Expanding Symptom Management Beyond the Clinic With AI-Assisted Virtual Care

Patients undergoing systemic treatment for cancer face stress and uncertainty when they are unable to easily consult their care team about symptoms they are experiencing. In these situations, they and their caregivers lack the information about whether symptoms are related to the disease or treatment, if they could worsen if not addressed, and if they should wait to speak to their oncologist or nurse or seek immediate medical attention.

Virtual cancer care is now giving patients a way to express their concerns and get help with symptom management on their own schedule, and with the integration of artificial intelligence (AI), it can take on a personal touch to communicate and interpret patient concerns on an individual level. At Moffitt Cancer Center in Tampa, Florida, an AI-enabled virtual oncology care program delivered by Reimagine Care is being expanded following an initial implementation phase that demonstrated sustained patient engagement, reduced acute care utilization, and high patient satisfaction scores. Patients are able to raise issues comfortably and be escalated to receive the best possible care if needed.

“People want their answers when they need it, and they want to be taken care of quickly,” Timothy Hembree, DO, PhD, vice president and Chief Quality Officer at Moffitt, said in an interview with Targeted Oncology.

The initial phase generated nearly 7000 patient interactions across proactive check-ins, electronic patient-reported outcomes (ePROs), symptom management pathways, and clinical escalations. The approach yielded a 97% rate of independent resolution of patient issues without escalation to Moffitt providers; 94% symptom management encounters were resolved virtually, and emergency department (ED) referrals were only needed in 2.4%.¹

The platform was launched with attention to both patient needs and reducing the workload of treating physicians, Hembree explained. In the interview, he discussed the effectiveness of the Reimagine Care program so far and how they are working to take advantage of the potential of AI without losing a human connection as the program expands.

Targeted Oncology: What are the issues in providing high-quality cancer care to patients, particularly between scheduled visits?

Timothy Hembree, DO, PhD: Cancer care is inherently complex and often challenging for patients and their families. I have been caring for oncology patients at Moffitt for over a decade in my role as an internist. My primary focus is the management of hospitalized patients who present with complications related to their cancer, its treatment, or a decline in their overall condition driven by underlying comorbidities.

A concept that has become increasingly important in oncology is “time toxicity.” When patients receiving active cancer treatment deteriorate at home, they are often admitted to local hospitals. These interruptions can delay treatment and negatively affect clinical outcomes. Recognizing this, we have become increasingly interested in expanding our ability to deliver care in the home setting, rather than relying solely on episodic interactions in the clinic.

The traditional care model places a significant burden on patients. After traveling to clinic, preparing for chemotherapy, and receiving multiple medications, patients are expected to return home and independently manage complex symptoms. This process can be overwhelming. Even in informal settings, when friends and family reach out for guidance, they often express uncertainty, asking whether they should seek emergency care or how best to manage their symptoms. These experiences highlight a critical gap in care delivery and are what drew me to the Reimagine Care program. It more closely reflects how patient-centered care should be delivered in real-world settings.

What are the key features of the virtual cancer care program that Moffitt has introduced?

The Reimagine Care platform has a couple of features that are great. It’s a two-way communication platform, so the patient can reach the provider or the care team very easily through texting. The technology is not complex. You do not have to download anything or go to a website or have a computer in front of you. You can text the digital component at any time of day or night, weekends included, and say, “I’m having trouble with this,” and then that triggers the digital piece of the Reimagine Care system, [the chatbot] which they call Remi.

Remi serves the purpose of asking good pointed medical questions to help us triage the severity of what we’re talking about, and then that system will do a severity ranking based on the CTCAE severity ranking for [adverse events] of chemotherapy, and if that ranking is moderate to severe, it will immediately promote that to someone in the virtual center, which is an oncology nurse or an oncology APP [advanced practice provider] to have a more direct conversation with the person to try to resolve the issue.

The other thing that I think is unique about this program is that it reaches out to the patient proactively 3 times a week: twice a week just to check on symptom clusters, and then a third time to do what we call [electronic] patient-reported outcomes (ePROs), which gives you a much broader take on how they’re doing overall. That proactive outreach keeps people thinking about their symptoms and how they’re doing, so that they keep things under control and…don’t end up in the hospital.

So, [it offers] easy communication in both directions, proactive outreach, [and] easy escalation to the virtual center. We did a pilot with the program over the last year, we had 97% resolution of the problem without escalation beyond that, and then we would have around 2.5% to 3% escalation back to the center, which we think is appropriate, because in some cases people should come back to the center.

Overall, the program has demonstrated the ability to effectively manage symptoms, maintain patients within mild to moderate levels of severity, reduce emergency department utilization, and ensure timely re-engagement with in-person care when necessary.

What are your takeaways on how the program is working so far?

We intentionally chose to operationalize the system with a relatively small patient cohort initially, allowing us to identify and address potential challenges in a controlled manner. Our evaluation focused on key metrics, including enrollment rates, longitudinal engagement, and patient satisfaction. Engagement, defined as the frequency of patient interaction with the platform over time, served as an important indicator of both usability and perceived value. Sustained use suggested that patients found the system acceptable and beneficial in managing their care.

…We wanted to see if we could reduce our inbox messaging volumes for our providers, [because it] went from somewhere around 40,000 the first year we did it to almost over 500,000 the last year. Instead of adding an exponential number of people to answer phones, could we send it through this process and be much more efficient? …If you send a message through the portal, it would say, “If this is an emergency, do not use the portal. But if it’s not an emergency, ask your question, we’ll get back to you within 48 to 72 hours.”

That’s fine for a lot of questions. …[But] if a person is struggling with symptoms like diarrhea, vomiting, incredible headache, that’s not something that can wait 48 to 72 hours, and…then the only recourse that a patient would have sometimes is to go to the local ED or drive into Moffitt and be seen in our urgent care, which eats into the patient’s time, eats into the patient’s cost, could disrupt their overall cancer care because of admissions and delays and those sort of things.

We wanted to make sure people didn’t have an issue interacting with Remi, and these systems are much better than they used to be. We started this process with Reimagine Care about 2 years ago, and even the evolution over those 2 years is just much better. The timing and the cadence in which we interact with the digital chatbot is much more like a natural conversation than it used to be, and…the AI component of it can remember things, so when you’re interacting with the system it’s not a brand new interaction every time you interact with it. Patients were very accepting of that.

Equally important is the system’s ability to rapidly escalate patients to the virtual care center when needed. Although the AI component is effective at gathering relevant clinical information and asking targeted questions, it is not intended to function as a treatment modality. Rather, its primary role is risk stratification. Clinical management remains firmly in the domain of trained providers. This approach represents a meaningful evolution in cancer care delivery, combining digital triage with clinician-led intervention. We have been encouraged by the results to date and are preparing to expand the program further in June.

How well are oncologists able to keep up with patients who engage with this system?

A critical objective was ensuring that the system could deliver clinically informed, patient-specific guidance. To enable this, we integrated the Reimagine Care platform with our electronic medical record (EMR). This allowed virtual nurses and advanced practice providers in the virtual care center to access the patient’s chart in real time, ensuring their assessments and recommendations were grounded in accurate clinical context. With this access, clinicians could review relevant details such as active treatment regimens, known adverse event profiles, and recent clinical history. For example, they could readily identify when a patient was receiving a chemotherapy agent commonly associated with diarrhea and tailor their response accordingly. This significantly improved the quality and precision of clinical decision-making within the virtual environment.

Equally important, we enabled bidirectional integration by allowing the virtual care team to document directly within the EMR. Following each interaction, a structured note was entered into the patient’s chart, summarizing the issue, the intervention provided, and the outcome. This ensured continuity of care by allowing the primary oncology team to remain fully informed. When providers opened the chart, they could immediately see that an interaction had occurred, along with the details of that encounter. Additionally, a clear indicator within the EMR identified patients enrolled in the home symptom management program, further enhancing situational awareness for the care team.

This documentation capability also facilitated direct collaboration between providers and the virtual care center. If a treating physician had questions about a specific interaction or disagreed with a management decision, they could directly engage with the virtual team. Establishing this feedback loop strengthened clinical alignment and trust in the model.

From an operational perspective, one of our primary goals was to reduce the burden on frontline providers. A key lesson from the pilot, however, was the importance of provider education and communication. Although the system functioned effectively, broader awareness and understanding among providers would have enhanced adoption and maximized its impact. For the pilot, we implemented an enrollment algorithm targeting select disease groups, including malignant hematology, gastrointestinal, and breast cancers. In addition, we built in flexibility by allowing providers to manually refer patients who were not part of the initial cohort but were experiencing challenges that could benefit from the program. This ensured that access to the service was not overly restrictive and could be applied in clinically appropriate situations.

Reflecting on the pilot, one important insight is that future implementations may benefit from focusing on a smaller, more concentrated group of providers. This approach would allow those providers to develop a deeper understanding of the program’s capabilities and value, rather than diffusing the experience across a broader group with variable engagement.

Many of the program’s benefits are realized at scale. Although we intentionally began with a limited rollout to confirm patient acceptability and ensure operational reliability, the full impact will become more evident as we expand. As we scale over the coming year, we anticipate a clearer demonstration of the program’s value in improving care delivery, enhancing patient experience, and reducing provider burden.

How can AI help assess treatment toxicity in patients?

Looking ahead, several upcoming enhancements to the platform are particularly promising. One area of significant advancement is the evolution of the triage system toward more disease-specific and treatment-specific symptom algorithms. For example, diarrhea in a patient receiving immunotherapy represents a very different clinical scenario compared with diarrhea in a patient treated with irinotecan-based chemotherapy. Each carries distinct risks and requires a different management approach. As the system becomes more refined in incorporating these nuances, it will further improve the accuracy and safety of symptom triage and intervention.

This is especially important in the current landscape of oncology, where combination regimens involving immunotherapy and chemotherapy are increasingly common. In these settings, clinical decision-making often requires a diagnosis-of-exclusion approach. The priority is to first rule out the most serious or high-risk etiologies and then stepwise narrow the differential diagnosis. As the platform continues to mature, we anticipate that its algorithms will become more sophisticated and better equipped to support this level of clinical reasoning.

We have also observed that the educational component of the system provides meaningful value to patients. It offers clear guidance on what symptoms to expect and practical strategies for managing common [adverse] effects, such as nausea. This proactive education helps patients feel more prepared and can contribute to improved symptom control.

Importantly, the AI-driven chatbot component draws from trusted, evidence-based sources, including NCCN guidelines and other reputable references. This allows it to function as a highly accessible, comprehensive knowledge resource that can support both patients and clinicians in real time. While no single clinician can rapidly access or recall the full breadth of this information at all times, the system can efficiently synthesize relevant guidance when needed.

That said, the role of the clinician remains essential. While the technology represents a meaningful advancement and a clear step in the right direction, it is not yet a substitute for clinical judgment. The human element, particularly through the virtual care center, continues to be a critical component of safe and effective patient care.

What have you learned and what new improvements are you excited about making?

During the pilot, we did encounter some challenges related to the clinical algorithms embedded within the system. These were not issues of clinical inaccuracy, but rather a lack of full alignment with our institution-specific standards of care. At comprehensive cancer centers, clinical workflows and management approaches are often highly tailored, and there is a strong preference for consistency with internal protocols.

One of our early and important lessons, therefore, was the need to harmonize the platform’s treatment algorithms with our own established clinical practices. Ensuring this alignment was critical not only for maintaining provider confidence in the system, but also for delivering care that reflects the specific expertise and standards of our institution.

Patient feedback highlighted an important opportunity for improvement: many patients felt that their providers were not sufficiently familiar with the program. This underscored the need for stronger internal communication and more comprehensive provider education to support awareness and adoption.

In addition, simplifying the enrollment process emerged as a key priority. During the pilot, the number of patients referred to the program was relatively small, and those who were enrolled through consultation pathways tended to be patients already experiencing significant symptom burden. Notably, this group demonstrated the highest levels of engagement, reinforcing the value of the program for patients with active needs.

In response to these insights, we expanded enrollment pathways to improve access. Patients are now able to enroll more easily, and we have also enabled nursing staff to directly enroll patients into the program. This change reflects the reality of clinical workflows, where nurses often spend more time with patients and have greater visibility into their symptom burden.

For example, patients presenting to the infusion center may not always see their oncologist during a visit but will routinely interact with nursing staff. By empowering nurses to identify and enroll patients who are struggling, we are better positioned to capture those who would most benefit from additional support and ensure more timely intervention.

Another important lesson from the pilot was the need to incorporate more objective measures into the system. Currently, the platform relies primarily on patient-reported symptom data, with twice-weekly symptom cluster assessments and a weekly ePRO survey. While these inputs are valuable, they are inherently subjective.

Looking ahead, we are interested in augmenting the platform with remote monitoring capabilities to capture objective clinical data. Integrating physiologic measures such as vital signs could provide a more comprehensive and precise view of a patient’s condition. Feeding this information back into the system would enhance its ability to detect early clinical deterioration, improve risk stratification, and support more informed decision-making.

The combination of subjective symptom reporting and objective physiologic data represents an important next step in the evolution of the program and has the potential to further strengthen its impact on patient care.

One challenge we increasingly see at Moffitt, particularly as we intentionally expand our geographic reach, is that patients who live farther from the center may experience a significant decline in their condition between visits. By the time they return for an infusion or clinic appointment, they may already have deteriorated substantially and, in some cases, may have required care at a local emergency department for issues such as dehydration.

Incorporating remote vital sign monitoring into the program could help address this gap. For example, if the system could identify that a patient’s blood pressure was declining, their heart rate was increasing, and they were experiencing positional dizziness, we would have an opportunity to intervene much earlier rather than waiting for further clinical decline. Earlier detection of these objective warning signs could support more timely intervention, improve symptom management, and potentially prevent avoidable acute care utilization.

We are planning to partner with Reimagine Care to bring this capability online. They have been strong collaborators, and their willingness to co-develop solutions alongside us has been a key factor in the success and continued evolution of the program.

What are your final thoughts about the need for programs like these?

Patients and families today increasingly expect care to be responsive, accessible, and timely. We are living in an on-demand, consumer-driven environment where people are accustomed to receiving what they need quickly, and those expectations naturally extend to healthcare. Patients want answers when concerns arise, and they want support at the moment they need it most.

This model helps bridge an important gap for patients who are at home and struggling with symptoms or uncertainty. Traditional systems often do not provide the speed or immediacy needed to address these concerns effectively. By creating a more responsive and accessible support structure, this approach moves care delivery in the direction patients increasingly expect and deserve.

At present, patients are highly appreciative of having this capability available to them. Looking ahead, I believe this type of support will move from being viewed as an added benefit to becoming an expected component of cancer care. In that sense, we are moving in the right direction.