Big Data, Small Devices: How Technology Is Advancing Public Health

Key Trends and Data Points

• Global Internet-connected devices have exploded to 50 billion, with projections of 7 devices per person by 2020, surpassing the trajectory of Moore's Law.
• The Human Genome Project (completed 2003) transitioned from megabytes to the zettabyte scale of data storage, with the Cancer Genome Atlas already generating petabytes of data migrated to the cloud.
• Nations are accelerating genomic sequencing plans: England and Saudi Arabia target 100,000 genomes, while China and the U.S. aim for 1 million, potentially reaching 25% of the world's population sequenced by 2025.
• Individual lifetime data generation is estimated at 300 million books per person, driven by a mix of 10% clinical data, 20% genomic data, 30% social/environmental data, and 40% behavioral data from wearables.
• Healthcare spending in the U.S. totals $3 trillion annually, with $1 trillion allocated to inpatient care, yet Electronic Health Record (EHR) data is described as the "most expensive data collected per byte" due to physician typing costs.
• A single study with University of Indiana employees using Fitbits resulted in a 60% reduction in A1C accounts for diabetics and a 40% reduction in BMI among participants.
• Global cancer cases are projected to rise from 14.5 million this year to 22 million within 15 years, described as a "pandemic of our time."
• Current data translation from research to practice takes an average of 17 years, a timeline IBM's Watson system aims to reduce to 17 seconds.

Decisions, Strategies, and Future Actions

• Data Integration Models: The panel distinguishes between federated data (keeping data within institutions while allowing access) and centralized data (moving all data to one location), with APIs serving as bridges for proprietary data.
• UC Health Initiative: The University of California is committing to providing all 14.1 million patient records back to patients via an integrated "Blue and Gold Button" system to replace fragmented hospital-specific portals.
• NQF Measure Incubator: The National Quality Forum is launching an agile "measure incubator" to rapidly develop and iterate health metrics by bringing together data experts, funders, and stakeholders.
• IBM's "Three Ps": IBM's strategic vision for public health focuses on Predict, Personalize, and Prevent, utilizing Watson to process structured and unstructured data from multiple silos.
• Public Health Modernization: Atul Butte advocates for modernizing public health to focus on population-level prevention rather than just individual care, noting that precision medicine must be viewed as a public health tool.
• Behavioral Intervention: Fitbit aims to use social engineering and real-time data to drive behavior change, such as prompting social interaction to increase exercise hours, rather than focusing solely on data volume.
• Equity Focus: Helen Burstyn emphasizes the need to address the digital divide, ensuring data collection methods (like EHR portals requiring desktops) do not exclude low-income populations who rely solely on handheld devices.

Disagreements and Challenges

• Business Case Viability: Atul Butte argues that the business case for using data to improve public health is not effectively made, hindering data sharing; conversely, Eric Friedman notes that consumers increasingly demand data sharing, provided robust privacy controls exist.
• Data Usability vs. Collection: A significant friction point exists where massive amounts of EHR data are collected but rarely analyzed or acted upon, with only the average byte of data being viewed more than once if a patient returns to the hospital.
• Data Veracity: The panel highlights issues with data accuracy in EHRs, where conflicting references to the same condition (e.g., diabetes diagnosis) occur, requiring patient empowerment to validate data truthfulness.
• Adoption Rates: Wearable retention is a challenge, with initial studies showing 50% of users stop using devices within six months; Fitbit counters with internal data suggesting better retention, emphasizing the need for interoperability and lifestyle integration.
• Paternalism vs. Patient Autonomy: The discussion identifies a conflict between paternalistic medical institutions that restrict data access and the reality that sick patients are often the most willing to share data to advance science.

Forward-Looking Statements

• Future of Care: Future healthcare models will shift from acute, episodic visits to continuous monitoring where wearables act as "future vital signs," triggering alerts when behavior trends (e.g., sleep drops, step counts decline) indicate emerging health risks.
• Cognitive Systems: Cognitive computing like Watson is expected to become a standard tool in every health decision, functioning similarly to how the stethoscope revolutionized physical exams 200 years ago.
• Public Health Definition: Public health will evolve to include digital phenotyping and genomic data integration, moving beyond infectious disease tracking to managing chronic conditions through early detection and prevention.
• Data Lakes: No single "massive lake" of data will serve all stakeholders; instead, a distributed ecosystem of data lakes (including Watson and other platforms) will allow data to flow based on user trust and specific institutional needs.
• Accountability: Hospital systems and providers will increasingly be held accountable for primary prevention outcomes, not just tertiary care delivery, driven by transparent data on quality and cost.
• Knowledge Translation: The gap between research and clinical practice will narrow as cognitive systems ingest literature (e.g., 80 million pages in one second) to provide real-time treatment and clinical trial matching.