A digital self-coaching program for obesity and diabetes with deep learning capability offered in a financial incentive-based model.
Significance: Diabetes Is A Preventable Global Health And Economic Challenge. By 2025 – in less than 10 years - the number of US adults with diabetes will exceed 50 million.1 The annual cost of diabetes alone in the US will soon surpass the total yearly healthcare expenditure of $3 trillion. Similar trends are emerging in other countries. This pandemic results from obesity or excess body fat. The Diabetes Prevention Program (DPP) proved that just 7% weight loss reduces the risk of diabetes by over 50% and is cost-effective over a ten-year period.2 However, implementation and dissemination of DPP-like programs have not been successful. Individualized DPP-like programs such as Medicare’s “Intensive Behavioral Therapy” are not scalable due to cost, shortage of healthcare professionals and delivery sites. Population-based approaches emphasizing education and group sessions lack customization and suffer from poor retention rates. Emerging Internet-based coaching approaches address these challenges by offering scalable, cost-effective and yet personalized lifestyle intervention solutions. However, they are often heavy on technology and light on clinically validated programmatic protocols.
Our solution is an affordable, scalable and automated digital self-coaching program for use anytime anywhere eliminating need for healthcare staff and overhead costs.
Specific Aim 1: Pilot a digital coaching platform using clinically validated algorithms for individualized weight loss and daily interactive quizzes called Health Lifecards to engage and support end users for long-term lifestyle change.
Specific Aim 2: Using a six-month randomized trial prove that participation in our automated digital self-coaching program results in significant weight loss and diabetes risk reduction comparable to DPP.
Specific Aim 3: Implement and disseminate our approach in an exponential financial incentive-based business model targeting end users: health consumers, employees and insurance members.
What are the key outcomes and impact of your solution?
Millions of people will have affordable access to a mobile, clinically validated digital platform to lose weight and reduce diabetes risk.
Outcomes to be reached in three phases:
In the pilot study the main outcome is feasibility. Parameters and metrics that will be studied will focus on potential technical, logistical and cost related problems. Mediators to efficacy (for use in phase II) such as health literacy, self-efficacy (goal setting and adherence), chronic stress, cultural preferences and socioeconomic status will be studied to better evaluate indicators of success and failure.
Weight Loss – In most DPP-like trials significant weight loss is measured as the number of subjects with >3%, >5% and > 7% of initial weight loss. At our center, the % weight loss at 6-months has been typically 12%. This study aims to reach weight loss efficacy with target 5% weigh loss at six months. Additional weight loss data such as maintenance at 6 months will be studied. In order to reach significance and not to underestimate the number of subjects the power of our study will be determined to reach significance in the secondary outcome HbA1c.
HbA1c - Normal values of HbA1c are less than 5.6%. 5.7 to 5.9% range is considered borderline or pre-diabetic and above 6.0% is considered diabetic. Most studies consider a change of 0.5% to be significant. We set mean treatment effect for each arm as the change from baseline to six months: the intervention arm = 1.0% reduction, and for the control arm = 0.5% possible drop. The mean treatment effect difference between arms is Sig = 1.0 – 0.5 = 0.5%. This is a clinically meaningful difference. It is also assumed that SD (baseline) = SD (24 weeks) = SD. A two sample t-test for the null hypothesis: Sig = 0 versus alternative hypothesis: Sig not = 0 requires 112 subjects per arm to reject the null hypothesis when Sig = 0.5, power = or > 0.80 and two-sided alpha = or < 0.025. Attrition rate is expected to be similar to other DPP-like trials including our own experience at 20%. Therefore a total number of 280 participants will be enrolled over a 12-month enrollment period.
In addition to the above metrics the following will be tracked:
· Body Fat %
· Body Fat Distribution
· Muscle Mass
· Nutrition intake including macronutrients
· Exercise type and minutes (and estimated METs)
· Quality of life indicators
· Stress level
· Prognosis score
· Health and wellness assessment scores
This phase corresponds to the rapid growth of end users. The focus will change to optimizing and data mining of responses to lifecards and consequently enhancing our automated self-coaching algorithms. Quizzes and answers will be tagged to provide sufficient granular data for machine learning purposes.
What actions do you propose to realize your stated goals?
Over the last two decades, Dr. Yavari and his team at Beyond Care have developed clinically validated algorithms and programs that are central to this proposal:
1- Metabolic Rehab App – partly published in peer reviewed publications and used daily in clinical settings and so far used for data mining of 14000 employees of a healthcare company and a database of over 20000 subjects.3
2- Beyond Weight Metabolic Rehab Program – used for the last 15 years by Dr. Yavari and his colleagues in various clinical settings and partly published in Dr. Yavari’s book It Must Be My Metabolism!
3- Digital Self-Coaching Health Lifecards – 365 + unexpected situations taken from daily experiences of Dr. Yavari’s patients have been turned into quizzes to engage, educate, support and coach end users.
In the following our tools are described for participants’ use. Initially, our strategy is to implement our digital self-coaching program and to prove feasibility using a limited number of volunteers. Next, efficacy of our digital self-coaching approach will be validated compared to DPP benchmarks. 2 Finally, our business model will be implemented as an incentive-based model for large numbers of users.
STUDY DESIGN & METHODS
1. Innovative Assessment Tools: Metabolic Rehab App (MRA)
Non-invasive Biometric Body Composition App Determines Disease Risk, Sets Personalized Objectives And Clinically Sound Goals.
Using six biometric parameters, age, gender, weight, height, waistline and hipline, our proprietary algorithm “Metabolic Rehab App or MRA”, measures body composition compartments such as total fat, abdominal fat and muscle mass accurately and comparable to total body x-ray scans (Fig.1.,2.) 3 It immediately sets weight loss goal, total muscle mass targets, timeline, recommends daily caloric intake and exercise regimen. Our algorithm also determines the risk of diabetes as accurately as the blood test HbA1c. 3 During the program MRA provides the following information:
- Diabetes risk
- Goal weight
- Estimated caloric burn rate
- Body composition measures such as body fat %, abdominal fat and muscle mass
- Predicted weight loss rate “curve”
Our app predicts an average weight loss rate as a curve, which can be used for monitoring, intervention and quality control. Participants are coached according to their success: those exceeding expectations progress below the average predicted rate (faster); those lagging will be above the curve (slower.) Finally, if an unexpected change happens, specific measures are adopted to change course (Fig.3.,4.)
Fig. 2. Fat, Fat Free (muscle), Trunk & Weight Predicted Vs. Measured
2. Beyond Weight Metabolic Rehab : An Innovative Therapeutic Lifestyle Change (TLC) Intervention
TLC interventions implemented in DPP and Look Ahead trials contained 3 components: nutrition intervention, exercise and behavior modification counseling. 2, 4 Our program also includes those 3 components. Prior to intervention pertinent current and historical medical and lifestyle related data are collected and carried through the duration of the program.
a. Screens and Data Collection:
· Weight History
· Stress Profile
· Depression Screen
· Quality of Life
· Eating Pattern
· Exercise History
· Motivators and Obstacles
· Medical Assessment
b. Nutrition Intervention - Low-fat, low-carb 1000–1800 calorie meal plans for home use. Guidance for eating prepared foods, fast foods, “take outs” and restaurant choices will be provided. Online free nutrition tracking apps such as myfitnesspal.com will be used. Educational quizzes focused on nutrition will be used to engage and educate users.
c. Exercise Training - Anywhere, Anytime Exercise
Customized exercise guidelines are provided based on body composition, fitness level and medical risks. For beginners, exercise begins with moderate aerobics and calisthenics. Later free weights and resistance (e.g. tubing) training and higher intensity more complex regimen will be provided.
d. Behavior Modification Coaching – Using didactic lessons, coaching tools such as Health Lifecards and other supporting materials, behavioral coaching intervention will address:
· Problem Solving
· Time Management
· Setting Boundaries
· Organizational Skills
· Rewards Systems
· Fun and Joy
e. Didactic Teaching Lessons and Surveys focused on:
· Stress Reduction
· Lifestyle Change
3. Digital Self-Coaching Using Health Lifecards:
Lifecards are proprietary coaching quizzes, which require picking the “best answer” among four choices. Answering the quiz will yield a score of 1 to 4 depending on the answer chosen (no penalties.) Lifecards are mostly situational: a hypothetical unexpected challenge has to be addressed and as a result users learn time management, problem solving and organizational skills, etc. Some lifecards are psychological questions helping users commit, engage and sustain efforts. Others are more practical addressing everyday obstacles, surprises or setbacks. Lifecards are sent to users by text and email in an order defined by the phase of the program. For example, some lifecards are sent early on for goal setting and commitment while others are sent later to enhance self-efficacy and prevent relapse.
Here is an example of “problem solving”:
Your refrigerator breaks just as you are starting a new weight loss diet.
- That is ok. Use bars and shakes. They do not need to be refrigerated.
- Buy canned or prepackaged “TV Dinners” and microwave them.
- Your new fridge will not arrive anytime soon; you have to eat out.
- Buy or borrow a mini fridge and use it till your new one arrives.
1. Reach - Using our own medical office and affiliated doctors’, 30 obese patients with diabetes will be recruited as volunteers to enroll in a 12-week program over a 3-month period.
2. Feasibility – Potential technical and logistical problems will be resolved. Mediators to efficacy such as health literacy, self-efficacy (goal setting and adherence), chronic stress, cultural preferences and socioeconomic status will be studied to better evaluate indicators of success and failure. Findings will be used for implementation in Phase II.
1. Efficacy- In DPP, diabetes prevention was the primary outcome. In most other DPP-like trials, weight loss has been the primary outcome.5 In our study our primary outcome is weight loss using our digital automated coaching intervention as compared to a traditional office-based DPP-like intervention. Our secondary outcome is diabetes risk reduction. Based on previously published DPP-like trials and our own clinical experience, we conservatively estimate an attrition rate of 20%. This projected attrition rate accounts for all dropouts and participants who will be excluded due to medical events such as surgery, disability, accidents or mental illness.
2. Research Design: Control and Intervention Arms
All clinically sound weight loss interventions resulting in a significant weight loss (5% or more) are at least 3 to 6 months long. We propose a prospective, six-month randomized study. The control arm will have access to our MRA app for screening, weight loss and time line targets, meal plans and didactic lessons. Control subjects will be reassessed as the intervention group at 3 and 6 months.
The intervention arm will in addition have access to our real-time automated digital self-coaching tool, Health Lifecards. Users will be monitored with logins and self-reporting. Compliance is monitored and users will be contacted if they fail to self-report. At the end of 6 months, participants return to enrollment sites for weight and body composition assessment and HbA1c for diabetes control. Participants’ weekly weights are plotted against their predicated weight loss curve presented in the Metabolic Rehab App. Each month, participants are to report waistline and hipline measurements to update body composition and calculate new diabetes risk score. HbA1c are tested every 3 months.
3. Recruitment, Enrollment and Sample Size
Participants are enrolled at 2 geographically diverse sites. Others and we have showed that % weight loss is independent of age, gender and likely race and depends predominantly on initial total body musculature. 6,7 Our study will recruit randomly amongst midlife women and men ages 40 to 70. A clinician trained to register participants and obtain informed consent will randomize subjects into study and control arms. They will be recruited over 6 months and the study duration will be one year.
A overview of DPP-like trials showed that most studies reached significant outcomes for weight loss of 5% or more in spite of small numbers of participants - ranging from less than a hundred to few hundreds.5 (At our center, the % weight loss at 6-months has been typically 12%.) To have a more precise power assessment we determine the number of participants based on our secondary outcome, which is reduction in HbA1c.
The calculation of sample size is for a 2-by-2 repeated measures design consisting of two arms, each subject measured at two time points (baseline, 24 weeks). Considering a conservative attrition rate consistent with DPP-like trials, a total of 280 users will be enrolled over a 12-month enrollment period (see “Key Outcomes…”)
PHASE III: Exponential Growth and Incentive-Based Model
Diabetes is responsible for a large part of the cost of heart disease, dialysis, sleep disorders, etc. On the average an average diabetic patient costs $200K - a conservative lifetime estimate.9 This cost is revenue for businesses involved in diabetes and is mostly paid for by patients with diabetes for a promise of future quality-life years. We believe patients should also benefit from this revenue and be financially rewarded for their efforts to get well. Phase III proposes an incentive-based model where users are paid a monthly cash incentive ($20) to lose weight and reduce diabetes risk. This incentive-based business model is supported by the following trends in employee benefits plans:
• 20% of employee population accounts for 80% of medical costs.
• Identification of “near-future disease” employees, with incentives and intervention can reduce future disease burden costs.
• To engage employees and motivate them to play an active role in their health 73% of midsize to large employers use incentives - mostly monetary. 10
• The average annual employee incentive is $460/year; and use of monetary incentives is growing: cash gift cards, premium discounts, etc.
• Companies offering remote coaching platforms or automated digital coaching, reduce cost enough to fit in employee benefits budgets while traditional “brick and mortar” models remain cost prohibitive. 11, 12
• Remote coaching platforms operate on a monthly subscription of $100 or less (often reimbursed by 3rd party payers such as insurance plans.) While automated digital self-coaching tools are either free or less than $10 per month.
To the end user, we offer a mobile, low-cost self-coaching program accessible through a user-specific website. To employers and payers we offer cost effective measures with scalable national and international outreach. Finally, better health for employees means savings for insurance companies.
The increasing number of digital health start-ups getting funding or being acquired reflects interest in this rapidly growing sector. 13 Since businesses involved in employee benefits and insurance plans face lengthy sales cycles, we also propose a parallel consumer-driven business model. This business model is based on free access to our apps and a monthly subscription fee of $5. Here revenues are not generated as much from fees as they are from marketing and data mining. This model will be developed in parallel to the above employee-driven model and is expected to have exponential growth.14
Who will take these actions?
This proposal is based on Dr. Yavari’s intellectual property originating over the last 15 years from his clinical practice, Beyond Care LLC, in Madison, Connecticut. The apps, situational coaching quizzes called Health Lifecards, clinical algorithms, the lifestyle change program Beyond Weight Metabolic Rehab, and educational contents used in this proposal have been created and completed at Beyond Care.
Dr. Yavari and his staff at Beyond Care will be responsible for roll-out, testing and analysis of the prototype in Phase I (feasibility.)
During Phase II (efficacy), enrollment of subjects, follow-up, coordination, analysis and fine-tuning will be implemented at Beyond Care in Connecticut as well as at the MedScience Research office in Florida (see below.) It is understood that participants will be recruited and enrolled at other locations in Florida and New York as well but the Beyond Care office will the principal clinical hub during Phases II. Data analysis and biostatistics will be done under the supervision of Michael Brines MD, PhD at Beyond Care.
Dr. Yavari has partnered with Troy Grogan, President of MedScience Research Group in West Palm, Florida for business development:
Mr. Grogan’s role is to recruit physicians, physician organizations, health insurance companies and large employers. MedScience has over 150 primary care physician clients in South Florida and New York. Jesus Davila, Clinical Coordinator and Elizabeth Magdaleno, Customer Service at MedScience Research will assist with physician recruitment and participant onboarding and support.
Ray Scofield, Software Engineer, at Viper Software Solutions in Port Saint Lucie, Florida is currently in charge of development of the mobile application, database and content management system of the prototype.
During Phase III, additional staff will be recruited for sales and marketing targeting large employers, insurance and healthcare companies. We will also partner with established media and marketing companies for expanded outreach to consumers in diverse geographies.
The Obesity Pandemic
Adult onset diabetes or type 2 is a global health and economic challenge related to the obesity pandemic, genetic predisposition and lifestyle choices. It is estimated that in less than 10 years that is by 2025, global obesity (BMI= or >30) prevalence will reach 18% in men and surpass 21% in women; severe obesity (BMI= or >35) will surpass 6% in men and 9% in women. The following is extracted from the landmark study of NCD Risk Factor Collaboration.15
Over the past four decades, there has been a dramatic change of body weight worldwide: the prevalence of underweight was more than double that of obesity in 1975 but now more people are obese than underweight. During the same period obesity among men has tripled and doubled in women. Today there are more than 266 million obese men and 375 million obese women worldwide. Although obesity is a global pandemic, more obese men and women live in China and the US than any other country. USA ranks number one in prevalence of severe obesity (BMI= or > 35): more than one in four severely obese men and almost one in five severely obese women in the world live in the USA. Overall obesity in absolute numbers (including severe obesity) is most prevalent in the following countries (+/- in descending order):
Fig. 1. Diabetes in Women 2025 (NCD Risk Factor Collaboration)
Fig. 2. Diabetes in Men 2025 (NCD Risk Factor Collaboration)
Diabetes Worldwide and USA
Worldwide and in the USA, the geographic distribution of obesity and the prevalence of diabetes overlap but are somewhat divergent. There are many reasons for this divergence but it is likely that the main causes of the disproportionate rise of diabetes in the Western Pacific and in particular South East Asia as well as in Southern and Midwest USA is a combination of factors related socio-economic status, education level and access to care.
The following is extracted from data provided from IDF and ADA and helps elucidate this conclusion. 16,17
Diabetes Facts Worldwide:
- Worldwide: In 2015, 415 million people with diabetes; 2040, 642 million
- Number of men with diabetes: In 2015, 199.5 million; 2040, 313.3 million
- Number of women with diabetes: In 2015, 269.7 million; 2040, 477.9 million
- Cost of diabetes: $673 billion worldwide in 2015
- 12% of global health expenditure was spent on diabetes in 2015
- 3/4 of people with diabetes live in low and middle income countries
The following is the spread of diabetes in various regions of the world in descending order:
1- Western Pacific: 2015, 153.2 million; 2040, 214.8 million
2- South East Asia: 2015, 78.3 million; 2040, 140.2 million
3- Europe: 2015, 59.8 million; 2040, 71.1 million
4- North America and Caribbean: 2015, 44.3 million; 2040, 60.5 million
5- Middle East and North Africa: 2015, 35.4 million; 2040, 72.1 million
6- South and Central America: 2015, 29.6 million; 2040, 48.8 million
7- Africa: 2015, 14.2 million; 2040, 34.2 million
Prevalence: In 2012, 29.1 million Americans, or 9.3% of the population, had diabetes. Undiagnosed: Of the 29.1 million, 21.0 million were diagnosed, and 8.1 million were undiagnosed. These numbers are expected to double by 2025.
Pre-diabetes: In 2012, 86 million Americans age 20 and older had pre-diabetes; this is up from 79 million in 2010.
Prevalence in seniors: The percentage of Americans age 65 and older remains high, at 25.9%, or 11.8 million seniors (diagnosed and undiagnosed). These numbers are expected to double by 2015.
Diabetes by Race/Ethnicity
The rates of diagnosed diabetes by race/ethnic background in descending order:
1- American Indians / Alaskan Natives: 15.9%
2- Non-Hispanic Blacks: 13.2%
3- Filipinos: 11.3%
4- Hispanics: 12.8%
5- Asian American: 9.0%
6- Non-Hispanic Whites: 7.6%
Fig. 3-5 USA Diabetes & Demographics 2015
Non-Hispanic White Adults
Non-Hispanic Black Adults
In conclusion, while obesity is a pandemic affecting hundreds of millions of men and women of all demographics and in all geographies, diabetes appears to be growing at a higher rate in regions with predominance of low-income and non-white populations both in the USA and worldwide. In these populations, lack of access to trained healthcare professionals or lifestyle coaches will not help reverse the trend of diabetes and its complications. Our solution offers a low-cost, scalable and clinically validated digital platform, which will empower individuals to self-coach to lose weight and reduce diabetes risk in spite of geographic and socio-economic constraints.
What do you expect are the costs associated with piloting and implementing the solution, and what is your business model?
Our solution is a for-profit venture with a B-to-C model taking advantage of millions of adults at risk of diabetes and its complications. However, we also offer solutions for employers and insurance companies. In our incentive-based model users’ benefits accounts are used for incentive payments. Cost projections are divided into the previously defined three phases of our business development:
Phase I: Pilot Study of Our Prototype – 6 months
This pilot phase entails enrolling 30 volunteers over a 3-month period in a 3-month version of the Beyond Weight Metabolic Rehab program. Subjects will be recruited through our medical office and other affiliated doctors’ offices.
· Cost of IT / Software upgrades, modification and support $75,000
· Cost of Dr. Yavari’s 20% FTE time for 6 months is $25,000
· Cost of part-time coordination and operations staff is $25,000
· Office misc. cost $ 5000
· Subtotal Phase I: $130,000
Phase II: Efficacy Validation and Implementation – 12 months, 280 users
· Study cost estimate (over 1 year) $552,960
· Advertisement and recruitment $50,000
· Cost of Dr. Yavari’s 20% FTE time for 12 months is $50,000
· Cost of part-time coordination and operations staff is $80,000
· Cost of IT / Software upgrades, modification and support $200,000
· Office misc. cost $ 10,000
· Statistical analysis $20,000
· Subtotal Phase II: $962,960
Phase III: Business Development, Dissemination and Exponential Growth – 2 to 5 years.
Dr. Yavari, his team at Beyond Care and his business partners have developed an array of clinical tools, software applications, assessments as well as over 365 quizzes to be used for digital self-coaching in our weight loss and diabetes risk reduction program.
These proprietary tools have been clinically tested, partially published in peer reviewed publications and are close to be assembled into a complete digital platform.
The time line for our solution to be fully operational is as follows:
· Months 1 to 6: Testing and “debugging” the prototype as part of a feasibility study of 30 volunteers to be enrolled in a 3-month version of the program over a 3 month period.
· Months 7 to 18: Implementation of an upgraded version to be used for validation of efficacy and data analysis of close to 300 participants randomly enrolled over 6 months in a controlled study of 6 months duration.
· Launch of the commercially viable digital coaching Beyond Weight Metabolic Rehab in year 2.
· Year 2 and after, business development, sales and marketing with focus on consumer-driven and employee subscription models. Our self-coaching program will be translated to various languages (beginning with Spanish) and adapted to diverse cultures.
As of 1/17/16, few of the submitted proposals focus on diabetes and obesity or their complications. Our proposal includes a self-coaching lifestyle change program, which would naturally benefit from innovative nutrition and exercise digital tools and public outreach approaches such as the following:
As long as proposals offer affordable, scalable and mobile access to healthcare providers or software apps, they could complement ours:
Finally, proposals offering data mining and machine learning of health data benefit ours:
1- Rowley W R, Bezold C. Creating public awareness: state 2025 diabetes forecasts. Popul Health Manag. 2012: 15(4) 194-200; Diabetes & Obesty 2025. Institute for Alternative Futures. www.altfutures.org/pubs/health/ Diabetes_Scenarios_June_1st.pdf
2- Diabetes Prevention Program Research Group, Knowler WC, Fowler SE, Hamman RF, Christophi CA, Hoffman HJ, Brenneman AT, Brown-Friday JO, Goldberg R, Venditti E, Nathan DM. 10-year follow-up of diabetes incidence and weight loss in the Diabetes Prevention Program Outcomes Study. Lancet. 2009;374(9702):1677-86.
3- Yavari R, McEntee E, McEntee M, Brines M. Anthropometric variables accurately predict dual energy x-rayabsorptiometric-derived body composition and can be used to screen for diabetes. PLoS ONE 2011;6(9):e24017.
4- Reduction in weight and cardiovascular disease risk factors in individuals with type 2 diabetes: one-year results of the look AHEAD trial. Look AHEAD Research Group Diabetes Care. 2007
5- Whittemore R. A systematic review of the translational research on the Diabetes Prevention Program. Transl Behav Med Sep 2011;1(3): 480–491.
7- Yavari R and Henderson K. Lean Mass Index (LMI): A new body composition index with strong predictive value for weight loss in obese men and women.
8- Phelan S, et al. Prevalence and predictors of weight-loss maintenance in a bi-racial cohort. Am J Prev Med 2010:39(6):546-554.
9- Zhuo et al., CDC, ADA 2010.
10- Miller S, Use of monetary incentives to promote wellness grows, SHRM, CEBS 8/14/12
11- See for example: Omadahealth.com and Zillion.com (disclaimer: Dr. Yavari is an early member and share holder of Zillion.com)
12- See for example: Lark.com and Noom.com
13- 44 digital health acquisitions from 2016; 2016 funding roundup. MobiHealthNews, Dec 12/2716.
14- Diamandis P. and Kotler S. Bold: How to Go Big, Create Wealth and Impact the World. Simon & Schuster (February 3, 2015)
How can we help people prevent, detect and manage chronic diseases, especially in resources-limited settings?