New Technologies for the Collection of Data for Tailored Messaging Kevin Patrick, MD, MS Professor of Family and Preventive Medicine School of Medicine Director, Center for Wireless and Population Health Systems, Calit2 Editor-in-Chief American Journal of Preventive Medicine Co-Founder Santech, Inc.

American Institute for Cancer Research November 6, 2009

Collaborating Investigators UCSD School of Medicine Kevin Patrick, MD, MS, Greg Norman, PhD, Fred Raab, Jacqueline Kerr, PhD Jeannie Huang, MD, MPH UCSD Department of Political Science James Fowler, PhD UCSD Jacobs School of Engineering Bill Griswold, PhD, Ingolf Krueger, PhD, Tajana Simunic Rosing, PhD SDSU Departments of Psychology & Exercise/Nutrition Science James Sallis, PhD, Simon Marshall, PhD PhD students and Post-doctoral Fellows (current) Barry Demchak, Priti Aghera, Ernesto Ramirez, Laura Pina, Jordan Carlson

Program of research on systems of wireless, clinical, and home technologies to measure and improve lifestyle and other health-related behaviors in:

-- Healthy adolescents -- Overweight and obese children and adults -- Depressed adults -- Adolescents risk for type 2 diabetes -- Older adults to promote successful aging -- Adolescents recovering from leukemia -- Young adults to prevent weight gain -- Post-partum women to reduce weight -- Adults with schizophrenia -- Exposure biology research

Several projects exploring the use of wireless, mobile and connected devices, sensors and systems: mDIET SMART PALMS CitiSense CYCORE

Mainly, it’s all about Sensors…

Sensors embedded in the environment Geocoded data on safety, location of recreation, food, hazards, etc

Psychological & Social sensors Mood, Social network (peers/family) Attention, voice analysis

Biological sensors BP, Resp, HR, Blood (e.g. glucose, electrolytes, pharmacological, hormone), Transdermal, Implants

Diet & Physical Activity sensors Physical activity (PAEE, type), sedentary Posture/orientation, diet intake (photo/bar code)

Wearable Environmental sensors Air quality (particulate, ozone, etc) Temperature, GPS, Sound, Video, Other devices & embedded sensors

Sensor data + Clinical & Personal Health Record Data + Ecological data on determinants of health + Analysis & comparison of parameters in near-real time (normative and ipsative) + Sufficient population-level data to comprehend trends, model them and predict health outcomes + Feedback in near real-time via SMS, audio, haptic or other cues for behavior or change in Rx device

= True Preventive Medicine!

Challenges: Many parts – how to get them to work together, especially when they become life-critical? Changing ecosystem of devices – how to make systems adaptable as new technologies come online? Usability, persuasiveness and fun (!) factor Data – lots and lots of data…

mDIET mobile Dietary Intervention Through Electronic Technology Research question: Can a behavioral intervention delivered primarily through text messages be effective in promoting short term weight loss in overweight/obese adults?

Funded by a grant from the National Cancer Institute R21 CA115615-01A1

Personalized Text Messages •

Eating Behaviors – 4 items on the EBI were emphasized (based on our own logic rules)

•

Name – Some text messages included their first name (e.g. Congrats, Maria! You continue to improve. You're clearly working hard and it shows).

•

Grocery Store – Participants were asked to identify the grocery store that they most frequently visit (e.g. Did you buy fruits and vegetables from Trader Joes this week?)

•

Social Supporter – Participants were asked to identify someone in their personal life (family member, friend, coworker) that could part of their social support system (e.g. Have you been telling Mark about your weight loss success?)

MMS used for Images and Graphs

Add a variety of colorful vegetable to your shopping list this week. Choose green, red, orange and yellow veggies.

A one-cup serving size is about the size of a tennis or baseball.

Copyright, Regents of the University of California and Santech, Inc, 2008

Nice progress. You're on your way to reaching your goal. It will take time, but you have the motivation to succeed.

Rule-based Dialogues Messages sent are based on the user’s previous responses

Study Design Randomized Controlled Trial: Participants were randomized to either an Intervention (mDIET) or Control group: mDIET - Weight loss program using text and MMS messages along with modest amount of adjunctive intervention elements Control - Usual care Participants completed 3 in-office measurement visits over a 4 month period

Copyright, Regents of the University of California and Santech, Inc, 2008

Study Sample Characteristics • • •

•

Sample N = 63 81% Women Mean age 45.9 years – Women ranged 26 – 55 yrs. – Men ranged 33 – 55 yrs. Race/Ethnicity – 76.2% Caucasian or White – 15.9% African American or Black – 3.2% Asian American/Pacific Islander/Native Hawaiian – 4.8% Prefer not to state – 22.2% Hispanic

Copyright, Regents of the University of California and Santech, Inc, 2008

At end of 4 months mDIET  group lost about 6.25 lbs and  Control group 2 lbs (p=.03) 

Patrick et al, J Med Int Res, 2009 Copyright, Regents of the University of California and Santech, Inc, 2008

Participant Feedback on mDIET • • • • • • •

“Steady reminder – keeping health on my mind” “Felt commitment every day – could not let myself forget my goals” “They served as an excellent reminder to watch what I ate” “Keeps me focused” “Constant reminders to believe in myself and make the right choices” “I found that texting your weight every week was extremely helpful” “I miss my 6am message!”

Overall satisfaction with mDIET program for weight loss 95.6% of participants would recommend mDIET to friends/family

Resulted in the first report in the literature of an RCT evaluating SMS/MMS for weight loss

SMART… – Social/Mobile Approach to Reduce Weight in Young Adults

SMART… – Social/Mobile Approach to Reduce Weight in Young Adults NHLBI-funded study to integrate mobile phone services & applications with Facebook apps to promote improved weightrelated behaviors

Collaborators and platforms:

SMART… – Social/Mobile Approach to Reduce Weight in Young Adults

Research aims – 24 month weight gain prevention or weight loss outcomes Focus on physical activity, sedentary behaviors and diet habits of young adults Develop behaviorally sound integrated mobile/social applications – developed and prototyped by young adults in iterative fashion Incorporate location/context aware elements for persuasive behavioral prompting

PALMS Physical Activity Location Measurement System An integrated suite of hardware and software to enable continuous capture and subsequent variable creation, visualization and analysis of data on the temporal and spatial characteristics of physical activity – combined with other sensor data if desired PALMS supports gathering data from multiple participants within studies and aggregating and comparing data between and among multiple studies Funded through the NIH Gene, Environment and Health Initiative Exposure Biology Program (NCI, NIEHS, NHLBI)

Exposure Biology Program: Deliverables FY07

FY08

FY09

Environmental Sensors • Diet/Physical Activity • Chemicals/Biologics • Psychosocial Stress/Addictive

FY10

FY11

DEVICES

Substances

APPLICATION

Biological Response • Biomarkers • Centers–biomarkers/biosensors • Inflammation • Oxidative stress • Programmed cell death • Epigenetic markers

Research on Genome Wide Associations FINGERPRINTS DEVICES

& Epigenetics

PALMS Overview

Sensors for Physical Activity & Location

• • • • •

HR+M - Actiheart HR+M – Actitrainer M – Actigraph M – Actical Bioharness

GPS Globalsat Datalogger – DG-100 – BT-335

PALMS outputs

CSV files for importing into GIS, Excel, SPSS or SAS

Google Earth KML files Data sorted by user / by date and organized into folders

PALMS mapping and charting functions allow the user to quickly view raw or processed data sets before exporting.

The PALMS KML generator organizes data into folders, enabling to user to view the data by subject id, by day and by data type

KML files can be generated for individual subjects or the entire study.

Example: Activity patterns and intensity levels of 5 individuals

Example: Periods Example: Sedentary of seden of same 5 patterns individuals

When PALMS results are imported into a GIS system, the GIS spatial analysis functions can be used. For example, once an activity intensity layer is added to a land use layer….

The GIS system can isolate activity (by intensity level) occurring in public parks.

Or create density mapping showing the amount of time spent in various locations.

PALMS built-in functions Activity bout detection light & moderate counts (on bicycle)

PALMS built-in functions Location detection

Trip start Trip end In transit Momentary pause

Shows trips between locations (grey circles) where subject spend significant amounts of time.

PALMS built-in functions Indoor / outdoor detection Outdoors Indoors / In-vehicle Locations shown in gray

Shows drive to Convention Center, walk to Seaport Village for lunch, drive from Center

Cyber-Physical Systems program of the National Science Foundation

CitiSense William Griswold, Kevin Patrick, Tajana Simunic Rosing, Ingolf Krueger, Sanjoy Dasgupta, Hovav Shacham

Air Pollution Case Study •

•

•

•

158 million live in counties violating air standards – cancer in Chula Vista, CA increased 140/million residents – largely due to diesel trucks and automobiles • particulates, benzene, sulfur dioxide, formaldehyde, etc. 30% of public schools are near highways – asthma rates 50% higher there – 350,000 – 1,300,000 respiratory events in children annually 5 EPA monitors in SD Co., 4000 sq. mi., 3.1M residents – but air pollution not uniformly distributed in space or time – hourly updates to web page; annual reports in PDF form Indoor air pollution is uncharted territory – second-hand smoke is major concern; also mold, radon

37

CitiSense Seacoast Sci. Seacoast Sci.

4oz 4oz 30 compounds 30 compounds

Intel MSP Intel MSP

CitiSense CitiSense

W

ve rie ret

se e ns

contribute

L

y” la isp “d

dis r cov e

C/A

EPA

S

F

distribute CitiSense Team PI: Bill Griswold Ingolf Krueger Tajana Simunic Rosing Sanjoy Dasgupta Hovav Shacham Kevin Patrick

CYberinfrastructure for COmparative effectiveness Research

CYCORE ARRA Grand Opportunity Grant, 2009-2011, NCI Partnership between MD Anderson Hospital and Calit2/CWPHS Plan is to “…design, and validate within a community of cancer investigators and patients, a prototype cyberinfrastructure that supports the acquisition, storage, visualization, analysis and sharing of clinical, genetic, physiologic and behavioral data for cancer-related trials.” -- multiple data sources -- two-way communication between care providers, patients and families -- use of novel brain-based devices for data analysis -- set the stage for longer-term investment and R&D

CYCORE Systems Architecture

Web UI for cancer researchers conducting comparative effectiveness research

Data from wearable sensors - behavioral - physiological - self-assessment

Home health hubs

Router/Interceptor/messenger Policy

Authentication

Encryption Logging

Repositories for tools for data visualization, integration analysis and export

Other Data Sources -- EMR (clinic, hospital) -- Research -- Public Health -- CaBIG

So how do we get there?

Wireless Persuasion

• Medical care • Public health • Personal health

Wireless Technology Health Science

* Persuasive Design

• Mobile phone apps • Body area networks • Wearable sensors • Ecosystem of external sensors • SMS/MMS • Server analytics

• Tracking • Goals • Reminders • Rewards • Tailoring • Social Support • Preference based • Attentive • Ecological • Cybernetic

Tracking Voluntary • Event-driven – e.g. record food after every meal • Practiced art – mastery…

Prompted • Time of day – routine or tailored • Event-driven – requires context/situation awareness via sensor (e.g. heart rate, location via GPS or place-based sensor)

Learned • Pattern recognition of selected location/behavioral constructs that may be highly personalized • Rule-based – e.g. continuous thread of interaction informed by expert system Population-level • Group assessment & intervention (“air traffic control”)

Reminders Supportive of self-monitoring and goal setting Can be set up beforehand based upon user preference Programmable based upon new needs • new medication adherence schedule provided at time of Rx

Context aware for relevant behaviors • e.g. time of day and prompt for physical activity or healthy lunch

Reminders can be coupled with instructive mobile apps • Relaxation prompts – visualization, physical break, thought task

Learning systems – reminders can remember how to remind

Attentive Proactive – knows when getting your attention is most relevant (e.g. when you are being exposed to something) Peripheral – shouldn’t divert attention during critical tasks (e.g. driving) Unobtrusive – shouldn’t cause social problems (e.g. annoying ring tones) Rich – multiple ways to get attention and inform (e.g. don’t have to get phone out of your pocket or purse) Adaptive – changes according to your needs and patterns of use Redundant – if you are busy, miss a notification or don’t understand it comes back later and/or finds another way to reach you

Cybernetic Sensor data are used in a feed-back system to change something: -- Dose of medication via auto-delivery systems -- Cognitions - thoughts, awareness, decision-critical information -- Feelings - enjoyment, reward, concern -- Behaviors - movement, diet, adherence -- Interactions with others - social contacts -- Environment – interactions with objects or places in the environment to promote healthy behaviors or deter unhealthy ones

“The future is already here. It’s just not evenly distributed.”

William Gibson, Science Fiction Writer

Mobile devices and gadgets Heart rate monitors Pedometers GPS devices – topo maps for hiking, biking, etc. All fit the definition of tools that support tailored behavior assessment and feedback and, in most instances fun and games…

Even the iPod…

cwphs.ucsd.edu