• Big Data, Consumer Technology, and the Obesity Epidemic

    Emerging Science and Ethical Considerations

    Big Data, Consumer Technology, and the Obesity Epidemic

    Emerging Science and Ethical Considerations

    Organizers: Jason Block (Harvard Medical School), Matthew Harding (University of California Irvine), John G. Kral (SUNY Downstate Medical Center), Kristina H. Lewis (Wake Forest School of Medicine), Satchidananda Panda (SALK Institute), Gilles Bergeron (New York Academy of Sciences), and Mireille Mclean (New York Academy of Sciences)Presented by: The Sackler Institute for Nutrition Science and the New York Academy of Sciences
    Reported by Elyssa Bernfeld | Posted January 26, 2017


    From smart phone applications to electronic medical records, sophisticated means for tracking our health have never been more accessible. But with all this new technology, comes new challenges—not the least of which being researchers and physicians confronted with the daunting task of processing an unprecedented trove of new information. On September 16, 2016, The Sackler Institute for Nutrition Science presented Big Data, Consumer Technology, and the Obesity Epidemic: Emerging Science and Ethical Considerations to address the uses, benefits, difficulties, and ethical implications associated with Big Data.

    The day's first session focused on three innovative uses of Big Data. Electronic health records (EHR) document millions of events each day, representing novel opportunities to cull new observations in realms such as disease biology and pharmacology. Nicholas Tatonetti of Columbia University uses data mining, to better understand and predict drug–drug interactions (DDI) in hopes of improving prescribing practices. DDI are responsible for up to 30% of adverse drug reactions, yet according to Tatonetti, are relatively understudied, with the FDA and clinical trials not requiring the analysis of such interactions. Instead, the FDA maintains an adverse event reporting system for healthcare professionals and patients to voluntarily report the adverse outcome experienced, the drugs taken and the diseases being treated for.

    Still, even with access to these reports, determining which of the drugs, if any, are actually causing the adverse effect is an arduous task. In response, Tatonetti developed an algorithm to infer DDI for two drugs. As proof-of-concept, Tatonetti looked for adverse events related to type 2 diabetes. Applying the algorithm, he identified two drugs frequently taken together: paroxetine, an antidepressant, and pravastatin, a cholesterol-lowering drug. Using EHR to compare the blood glucose values for patients on these two drugs, patients on both drugs had a 60 mg/dL increase in blood glucose levels compared to the control. Conversely, patients on either of these drugs individually had no change in blood glucose values. These results were validated in a mouse model: mice on both drugs had a blood glucose value 60 mg/dL higher than the control. This example demonstrates how data mining of EHR can be used to discover and validate new DDI to improve prescribing practices, and how EHR might predict experimental outcomes.

    Tooraj Mirshahi of the Geisinger Clinic discussed the use of EHR to predict the outcomes of bariatric surgery. Patients seek various forms of bariatric surgery to promote weight loss and to improve their co-morbidities, such as diabetes and cardiovascular disease (CVD). Using EHR, Mirshahi matched 1750 bariatric patients with non-surgery control patients according to BMI, age, gender, and co-morbidities. Mirshahi found that gastric bypass is effective at reducing BMI, mortality and CVD. But although many do improve post-surgery, there is a high degree of variability among patients. Using the Geisinger Clinic's data, Mirshahi set out to better understand what factors can be used to predict bariatric surgery outcomes.

    Mirshahi found that a high pre-surgery BMI was one of several conditions that were predictors of less weight loss post-surgery. However, patients with a BMI below 50 were more susceptible to the presence of four common genetic obesity markers controlling their weight loss progress, whereas patients with a BMI above 50 were unaffected by genetics. To predict the probability of diabetic remission post-surgery, Mirshahi developed a DiaRem scoring system, calculated from the patient's age, HbA1c, prescribed diabetes drugs, and insulin use. After validating in multiple cohorts, patients with a low DiaRem score correlated with a higher probability of diabetes remission. The Get 2 Goal mobile app, available for patients and providers, calculates a post-surgery DiaRem score and weight loss prediction to help motivate patients and track their progress.

    A DiaRem (diabetes remission) score can be calculated for patients undergoing bariatric surgery to predict their probability of diabetes remission post-surgery. The DiaRem score is calculated based on patient age, HbA1c, and prescribed diabetes drugs. A lower DiaRem score correlates with better probability of remission, as validated in 3 separate cohorts. (Image modified from Tooraj Mirshahi)

    In an effort to curb the obesity epidemic in the U.S., a provision under the Affordable Care Act requires chain restaurants to post calorie counts on their menus. While there is evidence for the effectiveness of calorie labels on weight loss, the reasons are not well understood. Michael K. Price of Georgia State University is using Big Data to understand this initiative's success and link the effects with social welfare.

    Price found that people from areas with local mandates already in place—like New York City—have a reduction in average BMI, even in people at healthy weight, indicating that the consumer was previously inattentive to calories and is now making more informed, healthier choices. However, people also reported a reduction in life satisfaction. Although these people end up consuming healthier foods and fewer calories, they are left with a feeling of guilt for wanting to choose the higher calorie option. Going forward, Price plans to collect more data following the federal mandate, for example, by analyzing people's restaurant visits and food choices before and after the mandate. Those feeling guilty in response to calorie postings may change where they eat to avoid restaurants that post calories. Conversely, people who want to be more informed may switch to restaurants that do post calories. Ultimately, a better understanding of how people are affected by the mandates can allow improvement in policies going forward.

    The second session focused on new implications for obesity intervention. Ruth E. Patterson of the University of California San Diego discussed maintaining a healthy weight by practicing time restricted feeding (TRF), which focuses on when subjects eat, rather than the amount of calories. Studies in rodent models have indicated that TRF improves body weight, fat mass, glucose tolerance, fatty liver, and lifespan, and can reduce the risk of cancer. However, there is very little data on the effect of TRF in humans. Using Big Data from 24-hour dietary recalls, Patterson looked at the effect of prolonged nightly fasting among women. Analysis revealed that increasing nightly fast duration from the typical 10 hours to the prolonged 13 hours reduced the risk of elevated HbA1c by almost 20%. Additionally, breast cancer survivors with a nighttime fasting of less than 13 hours per evening were 40% more likely to have more breast cancer events sooner, and about 20% more likely to die from breast cancer than survivors that followed prolonged fasting.

    Despite its apparent effectiveness, is prolonged nightly fasting a feasible public health recommendation? To test this, Patterson performed a one-month pilot study, in which 20 female participants were told to fast for at least 12 hours and were given a mobile app to record their meal time and remind them when to stop eating. The study showed modest weight loss among the women, and 90% of the participants found the fast to be easy and somewhat pleasant.

    Satchidananda Panda of the Salk Institute performed a similar feasibility pilot study with eight overweight participants with a baseline eating duration greater than 14 hours. Over 16 weeks, participants were asked to decrease their eating duration to any 10–11 hour window they choose, with no other dietary changes. TRF resulted in sustained weight loss for at least one year after intervention, and an improvement in sleep quality. There was also a 20% reduction in calorie intake, so TRF may also be an effective method of caloric restriction. Together, data presented by Patterson and Panda indicates the effectiveness and feasibility of limiting eating duration for weight management.

    One possible mechanism for the effectiveness of TRF is by syncing with the body's internal circadian clock, eating during the daylight hours when we are most active. However, Panda found that more calories are consumed at night between 6 pm – 9 pm, than during the daylight hours of 4 am – 12 pm. Circadian clocks regulate many aspects of human health and disruption of the clock through lifestyle may be the cause of many diseases, particularly those associated with obesity.

    The circadian clock regulates many different aspects of general human health. But overtime, aging, jet lag, and erratic schedules can disrupt the clock, resulting in dysregulation of these systems and ultimately, being a major contributor to human disease. (Image courtesy of Satchidananda Panda)

    Barbara E. Millen, founder and president of Millennium Prevention, Inc., discussed the importance of personalized prevention and early intervention when addressing patient health and wellness. Millennium looked at bodyweight change and obesity development in 4000 men and women over a 28-year period. These studies showed the importance of early intervention, as many women continued for a full decade in developing obesity. Additionally, more than half of the foods consumed by both men and women were nutritious, thus a complete overhaul of their current dietary patterns would not necessarily be effective. Millennium developed the HealthMain website, a platform for personalized medicine. After completing several brief surveys, evidence-based assessment tools benchmark the individual against current expert guidelines and suggest lifestyle changes that complement the healthful decisions the patient already makes. Six-month follow-up data from participants showed high engagement and impact on lifestyle changes.

    The final session of the day discussed the use of Big Data in public health, with a discussion of equity and ethical considerations. David S. Siscovick of the New York Academy of Medicine, discussed Health Data for New York City (HD4NYC), a platform using data sharing and collaboration to improve public health research by connecting researchers and datasets from the Department of Health and Mental Hygiene (DOHMH) with other institutions. Charon Gwynn of the DOHMH discussed the use of Big Data to implement and promote public health initiatives, such as Take Care New York (TCNY) 2020. Using baseline numbers for obesity, sugary drink consumption and physical activity, they set goals to curb obesity in all New Yorkers and priority populations in NYC by 2020. The DOHMH has outlined projects towards their 2020 goals including collaboration with Department of Transportation to identify additional areas for bike lanes in Brooklyn and East New York, and an initiative to bring food trucks with fresh produce to areas with limited access to healthy foods. Gary Bennett of Duke University advocated the use of digital approaches to improve equity and narrow racial disparities in obesity, since medically vulnerable populations are now tuned into mobile devices.

    Take Care New York (TCNY) 2020 is a program instituted by the New York City DOHMH to improve health and equity in New York City. Baseline numbers have been calculated using datasets collected by DOHMH. DOHMH has set a goal to decrease obesity and sugary drink consumption, and increase physical activity both citywide, as well as in at-risk populations. (Image courtesy of David S. Siscovick and Charon Gwynn)

    Lori Andrews of Chicago–Kent College of Law discussed ethical considerations surrounding Big Data. Much of the health information we share through email, social media, and medical apps is collected by data aggregators and sold to employers, insurers, and advertisers. For example, a teenager discussing in an online chat forum their plans for suicide using a certain chemical may result in being targeted with ads for the same chemical.

    Andrews analyzed 211 diabetes management apps and found that only 19% had privacy policies and 77% shared consumer information with data aggregators, including ones that sell to insurers and pharmaceutical companies. Even worse, having a privacy policy meant nothing: apps with a privacy policy were just as likely to share personal information to aggregators. According to Karandeep Singh of University of Michigan, consumers using health and medical apps are often unaware of the lack of privacy granted by these apps. Many consumers assume they are protected by HIPAA, but most health apps are not required to follow these compliances. Andrews warned that with 7% of physicians prescribing medical apps, there is a need to assess the effects of sharing personal health information with others. Likewise, the FDA does not regulate most apps, so potentially dangerous medical situations could ensue. For example, an app may make claims to alert a patient to seek medical attention when they input a dangerously high blood glucose value, but the app may fail to actually recognize these values. Consequently, although many mobile apps may offer significant benefits and are great for research and data collection, consumers and developers need greater awareness of the security and safety issues that may exist.

    Use the tabs above to find multimedia from this event.

    Presentations available from:
    Satchidananda Panda, PhD (SALK Institute)
    Lori B. Andrews, JD (Chicago-Kent College of Law, Illinois Institute of Technology)
    Gary Bennett, PhD (Duke University)
    Charon Gwynn, PhD (New York City Department of Health and Mental Hygiene)
    Barbara E. Millen, DrPH, RD, FADA (Millennium Prevention, Inc.)
    Ruth E. Patterson, PhD (University of California San Diego)
    Michael K. Price, PhD (Georgia State University)
    Karandeep Singh, MD, MMSc (University of Michigan)
    David Siscovick, MD, MPH (The New York Academy of Medicine)
    Nicholas Tatonetti, PhD (Columbia University)

    Presented by

    • The Sackler Institute for Nutrition Science
    • a program of the New York Academy of Sciences

    How to cite this eBriefing

    The New York Academy of Sciences. Big Data, Consumer Technology, and the Obesity Epidemic: Emerging Science and Ethical Considerations. Academy eBriefings. 2016. Available at: