By 2022, projections estimate that 18 million people will be cancer survivors and by 2023, 2.3 million people will be newly diagnosed with cancer. But significant challenges remain when it comes to ensuring high-quality, patient-centered, and proactive care for this growing population.
Most oncologists agree that social determinants of health (SDOH) – such as financial, housing, and food insecurity; social isolation; addiction; access to transportation and patient health literacy – can significantly impact cancer patient outcomes. In fact, a growing body of research has shown that SDOHs are equally, if not more important in determining health outcomes than a patient’s genes.
A recent survey by Cardinal Health Specialty Solutions supports this assertion. 68% of the 160 oncologists surveyed said that more than half of their patients are negatively impacted by SDOH. Yet oncologists feel challenged to address the issues. The survey showed that 81% don’t have enough time with patients to adequately understand or address their SDOH needs.
In addition to impacting patient outcomes, SDOH can also have a significant impact on the cost of care.
In an era of value-based care, if healthcare providers can identify patients early with SDOH risk factors that leave them vulnerable to hospital readmissions or emergency room visits, they can intervene early to improve outcomes and prevent unnecessary costs from being incurred.
The challenge, according to Bruce Feinberg, DO, Chief Medical Officer at Cardinal Health Specialty Solutions, is identifying the patients who are most at risk. Most oncology practices have no practical, useful, or actional way to make sense of the seemingly endless patient data in their EMR. However, recently developed artificial intelligence (AI) solutions are providing a novel way to help practices uncover these hidden SDOH factors.
Recent research has demonstrated that machine learning trained on SDOH factors could accurately predict healthcare utilization, both for inpatient and emergency department settings — in the absence of clinical data. The AI solution in question used only publicly available and purchasable SDOH data to make its predictions. These results show a promising future for AI in empowering clinicians to address SDOH risk at both the individual and neighborhood level to improve population health and patient outcomes.
Cardinal Health Specialty Solutions recently teamed up with clinical AI developer Jvion to deliver an AI decision support tool, CORE™, that helps oncology specialists reveal hidden SDOH risk factors in their oncology patients. The solution helps oncologists more quickly identify patients who are at risk of mortality, emergency department visits, avoidable and multiple admissions, and other adverse outcomes, so they can provide interventional care sooner to change the outcome. The decision support tool also recommends individualized interventions that can improve patient outcomes.
Here are 3 ways AI can help practices better address SDOH
# 1 Identify high-risk patients earlier — and with less cost
Barry Russo, CEO of the Center for Cancer and Blood Disorders (CCBD) in Fort Worth, Texas, says that when his practice reviewed its value-based care arrangements, they realized that they had more than 1,000 at-risk patients who might benefit from case management support to preemptively address rehospitalizations and other health events. The challenge, according to Russo, was that his practice simply couldn’t afford to hire enough staff to focus on that many patients — and there was not a simple way to hone in on the patients who were most at risk.
Applying Jvion’s clinical AI technology-enabled CCBD to narrow its list of high-risk patients down to 200. The tool first analyzes 4,500+ clinical and socioeconomic factors to reveal hidden risks among cancer patients, as defined by seven vectors, including 30-day mortality risk, 30-day readmission risk, 6-month deterioration risk, avoidable admission risk, patient experience, and pain management, depression risk, and no-show risk. It then recommends interventions that address their exogenous risk factors, as well as their clinical risk.
Russo says the tool has allowed his case management staff to be more effective and efficient by focusing on the most vulnerable patients, and by enabling proactive support and outreach targeted for their needs. The goal is to improve outcomes and deliver care for those patients who need it most while reducing the costs of their care, which can skyrocket without the right interventions at the right times.
#2 Inform oncology workflow adjustments to focus on optimizing patient outcomes.
Once practices understand which patients need the most immediate intervention, they can adjust their workflows to enable rapid intervention for patients at the greatest risk. For example, when Northwest Medical Specialties (NWMS) in Tacoma, WA, used the Jvion CORE clinical AI to better understand how many of its patients faced various risk factors, they recognized a need to provide same-day appointments and check-ins with their triage nurse. So they expanded their workflow to include an in-house urgent care clinic that operates five days a week. They also hired nurse practitioners and a full-time acute care provider to accommodate patients’ same-day appointment needs.
Similarly, the socioeconomic insights gained from clinical AI at CCBD led Russo’s team to hire a team of social workers to proactively reach out to patients who have been identified as high-risk, so they can determine whether intervention is necessary before the patient reaches the point of crisis.
Access to this type of nonclinical information allows oncology providers to take a more comprehensive approach to patient care, which in turn optimizes cancer treatment.
“Providing quality care to cancer patients entails much more than just treating the patient’s cancer,” said Feinberg. “It requires taking into consideration all of the socioeconomic impediments they are encountering in their everyday life, so oncologists can take action to help them overcome obstacles to successful cancer treatment.”
#3 Enable earlier end-of-life conversations and referrals
Managing palliative care for patients and referrals to hospice is one of the most challenging tasks that oncologists face. Research shows that patients are often referred too late, in part because oncologists struggle to determine when the time is right. As a result, many patients with terminal cancer undergo expensive and aggressive treatment late in their disease — often with little, if any, positive impact.
AI tools are helping oncology practices simplify — and take the influence of emotions out of — challenging palliative care decisions, by identifying the patients who would benefit most from early hospice or palliative care in their final stages of a terminal illness. By recommending a personalized end-of-life care plan that is based on the individual’s therapeutic needs and preferences, this functionality can help simplify the patient selection process for hospice, as well as palliative care.
Leveraging the power of AI to serve patients receiving end-of-life care has a particularly powerful opportunity to improve cancer care costs because costs of cancer care are highest as patients near the end of life.
After NWMS implemented clinical AI, its practice experienced an 81% increase in the number of patients it referred to palliative and hospice care.
Referring patients for hospice or palliative care is never an easy decision, but failing to do so when the patient could benefit ultimately worsens the patients’ end-of-life experience,” says Sibel Blau, MD, lead oncologist at NWMS. “AI helps us to make these difficult decisions, saving not only treatment costs but also unnecessary suffering and emotional turmoil for patients with terminal cancer and their caregivers.”
Addressing SDOH in oncology has arguably never been as important as it is now, as millions of patients are deferring cancer screenings due to the COVID-19 pandemic, and the resulting economic devastation has made SDOH factors even more acute for many Americans, according to Jvion Chief Medical Information Officer John Frownfelter, MD, FACP. Already, the U.S. has seen 37% fewer cancer diagnoses than at the same point last year, and experts are predicting 34,000 excess deaths in the U.S. due to reduced cancer care during the pandemic.
While AI tools may not be a silver bullet for addressing all of these care challenges, the potential to more effectively identify and support patients affected by SDOH is unquestionable. As oncology practices continue to focus on how they can deliver higher quality care to more patients at a lower cost, clinical AI technology will become increasingly essential.
 Institute of Medicine. Delivering High-Quality Cancer Care: Charting a New Course for a System in Crisis. The National Academies Press, Washington, DC, 2013.
 Chen S, Bergman D, Miller K, Kavanagh A, Frownfelter J, Showalter J. Using applied machine learning to predict healthcare utilization based on socioeconomic determinants of care. American Journal of Managed Care. 2020 Jan;26(1):26-31. doi: 10.37765/ajmc.2020.42142.
 Lai, A. G., Pasea, L., Banerjee, A., Denaxas, S., Katsoulis, M., Chang, W. H., … Hemingway, H. Estimating excess mortality in people with cancer and multimorbidity in the COVID-19 emergency. BMJ Open. 2020. https://doi.org/10.1101/2020.05.27.20083287