COVID in CT: The data that tell the story

The Connecticut Mirror is tracking the latest coronavirus data in Connecticut. These graphics are updated with the latest data as soon as the state publishes them.

Statewide case, death, hospitalization and positivity rate data are updated each weekday around 4 p.m. Some graphics are updated on different schedules, as noted. Unless otherwise specified, the data are sourced from Connecticut’s Open Data portal.

Follow CT Mirror’s COVID-19 coverage here. Questions or ideas about these data? Send us a note at data@ctmirror.org.

What test positivity tells you — and what it doesn’t

The positivity rate is a “fairly crude way to see if we’re doing enough testing,” according to Yale Professor of Epidemiology Ted Cohen. In Connecticut, the positivity rate is defined as the number of cases divided by the number of tests reported each day. It’s often talked about in the context of assessing COVID-19 spread, but it is highly influenced by fluctuations in testing patterns.

For instance, high positivity rates in March and April reflect the relative scarcity of testing over that period, and any comparison to present-day rates must keep that in mind, he said.

“The pool of people we were testing in the spring is totally different from who is being tested now. The tests were more restricted to people who were severely ill, and there were shortages of tests. Therefore, you can’t directly compare the numbers of positive tests between the spring and now,” said Yale Associate Professor of Epidemiology Dan Weinberger.

The World Health Organization’s 5% positivity threshold for reopening is somewhat arbitrary, Cohen said, but in general, a high positivity rate indicates that testing data isn’t capturing the full case count.

Daily case data is a snapshot of the past — not a reflection of the present

The governor’s briefings and the CT Mirror’s case-count charts report test data based on the day that the result comes back from a lab. As there is a variable time delay between getting tested and getting a result back, these data will be showing the past, not the present.

The case rate data by town, on the other hand, releases data by the date that a person was tested. Therefore, the case numbers in the town-by-town bubble chart reflect active COVID-cases over the previous two weeks. The data are updated weekly.

Deaths trail behind cases

While the case data on a given day is a snapshot of the past, it is also a clue to what people might eventually see in the death data.

In general, deaths trail behind cases by about two to three weeks, Cohen said. The delay is because of the nature of the disease and the time it takes to prove fatal; it is also influenced by the process of recording deaths. “This lag has been abused by many people” to question the severity of the disease, Cohen said, but only time can reveal the full impact of the pandemic.

That said, the extent of the spike in deaths following a spike in cases may not be the same as it was in March, Cohen said. In part, this is because doctors and nurses are better equipped to handle the virus now than they were then. Those infected today might also be less vulnerable, particularly if they are younger.

Case rates by town do not include nursing homes and correctional facilities

While this means that the town map does not provide a comprehensive picture of the virus in an area, separating out nursing home and prison data can provide most town residents with a more accurate understanding of their risk of exposure “if these are truly isolated populations within these community where there’s little to no interaction,” Cohen said.

Case rates by town don’t normalize for testing variation

The sheer number of tests might differ by town, and “that is a potential source of bias,” Cohen said. College towns, for instance, may see higher testing numbers because universities are testing communities on a regular basis. The town-by-town case rate data includes cases from schools and universities, confirmed Pauline Zaldonis, the dataset’s manager.

That means that some towns may be doing a better job of identifying cases than others, resulting in higher case rates.

The state of Connecticut’s COVID-19 website also features a forum where Connecticut residents can request COVID-19 datasets that are not already available, said Scott Gaul, the state’s chief data officer. “It’s been unexpected how much people have used it,” he said. “We’ve had more comments come in the last six months than in the first six years that the [Open Data] portal was online.”