The relatively low number of deaths could be due partly to data collection problems and data reporting delays, partly to improvements in COVID-19 treatment methods, and partly to the nature of COVID-19: Deaths from COVID-19 tend to occur about two to eight weeks after people start to have COVID-19 symptoms, according to researchers at Harvard.

The most recent week that appears to have complete death data is the week ending Sept. 12. In that week, 24 states had total death counts that were at least 10% higher than the expected level, compared with 35 states that had death counts at least 10% higher than the expected level in the week ending Aug. 22.

Related: Is the U.S. ready for a COVID-19 vaccine?

The CDC bases the expected death counts for a week on the average number of deaths reported in a state for the comparable week in the year in the previous three years.

For the week ending Sept. 12, the total number of deaths from all causes was about 6% higher than the expected number.

The United States normally averages about 2.7 million deaths per year.

The country has recorded about 221,000 deaths resulting directly from COVID-19 as of the time of this article. If mortality stayed at 6% over the usual level for the rest of the year, then the country might experience about 36,000 additional excess deaths as a result of COVID-19 and related problems, such as the effects of COVID-19 on patients' willingness to seek care for conditions other than COVID-19.

One CDC indicator that suggests that COVID-19 morality could stay high is the ratio of deaths caused by COVID-19 and illnesses that look like COVID-19 — pneumonia and influenza — to all deaths. That indicator continues to be well above the epidemic threshold for the week, despite data reporting problems that tend to make the ratio look lower than it really is for several weeks after the death numbers start coming in.

The Dynata data

Dynata, a data firm, powers many website instant survey programs. In place of its regular surveys, it serves some site users a COVID-19 symptom survey.

The company than feeds the survey data into a free, public — but copyrighted — tracking dashboard. The company can't be certain whether the people who report having COVID-19-like symptoms actually have COVID-19 or some other health problem, but, while COVID-19 appears to be a common viral infection, reports of people having those symptoms could be an early sign of the intensity of COVID-19 activity in a given region.

One question has been what the return of students to colleges, universities and other schools might do COVID-19 infection rates.

The percentage of U.S. Dynata survey participants ages 18 to 24 who said they had a dry cough with loss of the sense of smell or taste was 3% for the two-week period ending Oct. 16. That's the same as the percentage of U.S. college-age respondents who told Dynata they had COVID-19-like symptoms for the two-week period ending Sept. 25.

But the number of states where 3.5% or more of the college-age Dynata survey takers said they had COVID-19-like symptoms increased to at least 11 as of Oct. 16, from four for the two-week period ending Sept. 25.

That state count excludes some states for which Dynata has fewer than 50 responses for people ages 18 to 24.

When the minimum reporting count is set at one, the number of states with COVID-19-like symptom rates higher than 3.5% for college-age Dynata survey respondents increases to 13 for the latest period, from seven three weeks earlier.

The national percentage may have stayed about the same because the states where the symptom rate is spiking are low-population states: The college-age survey taker symptom rates were highest in South Dakota, Alaska and Oklahoma.

Some COVID-19 trackers have suggested that the 80th Sturgis Motorcycle Rally, which started Aug. 7 in Sturgis, South Dakota, may be leading to a spike in COVID-19 cases in the Dakotas and other states in the middle of the country.

Among U.S. states with at least 100 college-age Dynata respondents, the state with the highest COVID-19-like symptom rate is South Carolina. About 6.9% of the college-age Dynata respondents in South Carolina have been suffering from COVID-19-like symptoms. Only 0.6% of the South Carolina Dynata respondents in other age groups have been having COVID-19-like symptoms.