Right now, the state of Pennsylvania is reporting more than 3,100 deaths from COVID-19 but that could understate how many people have died during the pandemic.
One model from the Centers for Disease Control and Prevention [CDC] estimates that, during the first five weeks of the pandemic, there were around 7,000 more deaths than would be expected over a similar time period in the past.
This count above the norm would be among the highest for any state in the country, but PublicSource reporting shows that number is still quite uncertain and could be higher or lower.
The reason: The CDC’s estimate is based on the number of deaths reported so far and how quickly the state normally reports its data — with that information, the CDC adds how many deaths it expects to be eventually reported. But the estimate could be skewed because the state hasn’t been reporting its data at the same pace as it has in the past.
To be clear, the 7,000 deaths figure is based on the CDC estimate and, so far, the state of Pennsylvania has actually reported to the CDC about 500 fewer deaths than what would be considered normal during the weeks of March 21 through April 18. It takes a while to report all of the deaths, and the CDC believes what Pennsylvania has reported so far is an undercount but it’s not clear by how much. So far, Pennsylvania reported 13,579 deaths and the model predicts 21,416 for the five-week period.
One problem with the state’s official count of deaths is that not everyone who dies has been tested for coronavirus. Some people who officially died of one cause may have actually been impacted by COVID-19 instead of or in conjunction with the listed cause of death.
The second reason the total deaths may be undercounted is that there are people dying of other diseases at a higher rate than normal because there isn’t enough healthcare capacity to help them or they are afraid to seek help because of COVID-19.
“We are seeing a big increase in heart disease deaths, diabetes deaths and some other causes well above what we would’ve expected given previous periods,” said Bob Anderson, the chief of the mortality statistics branch at the CDC.
It’s also possible that the total number of deaths could be going down in some places, as the stay-at-home order may be decreasing traffic fatalities and homicides, for example.
On Tuesday, the New York Times published an article that tried to calculate how many excess deaths there might be in most states across the country. It found that in some places, like New York City and New Jersey, there were thousands more deaths than would be expected since the pandemic started, including thousands more in New York City than have been attributed to COVID-19.
But the Times left out Pennsylvania from its story: “In Pennsylvania and Ohio, death reporting seems to be lagging far behind the normal rate all year.”
The Times story relied on a CDC database that collects mortality data from states. Because there is a lag in how fast the deaths are reported, the CDC calculates how many deaths it expects to be reported eventually. The way it does that is by looking at how many deaths were reported at a similar time period in the past and predicts how many additional deaths it expects to eventually be reported later.
This works well in normal times when states are predictable about when they report. But it may not be as reliable now when some states are reporting data more slowly or more quickly than in the past. Anderson said that, in some cases, the CDC may be overweighting or underweighting the number of deaths it expects.
Undercounting or overcounting?
At the end of the five-week period at issue, Pennsylvania had only reported about 1,500 deaths directly attributable to COVID-19. The CDC's estimate of 7,000 deaths beyond the norm for the state is higher than any other state in the Times article except for New York City.
“That’s a lot,” Anderson admitted. “It’s a lot.”
The state of Pennsylvania is reporting deaths more slowly than usual, Anderson said. If the data is actually coming in more slowly than usual, that would mean there could have been even more than 7,000 excess deaths during that five-week period.
But the CDC’s model for predicting deaths could also be predicting too many deaths, Anderson said. That would happen if Pennsylvania reported its death total slower than usual in January and February and then started reporting its death total more quickly in March and April. Because the model is based on the state’s pattern of reporting deaths, he said, this unusual behavior would throw off the calculations.
“We’re trying to get a sense of excess deaths and recognize it’s not perfect and not a hard number at this point,” he said. “But the numbers will become better and better as more data flow in for these time periods.”
Anderson said only the state of Pennsylvania could say in more detail how its state death total may change based on the data it has reported so far.
In response to a PublicSource question on Wednesday, Pennsylvania Secretary of Health Dr. Rachel Levine said the state has had trouble reconciling several different reporting systems, particularly stemming from Philadelphia.
"Reporting the deaths and cases has certainly been a challenge in our department in real time almost,” Levine said. "We are going to be working every day now to reconcile our data with the Philadelphia County Health Department.”
Levine said the state hasn’t yet looked into the CDC’s report of excess deaths. "We certainly know there are over 3,000 specific deaths from COVID 19,” she said. "I’m aware of some of the calculations the question is referring to and we’ll be taking a look at that."
PublicSource requested the total number of deaths in 2020 by week from the state Department of Health in an effort to make a comparison to the CDC data. However, the state reported its data as a monthly total rather than a weekly total and didn’t report any data past March, so the data isn’t directly comparable to the CDC’s data. A spokesperson for the state wrote that the data only includes deaths from its electronic system and it expects there to be just under 2,000 additional deaths reported by paper certificates.
Deaths reported in Pennsylvania through the electronic system only:
- January 2020 – 12,498 deaths registered
- February 2020 – 11,309 deaths registered
- March 2020 – 10,783 deaths registered
Without including deaths from the paper certificates, these death totals are similar to the death totals reported last year. In 2019, the state reported a low of 10,302 deaths in September and a high of 12,123 in December.
“Comparing 2019 data to 2020 data would mean you were working with one data set with the paper records entered in, and one without,” wrote Nate Wardle, a spokesperson for the Department of Health. “You will not be comparing apples to apples, since paper records are not part of this count."
Oliver Morrison is PublicSource’s environment and health reporter. He can be reached at email@example.com or on Twitter @ORMorrison.
PublicSource has a special page dedicated to our reporting on COVID-19 for the Pittsburgh region. See it here and sign up for our newsletter to stay informed. We hope you are following the news and, if your situation allows, social distancing guidelines. Have a tip or an idea? Please email firstname.lastname@example.org.