Validez de distintos instrumentos de evaluación

de fragilidad en ancianos atendidos en diferentes

medios hospitalarios

Prof. Leocadio Rodríguez Mañas

Sº de Geriatría
Hospital Universitario de Getafe
Madrid, España
 TOPICOS
1. El nuevo escenario en la atención sanitaria

2. Fragilidad: Un nuevo concepto que responde a las necesidades

3. Fragilidad un reto clínico: problemas a resolver

4. Fragilidad ¿cómo funcionan los instrumentos de medida?

5. ¿Por dónde seguir?

6. Conclusiones
JAMDA, 2017

TRANSICION TRANSICION TRANSICION

DEMOGRÁFICA EPIDEMIOLOGICA CLINICA
 THE THIRD TRANSITION

BREAKING THE CLINICAL INERTIA

CURE CARE

DISEASE FUNCTION

SURVIVAL QUALITY OF LIFE

TO DO RISK TO BENEFIT RATIO (NOT TO DO)

LONG-TERM TIMELY INTERVENTIONS

Rodriguez-Mañas et al., JAMDA 2017

 The functional continuum

Isolated MULTYSYSTEMIC Multiple

Physiological Non-reversible
IMPAIRMENT
conditions
Vulnerability

SEVERE DEPENDENCY
LOW FUNCTIONAL RESERVE

DEATH
DISABILITY-DEPENDENCY
ROBUSTNESS

Whitson HE et al., 2007

 FRAGILIDAD y DM
Modelos de fragilidad

Fried LP

Rockwood K

Rockwood K. J Am Geriat Soc. 2006;54:975-979

Fried et al. J Gerontol Med Sci. 2001;56A:M146-M156

Lancet 2015; 385: e7-9
 OUR CHALLENGE
LONGEVITY
(AMOUNT OF LIFE)
HEALTH
Prevention Integrated
CHRONIC SYSTEMS
Risk manag. Coordinated.
Empowerment DISEASE + Continued
SOCIAL
SYSTEMS

QUALITY OF LIFE
(FUNCTION) TO MAINTAIN

OUR APPROACH
 Management of chronic disease oriented to avoid frailty and preserve function

 Management of frailty, as the phenotypic expression of disease in older adults

 Management of frailty, as the main predictive factor of adverse outcomes

 Promoting integrated, coordinated and continued care

 NECESITAMOS
DIAGNOSTICAR
LAS SITUACIONES
QUE PRECEDEN LA
DISCAPACIDAD
PARA INTERVENIR
“A TIEMPO”
 Lancet, 2015

1.- We need clinical pathways

2.- We need biomrakers

3.- We need treatments

 Our research projects as Coordinator

FRAILTOOLS A comprehensive validation of tools to screen

and diagnose frailty in different clinical and
social settings to provide instruments for
integrated care in older adults

FRAILCLINIC is aimed at assessing the feasibility

and effectiveness of programs designed to detect and
manage frail older patients in high risk clinical settings

The FRAILOMIC initiative is a large scale

research project aiming to identify the factors that
turn frailty into disability.

A randomized clinical trial to evaluate the effectiveness

of a multi-modal intervention in older people with type 2
diabetes on frailty and quality of life
 Tools to assess frailty
• Variability in different populations
• Variability in risk prediction
• Clinical usefulness
– Feasibility
– Clinical settings
– Biomarkers in risk determination, diagnosis
and prognosis
 Variability in classification and risk
prediction

Fried
ST-Fried
FRAIL
FI
FTSE X% X% Y% X% Y%
Groningen
etc

RISK A RISK A RISK A RISK B

 Variability in different
populations

Lourenço R et al., Age and Ageing 2014

 Raising misclassification with changing risks

Bouzon et al., JAMDA 2017

 *
*

*
*
* *

Bouzon et al., JAMDA 2017

 Time to event

Bouzon et al., JAMDA 2017

Feasibility and effectiveness of the
implementation of programs to
screen and manage frail older
patients in different clinical settings

Project Grants (HP-PJ)

2nd European Union Health Programme
 Objective

• To assess the feasibility and effectiveness of

programs designed to detect (observational phase) and
manage (interventional phase) frail older patients in
high risk clinical settings and to avoid functional
impairment and other associated adverse outcomes:
Emergency room
Cardiology
General surgery
Oncology
 Phases
PHASES: Observational and interventional
1.Observational phase:
•Feasibility and effectiveness of frailty screening
programs implemented in different clinical settings.

2.Interventional phase:
•Feasibility and effectiveness of the implementation of
programs to screen and manage frail older patients in
different clinical settings
 Participating centres
SPAIN
oHospital Universitario de Getafe
(Madrid, España)
oHospital Universitario Monte
Naranco (Asturias, España)

ITALY
oOspedale San Raffaele (Roma,
Italia)
oUniversita Cattolica del Sacro
Cuore (Roma, Italia)

UNITED KINGDOM
oDiabetes Frail Ltd (DIFRAIL)
oAston University
 Observational phase
Tools assessed for Frailty:
•Fried criteria
•FRAIL Scale
•Tilburg Frailty Indicator
•Gröningen Frailty Indicator
•CFS or Rockwood modified
•ISAR (Emergency room)
•Balducci criteria (Oncology)
•VES 13 (Oncology)
•G8 (Oncology)

•Time recruitment: 10 months

 Participants
• Inclusion criteria:
o Patients older than 75 years, assessed in several clinical settings (different
from Geriatry); Emergency Room, Cardiology, Elective and Urgent Surgery
and Oncology.
• Exclusion criteria:
o Do not able to give informed consent.

o Participants with Impairment cognitive moderate or severe according to

MMSE scale (score 18 or lower) and/or the GDS scale (score 5 or higher ).

o Those with physical disability according to Barthel Scale (lower to 40)

o Participants with critical acute disease.

o Life expectancy less than six months

o Patients living in nursing homes

 Sociodemographic characteristics
Emergency Elective Urgent
Variables Cardiology Oncology Agregate
Room Surgery Surgery

N 118 221 155 65 50 609

82,58(5,06
Age (DT) 83,71(5,66) 80,65(4,30) 79,50(3,04) 78,9(3,19) 81(4,62)
)
Gender (%
36,75 54,55 61,04 47,69 64,00 52,81
males)
White race 100,00 99,55 99,35 100 98,00 99,50

Civil Status (%)

Single 6,84 3,64 2,60 7,69 4,00 4,46
Married or
39,32 51,36 70,78 44,62 62,00 54,13
cohabiting
Widow 52,99 43,18 25,32 47,69 32,00 40,10
Necessity of principal caregiver
Profesional
12,62 4,78 2,92 13,85 2,00 6,56
Caregiver
Frailty classification by tool and setting

Fried FRAIL Tilbg Grng Rockw ISAR Bald G8 VES Total

Setting
(%) (%) (%) (%) (%) (%) (%) (%) 13 (%) (%)
Emergency
50,51 40,71 68,14 74,34 47,46 78,76 -- -- -- 60,00
Room
Cardiology 61,39 41,36 65,55 62,32 42,47 -- -- -- -- 54,61
Elective
24,67 15,48 30,32 30,72 5,16 -- -- -- -- 21,27
Surgery

Urgent
53,33 41,54 37,50 50,77 18,46 -- -- -- -- 40,32
Surgery
Oncology 47,92 30,00 36,00 40,00 6,00 -- 14,28 81,63 34,69 36,31
Agregate 47,43 33,67 51,27 53,23 28,34 -- -- -- -- 42,78
 Feasibility of scales
Emergen Elective Urgent T
Complet Cardiology Oncology
gy Room Surgery Surgery (min
e (%) (%) (%)
(%) (%) (%) )
Fried 68,64 60,17 76,92 87,74 12,31 66 5
FRAIL 98,52 94,07 99,55 99,35 100 100 4
I. F.Tilburg 92,45 85,59 90,05 96,13 98,46 100 5
I.F.Gröninge
91,13 85,59 90,05 96,13 98,46 100 5
n
Rockwood
99,67 100 99,10 100 100 100 3
modificada

ISAR* 93,22 93,22 -- -- -- -- 3

G8** 98,00 -- -- -- -- 98 5
Balducci** 98,00 -- -- -- -- 98
Ves13** 98,00 -- -- -- -- 98
• *Specífic of Emergency Room
• ** Specific of Oncology
 Feasibility Fried scale

Weight Physical Gait T

Fatigue Hand Grip Full Scale
lost Activity Speed (m)
Medido Medidos Medidos Medidos Medidos Medidos
Setting N
s (%) (%) (%) (%) (%) (%)
Emergency
118 98,31 98,31 98,31 62,71 91,53 60,17 5
Room
Cardiology 221 99,10 99,55 99,10 76,92 95,02 76,92 5
Elective
155 100,00 99,35 100,00 88,39 98,71 87,74 7,5
Surgery
Urgent
65 98,46 98,46 95,38 12,31 98,46 12,31 9
Surgery
Oncology 50 100,00 100,00 100,00 66,00 100,00 66,00 4
Agregate 609 99,18 99,18 98,85 69,29 96,06 68,64 5
Causes of non implementation:
Fried scale
Degree of concordance among
scales: Global
INTERPRETATION
FRAIL Fried Rockw Grng KAPPA
Tilbg
(Landis & Koch, 1977)
Tilbg
Kappa Index:

0.01–0.20 Slight
Grng agreement

0.21– 0.40 Fair

Rockw agreement

0.41–0.60 Moderate
agreement
Fried
0.61–0.80 Substantial
agreement
FRAIL 0.81–0.99 Almost perfect
agreement
 Emergency Room Cardiology
FRAIL Fried Rockw Grng Tilbg ISAR FRAIL Fried Rockw Grng
ISAR Tilbg
Tilbg
0,62
Tilbg
0,59
0,59 Grng
Grng

Rockw
Rocw

Fried
Fried

FRAIL
FRAIL

Oncology
Elective Surgery
FRAIL Fried Rock Grng Tilbg Bald G8
FRAIL Fried Rockw Grong Tilburg VES 13
VES 13
Tilbg
0,65 G8

Bald
Grng 0,07
Tilbg
0,66
Rockw 0,11 Grng

Rockw
Fried 0,02
Fried

FRAIL
0,66
FRAIL
 KAPPA INDEX AMONG SCALES:
URGENT SURGERY
INTERPRETATION
FRAIL Fried Rockw Grng
KAPPA
Tilbg
Tilbg (Landis & Koch, 1977)

Kappa Index:

Grng 0.01–0.20 Slight

agreement

0.21– 0.40 Fair

Rockw
agreement

0.41–0.60 Moderate
Fried agreement

0.61–0.80 Substantial
agreement
FRAIL
0.81–0.99 Almost perfect
agreement

In this setting, feasibility of Fried´s criteria is vrey low (12.8%)

 A comprehensive validation of tools to
screen
and diagnose frailty in different clinical
and social settings to provide instruments
for integrated care in older adults
Project Grants (HP-PJ)
3rd European Union Health Programme
 General objectives

1. To assess the usefulness as screening and

diagnosis tools of some selected instruments to
detect frailty in:
• Clinical
Hospitals (Geriatric settings: Acute Care Unit and
Outpatient Office)
Primary Care

• Social
Nursing homes

2. To provide sequential diagnostic

algorithms
 Overview of the Project

Participating centres
• SPAIN
Getafe University Hospital
• Fundación para la Investigación Biomédica del
Hospital Universitario de Getafe (FIBHUG)
• Servicio Madrileño de Salud (SERMAS)

• ITALY
Università Cattolica del Sacro Cuore (UCSC)

• FRANCE
Centre Hospitalier Universitaire de Toulouse
(CHUT)

• POLAND
Jagiellonian University Medical College (JUMC)

• UNITED KINGDOM
DIFRAIL
Aston University (since 1st December 2016)
 Methodological approach
PARTICIPANTS AND SETTINGS:
Sample size is established in 485 person by setting; total of 1940
persons (97 per setting)
• Inclusion criteria:
People 75 years or older.
Attended in 4 different settings:
 In-Hospital Geriatric wards
 Hospital outpatient offices
 Primary Care
 Nursing Homes
– Exclusion criteria:
Subjects unwilling or unable to consent or unable to participate
MMSE <20 points
Terminal illness (life expectancy <6 months)
In Hospital and Primary Care: Dependency in more than 2 IADL (Lawton)
In Nursing Homes: Barthel Index < 40
 Instruments assessed

• INSTRUMENTS/TOOLS:

1. Frailty Phenotype- Fried criteria

2. FRAIL scale
3. SHARE-FI
4. CHSA- Clinical Frailty Scale
5. The 35 item Rockwood Scale
6. Shorten version of the Frailty Trait Scale- FTS
7. The Gerontopôle Frailty Screening Tool- GFST
 Implementation of the Project

Preliminary Results (Spain)

RECRUITMENT STATUS
SETTING Recruited Goal To Go
Geriatric Ward 53 97 44
Outpatient 62 97 36
Consultation
General Practice 107 97 Reached
Nursing Homes 37 97 60
TOTAL 257 388 131

Recruitment started: February 2016

Recruitment scheduled until: February 2017 (month 22) - extend until July
2017?
 Implementation of the Project

Preliminary Results (Spain)

Rate of fulfilment and time spent on administration of six different frailty assessment tools in four clinical and social settings

SETTINGS

SCALES GW OC GP NH

Completion Time (sec) Completion Time (sec) Completion Time (sec) Completion Time (sec)
(%) Mean (SD) (%) Mean (SD) (%) Mean (SD) (%) Mean (SD)

Fried 44.7 185.5 (90.5) 96.4 186.5 (94.8) 95.6 227 (93.4) 77.8 164.6 (56.4)

FRAIL 100 61.8 (50) 100 114.9 (405.6) 100 70.5 (32.1) 100 41.2 (17.7)

Rockwood-35 8.5 211 (85.5) 7.3 152.4 (89.7) 1.1 332.1 (103.7) 3.7 156.9 (61.3)

CFS 100 7.9 (4.3) 100 9.4 (11.9) 100 4.1 (3.6) 100 9.4 (14)

SHARE-FI 95.7 66.5 (46.4) 96.4 78.3 (41.1) 98.9 90.9 (35.5) 77.8 67.1 (51.2)

GFST 100 34.3 (23.4) 100 40.1 (85.7) 100 19.7 (12.2) 100 30.4 (12.4)

GW: Geriatric Ward. OC: Outpatient Consultations. GP: Primary Care Centres. NH: Nursing Homes.
CFS: Clinical Frailty Scale.
 • Si no nos vale, probar 2-step uno
sencillo con sensibilidad y otro con
especificidad y biomarcadores

Garcia-Garcia FJ et al., JAMDA 2014

41
Partners of FRAILOMIC
 PHASE 1 PHASE 2
Prospective Cohorts with Validation Cohorts to
Follow-up of ≥ 3 yrs Potential BM Be Followed during 2.5 yrs Validated
+ +
Bio-banks - Risk factors Samples from participants - Risk factors
Lab Biomarkers - Diagnoses Potential Lab BMs - Diagnoses
- Genomic - Genomic
- Proteomic - Prognostic factors - Proteomic - Prognostic factors
- Metabolomic - Metabolomic
- Classical BM - Classical BM

Substudies in Special populations/conditions: Diabetes, CVD, CVRF,

Nutrition, Exercise

Best Fitted Models PHASE 3

ToolKits Dissemination Exploitation PHASE 4

5bii. University of Valencia
Main Frailomic results found during 2015.

• We carried out metabolomics plasma nuclear magnetic resonance (NMR)

spectral deconvolution and further analysis on the ETES cohort. We found
associations with frailty for total NMR cholesterol, isoleucine, lactate, acetate,
citrate, polyunsaturated fatty acids and total creatine.
• We are exploring other hypotheses like pH variations (determined by NMR),
lipid mobility (peak line width) and metabolic intercorrelations.
6b. Academic Hospital of Parma
Main Frailomic results found during 2015.

• STARTING FROM 14 ANALYTES: VCAM-1, ICAM-1, MMP-9, MMP-11,

ACTIVIN-A, ADIPONECTIN, MYOSTATIN, GALECTIN-3, TROPONIN-T,
PROBNP, PCT, ESTRADIOL, A.N.A., SUPAR.
• AFTER COVARIATES ANALYSIS, 7 BIOMARKERS DISPLAY
SIGNIFICANT CORRELATIONS PROBNP, TROPONIN-T, SUPAR,
VCAM-1, MMP-9, GALECTIN-3, ADIPONECTIN.
 Questions
Ques<ons to be answered
Addressed in Detail by FRAILOMIC

Ques<on: Ques<on:

QUESTION 2
QUESTION 1
DIAGNOSIS

To include: To include:

Covariates Covariates

Ques<on:
QUESTION

Covariates
RISK

To include
3

-
-

Ques<on: Ques< on:

QUESTION 4

QUESTION 5
PROGNOSIS

To include all To include:

Covariates Covariates

17
 LABORATORY DATA

PROCESS STATUS

Normaliza( on
Solving ID issues Raw-processed data Experimental data is:
Co-variates to consider
Batches
§ De-idenKfied
§ Added to the database

COHORT 1 LABS 1 To consider: mRNA-ETS (SERMAS

and SG) will be considered for
COHORT 2
COHORT 3
YOUHEALTH
LABS 2
LABS 3
Feature Selec<on: Step 2
the Phase 2. Not enough
Analysis
COHORT 4 of data: Associa<ons and Models
LABS … AIM METHODS
overlap with exisKng paKents.
RESULTS

FAST NOTES ON
HARMONIZATION LAB DATA FOR EACH COHORT

15

QUESTIONS STEP 1: SELECT VARIABLES STEP 2: IMPUTATION STEP 3: MINIMAL MODEL

ADDRESSED

Pre-selection Imputation IniKal SelecKon: Journal of Statistical Software

JSS
MXM package
§ Experimental variables. MMMMMM YYYY, Volume VV, Issue II. http:/ / www.jstatsoft.org/

FEATURE SELECTION FEATURE SELECTION § Cohort variables

MINIMAL MODELS MICE: Multivariate Imputation by Chained
STEP 1 STEP 2 Equations in R

Filter: Stef van Buuren Karin Groothuis-Oudshoorn

AIM AIM AIM § Those variability = 0 1 TNO Quality of Life, Leiden

2 University of Utrecht
1 Roessingh RD, Enschede
2 University Twente

§ Those with > 25% of NA Abstract

STATISTICS STATISTICS STATISTICS Multivariate Imputation by Chained Equations (MICE) is the name of software for
imputing incomplete multivariate data by Fully Conditional Speci cation (FCS). MICE
V1.0 appeared in theyear 2000asan S-PLUSlibrary, and in 2001asan Rpackage. MICE

250 ImputaKons Selected for each

V1.0 introduced predictor selection, passive imputation and automatic pooling. This
articlepresentsMICE V2.0, which extendsthefunctionality of MICE V1.0in several ways.
In MICE V2.0, theanalysisof imputed dataismadecompletely general, whereastherange

RESULTS RESULTS RESULTS of models under which pooling works is substantially extended. MICE V2.0 adds new

imputaKon
functionality for imputing multilevel data, automatic predictor selection, data handling,
post-processing imputed values, specialized pooling and model selection. Imputation of
categorical data is improved in order to bypass problems caused by perfect prediction.
Special attention to transformations, sum scores, indices and interactions using passive
37
imputation, and totheproper setup of thepredictor matrix. MICE V2.0isfreely available
from CRAN as an R package mice. This article provides a hands-on, stepwise approach
to using mice for solving incompletedata problemsin real data.

Keywords: multipleimputation, chained equations, fully conditional speci cation, gibbssam-

pler, predictor selection, passive imputation, R.

1. Introduction
4 Multipleimputation (Rubin 1987, 1996) isthemethod of choicefor complex incompletedata
problems. Missing data that occur in more than one variable presents a special challenge.
Two general approaches for imputing multivariate data have emerged: joint modeling (JM)
 Minimal
Feature model
Selec<on: Step 2 for diagnosis
Q1: DIAGNOSIS ALL
 ETS

Q1_FINAL

Use of Minimal model and Toledo cohort information variables in diagnosis (Q1) by PCA a
 CONCLUSIONES
1.- La implantación definitiva de la fragilidad en la práctica clínica
necesita disponer de procedimientos diagnósticos bien definidos

2.- Los instrumentos disponibles hoy dia presentan problemas

para su aplicación en la práctica diaria que afectan a su
factibilidad, variabilidad en diferentes medios clínicos y su
pobre correlación.

3.- La combinación de criterios clínicos y de laboratorio pueden

aportar una mayor precisión diagnóstica
 ¡Gracias por
su atención!

leocadio.rodriguez@salud.madrid.org