JR's Lifetime Advanced Analytics

Editor's Notes

1. Any discussion about analytics in healthcare needs to start with data Audience hand-raise poll… Who has heard about Big Data?
2. One septillion…had to look it up Google, Amazon, Apple and Facebook are the pioneers of Big Data and already understand the value of it for their business
3. All this stuff is interwoven into the concept of Big Data Requires a strategy
4. The major pools of big data that exist in healthcare today are currently poorly integrated with little or no overlap Its going to take a lot of work and it’s not easy, but the benefits of integrating this data are obvious and almost untapped
5. Reporting & Monitoring – EMRs; identifying patients misusing drugs; readmission rates; patient satisfaction Trending – healthcare businesses seem to worry about year-over-year and lack of long term trending distorts the picture Routine Analytics – Time Series analysis – New patients over time; Churn; Patients with certain conditions; Office utilization analysis; Brand/generic utilization Data Mining & Evaluation – Identifying patients with negative drug/drug interactions; evaluation of clinical pathways to determine a best course of action; Identifying patients with potential diseases that have been misdiagnosed/undiagnosed; performance evaluation of integrated care programs; Understanding the determinants of satisfaction and driving KPIs/visual dashboards Predictive modeling – predicting patients likely to frequent the ER; predicting patients/members likely to leave; risk adjustment; understanding the determinants of patient satisfaction; Visualization
6. Sometimes difficult to justify since you don’t see the rewards right away and you need to make a significant investment in time, capital and human resources 80% is really 95% for initial model development
7. Axiom, unstructured data Supply of healthcare data is catching up to the demand Another interesting point is that many healthcare organizations, typically healthcare providers, don’t have data integration strategies and technology in place.  In many instances they are supporting data integration requirements using a hodgepodge of primitive technical approaches that don’t provide the ability to change those solutions around changing needs.  Worse, they don’t provide the security and the governance subsystems required to remain compliant with the changing regulations.
8. Cluster Analysis – identification of subpopulations of complex patients (multiple conditions) who may benefit from targeted care management campaigns Anomaly detection – Finding patients or providers engaging in fraudulent behavior Association mining – (Discharged, Same-Day Prescription)  PCP follow-up correlated with lower readmission rates; (Onset of T2 Diabetes, Socioeconomic=High, Personal Trainer)  rapid improvements
9. All of these modeling techniques really look to provide insight into variables that cause differences across values of the dependent variable Regression – predicting cost or utilization for a person or population based on a set of health factors; improving rates of central line infections across hospitals Logistic Regression – Frequent Flyers; Hospital readmissions for certain diseases; Members likely to leave/shop Decision trees – determining the appropriate clinical pathway for migraine sufferers; determining a person likely to respond to a marketing campaign Neural Network – determining those likely to adhere to a care plan • Clinical Simulation: simulation is mainly used to study certain diseases • Operational Simulation is mainly used for capturing, analyzing, and studying healthcare operations, service delivery, scheduling, healthcare business processes, and patient flow. • Educational Simulation is used for training and educational purposes, where virtual environments and virtual and physical objects are extensively used to depict reality
10. Neural Network – determining those likely to adhere to a care plan Simulation- • Clinical Simulation: simulation is mainly used to study certain diseases • Operational Simulation is mainly used for capturing, analyzing, and studying healthcare operations, service delivery, scheduling, healthcare business processes, and patient flow. • Educational Simulation is used for training and educational purposes, where virtual environments and virtual and physical objects are extensively used to depict reality Text/Web Analytics – unstructured data like clinical notes, survey data, social media
11. It’s not really next generation. It’s here, but just not being used extensively in healthcare Text/Web Analytics – unstructured data like clinical notes, survey data, social media Sentiment Analysis – understanding how different people understand/accept a diagnosis Visualization – Turning rows and columns of data into pictures and graphs that provide for much better holistic understanding to support faster decisions Real-time – Currently the most prevalent application for real-time health care analytics is within Clinical Decision Support (CDS) software. These programs analyze clinical information at the point of care and support health providers as they make prescriptive decisions. These real-time systems are active knowledge systems, which use two or more items of patient data to generate case-specific advice. Alerts in a pcp office when a patient has been admitted to the ER; real-time concussion data to sideline doctor’s; out of control claims for the week for a payor You can interpret customers' opinions, improve products, optimize services, streamline processes and make proactive, fact-based decisions.
12. Exchanges Understanding and adapting to the individual marketplace faster than the competition ACOs Integrating EMRs and patient registries to cost and utilization data to improve care Commercial Risk Adjustment Handling the size and complexity of data requirements and streamlining processes for submitting and receiving data ICD-10 Adjusting to new requirements and analyzing disease information in a more robust way
13. 1/HIPPA, data security breaches, liability; Access to Data (Internal and External barriers). Liability – who is responsible when an inaccurate piece of data has a negative consequence? 2/Dependent on Initial data infrastructure 3/Must invest in IT (Technology, Storage and Computing Power) but must also recognize the need for IT to work collaboratively with Analytic functions (both learning/developing new skills and competencies) to capitalize on Big Data 4/Supporting human decision making with analysis 5/Some of this is already happening and if I can think of it, its possible and likely to happen 6/ Example of Amy and Under Armour mouthpieces
14. Weighted average of active patients booked was around 74%...now it’s 89%! This equated to $82K in extra revenue for the practice
15. Improving data operations to leverage existing ‘basic’ analytics more quickly ACOs, episodes of care, leaver/stayer Isolating outliers – High cost members/groups, providers or facilities that are higher cost/lower quality Employer Group Reporting Effective data capture, improved data quality structures and data governance Defining critical fields that are value drivers for the plan and members Building clear analytical methods to evaluate expected member value/satisfaction and system performance Identification of processes that could be made more efficient through big data such as provider authorization, referrals, evaluation of claims accuracy, auto-adjudication of claims Partnering with providers, manufacturers and government to implement monitoring and intervention programs supported by analytic frameworks Identifying positive trends such as providers, groups, health conditions and patient types that have much lower than expected costs Not only looking to maximize risk adjustment revenue, but also identify people/conditions where costs are more avoidable Creating incentives for best-in-class providers Working proactively with poorly trending groups or individuals on health and wellness programs and incentives
16. Standardized and comprehensive data capture Drive continued adoption of EMR Develop a strategy to capture data from ‘Smart’ devices and integrate into a holistic view of a patient Participation in HIEs and data sharing partnerships with private institutions Reinforce the culture of information sharing Explain to patients why their information could not only help themselves but others Simplifying the technical barriers to sharing information with appropriate parties Improving technology around clinical data Designing data architecture and governance models to manage and share key clinical data Eliminating gaps in patient health histories; Longitudinal patient records Improving technology around operations Creating decision bodies with joint clinical and IT representation that are responsible for defining and prioritizing key data needs Putting data to use in analytics to improve patient care and patient risk Developing informatics talent and predictive analytic capabilities Incorporating data into decisions and pilot programs that improve the overall quality of care Focusing on outcomes based protocols that balance cost and quality
17. 1/HIPPA, data security breaches, liability; Access to Data (Internal and External barriers). Liability – who is responsible when an inaccurate piece of data has a negative consequence? 2/Dependent on Initial data infrastructure 3/Must invest in IT (Technology, Storage and Computing Power) but must also recognize the need for IT to work collaboratively with Analytic functions (both learning/developing new skills and competencies) to capitalize on Big Data 4/Supporting human decision making with analysis
18. Move away from the notion that all data needs to be perfect. Advanced analysis and statistics supports strategy. It’s not used for reporting Develop cross-functional teams that understand data – how it is structured, where it exists, research into emerging sources, legacy system expertise, data quality, etc.