As we run the gauntlet of healthcare conferences and trade shows this hectic meeting season, one thing is clear: Artificial intelligence (AI) is top of mind. How do we mindfully, cost-effectively, and responsibly harness the power of this proliferating technology that is rapidly transforming our businesses and lives?
Having recently assumed a new, exciting role as Chief Marketing Officer at Lightbeam, my observation is that providers and payers are daunted by “all the noise” surrounding AI and are seeking relevant use cases in healthcare that deliver quantifiable value and ROI. All the hype seems to focus more on its flashy functionality than on what it delivers in real-world administrative or clinical contexts, but that is changing as I type. There has been disproportionate interest in generative AI and large language models (LLMs), but the focus is changing to explore data-driven predictive and prescriptive AI models, and automation of manual, repetitive and error-prone tasks, workflows, and processes.
Not All AI is Created Equal
Soy Chen, our Chief Data Scientist at Lightbeam, explains it best: “You want the right tool for the right job. Chat GPT, for example, the generative AI product that has captured the world’s attention, is based on an open large language model. Predictive AI for clinical applications is vastly different, as it is highly curated and data driven,” she says. “It works within hard guardrails and does not ‘hallucinate’ results, as generative AI models are known to do.”
Chen continues, “At Lightbeam, we are not incorporating large language models into our data-driven products, which have a tighter tolerance because we adhere to the highest performance standards. They are patient-facing, and the stakes are high.” (Read more.)
As a healthcare provider or payer executive, you might have already invested substantial money, time, and effort into GenAI with eager vendors (often without industry expertise) offering “solutions looking for a problem” that do not reliably demonstrate results or ROI. This is hardly surprising since GenAI is the least mature and has not proven its efficacy for use cases beyond ambient clinical intelligence (ACI) or “ambient listening.”
Given this scenario, we at Lightbeam recommend a Portfolio Management Approach to AI. Much like your tech stacks that include a full array of acronyms and abbreviations, such as EHRs, ERP and RCM tools, CRMs, and SCM software, an AI Portfolio, as illustrated in figure 1 below, consists of the AI technologies and modalities that can propel your productivity and optimize your mission-critical use cases.
An enterprise AI solution is a fully integrated portfolio of technologies and capabilities, some of which automate manual and often repetitive administrative tasks while others deliver in-depth analysis, predictions, and courses of action to maximize automation, outcomes, and value.
Figure 1: AI technology innovation portfolio for healthcare adoption.
Given the onrush of AI spending in healthcare, it is important that healthcare leaders understand exactly how to deploy AI strategically and effectively in their VBC environments, as the AI transformation has begun. The train has left the station. Providers and payers, as well as pharma, medical devices manufacturers, and diagnostic companies are already spending fortunes on AI software. According to Gartner, AI spending in healthcare and life sciences is projected to grow from $11.6 billion in 2024 to $19 billion by 2027, with a five-year CAGR of 16.6%, an astronomical investment that requires a keen awareness of available technologies and their applications.
To get you started, here is a guide to the various technologies that comprise the AI Portfolio for Healthcare:
- Machine Learning
The most mature technology in the AI portfolio, machine learning uses data and algorithms to imitate the way humans learn to perform tasks autonomously and to improve their performance and accuracy through experience and exposure to more data. Machine learning algorithms make predictions or classifications based on data patterns. This is a component of advanced solutions like Lightbeam AI that aggregate external and internal data repositories to stratify patients based on risk, identify gaps in care, and deliver personalized care recommendations to improve outcomes. ML also can automate scanning of medical images to help radiologists proactively identify patients at risk of a stroke or heart attack for intervention well before an acute event happens.
- Deep Learning
This is a subset of machine learning, one that replicates human reasoning. It uses deep neural networks to simulate human decision-making. Unlike machine learning models, which require structured and labeled input data to be effective, deep learning models can make accurate outputs from raw, unstructured data. One of the most common uses for this in healthcare is for image analysis.
- Natural Language Processing (NLP) and Natural Language Generation (NLG)
NLP/NLG uses machine learning to allow computers to understand and communicate with human language. It enables computers and digital devices to recognize, understand and generate text and speech by combining computational linguistics with statistical modeling, machine learning, and deep learning. In healthcare, it is used for computer-assisted coding to translate medical records into plain English, to analyze health records, and present a summary of the patient’s chart at the bedside/point-of-care (POC) for physicians and nurses to save time and enhance productivity.
- Generative AI/Large Language Models (LLMs)
NLP/NLG will soon be supplanted by LLM-powered generative AI. Like NLP/NLG, GenAI can create original content, including text, images, video, audio, and software code in response to a user query. It can power smart online chatbots for scheduling appointments, analyze patient sentiment from various sources, and more. One of the most compelling use cases for GenAI/LLMs is the seamless capture of clinician notes via a cell phone—then automatically synthesizing, sanitizing, and editing before populating EHRs.
- Prescriptive AI Models
With the increasing integration of embedded purpose-built models like Lightbeam AI, providers can turn diverse sources of data into actionable insights to hyper-personalize care at scale and identify rising-risk patients for early intervention in real time. Investment in prescriptive AI solutions improves patient engagement, enhances customer satisfaction, and minimizes costs. By synthesizing data from EHRs, ADTs, HIEs, claims information, and geodemographic/SDOH sources, providers can streamline operating efficiencies while maximizing patient outcomes to optimize VBC performance.
Other AI-powered clinical applications include robotics-assisted surgery, robotic process automation (RPA), virtual reality (VR), augmented reality (AR), machine vision, and medical robotics for patient monitoring, rehabilitation, and even companionship. For a detailed overview of medical robotics, please check out my personal blogpost, ‘How will Medical Robots enable Digital Patient Engagement in the post-pandemic new normal’?
Enter Agentic AI
Agentic AI acts independently to achieve objectives—planning actions, make decisions, and learn from their experiences. They can also perform a series of actions in response to a single request, called “chaining.” Agentic AI uses a combination of AI techniques, including:
- Large language models (LLMs): Allowing systems to understand and respond to natural language commands
- Machine learning algorithms: Enabling systems to analyze data and identify patterns
- Reinforcement learning: Allowing systems to learn from their actions and improve their decision making
We are approaching an era where AI will be synonymous with agentic AI/AI agents, along with hybrid and multi-modal AI within the portfolio management approach.
Hybrid AI for Healthcare
Given the rapid evolution of AI, it is not sufficient to deploy these various modalities individually but to combine one or more modalities as a hybrid AI solution. LLMs, for example, generate responses based on patterns in the data they’ve been trained on. This means they don’t “understand” the information in the way humans do; they just predict what’s likely to come next based on algorithms and training. They are perfect “virtual assistants” for drafting emails, summarizing documents, or even brainstorming creative ideas but have been known to hallucinate, which is very problematic in a healthcare context.
This is where hybrid AI demonstrates its true value—combining a traditional machine learning model trained on vast amounts of medical data with a generative AI component. The synergy of machine learning and gen AI leverages the strengths of both AI types while mitigating their weaknesses, producing a more accurate diagnosis along with better patient communication and understanding. It’s a win-win-win.
Harnessing Multimodal AI
Multimodal AI uses ML (machine learning) to process information from a full range of modalities, including images, videos, and text. For example, Gemini, Google’s multimodal model, can receive a photo of a plate of cookies and generate a written recipe as a response and vice versa. GenAI is an umbrella term for the use of ML models to create new content formats by transforming text, image, music, audio, or video inputs. Expanding on generative capabilities, multimodal AI can prompt virtually any input type to generate virtually any content type. Multimodal AI also delivers more advanced reasoning, problem-solving, and generation capabilities, which offers endless healthcare possibilities.
Key Considerations in Building Your AI Portfolio Management Strategy
- Challenge and Opportunity
Your AI strategy should start with identifying with what problem you are trying to solve. Then look at the potential for AI to automate the process or enable superior outcomes through actionable insights that will produce measurable enterprise value. Do not pursue AI simply because it is cool and happening. Analyze and prioritize the use cases with the highest value and ROI opportunity by using the portfolio management approach.
- Assess Organizational IT/AI Competencies
Do you want to build your own AI platform and apps from the ground up or do you want to leverage those provided by tech vendors like Lightbeam. Perhaps there is a specific technological component you need to complete your integrated solution? Lightbeam, a HITRUST-certified vendor, can guide you as your partner. Transparency and security are key considerations in any healthcare solution.
- Governance, Ethics, and Compliance
Oversight and responsible use have been evolving with accelerating AI adoption, and they will continue to do so. The industry and world will be continuously monitoring, measuring, analyzing, governing, and improving AI models in the days, months, and years ahead.
Build an Integrated AI Platform to Achieve Highest-Priority Objectives
As the market leader, Lightbeam’s proprietary AI-models and customized solutions can help your healthcare organization optimize patient outcomes with lower caregiver fatigue and cost of care with automated workflows, sense and respond capabilities, and proactive prescriptive insights. Lightbeam AI turbocharges VBC environments by:
Minimizing Avoidable Admissions:
- Predicts avoidable admissions within the next 30, 60, 90 days
Reducing Avoidable ED Visits:
- Predicts which members are at highest risk for avoidable ED visit in the next 30, 60, 90 days
Facilitating End of Life Support:
- Predicts mortality risk within specific timeframe (3 to 18 months)
- Supports setup of advanced directives
- Eases timely transition to appropriate palliative or hospice care.
Operationalizing Data in Real-Time Care Prioritization:
- Dual Identification: Ensures appropriate benefit eligibility to maximize MSSP performance
- HCC Suspecting: Harnesses coding intelligence to deliver timely proactive care
- Individual SDOH: Identifies patients at highest risk and why
- Community SDOH: Generates vulnerability scores to predict poor patient outcomes at geographic block-group level
Lightbeam AL Implementation Success Case: IDN Medicaid Population
A large Integrated Delivery Network (IDN) in the Southwest reduced avoidable admission events by 5.3%, and the risk of avoidable admissions by 43% among its Medicaid members by using Lightbeam AI’s 30-day avoidable admission model. Lightbeam AI enabled precise targeting, allowing care teams to focus on the most at-risk patients—resulting in better care coordination, 65 admissions prevented in one year, and a cost savings of $637,000.
Read more about this IDN success story here. Lightbeam AI solves your highest consequence problems by empowering your VBC organization to scale proactive outreach quickly, minimize avoidable admissions, maximize savings, and improve patient outcomes. Simultaneously, you can enhance provider experience, relieve administrative burden, and elevate care quality.
With Lightbeam AI, you move from prediction to action—automatically. Aggregate, identify risk, assess, and intervene.
Put AI to work for your organization. Speak to a Lightbeam AI expert to learn more.
#AI #AIportfolio4healthcare #AIportfolio4lifesciences #PrescriptiveAI #LightbeamAI #MachineLearning #NLP #NLG #GenAI #LLMs #HybridAI #MultimodalAI #MedicalRobotics #RPA
Andy Dé, Chief Marketing Officer, Lightbeam Health Solutions