AI in health care
Advancing the responsible application of artificial intelligence to help people live healthier lives and make the health system work better for everyone.
Using Al to solve real health care problems
Artificial intelligence (Al) has a lot of hype surrounding it in the health care industry, and for good reason. We have the keys to unlock its potential — skilled people and rich data. Applied practically, Al is one of the most important emerging technologies to the industry. It has the potential to bring together vast amounts of data to create the analytic insights and process efficiencies needed to advance population health, care delivery and operations.
Let's start with the definition of Al. It is not a singular technology, but rather an umbrella term. Al includes using methods such as deep learning, machine learning (ML) and natural language processing (NLP) to perform "smart" tasks we often associate with the human mind such as learning and reasoning.
Is artificial intelligence smart for health care?
Where is AI today?
Artificial intelligence (AI) has left behind its sci-fi legacy to become a transformative technology in a modern digital age. A range of possibilities exists for AI in health care. However, we need to get past the hype, such as robots fully replacing the expertise and services of doctors, and show how AI allows us to reimagine the status quo and sheds new light on real solutions.
60% of health professionals selected artificial intelligence as a top or high adoption priority among various technology categories. (Forrester study on behalf of Intel 2017)
Let’s start with defining what it’s not. AI isn’t a singular technology, but an umbrella term that includes using deep learning, machine learning and natural language processing, among other methods, to perform ‘smart’ tasks we often associate with the human mind such as learning and reasoning.
Deep Learning operates similar to how the brain’s neural synapses strengthen with repeated action becoming more efficient in adjusting for errors to modify its approach when faced with new data. (The Rise of Machine Learning, The Advisory Board, June 2017)
Natural Language Processing refers to the interpretation of speech and text. (The Rise of Machine Learning, The Advisory Board, June 2017)
Machine Learning uses advanced statistical techniques to identify patterns in data and then make predictions. (The Rise of Machine Learning, The Advisory Board, June 2017)
Is health care ready?
Al is actually already at work in health care. AI helps people with diabetes regulate their blood sugar. It automates prescription refills. It matches call center customers to the person most qualified to assist them. With the convergence of algorithmic advances, data proliferation and tremendous increases in computing power and storage, the opportunities for AI in health care will only just keep growing.
While only 4.7% of health care respondents are already using AI technologies, about 35% plan to leverage AI within two years — and more than half intend to do so within five years. (Adapted from a Healthcare IT News and HIMSS Analytics survey 2017)
What is the value?
Better performance: AI automates repetitive or tedious tasks and quickly helps discover patterns and anomalies to lower the total cost of care and help individuals work at the top of their field.
- Clinical Documentation Improvement
- Payment Integrity
- Prior Authorization
Better outcomes: AI can quickly process data, personalized for each patient, to provide appropriate medications, identify health issues and alert us to take preventative action.
- Early Diagnosis and Treatment
- Prescription Benefit Management
- Risk Adjustment
- Simplified Population Analysis
Better patient experience: AI simplifies and connects complex data to break down care silos, translates the complicated into easy-to-understand information and helps improve the patient care experience.
- Call Centers
- Care Coordination
- Employee Benefits
Better performance — increase operational efficiency, speed and consistency
Clinical documentation: Natural language processing (NLP) can identify clinical indicators across a patient’s entire electronic medical record, including lab results, notes and prescriptions. NLP unlocks this unstructured (free form) data to understand each case and identify those where information may be missing. The result is more complete and accurate documentation to support better coding and appropriate quality measures.
Coding: NLP can understand clinical documentation to capture comprehensive diagnosis and procedure codes, eliminating tedious, manual case reviews. Coders don’t have to start from scratch to provide more thorough, accurate coding results that lead to more appropriate revenue capture.
Payment integrity: Payment integrity is all about ensuring payments made from one organization to another are done correctly and appropriately. Data analysis and predictive modeling identifies problem areas and improper claims faster and more accurately than ever before.
Prior authorization: Physicians have all relevant prior authorization information, insurance coverage, formulary rules and potential drug-to-drug interactions in real time. This gets patients their medicines faster and with less uncertainty when they get to the pharmacy counter.
Better outcomes — increase early disease identification and preventative treatment
Early disease identification: Predictive models using regression techniques and newer machine learning approaches are helping in the early identification and preventative treatment of specific conditions. As an example, current models can identify signals of dementia five to eight years earlier than the first diagnosis.
Prescription benefit management: Predictive models are used to identify the right intervention at the right time. Algorithms use pharmacy data, lab results and other electronic health data to get to the right outcome.
Risk adjustment: NLP and AI are increasingly fueling prospective risk adjustment analytics that identify undiagnosed conditions and support the correct interventions leading to enhanced quality of care.
Simplified Population Analysis: Advanced analytics can be presented in a graphical or pictorial format helping quickly define and profile populations, simplifying and speeding time to analysis and evidence generation to lower risk and improve outcomes.
Better patient experience — personalize, simplify and deliver patient information at the right time
Call centers: AI helps call center agents make actionable, data-driven decisions at an individual member level. Interventions are prioritized according to clinical value and an individual's propensity to take action.
Employee benefits: AI provides reliable benefits data that is integrated and easy to understand to help translate health care data into actionable information.
Care coordination: With intuitive, configurable workflow designs, providers get the information critical to coordinating care, including key patient identifiers, gaps in care, and complications with chronic and complex populations.
What data does AI need?
If the foundational data is unorganized or limited, even the most advanced data analytic tools may simply get you to the wrong outcome faster. AI needs data that is standardized, tested and, in most cases, collected from multiple sources. For example, Optum Performance Analytics® uses longitudinal data and a range of sources from claims to clinical and across care silos. AI not only needs the right type of data, it needs the human component to understand the interrelationship of the varied data sets and how they can be applied to deliver real-world solutions.
Predictive accuracy of a system increases from below 1% to over 90% accuracy by including health care information from a variety of sources. (Artificial Intelligence and Advanced Analytics in Health Care, Frost and Sullivan, 2017)
Delivering on the promise
Delivering on the promise of AI, we infuse our products and services with OptumIQ™, a unique combination of data, analytics and applied expertise. At Optum, our over 24,000 data, analytics and technology experts focus on pragmatic applications of AI in health care with data that includes 5 billion medical procedures, 11 billion lab results and 4 billion diagnoses.
Optum invests approximately $3.3 billion annually in technology and innovation.
Planning for AI
Growing pains are inevitable. Set your AI project up for success using these fundamental steps:
a. Identify use cases. Understand what problems you want AI to solve before you begin.
b. Convene the right people. You’ll need data scientists to translate and expedite results. You’ll also need people in the field who can and will apply them.
c. Curate data. Data from disparate sources needs to be standardized and organized so it can uncover patterns and validate repeatable solutions.
d. Set benchmarks for success. Identify key benchmarks or milestones and be sure to document and share the results.
See how your organization can gain real value with AI. Visit us at optum.com/iqOR
AI: What does it really mean?
Artificial intelligence is at a tipping point in transforming the industry. Many factors make bringing Al to health care particularly attractive, starting with the complexity and rise of data in the industry. No matter where the start line seems most obvious, an unwavering reality is that Al requires lots of data. The better the data strategy, the greater impact Al makes in an organization's processes, products and services.
By curating and harnessing the vast information — traditional clinical and claims data, unstructured medical notes and nontraditional data sources — potential applications in the field are numerous:
- Transforming health care administration and operations, including health plan claim process redesign, fraud and waste identification, and physician network optimization and steerage.
- Empowering consumers with person-level engagement and recommendations, and enablement of lower-level or home-based care through remote monitoring technology.
- Enabling health care delivery that provides patient-level recommendations based on identification and stratification models, and can contribute to total cost of care reductions through integrated medical, pharmacy and behavioral interventions.
TECHNOLOGY SHOULD AUGMENT HUMAN SKILL
The question is not whether Al will change health care — it's already disrupting and changing health care. But the Al of today and the foreseeable future is trained to solve one particular problem at a time. We're a long way from the general Al that's often depicted in science fiction, and not sure that will ever be the right evolution for the health system.
At Optum, our team of over 25,000 health care experts seeks to solve the most complex problems in health care — through research, development and the application of emerging technologies — to improve the lives of the people we serve. We incubate emerging technologies and explore ways to apply them practically by working with clinicians, researchers and business professionals. Technology areas include Al, the internet of things (loT), deep learning, genomics, NLP, blockchain and graph databases.
The question we must address is how we can use Al, and other emerging technologies, in ways that help humans and systems work better and faster. Although there are many instances in which Al can perform tasks as well or better than humans, it's more likely we'll have an augmented age in health care to reach our goals for the health system. We'll put the insights and efficiency Al and other technologies help us create into professionals' hands to work at the top of their license or empower the consumer.
4th Annual Optum Survey on AI in Health Care
Four years into the Optum Artificial Intelligence (AI) in Health Care survey, and one trend remains clear: the state of health care-focused AI remains strong. See 500 health care executives' evolving views on artificial intelligence and its impact on the industry.
Al health care applications — now and in the future
Questions abound about the practical value of artificial intelligence in today's health care world.
Several types of Al are already being employed by payers and providers of care and life sciences companies. Systems engaging you with your health through your employer, financial institution, pharmacy or smartphone also use Al. The key categories of applications involve diagnosis and treatment recommendations, patient engagement and adherence, and administrative activities.
Combining multiple methods of Al can help providers identify patients at risk for chronic diseases and catastrophic medical events.
Health care revenue cycle
Clinical intelligence is having a transformative impact in how our care is reimbursed.
Applying Al methods behind the scenes better supports prescription management, consumer engagement and medication adherence.
Al is helping companies see a bigger picture to improve drug and therapy development through changing clinical trials.
Employee health benefits
Al and advanced analytics are helping improve savings rates in health savings accounts (HSAs).
Health care payment and reimbursement
Applying Al to reduce friction in the payment continuum could chip away at $200 billion in waste.
Sign up for updates
Receive fresh perspectives and expert advice on data, analytics and tech innovation in health care.