In cancer care, the medical journey already carries enough weight – the diagnoses, treatments, side effects, uncertainty. But under that surface often lie social realities: transportation constraints, food insecurity, housing stress, care responsibilities, emotional burden, financial strain. These social determinants of health (SDoH) can deeply influence outcomes, adherence, side-effect tolerance, and even treatment decisions. And yet, they often go unspoken.
At Reimagine Care, our model gives patients access to AI-enabled virtual support, wrapped in human oversight. In practice, patients sometimes reveal things to an AI assistant – in a text thread or symptom-checker – that they might hesitate to share face-to-face with a clinician. This dynamic opens a window for clinicians and support teams to proactively intervene on issues that matter just as much as their biology.
Why AI can surface what’s hidden
One reason: AI doesn’t carry judgment and can accept disclosure in discreet, private settings at any hour. A patient might feel safer saying, “I skipped my medication this week because I couldn’t afford the co-pay,” or “I live two hours from the clinic and struggle with transport” in a messaging interface than in a physician visit. That kind of disclosure matters. As studies show, some AI models trained to detect SDoH from text in medical records often identify far more than existing diagnostic codes alone – one tuned model detected 93.8% of adverse SDoH from notes, versus just 2% being captured in ICD-10 codes.
In the virtual cancer care domain, this could look like a patient texting to report worsening fatigue, alongside a note that they recently lost income or had to skip meals. The AI system can flag that as a risk signal. That flag triggers a clinician or social support team to follow up: “Tell me more about how you’re managing food or expenses?” or “Let me connect you to transport assistance or financial counseling.”
How Reimagine Care’s model is primed for this
Our virtual-first, AI-augmented workflow already thrives on continuous dialogue. We use a messaging-first assistant (Remi) to collect symptoms, guide basic triage, escalate clinically when needed, and keep instructions in thread form so patients and caregivers can revisit them. Because that infrastructure is always listening, we can flag SDoH like financial stress or transportation issues back to our provider partners immediately.
The promise and the caveats
Leveraging AI isn’t foolproof. It has the potential to misinterpret or overreach, especially when lexical nuance or context is missing. Ethical guardrails must be built in: patient consent, transparency about how disclosures are used, opt-outs, and human review of flagged content. Documentation must be handled sensitively to avoid stigmatizing patients.
Moreover, AI might surface the need, but the system must respond. Flagging food insecurity is only helpful if someone can connect the patient to a service or resource. That means operations and care teams must be aligned to act.
Finally, privacy, bias, and fairness must be front and center. Large language models have shown promise extracting SDoH across categories (employment, housing, transportation, social support) with macro-F1 scores above 0.7 when tuned correctly, outperforming generic models. However, they must be evaluated across demographics to ensure they don’t replicate or amplify disparities.
Why this matters in cancer care
Cancer patients already face mounting stressors: side effects, travel for infusions, nutrition challenges, caregiver demands, job changes. These social burdens can tip a patient from manageable toxicity to hospitalization. If AI helps capture those social stressors earlier, care teams can mitigate risk before it becomes a crisis.
In our own data, Reimagine Care patients show a 30% reduction in emergency visits versus national benchmarks, and up to 64% fewer among older patients. Integrating SDoH insights could boost that effect by surfacing nonclinical drivers that worsen symptoms or adherence.
Looking ahead
The “next frontier” in virtual care isn’t just better symptom triage or faster response – it’s holistic awareness. AI tools that detect social context ethically and smartly can enable more personalized, equitable care. For oncology care paradigms to evolve, we need to build systems that can see the whole person: their biology, their symptoms, and their life context.
If we do it right, AI can help unlock a layer of care that has long been hidden and, in doing so, make cancer care more responsive, fair, and human.