innovation for tomorrow: what the future holds falls...

© L. Chiari – University of Bologna EU FALLS FESTIVAL 2016 Bologna, 24 February 2016

Innovation for Tomorrow: what the future holds

Falls Prediction Lorenzo Chiari1,2

1 Personal Health Systems Lab, Dept. Electrical, Electronic and Information Engineering «Guglielmo Marconi»

2 Health Sciences and Technologies - Interdepartmental Center for Industrial Research University of Bologna, Italy {[email protected]}


Can we predict (future) falls?

Lorenzo Chiari1,2 1 Personal Health Systems Lab, Dept. Electrical, Electronic and Information Engineering

«Guglielmo Marconi» 2 Health Sciences and Technologies - Interdepartmental Center for Industrial Research

University of Bologna, Italy {[email protected]}


Pierpaolo Palumbo1, Luca Palmerini1, Sabato Mellone1 1 Personal Health Systems Lab, Dept. of Electrical, Electronic and Information

Engineering «Guglielmo Marconi» - University of Bologna, Italy

Carlo Tacconi, Alice Coni2

2 Health Sciences and Technologies - Interdepartmental Center for Industrial Research University of Bologna, Italy

Federico Chesani3, Luca Cattelani3 3 Dept of Computer Science and Engineering, University of Bologna, Italy

Stefania Bandinelli4, Marco Colpo4 4 Laboratory of Clinical Epidemiology, Azienda Sanitaria di Firenze, Italy

Acknowledgements


Acknowledgements

farseeingresearch.eu

Whatever the media are dealing with today, the university has already dealt with it twenty years ago, and what the university deals with today will be reported by the media in twenty years. Attending university means being decades ahead of your time. It's the same reason why literacy prolongs life. Who does not read has only his life, which, I assure you, is very little. On the contrary, when we shall die we will remember to have crossed the Rubicon with Caesar, fought with Napoleon at Waterloo, traveled with Gulliver and met dwarves and giants. A small compensation for the lack of immortality.

Umberto Eco (1932 - 2016)


The present


- Tools for falls prediction are needed as strategic components for fall prevention

RISK FACTORS FALL RISK ASSESSMENT

FALL PREVENTION

The present

- Clinical risk factors


J Am Geriatr Soc. 2011

1 • Screening for high risk

2 • Assessment of

multiple risk factors for those at high risk

3 • Implementation of a

tailored intervention

The guidelines (ask & evaluate)


Traditional tools (1st generation) - Often based on subjective

evaluations - No use of statistics, no

probabilistic meaning

2nd generation tools - (e.g. PPA)

Validation of traditional tools

Sensor-based tools (3rd generation)

-Proof of concept

1986 Get-Up and Go Test

time

2013 Howcroft’s review on sensor-based tools

2003 Lord’s Physiological Profile Assessment PPA

2010 Deandrea’s review on fall risk factors

1991 TUG 2008 Lamb’s screening tree

1986 POMA

Advent and diffusion of high-throughput technology: -inertial sensors

Advances in statistical learning for high-dimensional problems

Consciousness in statistics - development and validation

Predictive tools: a bit of history

1994 Guralnik’s SPPB


• Guidelines were tested on older disabled women and in community-dwelling older adults and found to be suboptimal with respect to other screening tool and of moderate clinical utility (Muir et al, J Geriatr Phys Ther, 2010).

• Predictive accuracy of current versions of these screening algorithms not currently reported.

• Many other screening tools have been proposed in the literature but few of them have been tested outside the derivation cohort.

• History of falls is a strong risk indicator for future falls (Ganz et al, JAMA, 2007), although it alone does not suffice for primary prevention.

• SPPB is a tool to assess physical performance, commonly included in comprehensive geriatric assessments. Its association with falls and injurious falls is documented (Veronese et al, Rejuvenation Res. 2014) but its prognostic performance is not reported.

Where are we with falls prediction?


Palumbo, PhD Thesis, 2015

• TUG has been judged inadequate in several studies (e.g. Barry et al, BMC Geriatr. 2014).

• Gait speed is an indicator of health state in geriatric populations. Its prognostic value for future falls has been shown to be equivalent to total time to perform the TUG (Viccaro et al, J Am Geriatr Soc. 2011).

Where are we with falls prediction?


Clinical knowledge

Clinical risk factors (CRF) for falls are well known

What if we fully exploit the knowledge generated by CRF?


J Am Geriatr Soc. 2011

1 • Screening for high risk

2 • Assessment of

multiple risk factors for those at high risk

3 • Implementation of a

tailored intervention

A possible operationalization of established 2nd generation tools


The FRAT-UP tool

• Assessment tool for evaluating the fall risk within a year

• Aimed to general practitioners and health organizations (per-subject evaluation vs. population wide)

• (So far) focused on community-dwelling older people • Based on 26 risk factors available in the literature • Exploits available clinical information about the

subject • Freely available as a web-based application at the url

ffrat.farseeingresearch.eu Cattelani et al., J. Med. Int. Res., 2015


User interface

-Allows the use of statistical prevalence of the risk factor -Support for dichotomic, scalar and synergy risk factors


© L. Chiari – University of Bologna I-DON’T-FALL WORKSHOP ON FALL PREVENTION AND MANAGEMENT -

Roma, 25 September 2015

CRF Coverage of guidelines domains

AGS/BGS FRAT-up Data-driven model (variables available for selection)

Data-driven model (variables most frequently selected)

Relevant medical history, physical examination, cognitive and functional assessment

Diabetes MMSE Self-perceived health status

Widely covered Self-perceived health status, CESD, familiarity to diabetes,

History of falls Yes Yes Yes Medications number of medications,

sedatives, anti-hypertensives, antiepileptics,

Domain accurately covered.

Number of medications, drug for dementia, anti-hypertensives, antidepressants

Gait, balance, and mobility Gait problems assessed via the Revised walking sub-score, walking aid use

Gait speed in different tests, FICSIT, SPPB, etc.

Gait speed, cautious attitude while walking

Visual acuity Visual acuity, contrast sensitivity, visual stereognosis

Visual acuity, contrast sensitivity, visual stereognosis

Other neurological impairment

Parkinson’s disease Parkinson’s disease, subclinical/non overt neurological signs

Muscle strength No Grip strength, lower limb muscle strength

Heart rate and rhythm No Yes Postural hypotension No No (There is information

about blood pressure before and after standing up but not explicity the difference…)

Feet and footware No Information about shoes Environmental hazards No No No


Clinical risk factors: can we do better?

Palumbo et al., PLOS One, 2015


AUC (95% CI) ActiFE ELSA InCHIANTI TILDA FRAT-up 0.562 (0.530-0.594) 0.699 (0.680-0.718) 0.636 (0.594-0.681) 0.675 (0.661-0.711) Model fitted on ActiFE 0.574 (0.541-0.604) 0.566 (0.545-0.585) 0.549 (0.505-0.594) 0.559 (0.532-0.584) Model fitted on ELSA 0.560 (0.527-0.593) 0.719 (0.699-0.739) 0.611 (0.570-0.654) 0.675 (0.648-0.704) Model fitted on InCHIANTI 0.530 (0.501-0.559) 0.664 (0.644-0.682) 0.571 (0.520-0.619) 0.633 (0.608-0.661) Model fitted on TILDA 0.561 (0.527-0.592) 0.661 (0.642-0.678) 0.600 (0.558-0.647) 0.686 (0.660-0.710)

Validation on epidemiologic DBs

Palumbo et al. 2016, in preparation


Predictive tools: FRAT-UP

Palumbo, PhD Thesis, 2015


A JUMP INTO THE FUTURE (THE FUTURE IS NOW)




Sensor-based assessment

Instrumented Functional Tests

Sensor-based multidimensional assessment

Timed Up and Go

Repeated Chair Standing

Balance

Gait




Sensor-based assessment

Instrumented Physical Activity Monitoring

Sensor-based multidimensional assessment (incl. exposure!)




State of the art

- A number of studies using the former approaches has been done now (e.g. Marschollek et al, 2011; Schwesig et al. 2013; Schwenk et al, 2014; van Schooten et al, 2015) which assessed fall risk prospectively.

- Results are very promising - Some commercial solutions are on the market but… - Datasets are relatively small, yet - They lack of external validation




Preliminary results

• Factor analysis on data from an instrumented TUG – global performance – smoothness of sit to walk transition (StW) – lateral weight shift control during the turn to sit transition (TtS) – lateral weight shift control during StW – forward weight shifting control during StW – smoothness of TtS

• Confounders: age, gender, MMSE, BMI, SPPB, CES-D, #Drugs • Logit multiple regression model • All factors are associated with falls at 6 months with AUC=0.74 (p=0,01) • All factors are associated with falls at 12 months with AUC=0.7327 (p=0,01) • Single factor, Smoothness of StW is associated with falls at 12 months with

AUC=0.708 (p=0,01)

Colpo et al, Proc. GSA, 2015 Colpo et al, Proc SIGG, 2015

Sensor-based assessment only




Prognostic tools: we can do better!


The added value of a measurement

Gait features computed from long-term recordings of PA


Looking into the future (As soon as reliable and validated tools

become available)

- Push the process back from secondary to primary prevention - Predicting time to first fall (e.g. Campanini et al, EUFF 2016) - Better stratified risk → phenotyping → tailored prevention - Link fall risk with exposure - Dynamic risk assessment

Open issues:

- Big data, data fusion, data integration - Standardization (e.g. see Redmond et al, 2014; Klenk et al,

2013) - Assessment of impact in clinical practice (e.g. see Meyer et al,

2009)


A future scenario?


Forecast

innovation for tomorrow: what the future holds falls...

Documents