In Smaller Hospitals, COVID-19 Patients At Triple The Risk Of Dying, National Study Finds

Research from Boston and around the country finds that critically ill COVID-19 patients are much likelier to survive if they’re treated at bigger hospitals.

The sweeping study just out in the journal JAMA Internal Medicine is the first to look at hospital mortality rates in COVID-19 patients nationwide. It includes detailed data on more than 2,200 patients in 65 hospitals.

"Patients who were admitted to hospitals with fewer than 50 ICU beds— so, smaller hospitals — had a more than threefold higher risk of death than patients admitted to larger hospitals," says senior author Dr. David Leaf from Harvard Medical School and Brigham and Women’s Hospital. The following are edited excerpts of our conversation.

How would you sum up what you're reporting in this paper?

We conducted a multi-center study across 65 hospitals in the United States where we collected detailed data on over 2,000 critically ill adult patients with COVID-19. We found that patients had a 35% risk of death within 28 days of ICU admission.

We also identified several independent patient- and hospital-level risk factors for death. Patient-level risk factors included characteristics like older age, being male, obesity and cancer.

On hospital-level risk factors, we found that patients who were admitted to hospitals with fewer than 50 ICU beds — so, smaller hospitals — had a more than threefold higher risk of death than patients admitted to larger hospitals.

We also found that treatments vary dramatically from hospital to hospital, whether we're talking about medications like hydroxychloroquine or steroids, or certain interventions like proning, which is when you flip a patient onto their belly, which helps with oxygenation.

Many people will hear that statistic about a three-fold higher risk and say, "If I get COVID, I want to make sure I go to a big hospital." Would you want to temper that reaction with any sort of caveats?

I'm glad you brought that up. We collected very detailed data in this study. We had over 800 data points per patient, each of which was manually chart-reviewed. Thus, with over 2,000 patients we had over a million discrete pieces of data. We adjusted for a very comprehensive set of potential confounders.

Nonetheless, we weren't able to capture everything imaginable. So, for example, we didn't have data on doctor and nursing staffing. We didn't have data on hospital strain. And we didn't have data on socioeconomic status of patients, which has emerged as an important risk factor in patients with COVID-19. So certainly there could be residual confounding. There could be factors that we didn't measure that could explain some of these findings.

But would you say that it's a reasonable takeaway, that it's probably a better bet to go to a hospital with more than 50 ICU beds?

That's really tough to answer. Larger hospitals do have more resources. And in settings outside of COVID-19 — for example, if you look at patients who require mechanical ventilation for respiratory failure from due to causes other than COVID-19 --  larger hospitals do seem to have better outcomes.

Could you comment on your finding that race didn't seem to be correlated with mortality?

We found that Blacks were actually much more represented in our cohort of critically ill patients than they are in the country: About 30% of our population was Black, whereas Blacks only represent about 13% of the US population. Thus, Blacks were over-represented in our cohort of critically ill patients by nearly threefold.

But among patients who were in our study, mortality did not differ according to race, which is similar to findings that were reported in a study from Louisiana that was in the New England Journal of Medicine. If you were Black, you were more likely to be admitted to the hospital or to an ICU, but once you're already there, you have similar outcomes compared to patients of other races.

How would you say our American mortality statistics are measuring up against other countries'?

It's really hard to compare studies because different studies have different lengths of follow-up. Some studies only follow patients for a week, and patients often don't die in the first week. There are also differences in admitting practices across countries, and even within the same country, across hospitals. I would say our mortality numbers are in the same ballpark as what other countries have reported.

Are your findings actionable for medical staffs or families?

We didn't develop a prognostic scoring system where you plug in the patient's age and gender and other elements, and it tells you that you have a 90% chance of death or a 10% chance of death, for example. We've done that in a subsequent paper that is currently under review.

But I think just having a sense of what the risk factors are for death might be helpful for patients and families in making decisions about how aggressive to be.

Were there any institutional red flags about outlier hospitals that looked really bad?

I can't get into specific names but the short answer is yes. The risk-adjusted mortality in the study ranged from 7% at the lowest-risk hospital to 80% at the highest-risk hospital. That was one of the key findings of the study, in fact: more than a tenfold change in your risk of death, after having accounted for many of the differences in patient characteristics across hospitals.

So again, doesn't that mean you really want to go to a bigger hospital if you can?

But you may not have a choice. You may have to just go to the closest hospital.

What's next?

What we're really interested in now is identifying treatments that can improve outcomes. Of course, randomized control trials are ongoing and those will always be the gold standard, but randomized controlled trials take a lot of time and they can't answer every single question.

So, for example, there are randomized controlled trials ongoing for certain drugs, but they're not specifically looking at critically ill patients; or they may only be enrolling a much smaller number of patients than what we had here, and thus will not have adequate statistical power to assess hard outcomes like mortality.

So we are trying to use our database, which now includes over 5,000 patients, to answer some of these questions about which treatments improve outcomes. We have several papers that are under review that are looking at this.