0
100 110
Histograms of PTOT values in the two groups (curves give normal fits). StdDv, standard deviation.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Validation of instrumented Tinetti test Panella et al. 69
et al., 2002). They were computed on the first 150 of the
163 healthy participants here considered, and on a group
of 130 patients, from which the sample of 111 here
considered was extracted after discarding the patients
whose parameter set was incomplete for any reason.
Those changes in the size of the samples did not
significantly change their statistics, and the interested
reader is then referred to Panella et al. (2002).
Owing to space limitations and the goals of this study,
only the standardized values of the parameters will be
considered here. Their descriptive statistics and that of
the PTOT are presented separately for the healthy
participants and the patients in Table 1. The differences
between the corresponding parameter values in the two
samples were all statistically significant (P < 0.001).
In Fig. 1 the distributions of PTOT values in the two
samples are shown.
The value of the Cronbach’s a, evaluating the reliability,
was of 0.88.
The correlations between single items and the PTOT are
shown in Table 2. Considering all participants together,
for 3/20 (15%) parameters an excellent correlation was
found (r Z 0,75), a good correlation (0.4 r r < 0.75)
for 12/20 (75%) parameters, a reasonable correlation
(0.25 r r < 0.4) for 4/20 (20%) parameters, and a poor
correlation (r < 0.25) for 1/20 (5%) parameter. Consider-
ing each sample individually, item scale correlation was
120 130 140 150 160 170 180 190
PTOT
 70 International Journal of Rehabilitation Research 2008, Vol 31 No 1

Table 2 Correlation coefficients between each normalized para- Fig. 2

meter and the PTOT
Item Normals Patients All partcipants PTOT

SD of lateral trunk sway when 0.01 0.24 0.25 100

sitting
SD of frontal trunk sway when 0.17 0.25 0.30
sitting
Time for trunk inclination during 0.40 0.66 0.67 80
sit-to-stand
Time for trunk straightening 0.54 0.62 0.68
during sit-to-stand
Overall time taken for sit-to- 0.60 0.73 0.77
stand manoeuvre 60

Sensitivity
Mean velocity of trunk inclination 0.53 0.52 0.65
during sit-to-stand
Mean velocity of trunk straigh- 0.65 0.55 0.75
tening during sit-to-stand 40
SD of lateral trunk sway during 0.12 0.32 0.26
immediate standing
SD of frontal trunk sway during 0.32 0.40 0.29
immediate standing PTOT cut-off = 153.9
SD of lateral trunk sway during – 0.09 0.32 0.45 20
AUC = 0.93 (95% Cl: 0.89–0.96)
prolonged standing: eyes
open
Sensitivity (88.3 [95% Cl: 80.8–93.6%])
SD of frontal trunk sway during 0.26 0.38 0.50 Specificity (84.7 [95% Cl: 78.2–89.8%])
prolonged standing: eyes 0
open
SD of lateral trunk sway during 0.11 0.35 0.24
0 20 40 60 80 100
prolonged standing: eyes 100-Specificity
closed
SD of frontal trunk sway during 0.28 0.41 0.46
Discrimination between healthy participants and patients: receiver
prolonged standing: eyes
operating characteristics curve for the PTOT corresponding to the test’s
closed
highest sensitivity and specificity (dashed curves are 95% confidence
Time taken to complete a 3601 0.34 0.53 0.67
intervals). CI, confidence interval; AUC, area under curve.
turn on the spot
Number of steps to complete a 0.34 0.36 0.52
3601 turn on the spot
Time for trunk inclination during 0.47 0.55 0.67
stand-to-sit
Time for trunk straightening 0.58 0.55 0.59 found in the healthy people sample (r = – 0.44), and a
during stand-to-sit reasonable one in the patients’ sample (r = – 0.38).
Overall time taken for stand-to- 0.70 0.68 0.78
sit manoeuvre
Nevertheless, the slopes of the two regression lines were
Mean velocity of trunk inclination 0.60 0.59 0.72 not statistically different from one another (P = 0.56).
during stand-to-sit
Mean velocity of trunk straigh- 0.66 0.45 0.60 The difference between PTOT means of males
tening during stand-to-sit
and females was statistically significant. (P = 0.009)
Figures in bold indicate excellent or good correlation.
in patients (mean ± SD; M: 142.7 ± 13.7 vs. F: 136.7
± 12.9), but not in normals (M: 163.3 ± 9.0 vs. F: 160.5
more consistent in patients (only one poorly correlated ± 9.2; P = 0.11).
parameter) than in healthy participants (five poorly
correlated parameters). The multivariate logistic model, with groups (control and
patients) as dependent variable and PTOT, age and sex as
The ROC curve of Fig. 2 gives the sensitivity [88.3 (95% independent variables, showed that PTOT (OR: 0.88,
CI: 80.8–93.6%)] and specificity [84.7 (95% CI: 78.2– 95% CI: 0.84–0.93; P < 0.001) and age (OR: 1.09, 95% CI:
89.8%)] of the PTOT. The area under the curve was of 1.056–1.14; P < 0.001) were statistically significant
0.93 (95% CI: 0.89–0.96). The corresponding PTOT cut- independent factors, whereas sex was not (OR: 0.72,
off value was of 153.9, that is, PTOT > 153.9 indicated 95% CI: 0.30–1.70; P = 0.457).
normal performance, and PTOT r 153.9 abnormal
performance. If only the subsets of both controls and In the patients’ sample, the correlation coefficient
patients aged 60–79 years (44 controls and 51 patients) between PTOT and the global score of the Tinetti test
were considered, the optimal threshold would have been was 0.39 (P = 0.014), and that between PTOT and
153.7, sensitivity 85.5% (95% CI: 73.3–93.5%), specificity the related to balance Tinetti scale partial score 0.41
75% (95% CI: 59.7–86.8%) and the area under the ROC (P = 0.009).
curve 0.85 (95% CI: 0.77–0.91).
The correlation between PTOT and BI was 0.39
Figure 3 illustrates the correlation between PTOT and (P = 0.013), whereas that with the MMS was 0.19
age in the two samples. A good negative correlation was (P = 0.25).

Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Fig. 3
200
180
160
PTOT
140
120
Healthy (r = − 0.44; P < 0.0001)
Patients (r = − 0.38; P < 0.0001)
100
10 20 30 40
Correlation between PTOT and age in the two samples (healthy participants and patients).
Discussion
The aim of this study was to validate an instrument
(PITT) designed to objectively measure a participant’s
performance during a test of balance. The results
obtained during administration of the partial version of
the Tinetti test related to the evaluation of balance may
be summarized as follows:
K PITT provided a global performance score (PTOT)
showing reasonable correlation (r = 0.39) with the
overall score of the Tinetti scale and good correlation
(r = 0.41) with its partial score related to the
evaluation of balance. It reasonably correlated also
with BI (r = 0.39), but not with the MMS score
(r = 0.19), indicating that cognitive disorders do not
necessarily influence patient’s balance.
K PITT provided reliable results, in terms of internal
consistency or homogeneity of items, for all the
domains (Cronbach’s a = 0.88).
K PITT provided repeatable and consistent measure-
ments. Repeatability is ensured by the use of a simple
measurement instrument and computer processing of
the data. An overall uncertainty, due to both test–retest
and interrater repeatability, of less than 7.5% had
already been estimated (Lombardi et al., 2001).
K PITT discriminated between the two considered
samples on the basis of their performance, with
sensitivity and specificity greater than 80%. Then, as
(i) all controls got the maximum score in Tinetti’s scale
and all patients did not, and (ii) PTOT statistics were
quite different in the two samples with generally lower
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Validation of instrumented Tinetti test Panella et al. 71
50 60 70 80 90 100
Age (years)
values in the patient than in the normal sample (Figs 1
and 2, and Table 1) it can reasonably be postulated that
the threshold that best discriminated between the two
samples (PTOT = 153.9) is also a good threshold for
discriminating between normal and abnormal
performance. In participants aged 60–79 years, an
almost identical threshold (PTOT = 153.7) resulted
in slightly lower sensitivity and specificity, but still
near 80%.
K The logistic regression analysis, in which PTOT, age
and sex were considered as possible discriminating
factors between the two samples, showed that the sex
did not significantly influence the possibility of
discrimination, and PTOT, although correlated with
age (Fig. 3), was still able to discriminate between
normal and abnormal performance.
In brief, as the PITT reproduces typical manoeuvres of
the Tinetti scale for evaluating balance, it can be
concluded that it provides almost the same information
as the considered Tinetti scale manoeuvres, with the
advantage of a more rigorous and objective quantification
provided by instrumental measurements.
The correlation between PTOT and Tinetti’s scale
suggests that also this scale is able to discriminate
between normal and abnormal performance. This repre-
sents an objective piece of evidence in favour of the
rationale of the Tinetti’s scale. The use of instrumental
measurements and a numerical evaluation is, however,
likely to significantly improve the sensitivity and
 72 International Journal of Rehabilitation Research 2008, Vol 31 No 1
specificity of the test, as suggested by the following
considerations.
All the healthy participants here considered scored 28/28
when examined by the Tinetti scale. Instead they got a
variety of scores when examined by the PITT
(PTOT = 128.5/200 to 181.4/200, see Fig. 1 and Table
1). This proves that the PITT (PTOT) makes it possible
to distinguish between different degrees of normality
(e.g. from barely normal to excellent) with fine resolution
and actually avoids the ceiling effect, which has been
suggested as possibly limiting the Tinetti test’s ability to
identify people at risk of falling (Raı̂che et al., 2000).
Instead, fine resolution, good sensitivity and good
specificity allow an accurate quantitative assessment of
how (ab)normal the participant’s performance is. In this
sense the PITT can be an effective instrument for follow-
up and evaluation of the outcome of any therapeutic
interventions or rehabilitation. As a confirmation, Fig. 3
shows that, both within the patients and the controls, the
PITT was able to document the well-known tendency of
postural stability to progressively deteriorate with age
(Potvin, 1980). This fact did not emerge from the Tinetti
scale, where, as already said, all controls scored 28/28,
regardless of their age.
The PITT and its test protocol were designed to fit the
needs of the use in a clinical or rehabilitation setting,
even in facilities with limited resources. Actually, its use
was easy, and did not give rise to any problem to both
medical personnel and patients. The whole test lasted
only a few minutes more than the corresponding
manoeuvres of the Tinetti scale, due to the need for
placing the instrumentation on the patient. The training
of the physiotherapists was fast and no discomfort was
reported by the participants.
This study showed the viability and the possible interest
of instrumental measurements applied to the Tinetti
test. The same principle, however, can be applied to the
measurement of other balance or motor tests. For
instance, immediate extensions of this study would be
(i) the application of the PITT to the items of the
Tinetti scale related to gait, (ii) the assessment of the
potential of the PITT during the follow-up and its ability
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
to document the outcomes of rehabilitative treatments
and (iii) the study of the sensitivity of the parameters
defining PTOT in predicting the probability of falls and
their correlation with any risk factors. The ability of the
Tinetti test in predicting the risk of falls is still a matter
for debate: ‘Although the Tinetti balance scale showed
acceptable characteristics [y] as a screening test
for falls, research should be done into improving
performance’ (Raı̂che et al., 2000). We believe that
objective, sensitive, instrumental measurements may
help to effectively tackle this problem.
Acknowledgements
This study was supported by grants from MURST/MIUR,
Rome, Italy, University of Pavia, Pavia, Italy and IRCCS
Policlinico S. Matteo, Pavia, Italy.
References
Camicioli R, Panzer VP, Kaye J (1997). Balance in the healthy elderly:
posturography and clinical assessment. Arch Neurol 54:976–981.
Cipriany-Dacko LM, Innerst D, Johannsen J, Rude V (1997). Interrater reliability of
the Tinetti balance scores in novice and experienced physical therapy
clinicians. Arch Phys Med Rehabil 78:1160–1164.
Colton T (1974). Statistics in medicine. Boston: Little, Brown and Co.
Elford RW (1994). Prevention of household and recreational injuries in the
elderly. In: Canadian Task Force on the Periodic Health Examination, editor.
The Canadian Guide to Clinical Preventive Health Care. Ottawa: Health
Canada. pp. 912–920.
Gill J, Allum JH, Carpenter MG, Held-Ziolkowska M, Adkin AL, Honegger F, et al.
(2001). Trunk sway measures of postural stability during clinical balance
tests: effects of age. J Gerontol A Biol Sci Med Sci 56:M438–M447.
Hinderer SR, Hinderer KA (1998). Principles and applications of measurement
methods. In: DeLisa JA, Gans BM, editors. Rehabilitation medicine:
principles and practice, 3rd ed. Philadelphia: Lippincott-Raven. pp. 109–136.
Lombardi R, Buizza A, Gandolfi R, Vignarelli C, Guaita A, Panella L (2001).
Measurement on Tinetti test: instrumentation and procedures. Technol Health
Care 9:403–415.
Metz CE (1978). Basic principles of ROC analysis. Semin Nucl Med 8:283–298.
Panella L, Lombardi R, Buizza A, Gandolfi R, Pizzagalli P (2002). Towards
objective quantification of the Tinetti test. Funct Neurol 17:25–30.
Potvin AR (1980). Human neurologic function and the aging process. J Am
Geriatr Soc 28:1.
Raı̂che M, Hébert R, Prince F, Corriveau H (2000). Screening older adults at risk
of falling with the Tinetti balance scale. Lancet 356:1001–1002.
Schoonjans F, Zalata A, Depuydt CE, Comhaire FH (1995). MedCalc: a new
computer program for medical statistics. Comput Methods Programs
Biomed 48:257–262.
Tinetti ME (1986). Performance-oriented assessment of mobility problems in
elderly patients. J Am Geriatr Soc 34:119–126.
Topper AK, Maki BE, Holliday PJ (1993). Are activity-based assessment of
balance and gait in the elderly predictive of risk of falling and/or type of falls?
J Am Geriatr Soc 41:479–487.