Our Data

Alabama combines PCR and antigen tests in the total tests figure reported on the state's dashboard.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

On October 21, 2020, Alabama did not update their COVID-19 dashboard due to "technical issues".

As of September 18, 2020, Alabama's total test results are drawn from our totalTestsViral field instead of calculated via positive+negative.

On September 1, 2020, Alabama began displaying the number of people tested by antibody, over 50,000 individuals, separately from its main total tested field, which it clarified represents antigen and PCR testing. Its main total tested number dropped by over 25,000 the same day. We have reached out to the Alabama Department of Public Health to confirm this drop was due to the removal of the individuals tested by antibody.

On July 15, 2020, we changed our data reporting to reflect that Alabama reports test totals in specimens rather than unique people tested, as we had initially believed. All test results were moved into the correct API field and metric.

On July 11, 2020, Alabama began using direct reports from hospitals for cumulative hospitalizations (which we report in our Ever hospitalized metric), rather than state case surveillance. Its number of cumulative hospitalizations increased by over 3,000 as a result.

Alaska combines PCR and antigen tests in the total tests figure reported on the state's dashboard.

Alaska reports a "Total count" of people who have been hospitalized, but we do not record that count in our Ever hospitalized metric, since it does not include people who are currently in the hospital.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

As of September 18, 2020, Alaska's total test results are drawn from our totalTestsViral field instead of calculated via positive+negative.

As of July 16, 2020, we are reporting non-resident cases as part of Cases or Positive PCR tests (people) for Alaska.

American Samoa does not have a regularly updated data source. We monitor their social media, emergency declarations, and local media for updates.

As of September 18, 2020, American Samoa's total test results are drawn from our totalTestsViral field instead of calculated via positive+negative.

Arizona combines PCR and antigen tests in the total tests figure reported on the state’s dashboard

Arizona regularly reviews and removes duplicate records which may occasionally result in minor decreases of cumulative figures.

The Recovered data point we report for Arizona reflects the number of COVID-19 patients discharged from the hospital and therefore does not represent the total number of people who have recovered from COVID-19, since many people with COVID are never hospitalized.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

On October 6, 2020, The Arizona Department of Health Services announced that they were removing cases who had been admitted to a hospital but had not been hospitalized from their total hospitalization counts. This results in a decrease of Arizona's cumulative hospitalizations.

As of September 25, 2020, Arizona’s total test timeseries is stored in our Total PCR Tests (People) field instead of Total PCR Tests (specimens), reflecting a response to our outreach to Arizona’s health department that it counts unique individuals tested. Arizona includes antigen tests in its totals figure, so if an individual received both an antigen and PCR test, they are counted only once. Additionally, as of September 25, 2020, Arizona's total test results are drawn from our Total PCR Tests (People) field instead of calculated via positive+negative.

On September 18, 2020, Arizona reported a policy change in the way they count people with positive antigen testing results to comply with CSTE case definitions. This appears to have resulted in a large increase in probable cases and also likely the reason where a decrease in confirmed cases was observed.

On September 16, 2020, Arizona added antigen testing into its main totals figure. This appears to have raised its Total Tests (PCR) number more than usual.

From July 18–August 5, 2020, Arizona’s dashboard stopped displaying confirmed and probable breakdowns for cases and deaths. We will backfill this data if Arizona provides historical numbers.

On June 23, 2020, we updated our historical data for Cumulative hospitalized to match Arizona's dashboard data for "hospitalized by date admitted." Data for this metric is not typically reported until several days after admittance. Our daily updates will continue to compile Arizona’s overall number as of cumulative hospitalizations, regardless of date admitted.

During the week of September 20, 2020, Arkansas made some continual adjustments to its antigen people tested numbers, resulting in some fluctuation in its negatives and totals.

As of September 18, 2020, Arkansas's total test results are drawn from our totalTestsViral field instead of calculated via positive+negative. The totals do not include repeat positives—counting individuals only for their first positive—but do count repeat negative tests.

On September 16, 2020, Arkansas’s Negative Tests (PCR) dropped by 26,801 tests, causing its Total Tests (PCR) to drop by 26,195 tests due to the reversal of a data dump of negative test results that Arkansas reversed. It also added 140 probable deaths, not all of which dated to September 16.

On September 3, 2020, Arkansas began reporting probable cases according to the most recent CSTE definitions, in addition to confirmed cases. This added 423 new cases to AR’s Total cases number.

On Aug 15, 2020, Arkansas data changed due to deduplication and the state's removal of out-of-state cases from its dataset. These changes resulted in reductions in Total cases, Ever hospitalized, and Ever in ICU data.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

As of September 18, 2020, California's total test results are drawn from our totalTestsViral field instead of calculated via positive+negative.

From July 23 through July 28, 2020, California reported that its Now hospitalized and Now in ICU were incomplete due to its transition to the HHS reporting system. We carried over hospitalization figures until July 28. We will backfill the complete numbers for this period if the complete data becomes available.

From June 29 to July 5, 2020, California made revisions to daily case numbers. On July 17, we updated the values based on the most recent data from the state dashboard.

On July 8, 2020, California data included a backlog from Los Angeles County, which had not reported for several days. Approximately 1,000 of the 11,000 new cases were attributed to this backlog data.

As of April 22, 2020, California has reported specimens tested instead of people tested. Because some people may be tested more than once, this number is probably higher than the number of people tested.

On April 4 and April 22, 2020, California reported large batches of backlogged negative test data.

Prior to April 1, 2020, we used county data sources for more reliably timely reporting of California's COVID-19 data. At the start of April, we began reporting primarily from the state data dashboard(s). This led to a modest drop in cases and deaths, as the state's data lags behind the county sources.

The Recovered data point we report for Colorado reflects the number of COVID-19 patients discharged from the hospital and therefore does not represent the total number of people who have recovered from COVID-19, since many people with COVID are never hospitalized.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

As of August 13, 2020, Colorado's total test results reflect total test encounters. The number of unique people tested is available in our API as totalTestsPeopleViral.

On July 27, 2020, Colorado posted a notice that, "Due to a server issue yesterday and today with our electronic laboratory reporting system, the number of tests completed and cases reported today, July 27, 2020, may be artificially low." We report Colorado data a day after posting due to timing, so this note applies to the data published on July 28.

On June 14, 2020, Colorado revised previous "daily discharged" numbers. This caused an overall decline in their Recovered data.

On May 15, Colorado started reporting deaths for which COVID-19 is listed on the death certificate, in addition to the figure it already provided, deaths among laboratory-confirmed and probable cases (whether or not COVID was a cause of death). On July 1, 2020, we revised our historical data to reflect the death-certificate metric instead of the deaths-among-cases metric. Because of this change in methodology, there is a drop from 1,091 deaths on May 15 to 878 on May 16 in our data. Colorado also reports subsets of confirmed and probable COVID-19 deaths, but these metrics correspond to the deaths-among-cases metric we no longer report, so we do not compile this data.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

The Recovered data point we report for Connecticut reflects the number of COVID-19 patients discharged from the hospital and therefore does not represent the total number of people who have recovered from COVID-19, since many people with COVID are never hospitalized.

On June 24, 2020, Connecticut corrected some errors in its data, resulting in a decrease in Negative PCR tests (specimens) (which can be found in our API and CSVs).

On June 18, 2020, Connecticut corrected some errors in its data, resulting in a decrease in Total PCR tests (specimens).

On May 29, 2020, Connecticut hospitalization data changed due to reporting laboratory-confirmed cases only. These changes resulted in a decrease in the numbers for Ever hospitalized and Recovered (which can be found in our API and CSVs).

On May 27, 2020, Connecticut deduplicated some of its data, resulting in a decrease in Cases and Total PCR tests (specimens).

On October 29, 2020, we moved the data in our Total Tests (PCR) field for Delaware to the Total Test Encounters (PCR) field after learning that it represents encounters from Delaware's state health department. However, Delaware appears to include antigen tests in its totals, so if an individual is tested using two different test types on the same day, it will get counted twice. Also as of October 29, 2020, Delaware's totalTestResults field represents encounters. The number of unique people ever tested remains available in the totalTestsPeopleViral field.

On September 9, 2020, Delaware reported that data from September 8, 2020 and September 9, 2020 would not be available until September 10, 2020 due to a maintenance issue with their electronic reporting system. We will backfill data for September 8, 2020 and September 9, 2020 to the best of our abilities when data is released

On September 6, 2020, the Division of Public Health reported a technical issue which caused between 175-200 positive cases to be excluded from Delaware's cumulative confirmed case count. These cases will be added in the coming days. As a result new cases reported may not have been diagnosed in the past 24 hour period.

On August 27, 2020, Delaware revised down total cases and confirmed positives. This is a semi-regular occurrence which will impact Delware's number occasionally.

On July 25, 2020, Delaware revised down total cases, confirmed positives, and total tests.

On April 11, 2020, Delaware revised down negative test results from 10,415 to 9,624.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people, confirmed + probable) from Total PCR tests (specimens) in the absence of better data.

As of September 18, 2020, District of Columbia's total test results are drawn from our totalTestEncountersViral field instead of calculated via positive+negative.

As of August 29, 2020, we store the “total overall tested” timeseries in our Total Test Encounters (PCR) field despite the following: According to our outreach, the District of Columbia’s “total overall tested” figure on its dashboard represents test encounters, the number of unique individuals tested per day, with one exception: if an individual receives a positive and a negative test on the same day, they are counted twice rather than once.

On May 25, 2020, the District of Columbia decreased the district's total tests number by 11,000 to reflect only DC residents.

Florida combines PCR and antigen tests in the total tests figure reported on the state's dashboard (first noted July 2, 2020). We include data about Florida non-residents in all metrics where possible. For some metrics, we use machine queries against the Florida Department of Health's raw data, which may not be displayed on its dashboard.

Per reporting, the number of pending tests (available in our API and CSVs) may include only results from public labs, and may therefore be an undercount.

On November 3, 2020, Florida’s Positive PCR Tests (specimens) dropped by over 200 with no explanation.

On October 30, 2020, the number of negative residents ('T_NegRes') remained unchanged on Florida's ArcGIS layer. The number of unique people tested in 'totalTestsPeopleViral' draws from 'T_NegRes' and 'C_AllResTypes' fields on Florida's ArcGIS layer.

On October 26, 2020, Florida started providing a full timeseries of total test encounters excluding tests from individuals after they test positive. We started capturing this data in our totalTestEncountersViral field. As of October 27, 2020, Florida's total test results are also drawn from our totalTestEncountersViral field instead of calculated via the number of unique individuals testing positive and negative. The number of unique people tested, now drawn from the T_NegRes and C_AllResTypes values on Florida's ArcGIS layer, is still available in totalTestsPeopleViral. The new encounters time-series only reflects Florida residents, so we have also revised our existing cases figures and unique individuals tested figure to reflect only residents. These changes resulted in a 1.28% decrease in Florida's cases and a 0.38% decrease in Florida's unique people tested.

On October 21, 2020, only some of Florida's data updated by our publication time. Data points such as Total PCR Tests (People) did not update.

As of October 13, 2020, Florida's Antigen positive timeseries is stored in our Probable cases field, reflecting a response to our outreach to Florida's health department that it switches Antigen positive cases to PCR positive cases in their daily state report upon confirmatory PCR testing.

On October 10, 2020, Florida did not report COVID-19 data. The state initially reported that this was because Helix Laboratory, a private lab, submitted 400,000 previously submitted results, which required extra deduplication time. On October 12, 2020, the Florida Department of Health announced that the error was not the fault of the private lab, but was due to "an unforeseen technical issue" that was "not the fault of Helix or the Department of Health." The state's data for October 11, 2020 appears to include data from both October 10 and 11.

On September 1, 2020, the Florida Department of Health announced that Quest Diagnostics had reported nearly 75,000 backlogged tests dating as far back as April.

On August 14, 2020, Florida began distinguishing between people with positive antigen tests and people with positive PCR tests in the Total Cases number given in their daily state report. As of September 30, 2020, we have removed these antigen positives from the values for Cases or Positive PCR tests (people) (positiveCasesViral in our API), so the values for this metric in our time series will be lower. Cases (confirmed plus probable) (positive in our API) will be unaffected.

On August 12, 2020, the Florida Department of Health reported via Twitter that it had received a large backlog of testing data from Niznik Lab Corp in Miami: "The lab reported over 4,000 cases occurring over the last 7 weeks, but which had not been reported to FDOH until today. Therefore, this backlog severely skews today's daily report for Miami-Dade and is not reflective of current trends."

As of July 10, 2020, Florida reports Now hospitalized data for those patients with a "primary diagnosis of COVID-19."

Since May 15, 2020, Florida reports total tests in unique people and in unique people tested per lab. For that latter number, if an individual is tested twice by one lab, they only get counted once, but if an individual gets tested twice at different labs, they get counted twice. We store the former value in Total PCR Tests (People), and the latter value, which does not match any of our totals units, in Total Tests (PCR), our specimens field, where we also store testing data with unclear units. This value does not represent the number of specimens.

On March 21, 2020, Florida began including tests of people not investigated as PUIs in its Negative PCR tests metric.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

On October 21, 2020, Georgia did not update their "COVID-19 Daily Status Report" by the time of our daily update due to technical issues. In order to provide the most up to date data possible, we updated only Current Hospitalizations, which are sourced from a different Georgia government website.

As of September 18, 2020, Georgia's total test results are drawn from our totalTestsViral field instead of calculated via positive+negative.

As of August 28, 2020, we started reporting patients Now hospitalized from the Georgia Geospatial Information Office. The new count includes both patients with a confirmed COVID-19 diagnosis and patients under investigation, as is our standard practice where possible.

As of May 27, 2020, Georgia reports antibody tests separately from PCR tests. This change reduced the number for total tests by approximately 74,000.

Guam's counts include 11 people tested at the San Diego naval base.

On September 3, 2020, Guam reported that it had shifted to using the more restrictive August 5 definition of probable cases released by the CSTE. From March 12, 2020 to September 2, 2020 Guam reported eight probable cases under the previous probable case definition. Due to the change, Guam has no cases that classify as probable cases as of September 3, 2020, and 8 cases have been removed from Guams's total case count.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people, confirmed + probable) from Total PCR tests (people) in the absence of better data.

As of October 9, 2020, Hawaii’s total test results are drawn from our totalTestEncountersViral field instead of calculated via positive+negative.

As of October 7th, 2020, Hawaii has stopped reporting the number of unique people tested, which is available in our API as totalTestsPeopleViral. We will carry forward the last value as a lower bound.

On September 3, 2020, Hawaii included cases which had been previously diagnosed in their daily update due to an electronic laboratory reporting error." As a result, new cases reported may not have been diagnosed in the previous 24 hours.

As of August 5, 2020, we are collecting Now hospitalized, Now in ICU, and Now on ventilator data from the Lt. Governor of Hawaii's Instagram account, as we have determined that these are official statistics not available elsewhere. We have backfilled our time series for all available data from this source.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (people) in the absence of better data.

As of September 18, 2020, Idaho's total test results are drawn from our totalTestsPeopleViral field instead of calculated via positive+negative.

On August 4, 2020, after another period of missing data for Hospitalized now, Idaho reported current hospitalizations of 106, down from the previous report of 223 on July 31. It reported a similar drop in ICU patients from 51 on July 31 to 22 on August 4.

On July 22, 2020, Idaho reported 126 total deaths but also reported 103 lab-confirmed deaths and 33 probable deaths, for a calculated total of 136 deaths. We treated the 33 probable deaths data point as an error and reported 126 total deaths for this date, including 103 confirmed and 23 probable deaths. See Idaho state page screenshots for July 22.

From July 13 through July 21, 2020, Idaho was unable to report Hospitalized now data due to HHS regulation changes. Idaho retroactively made figures available for these days on July 21, 2020, and we revised our figures to match.

On July 8, 2020, Idaho revised its historical numers for Hospitalized now. We updated our hospitalization figures for Idaho from March 23 to July 13, 2020 to match on July 17, 2020.

On July 3, 2020, Idaho revised down their Recovered data by about 1,500, noting that "an error in auto-calculation was discovered."

On June 5, 2020, Idaho reported a large batch of previously unreported tests. The state's Total tests number rose by approximately 2,500 more tests on June 5 than the seven-day average for new tests at that time.

As of April 9, 2020, Idaho reports probable and confirmed cases separately.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

As of October 15, 2020, Illinois combines PCR and antigen tests in the total tests figure reported on the state's dashboard.

On September 4, 2020, Illinois reported over 140,000 new Total PCR tests (in specimens) and over 5,000 new Cases. According to the Chicago Tribune, "The Illinois Department of Public Health said Friday there was significant slowdown in the agency's test reporting earlier this week that it is just catching up with."

Indiana combines PCR and antigen tests in the total tests figure reported on the state's dashboard.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people, confirmed + probable) from Total PCR tests (people) in the absence of better data.

As of October 16, 2020, Indiana’s total test results are drawn from our totalTestsViral field instead of calculated via positive+negative.

On October 16, 2020, Iowa removed 12,831 tests from the state’s PCR test total. On the same day, 19,308 tests were added to the state’s antigen test total. We were unable to locate an official explanation, but surmise that these changes were the result of data cleaning and reorganization work.

On September 26, 2020, Iowa announced that due to a data update, roughly 27,000 antigen tests from long-term care facilities will be added to their COVID-19 antigen test data. This update will occur from September 26, 2020 to September 27, 2020, and will result in an increased number of antigen test tests reported in our data for Iowa. (Antigen data is available in our CSVs and API.)

As of May 19, 2020, Iowa reports PCR and antibody tests separately.

Kansas combines PCR and antigen tests in the total tests figure reported on the state's dashboard.

Kansas reports much of its data on Mondays, Wednesdays, and Fridays only, so there are often large changes in the values, which can represent 2-3 days of data.

The Recovered data point we report for Kansas reflects the number of COVID-19 patients discharged from the hospital and therefore does not represent the total number of people who have recovered from COVID-19, since many people with COVID are never hospitalized. Kansas notes that this metric only includes cases with available discharge information reported to the Kansas Department of Health and Environment.

On October 21, 2020, the inIcuCurrently value for KS dropped from 81 to 1 on the state's dashboard. We recorded it, but expect it to be revised at another point.

On July 15, 2020, Kansas began reporting additional data from hospitals, and at that time, data for the Now hospitalized metric was only available through July 10.

From July 1–July 3, 2020, we reported incorrect numbers for Kansas Cases or Positive PCR tests (people). We have corrected the error.

As of May 7, 2020, confirmed and probable cases are included in the case totals.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

As of September 18, 2020, Kentucky's total test results are drawn from our totalTestsViral field instead of calculated via positive+negative.

As of May 27, 2020, Kentucky reports antibody tests separately from PCR tests.

On May 23, 2020, Kentucky revised down the state's figure for Total tests.

Louisiana reports probable deaths on a weekly basis.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people)* from Total PCR tests (people) in the absence of better data.

On August 30 Louisiana's reported data included 532 older cases that had not previously been reported; these cases had specimen collection dates ranging back to July 1.

On August 26, 2020, Louisiana started reporting Probable cases, this caused a notable increase in Total cases.

On June 19, 2020, Louisiana's data changed due to deduplication and the state's removal of out-of-state cases.

Between April 20 and April 24, 2020, Louisiana did not report testing data from commercial labs due to a state review of data.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

On November 2, 2020, Maine started reporting antigen and molecular tests separately on the state's dashboard.

On June 23, 2020, Maine began reporting the dates of cases, deaths, recoveries, and hospitalizations on the day they occurred instead of the day the event was reported. Maine also began providing separate figures for Deaths (probable) and Deaths (confirmed).

As of May 20, 2020, Maine reports PCR tests and antibody tests separately.

As of October 16, 2020, Maryland’s total test results are drawn from our totalTestsViral field instead of calculated via positive+negative. This will result in TotalTestResults only going back to 3/24 instead of 3/5.

On August 31 2020 , we updated Maryland's Total PCR tests (specimens) and Positive PCR tests (specimens) prior to 8/16 to match state-provided "Testing Volume" data.

On July 31, 2020, we began capturing Maryland's data for Positive PCR tests (specimens).

Between March 12 and March 28, 2020, Maryland did not report data for negative tests.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (people).

As of November 2, 2020, Massachusetts’ Currently Hospitalized, Currently in ICU, and Currently on Ventilator figures include only confirmed cases. Previously these numbers included confirmed and probable hospitalizations.

As of September 11, 2020, Massachusetts’ total test results reflect total specimens. The number of unique people tested is available in our API as totalTestsPeopleViral.

On September 2, 2020, MA reported that it had shifted to using the more restrictive August 6 definition of probable cases released by the CSTE. This change caused a significant decrease in several data points where probable cases or probable deaths were included, particularly the overall case count, which dropped by more than 7,000. Historical data was also affected.

On August 31, 2020 we updated Massachusetts' Total PCR Tests (specimens) before August 12, 2020 based on the state's "Tests by Date" dataset, extending the column back to January 22, 2020.

On August 12, 2020, Massachusetts began reporting probable cases weekly instead of daily. This can lead to apparent spikes in the data when the weekly numbers are incorporated, as happened on August 19. Beginning August 20, 2020, we will use the daily total case count from the race and ethnicity section of the dashboard, which includes both confirmed and probable cases, and will backfill the previous week's case counts.

On July 31, 2020, Massachusetts noted that "Delays in the reporting of test results to DPH from multiple laboratories, including large national commercial laboratories, are reflected in today’s case counts. However, these positive test results have been assigned retrospectively to the appropriate date of test." We will backfill case data for Massachusetts in the near future.

On June 30, 2020, Massachusetts decreased the number of cumulative hospitalizations and confirmed and probable deaths "due to ongoing data cleaning which identifies and removes duplicate reports."

On June 1, 2020, Massachusetts began to report probable cases and probable deaths, causing *Total cases and Total deaths to appear to increase sharply in our data.

Michigan updates its Recovered metric weekly and does not update any data on Sundays.

Our source for Now hospitalized and Now in ICU data in Michigan generally updates on Mondays and Fridays only.

Michigan combines PCR and antigen tests in the total tests figure reported on the state's dashboard.

On October 29, 2020, Michigan experienced connectivity issues resulting in some cases that would have been reported tomorrow to be reported today. This is due to the fact that the state has a strict 10AM cutoff for daily data reporting. This means the daily increase in cases encompasses more than one day's worth of data.

On October 15, 2020, Michigan announced that due to a laboratory result processing slowdown, several cases that would normally have been included in the October 14, 2020 update were reported on October 15, 2020. The issue has since been resolved. Aa a result, cases reported on October 15, 2020 may not have been reported within the last 24 hour period.

As of July 7, 2020, Michigan updates its death figures three times a week after conducting reviews of death certificates. As a result, "new" deaths may not always reflect deaths in the past 24 hours.

On June 11, 2020, we updated our historical data from a spreadsheet provided by Michigan's Department of Health and Human Services in order to reflect the lab-reported dates of probable and confirmed cases and probable and confirmed deaths by date of death.

Since March 10, 2020, Michigan has reported each day's Positive and Negative tests (people) by county but has never totaled these metrics for the state. We add new Positive and Negative tests (people) values daily to a running sum and report the total.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (people).

From September 27 to October 4, 2020, Minnesota did not report the current number of hospitalizations or patients in the ICU. We froze these values through October 2, 2020. On October 3, 2020, we removed the values in accordance with our policy of not carrying over current values for longer than 7 days. The values were restored on October 4, 2020.

As of September 18, 2020, Minnesota total test results are drawn from our totalTestEncountersViral field instead of calculated via positive+negative.

The number of unique people tested is available in our API as totalTestsPeopleViral.

On September 17, 2020, we switched Minnesota's specimens timeseries to Total Test Encounters (PCR) based on correspondence with the state, which revealed that their "total approximate number of completed tests" reflects individuals tested per day.

Mississippi started updating testing totals on a weekly schedule on September, 27, 2020, so each update includes many days worth of testing.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

On October 25, 2020, Mississippi announced that COVID-19 data for October 24, 2020 and October 25, 2020 would be published on Monday. As a result, Mississippi did not update on October 24, 2020 and October 25, 2020.

On September 15, 2020, Mississippi resumed reporting current hospitalizations, ICU, and ventilations on the source we have historically used for this metric. We are working to backfill these data points in our historical data.

On September 11, 2020, we discovered a new source for Mississippi current hospitalizations and current ICU admissions on the Mississippi State Department of Health website.

Between June 17 and June 22, 2020, Mississippi experienced technical difficulties with their reporting system. We filled in five days of case and death data on June 22, 2020.

On May 26, 2020, news reports indicated that 95 antibody tests had returned positive in Mississippi, but we have not found more recent data since then for this metric.

As of May 23, 2020, Mississippi is separating out PCR and antibody tests. We now report both separately.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (people).

On October 14, 2020, Missouri testing numbers by people decreased. This resulted in a decrease in all total testing numbers (people) and negative cases. Testing numbers by specimen are unaffected.

On October 11, 2020, Missouri announced that a database error had resulted in an “incorrect inflation” of cases in its October 10, 2020 report. When The COVID Tracking Project is able to obtain correct official data for October 10 from Missouri, we will update our dataset to reflect the correction.

On October 8, 2020, Missouri added current hospitalizations to their COVID-19 dashboard, and we began to capture this metric. Historical data is also available and we are working on backfilling our data for days where this metric was unavailable.

As of October 7, 2020, Missouri's total test results are drawn from our totalTestsViral field instead of calculated via positive+negative. The number of unique people tested is available in our API as totalTestsPeopleViral.

On September 5, 2020, the Missouri DHSS added an additional 72 deaths to their total count, "most of which occurred June-August".

As of May 22, 2020, Missouri reports PCR tests separately from antibody tests.

Negative test results reported in our API and CSVs are calculated by subtracting Positive PCR tests (people) from Total PCR tests (specimens) in the absence of better data.

Nebraska combines PCR and antigen tests in the total tests figure reported on the state's dashboard, and thus antigen tests are included in our "Total Tests (PCR)" metric.

As of October 30th, 2020,, our Total tests metrics include Nevada's Total tests (in specimens).

On October 30, 2020, we backfilled the full time series for Nebraska's Total Tests into our "Total Tests (PCR)" metric from a Nebraska GIS query. At the same time, we filled our Positive Tests (PCR) and Negative Tests (PCR) from the same query.

On September 20, 2020, Nebraska started publishing Total Tests (in specimens) in addition to People Tested (in people), and we started collecting the Total Tests daily into our Total Tests (PCR) metric.