The Empire Center’s February 18th Report, “COVID-positive Admissions Were Correlated with Higher Death Rates in New York Nursing Homes” shows an association between COVID-positive admissions and death rates in nursing homes in New York State, following the State Health Department order on March 25th, 2020 requiring that they accept recovering COVID patients. The Empire Center report finds that several hundred to more than a thousand deaths could have been associated with the Health Department policy.
However, a closer reading shows that the relationship is statistically significant only in areas of the state outside the New York City metropolitan area, perhaps because high community transmission levels were responsible for most deaths in the NYC area. It also shows that the authors may have treated deaths at nursing homes that sent patients to hospitals for COVID treatment and later received them in return after their infections were under control in a way that could have led to false conclusions about the relationship between the order and nursing home deaths.
The report raises a number of questions. One relates to the strength of the association between releases and death rates. Although the title of the report asserts that there is a correlation, it does not provide the commonly used statistical measure that shows the strength of the association between variables. It should be noted that the report notes 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.”
The authors of the report, Bill Hammond and Ian Kingsbury, conducted multiple regression analyses of the data that permit evaluations of the relative contributions of variables like admissions of COVID positive nursing home patients along with factors like rates of community COVID spread but did not report the relative contributions of the variables that they measured.
Without data that shows the strength of association between these variables and COVID death rates in nursing homes, we can’t understand whether the Health Department policy was strongly related to death rates, or was only weakly related to them.
Another question relates to the exclusion of cases where patients were readmitted to nursing homes after a COVID diagnosis. The authors report, “The admissions figures exclude 2,279 patients who were readmitted to nursing homes where they were already residents. Because such patients were not new to those facilities, they were seen as less likely to be the original cause of an outbreak.” But, the report included other deaths at these facilities, only excluding those of existing residents.
Excluding these cases from the analysis while retaining remaining cases from the affected nursing homes could well inflate the apparent relationship between the Health Department policy and mortality rates. Since patients who were sent to hospitals from nursing homes and then returned were most infectious prior to being symptomatic or before being admitted, these cases could be the cause of many or most of the later deaths at the facilities.
Segregating these nursing homes in the analysis and treating them as having outbreaks that were likely the result of internal infection spread would be helpful in understanding the causes of mortalities in them. Instead, the authors lumped the nursing homes that sent patients to hospitals and accepted them back with others that accepted new hospital discharges of COVID patients. Doing so could have skewed the overall analysis and affected the conclusion that the Health Department policy on admissions of hospital discharged COVID patients led to more COVID-related deaths in nursing homes.
I’ve been in touch with Bill Hammond of the Empire Center about the issues and hope to receive some clarifications of the data and the methodology used in the study.