MDS-PAS Comprehensive Care in Parkinson’s Disease June 5, 2021
Leveraging technology to meet
comprehensive care needs
Alberto J. Espay, MD, MSc, FAAN
Professor of Neurology
Director and Endowed Chair, James J. and Joan A. Gardner Family Center for Parkinson's Disease and Movement Disorders
University of Cincinnati
Disclosures
• Research: NIH and Michael J Fox Foundation
• Consultant/scientific advisory board: Abbvie, Neuroderm,
Neurocrine, Amneal, Adamas, Acadia, Acorda, Kyowa Kirin, Sunovion, Lundbeck, and USWorldMeds
• Honoraria: USWorldMeds, Acadia, and Sunovion
• Royalties: Lippincott Williams & Wilkins, Cambridge University Press, and Springer
Advantages of wearable sensor technologies
1. Objective and reliable measurements
˗ No concerns regarding inter- or intra-rater variability
2. Continuous data collection
˗ Assessing of patients across time instead of a single snapshot
3. High resolution of sensors
˗ Can detect smaller magnitudes of change compared to human observers
4. Unobtrusiveness of data collection
˗ Passive data collection while patients are in their natural environment
5. Patient empowerment
˗ Increase adherence to protocols and clinician’s directions
6. Minimal training required
˗ Easier to train in the use of a technology than to train a master clinician
Adapted from Kubota KJ. Mov Disord 2016;31(9):1314-26.
Sensor-based measures are more sensitive to change
Why technology in clinical trials?
2. Clinical measures by clinicians are not as sensitive as device’s
• More sensitive to small changes by Kinesia device
• Less variability in repetitive measures by Kinesia device
Held m an et al. Clinician versus machine: reliability and responsiveness of motor endpoints in Parkinson's disease. Parkinsonism Relat Disord.
2014 Jun;20(6):590-5
Heldman et al. Clinician versus machine: reliability and responsiveness of motor endpoints in Parkinson's disease. Parkinsonism Relat Disord. 2014 Jun;20(6):590-5
More sensitive measures: fewer patients needed
for clinical trials
Feature Clinician
ICC
Kinesia ICC
Number of subjects - clinician
Number of subjects - Kinesia
Percent fewer subjects
Rest tremor 0.63 0.68 100 93 7.5%
Postural tremor 0.68 0.71 100 96 3.9%
Speed 0.62 0.94 100 65 34.6%
Amplitude 0.72 0.94 100 77 23.3%
Rhythm 0.45 0.63 100 72 28.3%
Heldman et al. Parkinsonism Relat Disord. 2014 Jun;20(6):590-5
•More sensitive measures that vary less allow greater precision in trials
Portable Hardware / Software
• Plenty of digital health technologies
• Lots of data
• Any-time, continuous data
= Big data
Artusi et al, Neurotherapeutics 2020
Big data: the center of the health universe
https://iisc.talentsprint.com/digitalhealth/
Adapt patients to technologies or technologies to patients?
https://www.cbinsights.com/research/iot-healthcare-market-map-company-list/
Diagnosis
rejected
Diagnosis
confirmed ON No Dyskinesia OFF
Dyskinesia
The allure of making sense of Big Data
Non-fluctuator
Fluctuator
No Depression
Depression
No Anxiety
Anxiety
No OH
Labels we know OH
The allure of making sense of Big Data
Accelerometer
Gyroscope
Blood pressure
No OH
OH
ON
OFF
(Adapted from) Courtesy, Aristide Merola & Walter Maetzler
Active uses of mobile health technology
• All “channels” of information provided by sensors
require processing by algorithms
• The algorithms need to distinguish between
“background noise” and “clinically meaningful
signals”
• The validation process is anchored on the clinical
assessment by expert clinicians (“gold standard”)
From Big Data to Label = Validation
Input data Pre-process Feature extraction
Feature selection
Classification Patient with
suspected depression
“Depression”
Adapted from Bhat et all, Comput Biol Med. 2018: S0010-4825(18)30270-1
Diagnostic Concordance
Typical overflow for labeling procedures
Depression Scale
Validation
Depression Scale Patient withsuspected depression
Patient with confirmed depression
The “analog” era
Perfect agreement
Beyond threshold
Is digital health an opportunity to
validate our clinical categories?
Or to revisit them entirely from the
patient’s perspective?
Digital Health for
Personalized and Integrated Care
Medicalized?
Validated with prior scales?
Validated in large populations?
For our medical records?
For regulatory agencies?
For understanding populations?
First problem: Information irrelevance
Little information generated is
directly useful to the users
Source: Endeavour Partners, September 2013
Major problem: Losing the individual
The information generated
must be applicable to
larger populations
Source: Endeavour Partners, September 2013
The information generated
reflects that from a
previously validated scale
Digital Health Pathway
Analog
The PD diary
example
An example of a patient-completed symptom diary (Adapted from Hauser et al. 2000)
(throwback to)
Category of Problems Examples
Artificial language OFF, ON, ON with non-troublesome dyskinesia, ON with troublesome dyskinesia
Recall biases Retrospectively rating a 30-minute epoch can be cognitively challenging
Absent non-motor fluctuations
Non-motor symptoms and non-motor fluctuations neglected
Motor reductionism Dyskinesia assumed to be a peak-dose only, OFF- associated dystonia not captured, etc.
No partial states All-or-none duality for fluctuating motor behaviors, without any gradations to ON and OFF states
Averaging of behaviors If dyskinesia is experienced < 50% of a 30-minute epoch, state is marked “ON without dyskinesia”
No medication tracking Lack of medication tracking affects interpretation of symptom fluctuations in relation with levodopa cycles
Adapted from Vizcarra et al, Mov Disord 2019 May;34(5):676-681
Selected shortcomings of the analog diary
A conceptual e-Diary development
? Or
Other examples
Digitalization of the UPDRS or a patient-centered
assessment of motor function?
Digitalization of the Spiral drawing or a patient-centered assessment of tremor?
Digitalization of the finger tapping or a patient-centered assessment of speed?
A true PD diary…
• Would not have to match the language of neurologists or dichotomize their lives into OFF and ON half-hour epochs
• Would summarize the real-life individualized experience of a patient through good and bad times during the day, and good and bad days with spectrum thereof –in a manner that is directly valuable to
patients themselves, without need for “translation” by a neurologist.
Passive and active data merged
• A future MDS e-Diary is being designed to provide a digital interface (e.g., smartphone and/or watch) for logging personalized information and tracking data
• Passive data collection: capture through one or more sensory channels, adapted to as many different needs as possible
• Active data entry: a finger sliding across a “Smiley meter” as often as desired (Less need over time).
• The latter provides context to the former so that the artificial intelligence algorithm “learns” the patterns associated with good and bad times –patterns in
background activity, speech, heart-rate variability, etc.
Researchgate, James Michael Fisher
Perception
Capacity
Performance
CAPACITY
PERFORMANCE PERCEPTION
Addressed in Prof. Maetzler’s lecture
1. Objective and reliable measurements
˗ No concerns regarding inter- or intra- rater variability
2. Continuous data collection
˗ Assessing of patients across time instead of a single snapshot
3. High resolution of sensors
˗ Can detect smaller magnitudes of change compared to human
observers
4. Unobtrusiveness of data collection
˗ Passive data collection while patients are in their natural environment
5. Patient empowerment
˗ Increase adherence to protocols and clinician’s directions
6. Minimal training required
˗ Easier to train in the use of a technology than to train a master
clinician Adapted from Kubota KJ. Mov Disord 2016;31(9):1314-26.
Requirements for translating promise into action
Relevant to patients:
long term adherence
Endorsed by regulators:
accepted outcomes Data accurate and interpretable
Metadata refers to data that accompany and describe the primary data, can be utilized to better understand the context of the sensor data and assist in data management, data sharing, and data analysis.
Badawi et al, Digit Biomark 2019;3:116–132
Enter Metadata
The Tension in The Translation
Relevant to patients:
long term adherence
Endorsed by regulators:
accepted outcomes Data accurate and
interpretable
Individualized, need not be generalizable
Generalizable, exchangeable across platforms
Validation for an individual
Validation for a population
1. Big data does not equal meaningful data; context is critical
2. Direct patient relevance is indispensable for success of
personalization of Digital Health Pathway and the
integration into care
3. “Validation” anchored on existent (analog) instruments
(e.g., prior scales) will continue to be inadequate to
harness the value of digital technology
4. Metadata (data interpretability, reproducibility, and
suitability for regulatory acceptance) allows validating
technology to individuals –and generalizing to populations
5. The MDS e-Diary development stands to become the first
demonstration that technology is not just a replacement of
the paper world but a veritable form of Digital Health
Concluding remarks
Thank you for your attention!
Email: [email protected] Twitter: @AlbertoEspay