Conference Day 1: Monday, October 6, 2014 • Sessions will take place in Federal
8:00-8:30am • Room: Commonwealth Hall
Registration & Networking Breakfast
8:30-8:35am
Conference Founder Remarks
Eric SiegelFounding Chair
Predictive Analytics World
Conference Chair Welcome
Gold Sponsor Presentation
Applying Predictive Analytics to Improve Healthcare Delivery and Outcomes� Lessons Learned
Predictive modeling methods have transformed practically all industries and data-driven activities providing actionable predictive and prescriptive recommendations and delivering ROI from structured, unstructured, and Big Data. However, while it is true that the core predictive modeling algorithms are mostly universally applicable across various domains, the specific best practices that must be observed to achieve accurate, validated, and fair analytic outcomes and results vary by application. When predictive models are used to inform critical and sometimes life-changing decisions for patients, considerations of model validation, transparency, privacy, and monitoring of outcomes are primary concerns. The presentation will provide a brief overview of how the application of analytics to optimize healthcare delivery and outcomes is different from workflows and best practices commonly applied in other domains, and will discuss recommendations for best practices that should be observed.
9:05-10:00am
Keynote
Big Data and Clinical Decision Support
There is global recognition that a major change in the delivery of healthcare is necessary. If the goal of healthcare is population and individual health, then we are not doing as well as we should. There are numerous studies that report that many health care interventions have no value. Many decisions in health care are not based on solid evidence. The provision of healthcare is economically wasteful and of limited clinical effectiveness for a variety of reasons.
One concept that is advocated to improve healthcare is sometimes called personalized or precision medicine. It is the idea that we should be able to make decisions that are focused on the individual patient. We know that, for any disease, a potentially large fraction of patients will not respond to standard therapy. We usually learn that only after the patient has not responded to, or has worsened, after the standard therapy has been implemented. The goal of personalized healthcare is to learn enough about the individual patient, and all the medical options available, as to make it more likely that any intervention will be of value for the specific individual.
In essence, we have to make better, more evidence-supported decisions if we are to improve healthcare. The opportunity is great because we are surrounded by an ocean of healthcare data. However, that ocean of data is also the challenge, and might be better described as a confused sea. It qualifies as "Big Data" as it is huge in volume, accumulating very quickly, in a multitude of forms and formats and plagued with uncertainty and ambiguity. Big data can overwhelm human thought process and conventional programmatic computing. Effectively using big data in healthcare requires a sophisticated array of computer resources, sometimes called Cognitive Computing, to retrieve the relevant insights from all that data, and provide the insights to the decision-makers to achieve evidencesupported personalized healthcare. The intent is to augment the ability of patients and clinicians to use evidence, not to make decisions for them.
[ Top of this page ] [ Agenda overview ]
10:00-10:30am • Room: Commonwealth Hall
Exhibits & Morning Coffee Break
Track sponsored by: 
Track: Predicting Disease and Infection
Case Study: Baptist Health
Predicting the Invisible Patient: Using Predictive Analytics to Reduce Suffering, Save Lives, and Optimize Cost of Care
Patient-centered care can be improved exponentially if the caregiver can forecast illness or complications so that preventative measures can be applied. Recognizing this, leadership from Baptist Health collaborated with a leading software developer and clinical experts to implement a predictive solution that analyzed patient phenotypes to predict disease and infection across the system. Their objectives were to reduce suffering and save lives while optimizing cost of care. During this session, Katrina Belt, CFO for Baptist Health, will present predictive findings against one use case, CAUTI, including the business case for prevention; models used; and preliminary health and financial outcomes.
[ Top of this page ] [ Agenda overview ]
Track 2: Disease Modeling
Case Study: Southern Nazerene University
Developing a Mortality Prediction Model for Disseminated Intravascular Coagulation (DIC)
DIC occurs when the cascade of reactions responsible for regulation of hemostasis goes awry resulting in widespread hemorrhage and clotting. Although trauma or infection can be initiators, DIC may follow surgery or other systemic event. Some triggers are known; however, presentation is highly variable. Some children with sepsis develop coagulopathy and expire, while others do not. Prophylactic procedures cannot be standardized without better prediction of morbidity and mortality. Better prediction means better treatment. Better treatment leads to better outcomes. Better outcomes waste fewer resources. We seek to develop a model for DIC, to predict better outcomes with fewer resources expended.
[ Top of this page ] [ Agenda overview ]
12:05-1:30pm • Room: Commonwealth Hall
Lunch in the Exhibit Hall
Special Plenary Session
The Peril of Vast Search (and How Target Shuffling Can Save Science)
It's always possible to get lucky (or unlucky). When you mine data and find something, is it real, or chance? The central question in statistics is "How likely could this result have occurred by chance?" Ancient geniuses devised formulas to answer this question for multiple special-case scenarios. Yet, their calculus only applies to quaint, handmade analyses, where only a few hypotheses are considered. However, modern predictive analytic algorithms are hypothesis-generating machines, capable of testing millions of "ideas". The best result stumbled upon in its vast search has a much greater chance of being spurious. Such overfit is particularly dangerous, as it leads one to rely on a model molded to the data noise as well as signal, which usually is worse on new data than no model at all. The problem is so widespread that it is the chief reason for a crisis in experimental science, where most journal results have been discovered to resist replication; that is, to be wrong!
The good news is an antidote exists! Dr. Elder will explain the simple breakthrough solution -- still rarely employed, though newly being re-discovered in leading fields. John will illustrate how to use the resampling method he calls "Target Shuffling" in multiple learning scenarios, from model fitting to data exploration, showing how it calibrates results so they are reliable - essentially providing an honest "placebo effect" against which to test a new treatment (finding).
Bottom line: Honest Data Science can save Experimental Science!
[ Top of this page ] [ Agenda overview ]
2:15-2:35pm
Silver Sponsor Presentation
Preventable Hospital Readmissions are a Significant � yet Avoidable � Cost
In fact, these readmissions account for an estimated $25 billion in wasteful healthcare spending in the US alone. Learn how Carolinas Healthcare System is using predictive analytics to improve and manage their readmissions case-management and how bringing predictive analytics to the point of care has empowered their nurses and case managers to better manage the risk of readmissions and, at the same time, improve patient care.
Track: Predicting Insurance Costs
Case Study: Pennsylvania Department of Public Welfare
Superutilizers Made Simple. Identifying High-Cost Recipients Using a Model Any Medicaid Agency Could Implement
Medicaid agencies have many commercial products available to them to identify potential high-cost recipients. These products are often expensive and are not always tailor-made to the particularities of the individual state's Medicaid population. Creating an in-house predictive model is an attractive alternative however hiring the necessary statisticians, health-care experts and other staff needed to make an effective model is also expensive and resource intensive. In Pennsylvania we have created a logistic regression model using diagnosis, cost and utilization data that is very effective in correctly identifying potential high-cost recipients. It is hoped that other states might implement this approach.
Track: Personalized What-if Analysis
Case Study: Washington University St. Louis School of Medicine and PotentiaMED Using Predictive Analytics to Empower Cancer Patients through "MyCancerJourney"
Using Predictive Analytics to Empower Cancer Patients through "MyCancerJourney"
MyCancerJourney's MyNavigator tool is a cloud-based, interactive predictive tool utilizing advanced analytics approaches to empower newly diagnosed adult cancer patients with the ability to explore the potential outcomes of different treatments and combinations of treatments. Rather than attempting to infer a prognosis from clinical trials based on less than 5% of adult cancer patient participation, outcomes estimates are based on the largest hospital-based cancer registry in the world with detailed co-morbidity information based on collection of ACE-27 index measurement. The result is personalized estimates of survival derived from the real-life experiences of similar patients to support more informed treatment decisions.
Jay PiccirilloProfessor, Vice-Chair Research,
Director of the Clinical Outcomes Research Office
Washington University in St. Louis School of Medicine
[ Top of this page ] [ Agenda overview ]
3:25-3:55pm • Room: Commonwealth Hall
Exhibits & Afternoon Break
Track: Targeted Care Management
Case Study: ActiveHealth Management
Significant Improvements in Population Health Management
Health reform shifts our focus from wholesale population to retail individual consumer. Our care management improved patient care quality and affordability, but the new ecosystem required us to improve patient engagement, health outcomes and lower costs. Predictive analytics micro-segmented our member populations, and improved patient engagement. K-Means Clustering and CART classification trees were used with a 1.3 million person test population. Response rate lift was 99 percent, and a follow on study showed an additional 74 percent lift with targeted messaging. The improvement saves our clients an estimated $6 million in avoidable health costs, and reduced our labor costs.
[ Top of this page ] [ Agenda overview ]
Track: Readmission Risk
Case Study: UPMC Health Plan
Using Association Rule Mining to Identify Risks for Readmissions
A readmission within 30 days of a hospitalization is a commonly-used quality measure of hospital care. Under the Affordable Care Act, CMS is authorized to decrease Medicare payments to hospitals with higher than expected readmissions for acute myocardial infarction, heart failure, and pneumonia. Consequently, understanding the circumstances under which a readmission occurs is of increased importance to a provider. We use Association Rules Mining applied to discharge records from the Agency for Healthcare Research and Quality (AHRQ) to derive a set of easily applied rules that capture high frequency readmissions for the above conditions.
5:30-7:00pm • Room: Commonwealth Hall
Networking Reception
[ Top of this page ] [ Agenda overview ]
Conference Day 2: Tuesday, October 7, 2014 • Sessions will take place in Federal
8:00-8:45am • Room: Commonwealth Hall
Registration & Networking Breakfast
Conference Chair Welcome
Gold Sponsor Presentation
Real-Time Analytics: What is possible and how do you optimize performance
Analytics is top of the mind for every industry and touches multiple roles across an organization with the expectation being near real time. But what is the reality given all the technology available? During this session, you will learn about the challenges we face with data integration, MDM and big data to achieve top quality and performance with analytics. In addition, Mr. Moschella will share EMR and claims data use cases.
Keynote
Predictive Analytics: Advancing Precision and Population Medicine
An enormous chasm exists in chronic disease management between treatment benefits demonstrated under controlled clinical circumstances�efficacy�and results seen in the real world�effectiveness. Predictive analytics methods have a key role to play in bridging this efficacy-effectiveness gap; helping physicians, nurses and health systems deliver superior outcomes to patients at acceptable cost. However, in order to enjoy the benefits of Big Data in health care we must first ensure that we are collecting Good Data.
The sad fact is that medical data today is most meaningful when applied to populations, not individuals, and lags clinical events. A 40 percent tumor response rate in a population is 100 percent, one way or the other, for the individual; a low medication possession ratio explains organ rejection for a particular transplant recipient only after the fact.
Perhaps the population emphasis so essential for establishing therapeutic efficacy explains why only about 50 percent of patients use prescribed medications appropriately in the real world. And the lack of timely and actionable information about medication use and response accounts for the difficulty physicians have in optimizing therapeutic regimens to conform to evidence-based standards.
Bridging the efficacy-effectiveness gap requires shifting our analytic focus from drugs and diseases to individuals and families; understanding idiosyncratic preferences and behaviors in order to optimize individual therapy and response.
Inverting the paradigm�building Big Data one granule at a time�requires deployment of new medical device and wellness sensor platforms to close the feedback loop between lifestyle, therapy and response, leveraging mobile connectivity and the Internet of Things. The result of this is Good Data. Unprecedented insights that are precise, timely and meaningful to the individual.
[ Top of this page ] [ Agenda overview ]
Track sponsored by:
Track: Claims Analytics
Case Study: DentaQuest
Improving Provider Performance and Patient Outcomes with Evidence-Based Scoring
The healthcare/dental industry is developing performance measures to promote better patient outcomes and foster professional accountability. Learn how one dental insurer is experiencing great business value by applying analytics to the combination of transactional claims data and clinical best practices. In this talk, we demonstrate the power of combining multiple models to improve patient care and reduce costs.
10:45-11:15am • Room: Commonwealth Hall
Exhibits & Morning Coffee Break
Track 1: Resource Optimization
Using Predictive Analytics to Forecast Hospital Patient Volume for Hospital Resource Allocation and Staffing
In an era of soaring healthcare costs, healthcare reform efforts will increasingly favor organizations capable of achieving operational efficiencies that reduce the cost of the care they deliver. Human resources represent a significant proportion of these costs. This session gives an overview of the processes undertaken to predict patient census using predictive analytics models in order to more effectively manage hospital staffing. We review the process of data discovery, data collection, and model building in an attempt to create predictive models for hospital management. We also compare the effectiveness of such models to standard time series statistical models.
12:00-1:15pm • Room: Commonwealth Hall
Lunch in the Exhibit Hall
Keynote
Real-Time Modeling of Surgical Site Infections
Surgical site infections (SSI) are a major cause of hospital readmissions and, when preventable, are major contributors to waste in the current healthcare system. The majority of predictive modeling of readmissions has focused on administrative data in non-surgical populations. We have developed a system to predict SSI using real-time data from the electronic health record (EHR) that performs as good or better than non-real-time systems in discriminating that will develop SSI. Through predictive modeling, we are changing the surgical technique and postoperative care of patients at high risk, reducing SSI and associated readmissions.
[ Top of this page ] [ Agenda overview ]
Vendor Elevator Pitches
Expert Panel
Healthcare Analytics: Potential vs. Reality
The application of analytics in healthcare is already showing promising results from precision medicine advances to population health improvements. Still, success stories are often more anecdotal and limited to specific markets and classes of patients rather than to the broader population. Join our experts as they discuss current and opportunities in light of HIPAA, privacy concerns and a disjointed national health care system.
[ Top of this page ] [ Agenda overview ]
3:00-3:30pm • Room: Commonwealth Hall
Exhibits & Afternoon Break
Track: Deploying Risk Models
Case Study: State of Maine hospitals
Deploying Predictive Patient Risk Models through a Health Information Exchange (HIE)
HealthInfoNet (HIN), operates Maine's statewide health information exchange. HIN sought to bring greater value to its member hospitals by providing predictive risk models. With data from 92% of in-state hospitals and over 1.2 million patients, HIN provides near real time predictive patient risk information to any provider throughout the state. This includes identifying inpatient risk of a 30-day readmission; as well as population and individual patient risks of future emergency department visits, inpatient admissions, and high cost consumption. This helps hospitals reduce cost and improve quality by clearly highlighting the highest risk patients and targeting them for proactive care management.
Track: Healthcare Informatics
Case Study: University of California, Irvine
Healthcare 2020: How Emerging Technologies Will Advance Clinical Practice and Research
Big Data, Advanced Analytics, Wearable Sensors, Home Monitoring and Social Media Analytics are all emerging technologies that will shape future healthcare information systems. This presentation will explore emerging technologies as well as their current and future healthcare information applications. Emphasis will be on the use of predictive models to identify patients at risk in the inpatient, outpatient and home environments.
[ Top of this page ] [ Agenda overview ]
