Ohio

At the beginning of March, Ohio announced it would begin reporting COVID-19 mortality information based on death certificate data from the NCHS. We understand from statements made by public health officials that this was motivated by a desire to provide more accurate information about the pandemic’s impact in the state.

We notice that other states (Oklahoma, Iowa) have also begun using NCHS mortality data. Given that so many states have switched recently, we’re interested in understanding how states and the CDC interact when it comes to issues like this: Has the CDC been in communication with Ohio about switching to NCHS data?

When we made the switch, it’s a little less timely than the disease reporting system but we went for the accuracy of the data and went with the death certificate. We weren’t directed by the CDC to do this. The NCHS data is collected on all deaths in Ohio and the nation, so that data source has always existed but since it’s delayed, we haven’t used it for official numbers. The NCHS ICD10 coded data - that is the gold standard for death certification. Every death certificate is sent to NCHS and they have one underlying death code and they might also have multiple other death codes. They send it back to us and we do some data quality checks and then we post it on the dashboards. So there might be some lags since those are only updated twice a week for mortality. And i’d check to see if NCHS is looking at COVID and other causes of death. The COVID code is U07.1 and we’re including decedents where that is the underlying cause of death and we show it by OH residents and by decedents who died in OH regardless of residence. Each death has a single underlying cause of death and then they might have multiple other conditions but not the underlying cause of death. They only count the ones where U07.1 (COVID) is the underlying cause of death.

LTC: originally we reported the data from April 15th because that’s when the disease reporting system was stood up for residents and staff that put that out weekly. So we reported the deaths from that same time period to align those two. When we switched over to the NCHS, that’s a completely different system so we went back to the beginning of the pandemic. Would have to see what the earliest date we have there. There’s nothing on the death certificate that says this person was in a LTC facility. So we do a match between our Ohio disease reporting system and match that back to the Ohio. We show the data two different ways. The default is by folks who resided in Ohio. So regardless of whether they died in Ohio and then also the data by folks who died in Ohio regardless of their residence

(Context: OH added a download for total tests with a PCR/antigen breakdown and contacted us.) Right now, the download is a day behind the graph. This is causing a discrepancy of about ~7000 at the topline for confirmatory tests, and ~7000 missing presumptive tests, in the file vs. the graph even when we add the missing day from the graph, because of backfilling. We’d love for the download to match the graph, or for y’all to enable the download function on the graph, so that we can sync up the numbers to something we can use from their dashboard in data entry.

Our intention is for the download to exactly match the dashboard, and for both to be updated daily at 2PM. However, we found a data issue that caused the download to not refresh with updated data on certain days this week. It should be corrected Thursday (12/31). Once that happens, the numbers should sync up.

(Context: OH added a download with antigen tests). We’d love if you could display the topline number of presumptive and total instead of only having it accessible through right clicking after doing a select all of the graph, especially if they don’t enable a graph download. The dragging to select all the points is highly error prone and unscreenshottable. We’ve run into this kind of thing before and it scares me.

Certainly understand! I think once the download functionality is fixed, you shouldn’t have to get the numbers from the graph anymore. Does this meet your needs?

The sum of antigen positive tests and PCR positive tests on 12/26 is 678056 from this new file, vs. total case numbers including probables on the same day, which are 682570. We would expect the number of positive tests to be higher than the number of cases. We know there are reasons that it could be lower, though, from other states. A few questions about scenarios we’ve seen before: What exact case definition does OH use for probable cases? We know it is the “CDC expanded case definition” from the dashboard, but can’t tell if it’s April or August, or if they’re using all the criteria in that definition, including the non-testing ones which might explain the discrepancy. (In addition, could OH put this definition on the dashboard for the public to access?) Does OH use disparate data sources for the tests and cases? E.g. do the testing data only reflect electronic laboratory reporting or reporting through a certain submission form/program? Are there any lags in the testing data as compared to the case data? Is there any other reason the test positives data might be lower than the case data?

The testing results we show on the site are delayed a bit, based on how they are reported to us. The 678,056 positive tests figure you mentioned is from a reported results date of 12/26, whereas the 682,570 cases figure is through 12/29. So the testing numbers are behind by a few days. The testing data through a reported results date of 12/29 shows 711,384 PCR and antigen positives combined, which is higher than the cumulative cases to that same date.

ODH adopted the August CDC definition on 12/8/2020, and use the full definition.

Case data is collected in the Ohio Disease Reporting System, which is what ODH has used for Reportable Diseases for almost two decades. Disease reports are often submitted to ODRS via ELR. Testing volume data are reported separately in aggregate from the performing laboratories.

There can be lags in testing data reporting, in part due to the high volume of labs that are reporting COVID-19 testing.

No additional reasons that we have identified.

We’re also interested in testing data in units of “testing encounters”, which can be defined as “the number of unique people who have been tested per day.” Under this metric, if a person is being tested once today and once again the following week, that counts as two tests, however, if the same person has two samples tested at a testing center visit, then that only counts as one testing encounter. Are you willing to or are you planning to release testing encounters? If so, will you be able to provide historical data consistently under an automation friendly form, like an ArcGIS or CKAN API or a direct download of CSV or JSON?

On Ohio's COVID-19 dashboard, there's a category for "Total Tested in Ohio" – could you clarify what is meant by the term. Does it refer to the number of unique people tested or the number of specimens tested overall? If "Total Tested in Ohio" refers to total individuals tested, do you have plans to release the total number of tests? If "Total Tested in Ohio" refers to total specimens tested, do you have plans to release the number of total individuals tested?

Do you have plans to release a full historical time-series of all testing and results? We have been capturing the total tests in "Total Tested in Ohio" since 4/20. Do you plan to release a historical time series of that number? And will this historical data set be available consistently and in an automation friendly form, like an ArcGIS or CKAN API or a direct download of CSV or JSON?

Do you perform any deduplication to reach the "Total Tested in Ohio" ? If so, how? (i.e. instances swabbed per day/week; recording only one positive test per person, but multiple negative tests for the same person; etc.)? When reporting testing results, if the same person gets different test results on different days, would you report raw results as part of the daily positives or do you apply any kind of logic such as reporting only the first negative or positive result per individual?