Let's talk about policy.

Since the winter of 2020, The COVID-19 Data Project by BroadStreet has collected daily numbers of COVID-19 cases and deaths from states and their counties. We have noticed a disturbing pattern:

States are not reporting data for COVID-19 in a standardized process.

Having reliable data is crucial in decision-making processes, especially in times of Public Health emergencies such as the pandemic in which we are currently. In these situations, officials need to know what is happening within their jurisdiction. This is achievable if they are given accurate information.

Many people want to help find solutions. Infectious disease epidemiologists, researchers, and citizen scientists in The United States and around the world are willing and eager to help. Time is wasted if it is spent finding, assembling, and trouble-shooting missing and inconsistent data and methodologies. This leads to delays, and even inaccuracies, of critical information such as establishing true incidence, prevalence, and mortality rates. Ultimately, good data and clear, published methods would enable us to track the spread of the COVID-19 virus across the nation and to be proactive in creating effective strategies and policies.

Download our One-Pager

If you would like to share this 1-page and write to your representative

Download an editable letter here

Also, you can find your representatives’ contact information here: https://myreps.datamade.us/

Our Recommendations

We suggest a new policy for pandemic disease control. The policy will create a database where states and their counties report consistent information on:

Methods
Demographics: ethnicity, sex, and age of everyone tested.
Case numbers
Testing rates
Underlying conditions such as diabetes, cancer, asthma, etc.
Hospitalization rates
Recovery rates
Mortality rate
Utilization of limited resources such as: Equipment stock and ICU beds

All of these variables must have clear guidelines on their definition, how to collect the information, and how to report data. The database will also allow states to Indicate where data is sourced and provide the data to the public. This creates an open source of information for the public to analyze. This also gives scientists the opportunity to research new and emerging diseases in real-time so that they are able to provide critical new information to help slow the spread of our next pandemic.

We are not alone in thinking this.

Zylla and Hartman discuss the accuracy and reliability of state COVID19 data dashboards, common indicators, and best practices in their webpage State COVID-19 Data Dashboards. They warn that users should be cautious about comparing states with one another. Although state indicators may be common, there are discrepancies in the way these indicators are defined. For example, “hospitalizations” can be defined as the total number of cases hospitalized by one state, while, to another, it can be defined as the number of hospitalizations per day.

Commentary Regarding COVID-19 Data Inconsistencies

How Data Transparency Can Help Fight COVID-19
Scientists Warn CDC Testing Data Could Create Misleading Picture Of Pandemic Published as part of a series of controversies, The Centers for Disease Control and Prevention
How Virginia Juked Its COVID-19 Data
'COVID Tracking Project' Co-Founder On Misleading Virus Statistics
Comparison of Four Major News Sources
[Testing Data] Why It Matters | The COVID Tracking Project

CDC Resource on Data Collection:

The CDC Field Epidemiology Manual: Collecting Data

Examples of some inconsistencies:

Cherokee County in NC reports different numbers than the state. This is because Cherokee is reporting everyone physically located in the county, and North Carolina is reporting everyone who has their primary residence listed as being in Cherokee.
1. There is no consensus on what the definition of residency should be for this reporting

Some states include out-of-state cases in their figures somewhere (TN reports them separately, as an example), and other states report those cases to the home state of the carrier to include in their numbers.
1. The very first death reported by Alaska was an Alaskan resident who both acquired the virus and passed away while traveling.

There were also conflicting figures in PA deaths, detailed below. We actually used both figures for a while, meaning some cases and deaths were likely double-counted.
1. https://www.post-gazette.com/local/region/2020/05/25/Pennsylvania-coronavirus-deaths-county-coroners-death-counting-death-COVID-19/stories/202005220138

Various counties report probable cases even when their states don’t, resulting in two different pictures of how the virus is spreading. This occurs a few times in all states.
1. Benton-Franklin county in WA reported them for weeks (while the state did not), and then stopped in early June to be consistent with the state. This led to the appearance of a drop in a few hundred cases on 6/8.

Suffolk County in NY reported the sum of serum tests and PCR tests for several weeks without stating what they were doing, and without leaving a way for anyone in the public to interpret the numbers separately. Combining PCR tests and Serum tests creates an incredibly deceptive picture of how the virus is spreading by making it appear as though cases which were active months ago are currently active (e.g. it looks like there are 2,000 new cases on one day, but most of them were only contagious months ago). These should be reported as separate figures, not combined.

Connecticut removed nearly 400 duplicate cases from their state and county totals without updating historic data, meaning we have a drop of 400 cases in one day in the state without having any way of fixing historic totals. They presumably give the CDC the tools to sort them out, but the CDC has been moving fairly slowly (CDC has two sets of numbers, basically: fast, and slow. Slow numbers are more accurate, but they’re lagging a few weeks behind real-time because they need to check cases and determine the date of onset, get rid of duplicates, etc. Fast numbers are basically like what we’re recording.).

And Many More...

Authors:

Kathleen Lindsey

Bachelor’s degree in Public Health Science | University of Maryland, College Park

Master of Public Health: Epidemiology | University of Michigan, Ann Arbor

Miracle Onyeoziri

Bachelor’s degree in Nursing | University of Massachusetts, Dartmouth

Sarah Javaid

Bachelor’s degree in Neuroscience

Master of Public Health: Global Health Epidemiology | University of Michigan, Ann Arbor

Shelby Minas

Bachelor’s degree in Sociology | Portland State University, Portland