As COVID-19 hospitalizations continue to climb in many US regions, it’s important to clarify what each available hospitalization metric captures, and why some may be more immediately useful than others. We also want to explain why we recently removed one metric—national "cumulative hospitalizations"—from our website.
COVID-19 patients in the hospital now
The “current hospitalizations” number on our website and in our API measures how many people with COVID-19 are receiving care as patients staying in hospitals right now. What counts as a COVID-19 hospitalization varies across states and territories: some states only report hospitalizations from a confirmed COVID-19 case, others report confirmed and pending or suspected cases, and about half of the 56 jurisdictions we track either have an unclear definition or don’t offer any readily available definition. These differences in definition make it hard to compare current hospitalizations across states with precision—given that patients with outstanding test results are often excluded, we should usually think of it as representing the minimum number of COVID-19 patients in a state’s hospitals on a given day. Due to definitions differing between states, we also can’t assume that one state’s reporting of COVID-19 hospitalizations represents exactly the same category of patients as another state’s reporting. Of the metrics we can collect from states, this metric is the most useful for measuring what is happening in hospitals across the United States currently and over time.
Since state populations vary widely, it’s often helpful to use per-capita figures as context for hospitalization numbers. For example, at its November 10 hospitalization peak this fall, the state of South Dakota reported 607 current hospitalizations—a number that placed it in the middle of a ranking of US states; however, this corresponds to approximately 686 current hospitalizations per million people, far more than any other state at this point in the pandemic. In contrast, Texas (a state with a substantially higher population than South Dakota) reported a higher number of current hospitalizations on the same day, but the number of hospitalizations per million was substantially lower.
The biggest problem with current hospitalizations as a measure of a state’s outbreak is that it does not capture how many COVID-19 patients enter and leave the hospital on a given day; patients who are discharged or who die may obscure the number of newly admitted patients and fluctuations in patients’ length of stay add another element of complexity. As a result, there is another metric that’s more useful for understanding whether the situation in a given area is improving, worsening, or holding steady: new admissions.
COVID-19 patients entering the hospital
The “new hospital admissions” metric captures the number of people (adult and pediatric cases) entering into the hospital system on a particular date as a result of confirmed or suspected COVID-19 infection. In contrast to current hospitalizations—which can give an overall sense of how COVID-19 is impacting a state’s population—new hospitalizations can potentially pinpoint when and where the spread of a disease is becoming more severe. Most US states do not provide a new admissions number for COVID-19 patients, and therefore we don’t compile this metric from states, but these figures are now available in a public dataset from the Department of Health and Human Services (HHS) that we’ve written about here on the blog. We are evaluating how best to use this federal data in our work.
People ever hospitalized with COVID-19
The third hospitalization metric we work with, “cumulative hospitalizations,” measures the running total number of COVID-19 hospitalizations in an area since the beginning of the pandemic. Cumulative metrics can help us understand the overall impact of a disease on a population, but although we continue to collect this metric from states that offer it, we have found that it’s difficult to use in practice. Whereas cumulative cases and cumulative deaths are widely reported, only 36 states and territories report cumulative hospitalizations. Until recently, we rolled these 36 figures up into a national summary on our website, but as hospitalizations rose across the country, it became clear that this dramatically incomplete figure was both confusing and not reflective of reality at the national level.
If we had a complete historical dataset for every US state and territory of new COVID-19 hospital admissions since the beginning of the pandemic—and this dataset excluded all patients with suspected cases of COVID-19 or pending tests—it would be theoretically possible to stitch together a cumulative hospitalizations number for the states and territories that don’t report this figure. However, the HHS admissions data is only available from July onward and is very incomplete for the first few months of its time series. We have therefore removed the national cumulative hospitalization metric from our site, though we still report it as “ever hospitalized” at the state level wherever we can track it. Even at the state level data, we caution against using the cumulative hospitalization metric as it may not be recently or actively updated. For example, New York's reported cumulative hospitalization number has been stagnant since early June and Connecticut's since late October.
Understanding the differences between available COVID-19 hospitalization metrics is vital in understanding the current burden of COVID-19 on healthcare systems across the country. We will continue to report all official current and cumulative hospitalization data we can collect from US states and territories, and we will refer to the HHS new admissions metric in our analysis as well whenever it offers useful context on the state of the pandemic.
Catherine Pollack is a third year PhD candidate in the Quantitative Biomedical Sciences program at Dartmouth College. Her dissertation research combines data science, epidemiology, and public policy to combat online health misinformation.
Dave Luo has an MD/MBA and is a Data Science and Data Viz lead at The COVID Tracking Project.
The Department of Health and Human Services released a new public dataset on December 7 that includes data down to the facility level on where COVID-19 patients are hospitalized and how healthcare systems are coping. We explored how to use this dataset and what patterns it can reveal.
Despite a rocky beginning, the current hospitalization and new admissions metrics from the HHS Protect public dataset have stabilized—and they’re now largely harmonious with state-reported hospitalization metrics if we account for differences in data definitions and reporting lag time.