Federal Count of COVID-19 Nursing Home Deaths Drops After First Update

The federal government on Thursday made the first routine update to its official nursing home COVID-19 data, and while the number of confirmed and suspected cases increased as expected, the most recent figures show a drop in deaths of more than 2,000.

The Centers for Medicare & Medicaid Services (CMS) reported that there were more than 107,000 confirmed cases of COVID-19 infection in nursing homes, along with 71,278 suspected cases as of June 7.

Source: CMS.gov, captured June 18, current as of June 7

But the total number of deaths fell to 29,497 as of that date; earlier this month, when CMS posted its first round of data to its consumer-focused Nursing Home Compare site, that figure was 31,782 as of May 31.

Advertisement
Source: CMS.gov, captured June 4, current as of May 31

The declining official count came during the same week that an independent Wall Street Journal analysis of state-level data estimated the total number of COVID-19 deaths in long-term care above 50,000.

That would account for roughly 40% of the more than 116,000 coronavirus deaths in the United States tracked by Johns Hopkins University, according to the outlet.

When asked to clarify the discrepancy, a CMS spokesperson indicated that due to the large number of coronavirus-related media inquiries the agency is currently receiving, CMS could not provide an answer to SNN by press time Thursday. This story will be updated with a response when it arrives.

CMS has been upfront about the imperfections in its national COVID-19 information, which it began collecting in conjunction with the Centers for Disease Control & Prevention (CDC) in early May.

The self-reported data, which each of the nation’s more than 15,000 nursing homes must submit through the CDC’s National Healthcare Safety Network (NHSN) portal, had errors such as double-counting of cases and inconsistent reporting windows, administrator Seema Verma said earlier in June.

“As with any new program, some facilities are going to struggle as they come on line, and there’s going to be honest errors in data entry,” Verma said on a press call announcing the first round of data.

All three numbers were widely expected to rise with each subsequent update, given steadily improving access to testing in many parts of the country, greater compliance with the data-reporting rule, and CMS and CDC’s pledge to perform quality-control checks on potential errors.

For instance, three days before CMS posted the totals and building-by-building information to Nursing Home Compare, the federal government announced that nearly 26,000 people had died from COVID-19 in nursing homes, which rose to almost 32,000 later that same week.

“I want to be very clear, though, that this is not an increase of six thousand deaths in a few days,” Verma said in explaining the jump.

CMS and CDC also explicitly warned researchers not to use the first few weeks of data to conduct formal studies of death and infection patterns, given the potential for errors that will be corrected.

“CMS and CDC perform quality assurance checks on the data and may suppress data that appear to be erroneous,” the agencies wrote in a FAQ about the reporting system. “The data is not altered from what nursing homes report to the NHSN system. Data regarding numbers of new cases, suspected cases, or deaths are aggregated. Nursing homes have the ability to correct their data so future postings include corrected data.”

The agencies additionally cautioned that even entries that may seem incorrect — such as a facility reporting more COVID-19 cases than it has licensed beds — could be accurate.

“The nursing home data posted reflects cumulative cases and deaths of COVID-19 reported in a facility,” they noted in the FAQ. “A nursing home may report more confirmed COVID-19 cases than beds because they may encounter a large number of admissions and discharges of residents with COVID-19, resulting in more cumulative cases of COVID-19 over time than the number of beds at any one time.

Companies featured in this article:

,