Routine screening procedures for breast, colon, and cervical cancers in the first half of 2021 have failed to recover, falling by about a third below historical baselines, even as Americans are resuming normal activities.
Despite a brief recovery in late 2020, screening rates for common cancers continued to plummet in 2021, according to a recent study conducted by the Epic Health Research Network, an eponymous research entity within Epic Systems, a healthcare software giant and dominant vendor of electronic medical record products.
By June 2021—compared to pre-pandemic levels in early 2020—weekly volumes for breast cancer screenings were 29% below the historical average, colon cancer screenings were 36% below, and cervical cancer screenings were 35% below, a July update demonstrates. The screening rates in June were lower compared to March.
“This surprised me. I was really hopeful that all of the trends that we saw coming up from 2020 were going to continue and go on in 2021,” Jacqueline Gerhart, a clinical informaticist at Epic and an associate professor of medicine at the University of Wisconsin School of Medicine and Public Health, said to The Cancer Letter.
The study draws data from Cosmos, a massive health records database—arguably one of the largest in the world, given Epic’s market share—of over 113 million patients from 126 organizations. The company reported $3.3 billion in revenue for the 2020 fiscal year.
Breast cancer screening rates
Colon cancer screening rates
Cervical cancer screening rates
These data come from Cosmos, a HIPAA Limited Data Set of more than 107 million patients contributed by Epic customers. This study expanded on the data collected in two previous EHRN studies listed above. This study is based on data from 113 healthcare organizations representing 442 hospitals covering 70 million patients.
“Coming out of a very challenging 2020, where the drop in screening was precipitous due to the pandemic, we were hoping to see a turnaround in screening levels this year with a return closer to normal levels, thanks to vaccines and better COVID control,” NCI Director Ned Sharpless said to The Cancer Letter. “We have seen evidence that this recovery was actually underway, even in late 2020, but seeing these new trends now is quite concerning, especially given this is such a large, real-world evidence database.”
Epic has more than 250 million patient records, which means nearly half of those records are available in Cosmos.
“The whole goal of EHRN is to try to get good data out quickly,” Gerhart said. “The thing we most care about is getting the information in the hands of the people that can make a difference, whether that be the clinicians at the front line, Congress, people in public health, or thought leaders throughout the world.
“Over 66% of patients in the U.S. have a chart in Epic.”
Epic’s reach means that EHRN researchers can conduct population-level studies at a moment’s notice, and with increasing granularity, as the data continue to be aggregated and curated according to standardized definitions and common elements.
“I would encourage you and anyone who is reading this to consider the different things that influence whether [cancer patients] go to the physician or not, and when they’re doing an appointment,” Gerhart said.
With EHRN’s accelerated processing of Cosmos’s retrospective and real-world data, Gerhart and her colleagues are able to reduce research lag time and generate findings rapidly.
“Our goal is to really get it out in weeks rather than months,” Gerhart said. “And the concept is high-level, looking at large populations with the depth and breadth of data that we have. So, key findings are things like, did cancer screenings fall? Are people utilizing lung cancer screenings?
“Our study specifically looked at completed screenings,” Gerhart said. “As things were closed down, people weren’t getting them, which is what we saw—delayed cancer screenings. But now as things open up, we also have to think about, ‘Okay, all of health care has been shut down for a while, what are some of the most urgent and important things that we need to get in?’ Cancer screening absolutely is one of them.”
These findings from Epic are important, NCI’s Sharpless said.
“I think the next step is to look very closely at this data and compare it to what we are seeing from other similar sources to get a clearer picture of where we stand,” Sharpless said. “I believe we can learn a lot more through this study by comparing data from the NCI, as well as the CDC, and even from studies looking at Canada and the U.K.
“I think this study is a key snapshot of a much bigger picture we need to frame on pandemic-related cancer care and screening.”
In a separate study, Gerhart and her team found that an overwhelming majority of patients—nearly 9 in 10—who are eligible for lung cancer screening aren’t being screened, despite having insurance coverage.
According to the study, among Medicare and Medicaid patients who are eligible for lung cancer screening, only 12.9% received the procedure in 2019 and 2020. The Medicaid rate is marginally higher, at 15.7% compared to 12.5% for Medicare patients.
Lung cancer screening rates for eligible patients with coverage through Medicaid or Medicare
It’s unclear how the COVID-19 pandemic affected lung cancer screening rates, given that uptake of lung cancer screening has been historically low. Also, authoritative guidelines recommending screening—and reimbursement—for lung cancer are relatively recent.
In July 2020, the United States Preventive Services Task Force published a draft recommending lung cancer screening for people ages 50-80 years, who have a 20 pack-year smoking history and currently smoke or have quit within the past 15 years (The Cancer Letter, July 10, 2020). The task force finalized this guideline, assigning it a grade “B” recommendation in March 2021.
As the decrease in lung cancer deaths continues to drive annual drops in overall cancer mortality rates, it is imperative for researchers to understand the role of screening in improving survival in the real world.
As it stands, the relationship between lung cancer survival and mortality is complex, because, aside from recent survival gains resulting from new treatments, screening is likely to have some impact on mortality rates—a calculus that will change as screening rates go up (The Cancer Letter, Feb. 7, Sept. 4, 2020; Jan. 15, 2021).
“I do think [our study] is one of the most comprehensive ones that has looked at lung cancer to date,” Gerhart said. “We are trying to add value in the public health space—that is, observing what trends have happened.”
Epic, EHRN, and Cosmos
The study on breast, colon, and cervical cancer screening rates derived data from 113 healthcare organizations representing 442 hospitals covering 70 million patients. The lung cancer screening study includes 381,849 patients eligible for the procedure.
“EHRN is a place where we publish studies from Cosmos,” Gerhart said. “Those patients are from healthcare customers of Epic, and they choose to voluntarily send a HIPAA limited dataset into Cosmos. And then, that data, usually a few weeks after it’s been submitted, is able to be reviewed, not only by the data team within Epic, but also, our customers who have access to Cosmos can review it themselves as well.”
According to 2019 numbers from KLAS and HIMSS Analytics, Epic’s market share includes:
- 31% of multispecialty acute care hospitals,
- 42% of acute care beds in the U.S., and
- 67% of ambulatory physicians that are part of health systems.
According to Epic, the majority of U.S. News & World Report’s top-ranked hospitals and medical schools use Epic. Over 90% of medical students and residents train on Epic, and the top 15 recipients of NIH grants by funding are Epic community members. The company doesn’t share specific information about the academic-to-community ratio of its member hospitals.
Cosmos, established Aug. 13, 2018, contains about 33,000 cancer diagnoses and data from about 5.8 million cancer patients. According to Epic, up to 3.7 million of these patients have active disease. The company doesn’t track the proportion of the U.S. cancer population that seek treatment at member hospitals.
Experts who use simulation modeling predict significant increased mortality from the delays in screening and treatment caused by COVID-19—NCI researchers anticipate at least 10,000 excess deaths, for instance, from breast and colorectal cancers over 10 years (The Cancer Letter, June 9, 2020).
For illustration, early 2020 data compiled by Flatiron Health and made available exclusively to The Cancer Letter showed that visits from new patients, per week, decreased by about 3,000 from over 8,000 in early February to less than 5,000 toward the end of April last year, a nearly 40% drop (The Cancer Letter, May 1, 2020).
Although the 2021 cancer screening rates have significantly improved since the precipitous drop going into May 2020—94% decrease in colon and breast cancer screenings, and 86% decrease in cervical cancer screenings—researchers expect additional mortality if cancer screening rates continue to remain below the historical baseline.
“As I have pointed out, delayed diagnosis and delayed care from the pandemic will increase mortality for years to come and, if we neglect cancer care, we may end up trading one health crisis for another,” NCI’s Sharpless said. “We have to find ways now to get people back on course with cancer screening and treatment to prevent a new public health crisis.”
Focusing on clinical decision support
In the interest of rapid dissemination of findings, EHRN papers don’t go through a formal peer review process before publication.
To reduce errors, EHRN researchers use a two-team process to compare conclusions that are reached independently.
“We have a team of about 10 physicians that work on EHRN studies; we also have data scientists and software developers,” Gerhart said. “And whenever we do a study, we always use a two-team approach. In peer-reviewed literature, we often will have a randomized controlled trial or so forth. We don’t have the ability to do that, we’re really looking at retrospective data, patient data.
“The goal is to make sure that when we’re pulling that code and pulling that information, that we’re doing it accurately,” Gerhart said. “So, we ask the same question of two independent teams. They research that question completely independently, and then we come back together to validate the findings—is there some bias that we might be missing? Is there something else that we didn’t look at?”
Gerhart hopes to build collaborations between EHRN and other healthcare organizations, providers, and patient groups who are interested in answering specific research questions.
Seeing these new trends now is quite concerning, especially given this is such a large, real-world evidence database. We have to find ways now to get people back on course with cancer screening and treatment to prevent a new public health crisis.Ned Sharpless
“The information that we have in Cosmos is being used by other organizations for research,” Gerhart said. “For example, just recently, we worked with Penn and Yale to publish in the Journal of the American College of Gynecology and Obstetrics.
“We do not have a budget. It’s just not part of our process,” Gerhart said. “If we feel something is an ethical approach, if we feel like something really needs the extra resources and time and energy, we put our money there—83% of Epic’s operating budget is staff.”
As a clinician at Epic and a researcher at EHRN, building and improving the organization’s clinical decision support infrastructure is Gerhart’s main priority.
“That’s the whole point here,” she said. “It’s not only just publishing these studies on EHRN, it’s also helping to show the power of Cosmos and the power of just having a large data set that can actually do good.”
Epic’s EMR suite can identify patients who meet the eligibility criteria for a lung CT scan, including with data on pack-year smoking history, and alert providers to take action.
“With this kind of data, I can say, ‘Hey, all you health systems who have decided to not turn on the alert, or not link the order directly to the alert, or not check the patient’s insurance beforehand, do it so that you can make it easier on the lives of patients and clinicians.’ The patient might not even know that they are eligible for screening.
“There’s always a signal-to-noise ratio,” she said. “You don’t want to have clinicians get multiple alerts because they’ll start to ignore them, but we want them to have the ones that are most valuable to actually care for their patients.
“So, in this case, this would be very valuable to me. I’d be potentially helping to screen for cancer that could potentially save a life.”