Cancer and Race: Capturing COVID-19 Risk Factors Using the COVID-19 Symptom Study App

The emergence of SARS-CoV-2 and its ability to rapidly spread across the globe has made the collection of population-scale data, to help tackle the pandemic, exceptionally challenging. The COronavirus Pandemic Epidemiology (COPE) Consortium was established to unite scientists with big-data and epidemiology expertise and subsequently develop the COVID-19 Symptom Study app.

During the recent American Association for Cancer Research (AACR) three-day virtual COVID-19 and cancer meeting, David A. Drew, of Massachusetts General Hospital and Harvard Medical School, presented on how smartphone applications are able to facilitate the collection of self-reported data at scale. He discusses how data collected using the COVID-19 Symptom Study app has helped investigate COVID-19’s impact on racial and ethnic minority communities, and those living with cancer.

A closer look at the COVID-19 Symptom Study app

The COVID-19 Symptom Study app was launched on March 24, 2020, and is now used by ~4 million people in the United States, United Kingdom and Sweden. Users can self-report on their general health daily. This allows for the large-scale collection of data which can be used to help build a more comprehensive scientific understanding of SARS-CoV-2, track the symptoms related to COVID-19 and identify the risk factors and disparities related to infection.

“The COVID symptoms study uses a smartphone app previously known as the COVID symptom tracker to allow individuals to self-report a range of factors relevant to COVID-19 epidemiologic research, and this app was developed in conjunction with our collaborators at King's College London,” – David A. Drew, Massachusetts General Hospital and Harvard Medical School.

Drew explains that the users follow a logistical survey flow. They begin by providing informed consent and key baseline characteristics including information about existing medical conditions (e.g. cancer) that have been associated as significant comorbidities, affecting COVID-19 and related outcomes.

The app also asks users to share basic socio-demographics (e.g. age, weight, height, ethnicity and gender). Users can provide data on whether they work in a healthcare setting and can share on a daily basis any possible COVID-19-related symptoms they are experiencing.

“They're also queried about any testing they've received, any related results, any possible treatments that they've received or sought, and how much they are socially distancing, or any other possible community exposures or preventative measures they've taken,” said Drew.

Predicting COVID-19 infection and geographical “hotspots” with real-time tracking of self-reported symptoms

Drew and colleagues recently published in Nature Medicine on the ability of the app to determine whether loss of smell and taste is specific to COVID-19. Of the 18,401 app users that had been tested for SARS-CoV-2, the proportion who reported loss of smell and taste was higher in those users with a positive result (65.03% of individuals) compared to those with a negative result (21.71% of individuals) (odds ratio = 6.74; 95% confidence interval = 6.31–7.21). Statistical modeling was used to determine that 17.42% of participating app users were likely to have COVID-19.

“Largely this [research] has been driven by the strong association with nausea or the loss of sense of taste and smell with COVID-19 incidents,” explained Drew.

During his talk, Drew also touched on their second study published in Science, as it details the methodology used. The study reported that “the prevalence of combinations of symptoms (three or more), including fatigue and cough, followed by diarrhea, fever, and/or anosmia, was predictive of a positive test verification for SARS-CoV-2.” The mathematical modeling approach used in the study was able to predict geographical hotspots of incidence 5 to 7 days before official public health reports.

“It [the Science study] demonstrates a proof of principle that data collected from a mobile phone-based population-scale survey, like ours, can allow for real-time identification of hotspots in regions ahead of standard testing, especially when this testing is limited or hard to mobilize,” said Drew.

The COVID-19 Symptom Study app and participants with cancer

Users of the app were first asked about whether they were living with cancer on March 29, 2020. Between March 29 and May 8, app users’ data was collected and analyzed to establish the baseline characteristics of the study population across three countries – the United Kingdom, United States and Sweden.

“The majority of app users are female and span a large range of age and weight ranges with the median age hovering around 44 years of age,” explained Drew.

Drew notes that currently they “have modest representation on racial and ethnic minorities and have been working to improve recruitment of these critically under-served populations.”

However, recently the app was launched in Spanish. Since this update (statistical model two – which is the fully adjusted model), Drew says that they have > 20,000 users who have reported living with cancer at baseline.

“Here we adjust for personal characteristics, including BMI, sex, smoking status, comorbidities, at baseline, and frailty and additional risk factors including community interactions with known or suspected COVID-19 cases.”

He also explains that they adjusted for “whether an individual was a healthcare worker, which we and others, have previously shown to be a very strong risk factor for COVID-19.”

An increased risk of ~88% for testing positive for COVID-19 was observed among app users with cancer, compared to individuals that were cancer free. “Similarly, we see greatly increased risks among those going to be receiving chemotherapy or immunotherapy among cancer patients,” says Drew. In fact, chemotherapy and/or immunotherapy was associated with a two-fold increased risk of COVID-19 and risk of COVID-related hospitalization.

Drew also notes that the link between having a diagnosis of cancer and the odds of testing positive for COVID-19 was stronger amongst app users that were ≥ 65 years of age, compared to younger participants. Data also indicated that males were more likely to test positive for COVID-19 compared to females.

“Those that are living with cancer were simply more likely to seek treatment for COVID-19 in a hospital or clinic setting,” said Drew.

Exploring COVID-19’s impact on racial and ethnic minority communities

Drew and colleagues used the data provided by app users to investigate whether the risk of COVID-19 infection was higher among ethnic minorities.

Drew explains that “with the caveat that our cohort is predominantly non-Hispanic white, we similarly see strong risks for testing positive for COVID-19 in the US, among the Hispanic and Latinx, Black and Asian communities compared to non-Hispanic white participants.”

“We're increasingly aware of the disproportionate impacts COVID-19 is having on our racial and ethnic minority communities… as more and more data is collected, the disparities are staggering,” – David A. Drew.

The following self-reported cases of positive COVID-19 testing were documented:

• 8,990 cases among 2,304,472 non-Hispanic white participants
• ·3 cases among 19,498 Hispanic participants
• 204 cases among 19,498 black participants
• 608 cases among 64,429 Asian participants
• 352 cases among 65,046 mixed race/other racial minorities

“We observe a greater than two-fold likelihood of reporting being COVID-19 positive among minority groups, and the highest risks are observed among black individuals,” said Drew.

Drew noted that whilst other groups have reported, in hypothesis, that the supposed risks among minority groups may be due to the increased rates of comorbidities among these populations in the United States, Drew and colleagues saw very little impact on the risk estimates after adjustment for diabetes and heart disease, which are typically higher among black, Hispanic and/or Latinx populations.

Compared with non-Hispanic white app users, after accounting for risk factors (infection, comorbidities, and sociodemographic characteristics), the adjusted odds ratios for racial minorities were 1.37 (95% CI 1.09–1.72) for Hispanic participants, 1.42 (95% CI 1.23–1.64) for black participants, 1.44 (95% CI 1.33–1.57) for Asian participants, and 1.18 (95% CI 1.06–1.32) for the mixed-race/other minority group.

Due to differences in the way race and ethnicity questions are typically asked in the United Kingdom Drew explains that “we also were able to observe significant risks among Middle Eastern and South Asian individuals.”

Study design and limitations

“Our study has several strengths and limitations,” noted Drew. He explains that the smartphone app has enabled the rapid development of a large, prospectively collected longitudinal dataset.

“Beyond the general public, we also recruited individuals and ongoing cohort studies and clinical trials,” said Drew. Where possible the team has tried to leverage studies that are particularly enriched for minorities or participants with cancer.

“The mobile app allows us to be nimble and adaptable to address emerging questions during a rapidly evolving pandemic,” – David A. Drew.

The team has had to be mindful of participant burden and strategies have been developed, such as pushing regular software updates to participants, to encourage reporting.

“Importantly, we are aware that our study may not be generalizable sample. The population may have limited access to smartphone applications, which could be a barrier to steady access and the individuals who are likely to volunteer for a survey-based study may impact our results. We hope to better address these limitations in our ongoing and future work,” concluded Drew.

References

Drew, et al. (2020). Abstract: S09-01 Cancer and race: Two important risk factors for COVID-19 incidence as captured by the COVID Symptom Study real-time epidemiology tool. AACR COVID-19 and Cancer Meeting.

Menni, et al.(2020). Real-time tracking of self-reported symptoms to predict potential COVID-19.Nat Med. DOI: https://doi.org/10.1038/s41591-020-0916-2

Drew, et al. (2020). Rapid implementation of mobile technology for real-time epidemiology of COVID-19. Science. DOI: https://doi.org/10.1126/science.abc0473