I recently wrote about the relationship between Governor Cuomo’s policy requiring nursing homes to accept new residents discharged from hospitals with positive COVID-19 test results. The policy was in place from March to early May 2020. I compared my results with those of a study by Bill Hammond and Ian Kingsbury of the Empire Center, “COVID Positive Admissions Were Correlated with Higher Death Rates in New York Nursing Homes.”
Hammond and Kingsbury concluded that the Governor’s policy was associated with several hundred to more than one thousand additional deaths. My analysis agreed with Hammond and Kingsbury’s conclusion that the policy was related to additional deaths but found that the number was likely much lower than their estimates.
The Empire Center report found that “the data indicate that the March 25 memo was not the sole or primary cause of the heavy death toll in nursing homes, which stood at approximately 13,200 as of early this month.” My findings are consistent with Hammond & Kingsbury’s in that regard.
I recently reviewed the data in my analysis and found that about 25% of COVID-positive admissions and deaths were inadvertently omitted. For that reason, I rebuilt the dataset and reanalyzed the data. The reanalysis led to results closer to Hammond and Kingsbury’s regional findings. However, I believe that the estimates of deaths associated with the policy generated from separate regional analyses of data from the New York City Metropolitan area and Upstate provide more accurate estimates of the policy’s impact. Consequently, I believe that deaths associated with the policy are lower than Hammond and Kingsbury estimated.
Because Hammond & Kingsbury’s research design did not isolate deaths of residents who were COVID-positive at the time of admission from those who died as a result of the spread of infection from COVID-positive residents to others, their estimates are larger than the actual impact of the policy on COVID nursing home resident deaths. The concern that the Health Department’s mandate was associated with additional deaths was primarily based on the possibility that the newly admitted COVID patients might infect other nursing home residents, despite mandated infection control requirements.
I am indebted to Bill Hammond and Ian Kingsley of the Empire Center for being transparent about their approach, providing the data from their successful Freedom of Information suit, and their willingness to discuss their findings with me.
When the Health Department required nursing homes to admit new residents who had tested positive for COVID-19 from hospitals, nearly 9,000 COVID-positive residents were admitted. The Empire Center analysis examined deaths in April and May because of the delay between COVID infections and resulting deaths. About 8,000 COVID-related deaths occurred in New York nursing homes during that period.
Admissions of COVID-positive residents and COVID-related deaths were concentrated in the New York Metropolitan area. Although about six in ten nursing home residents in New York State were located in the New York Metropolitan area, more than 90% of COVID resident admissions and COVID-related deaths in nursing homes occurred there.
New York Metropolitan Area Findings
With the large numbers of infected metropolitan area residents, it is unsurprising that nursing homes in the region had large numbers of COVID-related deaths during the Spring of 2020. During the short seven-week period, COVID-related deaths in downstate nursing homes were equal to 11% of the region’s nursing home population.
Positive test results per thousand residents in the New York City metropolitan area were more than ten times higher than outside the metro area. With more than 90% of the state’s COVID-positive admissions to nursing homes and COVID-19-related deaths, conditions in New York Metropolitan area nursing homes reflected the high infection levels in the metropolitan area.
Because of the large number of infections in the metropolitan area, hospitals faced extreme crowding. More than 8,200 COVID-positive patients were discharged from metropolitan area hospitals to nursing homes during the March 25th to May 8th period.
The trend-line model included new COVID-19 admissions to nursing homes, county-level positive test rates, and the number of nursing home residents to predict the number of COVID-related deaths. In about 300 nursing homes in the New York City metropolitan area, the relationship between COVID admissions and deaths was very weak, associated with only 1% of the variables’ variation.
In Hammond and Kingsbury’s and my earlier analyses, the relationship between COVID-positive admissions and COVID-related deaths in New York City nursing homes was not statistically significant at the 95% confidence level.
Given that the strength of the association between COVID-positive admissions is so weak — only one percent of deaths are associated with COVID-positive admissions — and that the relationship was not statistically significant, I conclude that the Health Department mandate that nursing homes accept COVID-positive residents did not result in more deaths in the New York metropolitan area.
Findings Outside the New York Metropolitan Area
The COVID-19 pandemic had a much smaller impact on upstate New York during the Spring of 2020 than in the New York City metropolitan area. Upstate nursing homes were also less affected as well. While the COVID-related death rate in the metropolitan areas was 11.2%, the rate was 1.9% in upstate New York. The state Health Department reported 730 deaths in the Upstate nursing homes studied here. Transfers of COVID-related patients were also less common. Only 532 COVID-positive patients were transferred to Upstate nursing homes during the study period, compared to 8,200 in the metropolitan area.
Perhaps because community infection rates were much lower in Upstate New York than in the New York City Metropolitan Area, the relationship between COVID-positive admissions to nursing homes and COVID-related deaths is much more robust in Upstate nursing homes. A correlation measure between COVID admissions and COVID-related deaths shows that admissions are associated with 24% of the regional variation in fatalities.
This relationship is much stronger than I found in my earlier draft from the incomplete dataset. The model estimates 283 additional deaths from the Health Department policy in upstate New York. The model predicts that the range of additional deaths with 95% certainty is 225 to 341.
The differing findings in the New York Metropolitan area and the rest of the state point to two different modes of transmission in nursing homes. In the New York metropolitan area, where more than 90% of the COVID-positive admissions and deaths took place, COVID-related deaths were unrelated to COVID admissions, pointing to staff and visitors as likely the principal means by which infection was transmitted. In upstate New York, the role of new COVID admissions was significant, with about one-third of the 730 regional deaths associated with the Health Department mandate.
In addition to building a regional trend-line model, Hammond and Kingsbury also created a statewide model. That model was the basis of their estimate that COVID-related deaths associated with the Health Department mandate ranged from several hundred to more than 1,000. Using the revised, corrected data and the variables I included in my analyses — number of nursing home residents, county positive tests per thousand residents, and COVID-positive admissions, the model estimated that between 1,610 and 2,463 additional deaths were associated with the State Department of Health’s policy. Twelve percent of the variation in COVID-related deaths was related to COVID-positive nursing home admissions — a relatively weak relationship.
The corrected statewide data estimates that a larger number of COVID-related deaths were associated with the Health Department COVID-positive nursing home admission mandate than my earlier analysis showed. This estimate is also higher than the estimate published by Hammond and Kingsbury.
Because of the distinctive pattern of infections in New York State in the Spring of 2020, infections were highly concentrated in the New York metropolitan area. The regional models point to different transmission patterns in each region. In the downstate area, where more than 90% of COVID-related nursing home deaths occurred, the data does not show an association between the admission of COVID-positive new nursing home residents and COVID-related deaths. Upstate, a relatively strong relationship is present, with about 30% of regional deaths associated with the entry of COVID-positive residents, based on the model.
The large number of deaths that the statewide model estimates are associated with COVID-positive admissions — more than 2,000 is unlikely– given that in the region where a statistically significant relationship between admissions and deaths was present, only 730 deaths occurred.
Deaths of COVID-positive Nursing Home Residents Were Included in Study Estimates
The concerns about additional deaths that resulted from the Health Department’s mandate arose from the fear that introducing residents with a transmissible disease into a nursing home setting could spread to other residents.
But, Hammond and Kingsbury’s analysis does not subtract the deaths of newly admitted COVID-positive residents from their estimate of COVID deaths at the facilities. The deaths of people who were already COVID-positive could not be caused by the potentially COVID spreading effect of the Health Department mandate.
Data that breaks out the deaths of patients who were COVID-positive when admitted was not included in the information provided to the Empire Center by the State Health Department. Without that data, it is impossible to accurately measure the number of deaths of newly admitted COVID-positive residents at each nursing home.
However, there are a few published studies on the death rates of COVID patients after discharge from hospitals. A recent study published in the Journal of the American Medical Association, which dealt with patients in Veterans Administration Hospitals whose median age was between 70 and 80 years old and who were admitted before June 1st, 2020, found a death rate within sixty days of discharge of 9.1% of patients — an estimate that is almost identical to the increase of 0.09 deaths for each COVID-positive new resident admitted found by Hammond and Kingbury.
Although it is not possible to directly measure the net number of deaths of nursing home residents after excluding deaths that took place among residents who were COVID-positive when admitted, it is possible to estimate the net number of deaths by subtracting 9.1% of the number of COVID positive admissions from the number of deaths from each facility. This estimate is crude because we lack information about the actual number of deaths of COVID-positive residents, but it allows us to understand the potential impact of excluding them.
The exclusion of possible deaths of newly admitted COVID-positive residents slightly reduces the estimated deaths in Upstate nursing homes, from 283 to 258, with a range within 95% confidence limits of 205 to 310.
With corrected data, evidence that the Health Department’s requirement that nursing homes accept new COVID-positive residents from nursing homes is stronger than I initially concluded. Based on my corrected analysis, 205 to 310 residents in upstate New York may have died from COVID-related causes when controlling for deaths of COVID-positive admitted residents. Although the number is a relatively small percentage of COVID-related nursing home deaths statewide, it is 30% of the of the approximately 700 Upstate nursing home COVID-related deaths.
While this analysis answers the question of whether COVID-related deaths in nursing homes were related to mandated COVID-positive admissions, it does not provide information about the net impact of the policy on COVID-related deaths inside and outside nursing homes.
In the Spring of 2020, hospital facilities in the New York metropolitan area faced severe challenges related to facility, staffing, and personal protective equipment availability. Conditions were so extreme that it was difficult for COVID-positive patients to gain admission to hospitals in the New York City area.
The State Health Department’s COVID admission mandate on nursing homes was a response to the hospital capacity crisis — setting up a potential policy tradeoff between deaths in nursing homes and deaths outside them.
Few policy decisions involve unambiguous benefits and no costs. This policy decision, like many others, did have both. But, at the time, it was clear that the Health Commissioner was well aware of the crisis conditions in hospitals in the New York Metropolitan area and that people were dying because of them. We still do not know whether his decision cost or saved lives overall.
Unfortunately, press coverage of the nursing home issue has focused only on the policy’s effect on the number of nursing home residents’ deaths. But by looking only at these deaths, press coverage, like the response of many of the state’s political leaders, did not reflect the net impact of deaths inside and outside nursing homes.