MACHINE LEARNING & AI: A HIMSS EVENT
LAS VEGAS, NV - MARCH 5, 2018
Breakfast will be served in the ballroom so make sure to stop by the sponsor tables.
One of the biggest mistakes healthcare stakeholders make is believing that technology, include machine learning and AI, solves problems. It does not. Technology is a tool, part of the solution. And the real challenge is implementing that tool into workflows – existing or new – in a way that makes sense, and most importantly, drives the desired results.
In this opening keynote, renowned physician and scientist Lynda Chin takes a hard look at what must happen for machine learning and AI to move beyond hype and deliver on its promise to transform healthcare.
At the end of the day, it boils down to:
These days, it seems, everyone wants to jump on the machine learning bandwagon. But real success, real change, as Chin will explain, depends on healthcare laying the proper foundation.
There’s a ton of buzz around machine learning and artificial intelligence and the role they’ll play in revolutionizing and improving healthcare, but what are these evolving technologies, how are they different, and what are the multiple layers of use?
No complex health IT ever created has been plug-and-play. The same goes for ML and AI. To take full advantage of its potential, will require a lot of work.
In this morning leadership panel, our speakers look at the current state of machine learning and AI in healthcare and address where we are, where we are going, and what we need to do to get their faster.
What education do stakeholders need? What new vocabulary is required? How do you integrate ML and AI into operational and clinical processes? How do you show ROI?
Our speakers will address these questions and others head-on in a fascinating and insightful discussion that sets the stage for the speakers to follow.
Over the past several years, artificial intelligence and machine learning, with their ability to analyze, process, and adapt to large volumes of data, information, and knowledge to solve complex problems, have garnered significant investment among the world’s most successful corporations, including Google, Apple, Facebook, Amazon, IBM, GM and Uber.
Healthcare is also now stepping into the AI spotlight in a variety of application areas, including natural language processing (NLP), medical image recognition, clinical decision support, diagnosis, and prediction. But to what degree is healthcare embracing these technologies? What are the perceptions, expectations, sourcing approaches, and initiatives?
In this session, Chilmark Research, a health IT research firm, will shed light on those questions by sharing the results of a survey that included 100+ senior healthcare executives.
Key areas addressed by the survey:
Take this opportunity to mingle with your peers in a relaxed setting to build relationships and establish future partnerships. Coffee will be served in the ballroom area so make sure to stop by our sponsor tables.
This session will discuss the results of a study that used machine learning to analyze EMR data to identify individuals at high risk of colorectal cancer (CRC).
The study included 9,108 controls and 900 cancer cases among Kaiser Permanente Northwest adults, ages 40-89. The model identified individuals with a tenfold higher risk of undiagnosed colorectal cancer at curable stages, and flagged colorectal tumors 180-360 days prior to usual clinical diagnosis.
The detection model can be applied to broad populations to identify people at increased risk of CRC (in particular, right-sided CRC), and enables health systems to more effectively target colonoscopy resources.
This study also demonstrated the feasibility for the model’s use in a U.S.-based HMO adult population.
This CRC detection model narrows the screening gaps associated with people who decline fecal tests and/or colonoscopies, and instead opportunistically analyzes existing demographic data and CBC tests.
Key discussion points:
The nature of clinical medicine is highly nuanced and often localized. While artificial intelligence algorithms can be leveraged across health systems, human oversight by a clinician is necessary in order to achieve high model performance and accuracy and reduce alert fatigue. This unique approach to machine learning allows AI systems to achieve a positive predictive value (PPV) and sensitivity close, if not at, a level as that of a physician.
As Atlanta’s preeminent safety net hospital, Grady Memorial Hospital is an essential resource for many who otherwise would have little to no access to medical services. In an attempt to lower rates of patient illness, complications, and deliver personalized care, a little over a year ago the hospital incorporated artificial intelligence with the system’s mobile integrated health program.
The results have been impressive: Grady has seen a 10% decrease in readmissions for the targeted population and saved almost $700,000 in direct costs - a greater than 500% return on the program. Additionally, included within the intelligence of the machine are the socioeconomic factors such as literacy, income, access to transportation, and proximity to a food desert that drive more than half a patient’s risk.
In this session, among other things, attendees will learn how Grady has integrated AI to prioritize patient visits and to identify:
Excess unplanned hospital readmissions are a quality of care indicator, and pose a financial burden to hospitals.
In this session, attendees will learn how Children’s Hospital of Pittsburgh at UPMC developed an innovative real-time tool that calculates every inpatient's unique readmission risk, at point-of-discharge, from structured and unstructured elements in the EHR. The hospital integrated this predictor into its EHR, and when run in silent mode (Jan-March '17), it accurately predicted 80% of discharges with a high-risk of readmission.
The hospital now has an intervention plan (based on nurse calls, home health visits) that has already shown a reduction in preventable readmissions.
Key discussion points:
Despite excitement surrounding machine learning in healthcare, examples of full-fledged integration in day-to-day health system operations remains the exception, not the rule. Duke’s Institute for Health Innovation is in its third year developing and piloting machine learning technologies in clinical care.
To benefit from the full potential of machine learning, healthcare must step back from the trenches to acknowledge breakthroughs in technology, address barriers to progress, and critically reflect on the strategic priorities necessary to bring healthcare into a new digital age.
In this session, speakers will address the good, the bad, and the ugly when it comes to machine learning in healthcare.
Key discussion points:
One of the best ways learn is to network with your peers. This session will provide an opportunity to meet speakers and attendees who have similar privacy and security challenges and discuss solutions to those challenges.
Here's how it works:
Speakers will be stationed at different tables in the ballroom, and attendees can circulate and speak one-on-one or in groups with individual speakers. The speakers have been assigned different topics, but other topics can also be addressed.
Mingle, share and learn in this interactive environment.
Take this opportunity to mingle with your peers in a relaxed setting to build relationships and establish future partnerships. Coffee will be served in the exhibit area so make sure to stop by our sponsor tables.
This session discusses how to accelerate AI within your organization by bringing it out of the dark and being honest with what is and is not realistic, and where gaps exist that inhibit progress.
By walking through several use cases, attendees will learn how to solve for enterprise strategy development, use case prioritization, and key talent requirements.
Key discussion points:
This session will discuss considerations for selecting a vendor offering machine learning based products and services.
Key discussion points include:
Research literature indicates that human limitations, such as cognitive bias, can interfere with clinical decision-making and that artificial intelligence has the potential to improve the process.
The research also indicates that AI technologies may be viewed as a disruptive innovation to workflow and clinical practice. This mean that to implement AI effectively often requires a cultural change and strategy to promote clinician engagement.
This session will discuss an approach to implementing AI at the point of care that does just that.
Key discussion points:
Machine Learning for healthcare differs significantly from other application domains in terms of the use cases, data resources, operational constraints and evaluation metrics. Despite these differences, the majority of existing machine learning solutions for healthcare applications remain grounded in conventional methodologies that are re-purposed from unrelated application domains.
This session addresses the specific characteristics unique to healthcare that distinguish it from other application domains and creates the need for specialized solutions. It will also discuss the requirements created by these characteristics and how healthcare-focused machine learning technologies should meet these requirements to maximize practical impact.
Key discussion points:
As the healthcare industry moves to implement precision medicine in care delivery, healthcare systems are analyzing massive volumes of data toward the goal of increasing the accuracy and speed of clinical care. During this transition, healthcare IT managers are facing the challenge of managing different types of data to build systems that provide clear results to clinicians.
Penn Medicine's Chief Data Scientist Michael Draugelis will share the lessons learned from programs that have been developed at Penn to deliver clear benefits in improving outcomes.
Key highlights of the presentation will include how artificial intelligence (AI) can advance treatment and prevention as well as scientific and technical challenges for AI to be successful
After decades of lagging behind other industries in the use of data, healthcare is poised for its own data-driven transformation. Journalists describe a not-too-distant future where patients, phones, Fitbits, and physicians march hand-in-hand toward a healthier tomorrow.
But not everyone is jumping for joy.
Some healthcare organizations are so frustrated with the state of health IT that the American Medical Association’s CEO James Madara, MD, last year called digital health the “snake oil of the early 21st century.” Rather than improving care and boosting professional satisfaction, many digital tools, he wrote, don’t work that well, and actually impede care, confuse patients, and waste everyone’s time.
And then there’s machine learning. Or is it artificial intelligence? Or cognitive computing? Which is which or what is what?
In his closing keynote, Harvard Medical School professor Leonard D’Avolio, who previously led informatics at the VA, gets to the bottom of all this confusion and disgruntlement and examines what’s hype and what’s not. And how together, once a few very real barriers are eliminated, big data and machine learning will better serve doctors, patients and families, and contribute to improved healthcare.