When a major electronic health record vendor quietly rolls out an algorithm that predicts a patient's risk of dying, it is more than a technical milestone. It is a signal that artificial intelligence is moving from the back office into the exam room, and in this case, into the most sensitive conversation a clinician can have. Epic Systems, whose software holds the medical records of millions of Americans, has been expanding its mortality prediction model, and the development deserves careful scrutiny. Meanwhile, Omada Health, a company known for its digital diabetes prevention program, is planning to broaden its product line. Together, these two stories illustrate the twin forces shaping health tech: the push to predict and prevent, and the challenge of doing so responsibly.

What Epic's Mortality Model Actually Does

Epic's mortality model is not a crystal ball. It is a statistical tool that analyzes data already sitting in the electronic health record, such as age, diagnoses, lab results, medication lists, and prior hospitalizations, to estimate the probability that a patient will die within a certain timeframe. The goal is not to label patients but to prompt clinicians to have earlier, more honest discussions about goals of care. If a patient with advanced heart failure has a high risk score, for example, a care team might prioritize a palliative care consult or discuss advance directives before a crisis hits.

That is the promise. The reality is messier. Mortality prediction is notoriously difficult because death is influenced by countless factors, many of which are not captured in structured data. A patient's will to live, their support system, their access to nutritious food, and their mental health all play a role. An algorithm that relies on billing codes and lab values can miss these nuances. Worse, it can perpetuate biases if the training data reflects historical disparities in care. If a model learns from a dataset where Black patients were less likely to receive aggressive treatment, it might underestimate their survival chances, leading to self-fulfilling prophecies.

The Geriatric Care Dilemma

Geriatricians have long practiced the art of prognostication, but they do so with years of experience and a deep understanding of the individual. An algorithm cannot replicate that relationship. However, it can serve as a safety net, catching patients who might otherwise slip through the cracks. The key is how the tool is used. If it is presented as a definitive answer, it can cause harm. If it is presented as one data point among many, it can enhance care.

Consider a scenario: An 82-year-old woman with mild cognitive impairment and multiple chronic conditions comes to her primary care doctor for a routine visit. The Epic model flags her as high risk for mortality within 12 months. Without context, that flag might lead to unnecessary tests or a sense of hopelessness. But if the doctor uses it as a conversation starter, asking about the patient's values and what matters most to her, the tool can be transformative. It can shift the focus from treating diseases to treating the person.

That shift is exactly what many geriatricians advocate. They argue that AI in geriatrics should augment, not replace, clinical judgment. The danger is when health systems use mortality scores to ration care or to justify denying expensive treatments. Already, there are reports of algorithms being used to identify patients for palliative care programs, which can be beneficial, but also to flag patients for discharge planning in ways that feel coercive. The line between support and coercion is thin.

Omada Health's Expansion Plans

While Epic focuses on prediction, Omada Health is betting on prevention and management. The company started with a digital Diabetes Prevention Program (DPP) that combined connected scales, activity trackers, and human coaches to help people at risk of type 2 diabetes lose weight and improve their health. Now, Omada is expanding into new product areas, including hypertension, musculoskeletal conditions, and behavioral health. The idea is to become a one-stop shop for chronic condition management, leveraging the same behavior change platform across multiple diseases.

That strategy makes sense on paper. Many people with diabetes also have high blood pressure or joint pain, and treating them in silos is inefficient. A unified platform could offer a more holistic approach, addressing diet, exercise, sleep, and stress in one place. Employers and health plans, who are Omada's primary customers, like the idea of a single vendor that can manage a broad population. But execution is everything. Each condition has its own clinical nuances, and a generic coaching model may not work for all.

What Omada's Future Products Mean for Patients

For patients, Omada's expansion could mean more convenient access to care. Instead of juggling multiple apps and programs, they might get a single dashboard that tracks their progress across conditions. That is a clear win if the platform is well-designed and the coaching is personalized. But there is a risk of oversimplification. Chronic conditions are complex, and a one-size-fits-all approach can feel impersonal. The best digital health programs are those that adapt to the individual, offering tailored content and human support when needed.

Omada has a track record of investing in clinical research, which is a good sign. Its DPP has been validated in peer-reviewed studies, showing meaningful weight loss and reduced diabetes risk. If the company can replicate that rigor for new conditions, it could build trust with clinicians and patients alike. However, scaling a research-backed program is expensive, and Omada will need to prove that its expanded offerings deliver a return on investment for payers. That is a high bar, especially as the digital health market cools and investors demand profitability.

The Bigger Picture: AI and Chronic Care

Epic's mortality model and Omada's expansion are two sides of the same coin. Both rely on data and algorithms to improve health outcomes. Both raise questions about equity, transparency, and the human touch. The most successful health tech will be the kind that keeps the patient at the center, using technology to enhance rather than replace human relationships.

For Epic, that means being transparent about how its mortality model works and ensuring it does not exacerbate disparities. For Omada, it means maintaining the quality of coaching as it scales. For the rest of us, it means staying informed and asking hard questions. Health tech is not inherently good or bad; it depends on how it is designed and deployed. As these tools become more integrated into care, patients and providers must have a seat at the table.

What to Watch Next

  • Whether Epic releases validation data for its mortality model, including performance across different demographic groups.
  • How Omada integrates new conditions without diluting its core behavior change methodology.
  • Regulatory developments, particularly around AI transparency and accountability in health care.
  • Adoption rates among clinicians and patients, which will ultimately determine the impact of these technologies.

Frequently Asked Questions

What is Epic's mortality prediction model used for?

It is designed to help clinicians identify patients at high risk of death within a certain period, prompting earlier discussions about goals of care, palliative care, and advance directives. It is not meant to be a definitive prognosis but a tool to support clinical judgment.

Is Omada Health only for diabetes prevention?

No. Omada Health started with a diabetes prevention program but is expanding into hypertension, musculoskeletal health, and behavioral health. The goal is to provide a comprehensive platform for managing multiple chronic conditions.

How accurate are mortality prediction algorithms?

Accuracy varies widely depending on the population and the data used. They can be useful for identifying broad trends but are not reliable for individual predictions. They should always be used in conjunction with clinical expertise and patient preferences.

What are the risks of using AI in geriatric care?

Risks include bias in the training data, over-reliance on scores that may not capture a patient's full picture, and potential for rationing care. Ethical implementation requires transparency, validation, and a focus on augmenting rather than replacing human decision-making.

Will Omada's expansion lead to better outcomes for patients?

It has the potential to, if the company maintains clinical rigor and personalization. However, scaling digital health programs is challenging, and success will depend on evidence, user engagement, and payer support.