Tracking Covid-19 in Connecticut

New Covid-19 Cases in Connecticut

The plot shows the history of new cases and also shows the actions required by some of Governor Lamont’s executive orders.

Covid-19 Deaths in Connecticut

Connecticut Covid-19 Patients in Hospital

The number of new cases peaked in mid April and then began to decrease rapidly, although not as rapidly as it increased. A key issue is how fast epidemic is expanding or contracting. One indicator of that is to estimate the parameter R_t, the average number of people who become infected by an infectious person. When R_t is greater than 1 the number of daily new cases is increasing, If R_t is less than 1, it is decreasing. See the site covidestim · COVID-19 nowcasting for an estimate by state of the effective value for R_t. The site is described in this article at Vox.com. I don’t claim to have the expertise to evaluate the quality of the calculation of R at the covidestim site, but it seems to jive with other information about the pattern of the epidemic among states.

As of June (and a month previous), Connecticut, New York, and New Jersey have among the lowest values for R_t in the country. This is good news for us in Connecticut. I have included R_t for Arizona as well as Connecticut to provide a contrast with a state that has lately shown signs of a growing outbreak. The goal in Connecticut and in the entire New York City area is to keep R_t well under 1. As of 04/28/2021 R_t is 0.79. On the other hand, in Arizona R_t is about 0.86.

Estimate of R_t by State (from covidestim · COVID-19 nowcasting)

The next display shows the trend of R_t in each state arranged according to geographic position in the US. Remember that R_t is a measure of the rate of increase. When R_t is greater than 1 the number of daily new cases is increasing, If R_t is less than 1, it is decreasing. The color of the R line shows whether the most recent estimate of R for a state is above 1 (purple line) or equal or below 1 (green line).

R is estimated from the state data and the gray area around the line indicates the statistical uncertainty of the estimate showing an 95% confidence interval. Note that this map shows only whether transmission seems to be growing or shrinking. How concerned one should be about the epidemic in a particular state also depends on the prevalence of cases.

Trends in New Cases by State

The next map shows the history of new cases in each state per 100K of population. The lines are color coded the same as in the previous map to indicate whether the most recent estimate of R for a state is above 1 (purple line) or equal or below 1 (green line).

The tallest graphs of new cases per 100K of population are New York and Arizona.. From the point of view of someone living in Connecticut, the striking thing is how thoroughly Connecticut (and New York, New Jersey, Rhode Island, and Massachusetts) have recovered from the peak. New cases are low and the low value of the estimate of R confirms that they are likely to decrease even more. The great contrast is with the increase of cases in some of the southern states.

By County and Type of Town

There are 169 towns in Connecticut, which makes it hard to examine variations by town. To help examine variations within counties, I’ll use a typology of towns called The Five Connecticuts that divides towns into five categories based on census variables: Urban Core, Urban Periphery, Suburban, Rural, and a fifth category for Wealthy Suburbs used for some towns in Fairfield County. Adjusting for population, the number of cases has been greater in the counties closer to New York City. In the figure below I’ll fold the wealthy towns into the Suburban category. I have excluded Montville and Somers because the prisons in those towns complicates interpretation of the town statistics.

Counties are laid out from top to bottom and the counties closer to New York are toward the top. The columns show the four categories. Urban towns have been hit harder than suburban and rural towns. While density may have some effect on case rates, the report Towards Health Equity in Connecticut: The Role of Social Inequality and the Impact of COVID-19 by DataHaven documents how the epidemic interacts with existing social inequality in Connecticut.

Details by Town

The links below take you to a separate HTML page for each Connecticut county. That page contains a table with one row for each town in the county.

For each town it shows the category, the population, the total number of cases to date, the number of recent cases per 100K of population (total new cases in the last two weeks), total deaths, and the percentage of total deaths in that town attributed to nursing home residents.

Next there are two sparklines that show the trend in daily new cases and daily deaths (based on 14 day moving average). Cases and deaths are shown as a ratio to population in the town.

The purpose of these tables is to make it easier to quickly scan for trends in local towns. Here is sample (based on my home of Guilford) showing what is available for each town.

For each county there is also a map showing the towns in that county and the relative number of total cases in the town and a second map that shows the percentage of individuals who are below the poverty line.

The town sparklines at the right show the trends over time. The vertical scale is different for new cases and for deaths. For new cases, the maximum height shown on the scale is 100 while for deaths, the maximum value shown is 23.7. For both new cases and deaths, the sparkline shows the rolling seven day average of new cases or deaths.

Links to Summary Page for Each Connecticut County

Below are links to a separate web page by for each of the eight counties in Connecticut. Click on the county name for detail about individual towns within that county.

Litchfield County	Hartford County	Tolland County	Windham County
Fairfield County	New Haven County	Middlesex County	New London County

---

Code Used to Produce This Post

The R code used to download and process the data from Connecticut and elsewhere is in the file daily_ct_stats. This post was created using RMarkdown so the code to create most of the figures is in the .Rmd file for this post.

The separate HTML files for each Connecticut county were created using the rmarkdown::render function. The RMarkdown document that actually formatted the town details (including sparklines and county maps) is here.

Some noteworthy packages used here:

• geofacet used to create the map of US states in Figures 5 and 6.

• tidycensus used to create the Connecticut maps and retrieve Census data (most of which I didn’t use).

• sparkline

• formattable – Used to create the town-by-town detail tables primarily because this was the only tool I found that would readily add a sparkline to a table.

Some Notes on the Data Available From the Department of Public Health

Sometimes there is a lag before cases and deaths end up in the daily reports. As a result, there tends to be a “day of the week” effect. For that reason, observers of the Covid-19 statistics generally focus on a 7-day rolling average of the daily counts. Some of the unusually large peaks and valleys in these charts are due to reporting process. The DPH has also begun producing reports based on the date a sample was taken for a test and by the date of death. That’s a more accurate way to look at the change over time, but it means the data for the most recent days are hard to interpret because some data may be “in process” and not yet reported. In these charts I have used “date of report” rather than “date of sample taken” or “date of death” because that gives me all of the recent data that is available and because that is what most other data projects (such as the COVID Tracking Project or the New York Times) have been using.

The DPH reports always include a comment that “all data are preliminary and subject to change.” But when they make and adjustment to correct errors, as far as I can tell they do not go back and adjust the earlier reports. That can lead to a misleading report of recent changes. For example, in one note below they removed 70 cases because of errors on a day in which there were 81 new cases. In the data series that I download, that shows up as having been 11 new cases that day, not 81, because 70 were removed. Earlier days are not corrected.

Note as of June 1

*In Connecticut during the early months of this pandemic, it became increasingly clear that it would be necessary to track probable COVID-19 cases and deaths, in addition to laboratory-confirmed (RT- PCR) cases and deaths. This was needed to better measure the burden and impact of this disease in our communities and is now part of the national surveillance case definition for COVID-19. Today for the first time, DPH is reporting cases and deaths as “confirmed” or “probable.” Previous reports reported these as a combined number. The only change today is that they are being separated to conform with CDC reporting guidance. Probable cases of COVID-19 involve persons who have not had confirmatory laboratory testing (RT-PCR) performed for COVID-19, but whose symptoms indicate they are very likely to have a COVID-19 infection. In Connecticut, most of the probable COVID-19 cases involve persons whose death certificates list COVID-19 disease or SARS-CoV-2 as a cause of death or a significant condition contributing to death.

Note as of May 27

The staff at the Department of Public Health have removed 356 cases and 808 tests in the past 24 hours, which were identified as duplicates in the system, affecting both test and overall case numbers. Since yesterday, there have been 341 new positive cases, and 5,215 new tests were reported.

Note as of June 18

Please note that 81 new cases were reported in the past 24-hours; 70 previously reported cases were removed from the total counts due to correction of data errors.

In the data portal, the number of cases reported for Montville was 381 on June 17. As of June 18 it reports 293 cases, a reduction of 88 cases. That’s more than the 70 cases removed. Perhaps the case counts for Montville were affected by inmates being moved within the Connecticut prison system.

Note as of June 24, 2020

1175 new test results were reported since the last report and 2770 previously reported PCR tests were removed due to correction of data errors.

Note as of July 23, 2020

*Please note 83 new cases were reported to DPH since yesterday. In addition, 74 previously reported cases were removed due to updated laboratory findings of false positive results.

July 24, 2020: Governor’s press release

*NOTE: Today’s update includes a large set of data provided by an out-of-state lab on tests that were conducted on Connecticut residents between May 23 and July 20 and not reported to the State of Connecticut until today. This data set provided by the out-of-state lab includes approximately 12,000 tests, 440 of which were positive. The remaining 104 positive cases in today’s report are newly reported cases in the day-to-day update, giving a 0.79% positivity rate for the day.

See also Hartford Courant, July 24, 2020 I don’t see any explanation of this in the Department of Public Health PDF, but it was clear that an increase of 544 cases in one day would be remarkable.

On Friday, the state also announced a backlog of unreported test results for Connecticut residents dating back to mid-May. Among these approximately 12,000 tests, 440 were positive. On Thursday, the state reported an additional 13,000 tests. With the 544 new cases, the state has now recorded 48,776 coronavirus cases.

July 29, 2020 press release

*In addition to the 79 recently diagnosed cases and 12,367 tests, 384 cases and 750 tests performed between April and June were newly reported to the Department of Public Health in connection with a transition to electronic reporting by an out-of-state regional laboratory. For surveillance purposes, that data has been added to the total case and total test counts.

Note as of August 18, 2020 DPH daily update:

*Forty new cases were reported to CT DPH since yesterday; in addition, DPH removed 52 previously reported cases because of newly identified data errors.

Note as of October 12, 2020:

As of Monday, there were additional 1,066 positive cases from Friday. That included over 270 positive cases out of 23,130 tests conducted between September 26 and October 8 that are newly reported as part of catch-up reporting.