1 Big Data
Names Michael Willumsen Eficode Sales Executive, Big Data Jesper Karup Eficode Lead Architect, Information Management Mek Nielsen Agfa HealthCare BU IT Nordic Manager HE/Sales Nordics 2
What is Big Data? Big data is a broad term for data sets so large or complex that traditional data processing applications are inadequate. - Wikipidia 3
Leveraging healthcare s Big Data opportunity Revenue management Claims EMRs ICD 9-10 Meaningful use Lab/radiology notes P4P reporting Quality reporting Clinical quality measures Transcription Population health mgmt Billions of daily interactions Millions of daily transactions Video conferences Downloads Call notes SMS Web chat Blogs Social networks Mobile apps Sensors Survey response Emails Enterprise information that comes from line of business systems that provide structured database information that is used to run the business & Global information that comes from internal and external unstructured sources that is used to gain insight on the business drivers 4
Big Data Challenges? Disperse data source locations Inaccessibility Relevance Hidden unknowns Multitude of data types Text Images Audio how to use all this information is a big challenge 5
6 Challenges in clinical data A large portion of clinical records are unstructured, free-text. Data is typically in multiple systems and not normalized. Unstructured clinical data not being leveraged optimally at point of care or for research. Free-text is not standardized nomenclature - users have individual styles and terminology preferences. Clinicians are insulated from their data and must depend on Analysts for access. The processing and retrieval of human information must be automated.
An Industry Facing Substantial Challenges 20% 75% 80% 20%-40% Estimated increase in healthcare spend worldwide by 2015. Worldwide per capita healthcare spending is outpacing per capita income. (Source: Frost & Sullivan, Health Spending Projections Thru 2015: Changes on the Horizon) Amount of healthcare budget spent on managing chronic conditions. (Source: Centers for Disease Control) Of chronic diseases (premature heart disease, stroke, diabetes, etc.) can be prevented. (Source: The World Health Organization, Global Report on Noncommunicable Diseases, 2010) WHO estimate of expenses wasted due to inefficiency in the systems (Source: The World Health Organization, 2010 World Health Report: Health Systems Financing the path to universal coverage) Around the globe, countries are reforming outcomes-based reimbursement compliance and security (ICD-10 and 5010) patient centricity & wellness population health preparedness efficiency & effectiveness improvement care network integration new partnership models (ACOs) incoming participants 7
What makes a good professional? Experience Intuition Knowledge Empathy Observant Deductive Conclusive Updated Predictive 8
Identifying patterns in Big Data Singular data sources Radiology systems RIS/PACS/VNA EHR Opusplatform/EPIC Multiple data sources Longitudinal patient data history Optimized data patterns Increased specialty areas IDOL algorithm proven technology Low footprint with high efficiency Language agnostic Self learning technology 9
What if? Access to data was not an issue Patterns automatically could be detected in the unstructured data Hidden data could be used in the analysis Enormous amounts of historical data could be processed on-the-fly Time and data was not an issue Research could include much more disperse data Optimized evidence based patient treatment = better cost efficiency 10
Big data Analysis How? P θ x) = P x θ P(θ) θ εθ P x θ P(θ ) Bayesian inference Thomas Bayes 1701-1761 11
IDOL (Intelligent Data Operating Layer) Started as a research project at Cambridge University in the 1970 ies This research project aimed at identifying fingerprints using pattern recognition Pattern recognition from fingerprint analysis was later used for pattern recognition in unstructured written data A conceptual interpretation of written information based on pattern recognition, supports functionalities like automatic catagorization and grouping of information in clusters The altorithms behind the pattern recognition are based on Bayesisk Inferens and Shannons information theory 12
IDOL unstructured pattern algorithm Data information clustering: Autonomy can show patterns in data to allow business users to find actionable insight
Where to apply? Visitation-robot Decision support Automatization Individual cancer treatment Qualified treatment options Evidence based choice Research projects Cohort data volumes Pattern identification Chronical disease early discovery 14
Where to apply Limitations? 15