Combining real world evidence with artificial intelligence may improve visibility of treatment options for oncologists
An estimated 2.1 million new cancer cases have cropped up in the US in 2018—and that number is expected to jump to 3.1 million by 2040.1 Providing care for these patients is an estimated 11,700 clinical oncologists—that’s 179 patients per oncologist.2 There is already a maldistribution of oncologists within the US in that many rural areas are underserved. An overall shortage of more than 2,300 medical oncologists in the United States is anticipated in 2025.3
With an impending shortage of oncologists, an exponentially growing body of literature, and a steady stream of new cancer cases, it’s no surprise that one study found significant cancer health disparities in the US.3 It is simply getting more challenging for oncologists to keep up with best practices while caring for an overflowing roster of patients.
Keeping up with best practices
In 2013, IBM® introduced Watson for Oncology to provide multiple forms of information to oncologists at the point of care. Watson for Oncology provides decision support for 13 cancers. These recommendations are the result of training by expert physicians at Memorial Sloan Kettering Cancer Center. In addition, Watson for Oncology provides an analysis of medical literature pertinent to the patient’s clinical scenario, information on drug dosing and toxicity, and a growing number of National Comprehensive Cancer Network Guidelines. Watson for Clinical Trial Matching can also reduce the time and effort involved in referring a patient for participation in a clinical trial. It is important to remember that Watson for Oncology does not diagnose patients, and it does not make treatment decisions. Rather, Watson for Oncology supports clinicians as they make the best possible therapeutic recommendation for that patient.
Many concordance studies have been conducted worldwide, demonstrating a variable level of concordance between Watson for Oncology recommendations and those of local multi-disciplinary teams. When concordance is not high, this can be the result of differences in local drug availability, differences in the scenarios with which patients typically present, and/or differences in the natural history of certain diseases among geographic regions.
Enhancing Watson for Oncology with real world evidence
Earlier this year, IBM Watson HealthTM began working with Cota Healthcare, a real-world data and analytics company, supplementing Watson for Oncology with real world evidence. This new source of information enables the comparison of clinical outcomes, toxicity, and cost data across clinically similar patients—to further guide analysis and individualized treatment recommendations.
This functionality is being piloted currently at John Theurer Cancer Center (JTCC) in Hackensack, NJ, and IBM hopes to expand this to other centers in 2019. Eighty-eight early stage post-menopausal breast cancer cases were presented to three JTCC breast cancer experts without using Watson for Oncology.
The physician’s findings were then compared with Watson for Oncology recommendations. In 78 percent of the cases, the recommended treatment option from Watson for Oncology was concordant with that of the breast cancer experts. The “for consideration” option from Watson for Oncology was the same as the physicians’ in 9.4 percent of the cases. In the Cota historical database 19.3% had received treatment that was “not recommended” by Watson for Oncology.4
Watson for Oncology has the potential to educate physicians to help them provide state-of-the-art care, and to reduce practice variation. To make sure that Watson for Oncology accommodates the unique attributes of physician practices around the globe, there is a close partnership between Watson for Oncology and the physician using these tools. We value this feedback and continually modify the product to assure its ongoing adaptation to the real-world practice of cancer care.
 International Agency for Research on Cancer, Cancer Today Database. Accessed at http://gco.iarc.fr/tomorrow/home November 9, 2018.
 Aju Mathew. Global Survey of Clinical Oncology Workforce. Journal of Global Oncology 2018 :4, 1-12. Accessed at http://ascopubs.org/doi/abs/10.1200/JGO.17.00188 November 9, 2018.
 Yang W, Williams JH, Hogan PF, et al: Projected supply of and demand for oncologists and radiation oncologists through 2025: An aging, better-insured population will result in shortage. J Oncol Pract 10:39-45, 2014
 D M Atieh Graham, D M McNamara, S E Waintraub, S L Goldberg, A D Norden, J Hervey, A L Pecora, C Landstrom, J L Snowdon, P M Francis, N Jungbluth, C K Wang, L Latts; 1589P. Are treatment recommendations provided by cognitive computing supported by real world data (Watson for Oncology with Cota RWE) concordant with expert opinions?, Annals of Oncology, Volume 29, Issue suppl_8, 1 October 2018, mdy297.031, https://doi.org/10.1093/annonc/mdy297.031