Use of Real-World Electronic Health Records to Estimate Risk, Risk Factors, and Disparities for COVID-19 in Patients With Cancer

The coronavirus disease 2019 (COVID-19) pandemic has introduced additional serious risks for patients with cancer and is deeply affecting current cancer care. Early data from observational studies have suggested high rates of hospitalization and early mortality.¹ However, even more granular data are needed to identify specific populations with cancer at greatest risk of serious outcomes from COVID-19, as well as to understand the risk factors that predispose patients to develop COVID-19.

In this issue of JAMA Oncology, Wang et al² present results from a retrospective case-control study using IBM Watson Health Explorys electronic health record (EHR) data to assess the association of cancer and demographic factors with COVID-19 outcomes. Among the 73 449 510 patients, 2 523 920 were identified with any cancer diagnosis and 273 140 with a recent cancer diagnosis (within 1 year before infection). Included patients had 1 of 13 common malignant neoplasms. Among 16 570 patients identified with COVID-19 as of August 14, 2020, 1200 had any cancer diagnosis and 690 had a recent cancer diagnosis. The study found that patients with any or a recent cancer diagnosis were at substantially increased risk for COVID-19 compared with patients without cancer (adjusted odds ratios [AORs], 1.46 [95% CI, 1.42-1.50] and 7.14 [95% CI, 6.91-7.39], respectively, adjusted for major COVID-19 risk factors). The recent cancer diagnoses with the greatest risk were leukemia (AOR, 12.16 [95% CI, 11.03-13.40]), non-Hodgkin lymphoma (AOR, 8.54 [95% CI, 7.80-9.36]), and lung cancer (AOR, 7.66 [95% CI, 7.07-8.29]). Higher rates of hospitalization (47.46%) and mortality (14.93%) were observed in those with concurrent diagnoses of recent cancer and COVID-19 than expected with either diagnosis alone, suggesting a potential synergistic effect.²

The first large-scale cancer cohort studies—the COVID-19 and Cancer Consortium (CCC-19)¹ and the UK Coronavirus Cancer Monitoring Project (UKCCMP)³—estimate mortality to be 17% to 28% across all patients with cancer and diagnosed with COVID-19, depending on the period of follow-up, proportion of long-term survivors, and pandemic-related health care challenges. Mortality in patients with lung cancer or hematologic malignant neoplasms and COVID-19 has been substantially higher. Currently confirmed risk factors for mortality among populations with COVID-19 and cancer include older age, male sex, a history of smoking, presence of multiple active comorbidities, performance status, and progressive cancer.^1,3 Larger, more granular data sets for each cancer type are needed to definitively assess the effect size of specific cancer therapies, abnormal laboratory findings, individual comorbid conditions, race, and high-risk metastatic sites (particularly the effect of lung metastases). Among patients with COVID-19 without cancer, obesity and certain comorbid conditions have also been reported as likely clinical risk factors for mortality.

Wang et al² report higher mortality among Black patients with COVID-19 without cancer (6.55% vs 4.43% for White patients; P < .001) after adjusting for comorbidities but without the ability to adjust for differences in socioeconomic parameters or health care access. Mortality difference for Black patients with cancer and COVID-19 did not reach statistical significance (18.52% vs 13.51% for White patients; P = .11), but the results are limited by small numbers. Interestingly, a previous study of hospitalized patients with COVID-19 without cancer⁴ demonstrated that mortality rates for Black patients were comparable to those for White patients after adjustment for both comorbidities and deprivation index, suggesting that observed differences are mainly owing to societal disparities rather than biology. In general, clinicians and scientists alike must be cautioned against inferring biology in studies without the proper adjustment of all major health determinants, which go beyond the standard socioeconomic indicators assessed in most studies.⁵

The authors also found a greater risk of developing COVID-19 among Black patients compared with White patients with cancer (AOR range, 1.58 [95% CI, 1.17-2.14] to 5.44 [95% CI, 4.69-6.31] across cancers). This outcome is comparable to findings reported by the Centers for Disease Control and Prevention, which estimates a several-fold increase among Black patients. These higher rates of contracting severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection may also largely be explained by social determinants, such as increased burden of comorbidities, crowded living conditions (inner cities, multigenerational homes, etc), dependence on public transportation or child care, and higher work-related exposures.⁶ Until such societal disparities are accounted for, we cannot presume these findings are caused by any inherent differences among racial groups.

The investigators also reported that women with colorectal cancer and non-Hodgkin lymphoma had a higher risk than men of developing COVID-19. In the general and cancer populations alike, men have been found to be disproportionately affected by higher infection rates and more severe COVID-19 outcomes compared with women. However, the Centers for Disease Control and Prevention reports that women dominate infection risk among US health care workers, suggesting that occupational exposure may partly account for the observed sex differences.

Numerous studies have established older age as a risk factor for COVID-19 mortality in populations with and without cancer, including devastating risks of developing and dying from COVID-19 among nursing home residents. In contrast, Wang and coauthors² suggest that age may not necessarily be associated with an increased risk of developing COVID-19. It remains unclear whether this risk may be related to older patients with cancer taking particular precautions to avoid SARS-CoV-2 exposure.

We commend the authors for this study, which furthers our understanding of the risk factors for COVID-19 and its complications among patients with cancer. The investigation of risk factors associated with diagnosis and the large sample size of the EHR data set are notable strengths of this study. However, the study findings are limited by the retrospective case-control study design and EHR data with limited information on the specific types of data (billing, laboratory, or other) used to establish COVID-19 diagnosis as well other medical conditions, data quality, and missingness. In addition, although the data set has good representation of the US population with cancer (approximately 15% of new cancer diagnoses), the representation of patients with COVID-19 was surprisingly low (approximately 0.3% of all patients). This may be related to lack of Systematized Nomenclature of Medicine–Clinical Terms COVID-19 incorporation into IBM Watson Health Explorys and use of the parent term coronavirus infection until recently, as well as yet-unrecognized limitations for COVID-19–related EHR data. It is also worth noting that race ascertainment with big data, especially Black race, can be limited by moderate specificity (approximately 70%) when compared with self-report.

Few racial disparities in health outcomes have been explained by germline genetic findings. Because deprivation and environmental factors can influence biology, any observed differences beyond the germline may only be attributed to race if they were to persist even within the same socioeconomic strata after adjustment for major risk factors for ill health.^5,7,8 Polite et al⁷ implore us when discussing health disparities that “What should no longer be tolerated is the misguided belief that the problem is too difficult to solve, cannot be solved, or that it is due to the affected person’s genes or inaction.”^7(p1664)

It is becoming quite clear that the solutions for disparities observed with COVID-19 relate primarily to public policy. Needed solutions require (1) large-scale, high-quality epidemiologic data; (2) policies that mitigate socioeconomic risk factors as well as health care access disparities; and (3) validated risk prediction tools to identify patients at greatest risk from COVID-19 and its complications. Clearly, the haunting spotlight of COVID-19 has dramatically illuminated known US health care and societal disparities. This situation should be a wake-up call that brings much-needed improvements in US equity policies, including but not limited to better health care access.⁸ Nothing appears more critical for alleviating these disparate clinical outcomes in this time of crisis and beyond.

Corresponding Author: Nicole M. Kuderer, MD, Advanced Cancer Research Group, 210 Market Street, Suite 303, Kirkland, WA 98033 (kuderern@gmail.com).

Published Online: December 10, 2020. doi:10.1001/jamaoncol.2020.5461

Author Contributions: Mr Desai and Dr Khaki contributed equally to this work.

Conflict of Interest Disclosures: Dr Khaki reported owning stock in Merck & Co and Sanofi SA within the last 2 years. Dr Kuderer reported receiving research consultant fees from G1 Therapeutics, Inc, Invitae Corporation, BeyondSpring Inc, Spectrum Pharmaceuticals, Inc, Bristol Myers Squibb, Janssen Pharmaceutica, and Total Health outside the submitted work and being a founding member of the COVID-19 and Cancer Consortium (CCC19) and former founding member of the CCC19 steering committee. No other disclosures were reported.

Funding Support: Dr Khaki was supported by training grant T32CA009515 from the National Cancer Institute.

Role of the Funder/Sponsor: The National Cancer Institute had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.