The oncology community is becoming increasingly aware of the effect of social determinants of health on cancer outcomes, and on the potential of big data to improve health and reduce disparities in care. To explore how the application of big data can help improve care and outcomes for vulnerable cancer patients with social needs, the National Cancer Policy Forum (NCPF) of the National Academies of Sciences, Engineering, and Medicine (NAS) held a workshop with invited speakers in Washington, D.C. The workshop, which also explored bias and the downside of big data, was hosted in collaboration with the NAS Committee on Applied and Theoretical Statistics. A written summary report on the workshop will follow.
The workshop followed the release of a report from a related NAS consensus study, "Integrating Social Care into the Delivery of Health Care: Moving Upstream to Improve the Nation's Health." That consensus study noted that health care systems are paying increased attention to social factors, such as access to stable housing, reliable transportation, and nutritious food. Other social factors include income, education, availability of safe areas for recreation and exercise, and family support. The study examined what factors are needed to integrate social services into the infrastructure of clinical care in order to improve care delivery. It also recommended that the federal government establish a 21st century social care digital infrastructure.
"We understand the power of big data, but we also understand the negatives of big data," said NCPF member and workshop chair Robert A.Winn, MD, who has just been named Director of the Virginia Commonwealth University Massey Cancer Center. Prior to his new appointment, he served as Professor of Medicine in the Division of Pulmonary and Critical Care Medicine; Director of the University of Illinois Cancer Center; and Associate Vice Chancellor for Community-Based Practices at the University of Illinois at Chicago. While social determinants of health can be used to correct health inequities, they can also be used to create stigma, he noted.
Asked by Oncology Times what his goals for the NCPF workshop were, Winn said they were threefold: 1. To identify how social determinants of health can be included in discussions of health care; 2. To identify how data on social determinants of health can be used as a catalyst in communities to improve health care; and 3. To emphasize the need for caution in how data on social determinants of health are used in order to protect people from negative consequences.
For example, Winn mentioned the practice of redlining discrimination, in which a health insurer could deny insurance policies to whole neighborhoods based on geography, race, and ethnicity-or sell insurance policies to those in such risk-targeted neighborhoods at higher rates.
"Social determinants have a significant impact on oncologic outcomes," stressed Loretta Erhunmwunsee, MD, Assistant Professor in the Division of Thoracic Surgery in the Department of Surgery at City of Hope. Social risk factors can have an impact on the aggressive aspect of cancers, she noted. Erhunmwunsee, whose research focuses on eliminating health inequity in thoracic oncology patients, said she and her colleagues have leveraged publicly available national statistical databases to understand neighborhood social factors and risk factors in lung cancer, such as income, education, and housing. She said that disparities in lung cancer incidence and outcomes persist along racial lines, especially among African Americans.
Erhunmwunsee's research using data from the CDC has shown the following results with regard to lung cancer:
* black men have the highest incidence and mortality rate of lung cancer;
* blacks are more likely to be diagnosed with lung cancer at later stages than whites;
* African American patients, both male and female, are less likely than whites to receive stage-appropriate cancer care, including surgery, radiation, and systemic therapy; and
* blacks are more likely to die from lung cancer than whites.
She has found that socioeconomic status is tied to tobacco addiction and poorer outcomes in lung cancer. Low income increases the risk of lung cancer and heightens the chance of dying from it-likely due to lack of access to care.
In addition to data from the CDC, she said she has also used data from the U.S. Census; American Cancer Society; EJSCREEN, a social justice mapping and screening tool of the Environmental Protection Agency that can pinpoint potential areas that may need further analysis and outreach; and the U.S. Department of Agriculture. Erhunmwunsee noted that, in the future, health providers need to bring together neighborhood risk factors with individual-level risk factors to get a complete picture of a cancer patient's status.
"We have separated the individual patient from the disease," said Sean Khozin, MD, MPH, Associate Director of the Oncology Center of Excellence at the FDA. Doing so, he said, has resulted in a kind of "molecular reductionism" which results in a narrow understanding underlying the mechanisms of oncogenesis. "Some of the endpoints we're measuring may not be those that mean the most to patients."
Khozin advocated using data on social determinants of health to achieve a more holistic profile of a cancer patient-the patient's exposome, which represents the totality of human environmental (non-genetic) exposures, such as stress and diet, from conception onwards. He added that new clinical trial designs are needed that use technology to go beyond the boundaries of traditional clinical trials in order to capture real-world data of importance to patients, including social risk factors. For example, he said, the use of biosensors can capture facial and voice recognition to help assess pain and mood.
The issue of trust regarding social determinants of health came up repeatedly during the workshop. Machine learning in health care-which has the capacity to create smart systems through big data-requires the trust of patients, communities, and health care providers, said David Steiner, MD, PhD, Senior Clinical Research Scientist with Google Health and Adjunct Clinical Assistant Professor at Stanford University. Actionable information from artificial intelligence (AI) used in clinical practice, including social data, must be considered accurate, representative, and free of bias, he said.
Agreeing was three-time cancer survivor Gwen Darien, Executive Vice President for Patient Advocacy and Engagement at the Patient Advocate Foundation. "Trust is incredibly important," said Darien, noting that what is needed is a system that trusts patients and vice versa. Some patients may be so suspicious of the misuse of social data that they don't want to give their zip code, for example. She said that, in addition to trust, using big data to correct social disparities takes political and social will and strong privacy protections.
Eighteen posters presented at the NCPF meeting represented a variety of approaches to using big data on social determinants of health to decrease health disparities. They included "Disparities in toxic heavy metal burden and breast cancer risk: findings from the Metropolitan Chicago Breast Cancer Registry"; "Hidden figures: An example of using machine learning to prioritize cervical cancer screening outreach"; "Lessons from Dana-Farber Cancer Institute data about the double burden of cancer and social determinants of health: Aligning big data with local data to promote health equity"; "What can we learn about health inequity from satellite imagery?"; and "Addressing cancer disparity using artificial intelligence: Randomized clinical trial to increase physical activity in cancer survivors using intelligent voice assist (Amazon Alexa) for patient coaching."
The following issues emerged in discussions at the workshop.
* Big data on social factors can be valuable if used in ways that help lower cancer risk in certain groups and thus help prevent cancers. Examples would be increasing access to healthy food and creating places for walking and exercise in low-resource neighborhoods. "Our goal is to find modifiable risk factors," said Beth Virnig, PhD, MPH, Professor of Health Policy and Management and Senior Associate Dean for Academic Affairs and Research at the University of Minnesota's School of Public Health.
* If big data on social determinants of health are to be integrated into clinical care, it must be done in a way that does not overload oncologists who are already overworked and pressed for time with one more administrative burden. For example, inclusion of such data in the electronic health record must be simple and user-friendly. On the other hand, the FDA's Khozin noted that one of the causes of physician burnout is socioeconomic factors of patients he or she can do nothing about; so being able to use social data to take action could actually reduce burnout. In a separate project, the National Academy of Medicine's Action Collaborative on Clinician Well-Being and Resilience is studying the increasingly recognized problem of clinician burnout.
* The use of an AI algorithm on social determinants of health should never conflict with established clinical practice guidelines in oncology. AI democratizes expertise but it does not understand context, noted Google Health's Steiner, so it cannot replace the knowledge in a human brain gained from clinical experience with multiple patients.
* The use of social workers who are an accepted part of the cancer care team can help integrate the social determinants of health into cancer clinical care. Their services should be considered of value and eligible for reimbursement by payers.
* Patient navigators can be helpful in guiding vulnerable cancer patients with social needs through the complex cancer care system. As previously reported by Oncology Times, navigators are especially useful in working with patients who have low health literacy.
* When data on social determinants of health identify cancer patients with social needs, communities and neighborhoods should be engaged to help them-for example, through referrals that create linkages to social services. There should be interventions to provide needy patients with stable housing, low-cost or free transportation to medical appointments, and access to nourishing food.
* The integration of social determinants of health into clinical cancer care must always honor the ethnic and cultural background of the individual patient. For example, Native Americans respect tribal medicine and the role of tribal elders, and these factors must be taken into account in their treatment, said Nicole Stern, MD, FACP, a member of the Mescalero Apache tribe who is a staff physician and shareholder at a large non-profit outpatient health care organization in Santa Barbara, Calif., and Past President of the Association of American Indian Physicians (AAIP). The AAIP holds an annual cross-cultural medicine workshop to foster understanding and collaboration between western and traditional medicine.
Peggy Eastman is a contributing writer.