Oklahoma

At the beginning of March, Oklahoma announced it would begin reporting COVID-19 mortality information from the NCHS in addition to the tally reported on the state dashboard. We understand from statements made by public health officials that this was motivated by a desire to provide more timely information about the pandemic’s impact in the state.

We notice that other states (Ohio, 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 Oklahoma about switching to NCHS data?

This is a list of follow ups for OK's response to this question: https://covid-tracking.slack.com/archives/C0124MP2E5C/p1614848464021500

Has the HHS not used your CELR submissions because of the problems with onboarding labs to SpringML? Using your own aggregate submissions to estimate, approximately what percent of lab results in OK are coming in through SpringML? Do the aggregate submissions you get contain information that would let you separate PCR from antigen tests? Other states that have been unable to submit line level data were given the option of submitting aggregate counts that would be more complete than the federal government’s data directly from labs. Has the HHS given you the option of submitting via aggregate processes with the aggregate counts you collect from labs? If not, did it ever explain why it wouldn’t use that aggregate data? To make sure we’re interpreting correctly, do the numbers on your dashboard come from that voluntary aggregate submission process?

Unfortunately, the answer to lots of these questions is "I don't know." I can check with the surveillance team that submits data to the CDC, but I'm uncertain they will know. The section chief was out this week.

1. I don't know.
2. We would tentatively speculate that ~50% of submissions are coming as ELR via SpringML. However, it is highly variable and the counts do not align as closely as they should. We have been investigating this and will hopefully have greater clarity in the next couple of weeks.
3. No, the aggregate submissions do not allow distinguishing between PCR and antigen. Reporters are asked and expected to distinguish between agent testing (PCR or antigen) vs. serologic testing (antibody). However, we can't confirm that all reporters appropriately make such distinction.
4. I do not know.
5. I'm uncertain what numbers on the dashboard you are referring to. Our case counts are based upon individual results. Reports for "Specimens received," and "Percent positivity" are based upon aggregate submissions. We have been working to align our investigated fatality numbers with death certificates while still maintaining basic daily reporting. Once we get our system streamlined on that process, we intend to tackle the aggregate reporting vs. SpringML questions.

Currently, the HHS does not use data directly provided by OK in its PCR testing dataset (accessible here: https://healthdata.gov/dataset/covid-19-diagnostic-laboratory-testing-pcr-testing-time-series), as it does for most jurisdictions. Instead, it uses data submitted directly by laboratories in OK, which is very incomplete data—about 80% lower than OK’s total tests on its dashboard.

Does OK submit any COVID-19 testing data to the federal government, either in aggregate or the line level? If so, do you know why the CDC does not use OK’s data? If not, why doesn’t OK submit data to the federal government?

The issue resides in the fact that the CDC/HHS data reporting relies primarily upon the electronic laboratory reporting (CELR) system. Beecause Oklahoma has historically had a relatively low percentage of labs reporting via HL7, our CELR data has been deficient in representing the full status of testing in the state. As we've transitioned to SpringML and continued onboarding labs, the completeness of our CELR reporting should improve in terms of completeness. However, we're still far away from 100%.

A couple of important items to note:

1. We are assuredly not the only state in this situation. The email below implies Oklahoma is unique in having a difference between CELR and state-reported numbers. I am 100% certain that other states also have deficiencies in their CELR reporting, and I'm nearly 100% certain that those states are also reporting non-CELR data in their state dashboards, the same as wee are.
2. While we consider our data reporting to be more complete and reliable than CELR (because of the incompleteness of electronic laboratory reporting), it should be noted that our numbers of specimens reported are dependent on voluntary aggregate reporting by the labs themselves, with no ability for us to verify the veracity of the data.

Thee best situation will be achieved when we reach near 100% reporting via CELR. This will be achieved in the foreseeable future. SpringML is now fully functional, including with webform reporting for labs doing small amounts of reporting. Inductive Health is now contractually engaged for assisting with onboarding of labs into the SpringML reporting system. We have nearly 500 labs that have at least been substantively exposed to the SpringML system, with ~300 labs into stages of either active reporting or beta testing/troubleshooting. I don't want to over-promise, but I anticipate that we can have the majority of lab testing being reported via electronic processes in the very near future, enabling us to align our national data better with our state-reported data, but also facilitating our own data querying and Q/A processes.

Looking at Oklahoma’s “total number of specimens tested to date,” we can only find a record of your test numbers going back to April 29.

Beginning with the April 6th Executive Order Report that data element was included and can be found on our website htpps://coronavirus.health.ok.gov/executive-order-reports.

On what date did OK start tracking tests across the state? Do you have plans to release a full historical time-series for this metric?

Data that has been collected has been included in the EO report mentioned above.

Do you also record testing numbers by total unique individuals tested? If so, would you be able to release a full historical time-series for this metric as well?

No to 1st question. 2nd question: There are no plans at this time.

Many states provide their COVID-19 testing and outcomes in machine-readable format such as .csv or .json; but as far as we are aware, Oklahoma hasn’t been doing this. Is it possible to provide OK’s COVID-19 testing data as machine-readable downloads?

As mentioned above, data are available in the EO report (beginning 4/6). We have no plans to develop another dataset/download.

Do you perform any deduplication to reach the “total number of specimens tested to date” value?

None of the test data reported are person specific (all case data is person specific).

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 a specific protocol such as reporting only the first negative or positive result per individual? We’re interested in gathering test data in units of “testing encounters,” which can be defined as “the number of people who have been tested per day.” Under this metric, if a person gets tested once today and once again next week, that counts as two tests, however, if the same person has two samples collected at a testing center visit, then that only counts as one testing encounter.

Would Oklahoma be willing to release numbers on testing encounters?

Test data are specimen specific (not person). That information is not available. We do not have this level of detail.