Potentials and Risks of Digital Health Interventions · multiple chronic conditions (MCC) and the proportion of MCC is increasing with age: 80% of people older than 60. Gerteis et

Potentials and Risks of Digital Health Interventions

Prof. Dr. Tobias KowatschAssistant Professor for Digital Health, University of St.GallenScientific Director, Center for Digital Health Interventions, ETH Zurich & University of St.Gallen

Tuesday | November 12 | 2019

© University of St. Gallen & ETH Zurich & CSSSlide 2 | ETH Zurich | Potentials and Risks of DHIs | November 12, 2019

What is the problem?

1. Chronic diseases are the leading cause of death:70% worldwide and 85% in Europe. 86% of all healthcare spending in US.WHO 2017, Brennan et al 2017, Kvedar et al 2016

2. 31.5% of the US population are living with multiple chronic conditions (MCC) and the proportion of MCC is increasing with age: 80% of people older than 60.Gerteis et al 2014

3. Health care systems of the 20th century focused on acute and episodic diseases in hospitals and not (!) on behavioral support in the everyday life of individuals.Gerteis et al 2014, Kvedar et al 2016


One approach to tackle the problem

Design of evidence-based digital health interventions that allow caregivers to scale and tailor long-term treatments toindividuals (in their everyday life) at sustainable costs.


Digital Health Interventions

Anatomy of Digital Health Interventions

Body Mass Index

Distal Outcome

Proximal Outcomes

Physical activity Stress Diet

1. Vulnerability 2. Receptivity 3. Support

Example: Obesity

Evidence-based

knowledge


Alex

A Selection of Our Digital Health Interventions


https://voicebot.ai/2019/02/04/mercedes-focuses-entire-super-bowl-ad-around-hey-mercedes-voice-assistant/

Computers are social actorsThe “healing” car

Hey Mercedes, make it cooler and

play my music!

Sour

ce: y

outu

.be/

sgF4

jj2FG

lw

Hey Joe, I recommend you to take in some

glucose!

Mercedes text was developed with a type-1 diabetes patient.

Sinergia

https://voicebot.ai/2019/02/04/mercedes-focuses-entire-super-bowl-ad-around-hey-mercedes-voice-assistant/

https://youtu.be/sgF4jj2FGlw


Overview of adherence and (medical) outcomes

# Project Population Duration N Adherence Outcome1 PathMate2 Obese children 24 weeks 26 56% daily goal

achievementMuscle mass and physical capacities increased sig., fat mass decreased sig.

2 Ally Adults (public health)

6 weeks 274 54% daily goal achievement

Physical activity increased (pre/post), data collected to increase adherence

3 Clara Adult asthmatics

4 weeks 93 97% adherence to data collection protocol

Data for an asthma early-warning system (e.g., ca. 23K cough samples, self-reports)

4 Max Children with asthma

3 weeks / 14 lessons

49 80% intervention completion rate

Asthma knowledge increased sig. (pre/post) and inhalation mistakes were reduced

5 Alex Chronic back pain patients

4 weeks 1 92% exercise completion rate

Preference of Alex to video- and paper-based instructions / improved exercise performance

6 Selma Individuals with chronic pain

8 weeks 59 52% adherence to conversations

Pain intensity was reduced sig. and well-being increased sig. (no delta to control group)

7 PEACH Pilot

Primarily students

2 weeks 185 59.5% completed intervention

Initial evidence for intentional change of self-discipline and openness

8 PEACH Primarily students

10 weeks 1523 32.8% completed intervention

Approx. effect size of ca. 0.5 in personality change, final analyses ongoing

9 CAMP Cardiac rehabilitation patients

4 weeks 114 56% adherence to data collection protocol

Observational study (final analyses ongoing)

Health literacy intervention for children with asthma

www.c4dhi.org/projects/health-literacy-children-asthma/

https://www.c4dhi.org/projects/health-literacy-children-asthma/


Conceptual Model of MAX

Individual feedback by asthma expert

Support by family member

Gamified scalable digital assistant of

asthma expert

Inhalation technique

Intervention components

Monitoring ofasthma symptoms

Health literacy in asthma

Preparation of an action plan for asthma attacks

Asthma control(e.g., # of asthma attacks)

Future workProximal outcomes Distal outcome


Interaction with MAXSmartphone

Mobile AppWeb-based cockpitAsthma Expert

Digital HealthAssistant

Patient

MAX

Family Member

SMS

eMail

App chat with expertor face-to-face

Chat with MAX

SMS, phone call or face-to-face face-to-face


Inhalation Assessment of Norah, 12Informed consent was received from patient and parent to use video, name and age for presentation purposes

1. Video recording by family member 2. Expert rating

+Automated

feedback generation based on inhalation

guidelines

3. Feedback to Norah

1


Main results of the pilot study

1. Overall, 99 children with asthma have been approached by asthma experts from 6 study centers, within 4 months in 2019; 49 (49.5%)subjects started to interact with MAX.

2. The average adherence rate of the 49 subjects was 80.4%.

3. The result of a pre-post test shows that asthma knowledge was improved significantly with a large effect size (d=0.9).

4. On average, 1 inhalation mistake was identified in each video clip; 3 serious inhalation mistakes could be directly addressed and eliminated by the experts’ feedback in this trial.

Design of a prognostic digital biomarker for asthma control

https://www.c4dhi.org/projects/css-mobile-asthma-companion/

https://www.c4dhi.org/projects/css-mobile-asthma-companion/


Conceptual Model

Individual feedback by asthma expert

Support by family member

Gamified scalable digital assistant of

asthma expert

Inhalation technique

Intervention components

Monitoring of asthma symptoms

Health literacy in asthma

Preparation of an action plan for asthma attacks

Asthma control e.g., # of asthma attacks

Future workProximal outcomes Distal outcome (not measured currently)

Effort in timehighmiddlelow


Overview of the asthma study with Clara

https://vimeo.com/258412196

https://vimeo.com/258412196


The Clara Application


# Item Scale Mean SD

1 How satisfied were you with Clara? 1 = Not at all7 = Very satisfied 5,95 1,23

2 How easy was messaging with Clara? 1 = Easy7 = Difficult 1,90 1,61

3 How much would you like to continue working with Clara? (n = 87)

1 = Not at all7 = Very much 4,87 1,72

4 How likely is it that you will follow Clara’s advice?

1 = Not at all likely7 = Very likely 5,59 1,56

Clara assessment by 88 adults with asthma


Preliminary results of the study

Response rate to daily self-reported asthma control tests of 93 participants: 97.3% (2487 / 2557)Paper N subjects N coughsVizel et al. 2010 12 n/aMcGuiness et al. 2012 10 n/aCasaseca-de-La-Higuera et al. 2015 9 n/aMonge-Alvarez et al. 2018 13 n/aSwarnkar et al. 2013 3 342Amoh et al. 2016 14 627Birring et al. 2008 15 1.836Barry et al. 2006 15 2.000Amrulloh et al. 2015 24 2.090Drugman et al. 2011 22 2.304Liu et al. 2014 20 2.549Larson et al. 2011 17 2.558Coyle et al. 2005 8 3.645Klco et al. 2018 18 5.200Kadambi et al 2018 9 5.670Our Clara study > 77 23.488


1. Delivery of interventionsFrame digital health interventions as scalable assistants of physicians and other caregivers that “live” in the pockets of patients and have the potentialto improve the doctor-patient relationship,intervention adherence and health outcomes.

2. Medical researchUse digital health interventions for ecological momentary assessments(a) to better understand the development of diseases and(b) to develop diagnostic or prognostic digital biomarkers.

Potentials of Digital Health Interventions

Risks ofDigital Health Interventions


Smartphone addiction

https://www.ncbi.nlm.nih.gov/pubmed/26690625

https://www.ncbi.nlm.nih.gov/pubmed/26690625


Outdoor and indoor network coverage is still an issue

https://scmplc.begasoft.ch/plcapp/pages/gis/netzabdeckung.jsf?netztyp=lte

https://scmplc.begasoft.ch/plcapp/pages/gis/netzabdeckung.jsf?netztyp=lte


Android may ask patients to “shut down” their digital health intervention app

https://www.androidpit.com/how-to-stop-apps-running-in-the-background-on-android

https://www.androidpit.com/how-to-stop-apps-running-in-the-background-on-android


Apple’s iOS may ask patients to stop the collection of data relevant to their digital health intervention

https://techcrunch.com/2019/09/19/ios-13-security-privacy/

https://techcrunch.com/2019/09/19/ios-13-security-privacy/


We must deal with sensor data from various devices that differ in quality (e.g., signal-to-noise ratio)

Shih, I., Tomita, N., Lukic, Y., Hernández, Á., Fleisch, E., Kowatsch, T., Breeze: Smartphone-based Acoustic Real-time Detection of Breathing Phases for a Gamified Biofeedback Breathing Training, Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (Issue 4, December) (forthcoming).

Microphones

sound sound sound


Computers are social actorsVoice assistants and chatbots are all around us

Apple.com Microsoft.comGoogle.com

+https://www.amazon.com/dp/B06WP3HFCK

Alexa, I think I am having a Heart Attack.

“These first aid skills are for informational and educational purposes only”

https://www.amazon.com/dp/B06WP3HFCK


Using acoustic signal injection to attack voice-controllable systems (home automation, health assistant)

Fig. 5 from Sugawara et al. 2019 Light Commands: Laser-Based Audio Injection Attacks on Voice-Controllable Systems https://lightcommands.com/20191104-Light-Commands.pdf

https://lightcommands.com/20191104-Light-Commands.pdf


A selection of further risks linked to data collection and processing

1. Configuration of sensor parameters (e.g., sampling frequency)> we could miss life-threatening events

2. Burden of data collection (e.g., non-technical savvy patients)> we could significantly decrease acceptance

3. Behavioral reactivity of individuals> too many false positive alarms may reduce acceptance

4. Limited quality of data and algorithms (e.g., quality of data labeling)> we may count a cough by a dog as an asthmatic cough of a patient

5. Dependency on software and hardware (e.g., bugs, regular updates)> we have to deal with “moving targets”

6. Data leakage of messages with digital assistants and personal health data> this data could be used against an individual

7. Misinterpretation of chat messages by digital assistants> we may trigger the wrong intervention


Further risks related to digital health interventions

1. Motivational aid for patient engagement (e.g., sleep-tracking)> we may trigger data-driven perfectionism (e.g., orthosomnia from sleep tracking)

2. Cognitive aid to patients (e.g., visualization of blood glucose level)> a patient could misinterpret the aid

3. Decision aid to patients (e.g., medication reminder)> we may reduce medication nonadherence to a problem of forgetting

4. Cognitive aid to health professionals (e.g., symptom profiles)> we could increase health disparities as we have less insight into health states of patients who are not willing or not able to track themselves

5. Decision aid to health professionals (e.g., alert triggered by a digital health app)> we have to consider medical liability for ensuring real-time response

Note: Risks are derived from Sim, I. 2019. "Mobile Devices and Health," N Engl J Med (381:10), pp. 956-968.


One asthma expert received a lovely “Thank You” post card from a young patient after the interventionLeverage potentials and anticipate risksto build robust digital health interventions.


One asthma expert received a lovely “Thank You” post card from a young patient after the intervention


Advisory BoardProf. Dr. Elgar FleischLisa Marsch, PhDProf. Dr. Urte ScholzProf. Dr. Florian von Wangenheim

Center LeadershipProf. Dr. Tobias KowatschMatthias Heuberger, M.A.

Research TeamAlina Asisof, M.Sc.Filipe Barata, M.Sc.Caterina Bérubé, M.Sc.George Boateng, M.Sc.Christoph Gross, M.Sc.Jan-Niklas Kramer, M.Sc.Yanick Lukic, M.Sc.

Florian Künzler, M.Sc.Marcia Nißen, M.Sc.Joseph Ollier, M.Sc.Dominik Rüegger, M.Sc.Prabhakaran Santhanam, M.Sc.Iris Shih, M.Sc.Peter Tinschert, M.Sc.and many more …

Prof. Dr. Tobias KowatschAssistant Professor for Digital Health, University of St.GallenScientific Director, Center for Digital Health Interventions at the University of St.Gallen & ETH Zürichhttps://www.c4dhi.org | https://www.mobile-coach.eu | [email protected]

https://www.c4dhi.org/

https://www.mobile-coach.eu/

mailto:[email protected]

Documents

Potentials and Risks of Digital Health Interventions · multiple chronic conditions (MCC) and the proportion of MCC is increasing with age: 80% of people older than 60. Gerteis et