We recently published a post related to the beginnings of a current research project titled “The Unintended Consequences of COVID-19”. The focus of this research is on the medical consequences of the virus in terms of declining care in America resulting from an almost single-minded focus on COVID-19. The goal is to provide a model that can articulate the possible fatalities caused by myopic view of public health, and to highlight the fact that many more deaths might in fact come from the national response to the pandemic, rather than the pandemic itself. Mitigating actions must take place quickly to forestall such consequences.
This article focuses on a different set of consequences…. data accuracy and statistical analysis.
Data and statistics are the lifeblood of any medical research. They are used by a broad sweeping set of organizations that focus on research for diseases such as the American Cancer Society (ACS), the American Heart Association (AHA), and the American Diabetes Association, to name a few. Additionally, other research organizations, hospitals, and even pharmaceutical companies rely on the data to validate results of a range of therapies and protocols put into place. And of course, this is also a focus on the impacts of disease in underserved markets, within targeted communities, and across demographic social determinants.
But what happens when their data is wrong? Is COVID-19 skewing this data? If so, to what extent, and with what consequences?
To be sure, this is a very complex question.
Prior to COVID-19 there have been ongoing challenges in the data required for highly accurate analysis. Reporting is slow for many critical events, classifications of individuals vary by state (for example, the Native American community is sometimes listed separately and sometimes under a broader bucket of “other”. The level of granularity is also inconsistent. Some reporting is at the state level, some at the county level, some at the Census tract level. There are even different medical coding systems for inpatient and outpatient care (ICD and CPT). While there have been ongoing efforts attempting to address some of these issues, there is still a very long way to go.
As we all know by now, COVID-19 has impacted the elderly and those with chronic conditions (comorbid) the most. It is estimated that over 90% of all deaths have been to individuals with underlying comorbidities. Those individuals that would be associated with the ongoing research studies described above. Furthermore, while likely low as a percentage of the total number, one can assume that some portion of these deaths might have occurred regardless of a COVID-19 infection. But to what extent this is the case will never be known.
Guidance for identifying the cause of death related to COVID-19 has been evolving. While somewhat specific rules are in place today, these have not existed all along. And in fact, the general “guidance” has been inconsistent. Dr. Birx of the Coronavirus Task Force specifically announced that “If someone dies with COVID-19 we are counting that as a COVID-19 death”.
While this is not technically the guidance of the CDC, which lays out the need to define the sequence of conditions leading to death with COVID-19 being specified as the last item on the list (i.e. condition that began the chain reaction), many have argued that even this is not consistent with the prior reporting. This is in part since virtually every patient is now being tested for COVID-19, whereas in the case of the seasonal flu this would not have been the case. So, while it highlights the possible under-reporting of flu deaths each year (again, assuming this is the agreed upon method), it further highlights the potential disparate reporting statistics. Furthermore, deaths that took place prior to effective testing are now being adjusted to reflect COVID-19 if suspected to be caused by COVID-19.
Is there a financial incentive to identify hospital cases as COVID-19?
This has also become a topic of conversation. Medicare, who covers a vast majority of those who have suffered the greatest impact, does in fact pay more for COVID-19 cases in comparison to similar interventions for non-COVID-19 cases.
And, under the CARES act recently signed into law, the U.S. Department of Health and Human Services (HHS), will provide claims reimbursement to health care providers generally at Medicare rates for testing uninsured individuals for COVID-19 and treating uninsured individuals with COVID-19 diagnosis.
As a result, there are unintended financial incentives to improperly report COVID-19 cases. To be very clear, this is not to state that this is or has occurred, but to merely highlight the incentives to do so.
On the flip side, there is growing concern from others that the COVID-19 deaths are being underreported. The CDC has been providing a report of the number of deaths across the nation compared to the expected deaths based on historical trends. Over a 2-month period the actual death toll exceeded the expected number of deaths by over 10,000. Many posit that these deaths may well be undiagnosed COVID-19 deaths. Although there are other possible reasons for this increase in deaths due to the impact to our traditional healthcare services caused by the coronavirus pandemic. These other possible causes are highlighted in our study of the “Hidden Casualties of COVID-19”.
As highlighted above, the ability to attest to the accuracy of COVID-19 data is a challenge. This is not surprising while in the midst of a pandemic, and some of these data are likely to be modified in some manner in the future. But it does serve to highlight the fact that data is critical to how we approach public health and medicine in general. And the impacts of COVID-19 will be felt throughout the healthcare system well beyond the pandemic itself.
As of the writing of this article the death toll attributed to COVID-19 has just surpassed 100,000. A devastating number. While we will not know the total deaths in the US for 2020 for quite some time, the average number of deaths in the US approximates 2,800,000 individuals a year (or 860 deaths per 100,000).
While projections anticipated an increase in the number of deaths due to the aging of the US society, we will likely see an increase over projections due to the pandemic.
So, will this impact our nation’s ability to accurately use the data for long-term disease research? Will these numbers skew the overall statistics being tracked by researchers focused on specific diseases, targeted demographics, or underserved markets? The answer is not clear. However, the extent to which this is the case may depend on how researchers choose to use the data they receive.
It is important to note, however, that as a percent of total deaths the COVID-19 pandemic is still rather small. 100,000 deaths compared to the average of 2.8 million represents 3.6% of the overall deaths we anticipate, and roughly 31 deaths per 100,000. A large number to be sure, but perhaps less statistically impactful than one might initially fear. As we learn more of the specific comorbidities involved, we can further identify those conditions that will be most impacted. Currently it would appear that those with diabetes and hypertension have been impacted the most. Additionally, for those studying the impact to race and ethnicity it is also clear that deaths have impacted African Americans at a higher rate than others, and may have in fact impacted Native Americans at a rate considerably higher (although tracking this data can be challenging for a host of reasons).
Ultimately time will tell as to the impact of COVID-19 on the cause of death reporting, and the downstream impact on long-term studies. There will be an impact for sure, but in many cases, it may be somewhat negligible. In others, however, the impacts may be much more significant.