The evaluation of quality for hospitals is complex. One of the routine and key measures of quality is the re-admission rate. This is how often does a population with a certain condition who gets discharged from a hospital return for another inpatient stay within 30 days. The idea is that at least some re-admissions are preventable through better post-discharge planning, support and interventions. From this idea, preventing the preventable re-admissions leads to better care and quality while also plausibly saving money. Facially this makes sense.
There are a lot of issues with this measure for this construct of quality, but I want to probe at something that has been bugging me for a year or more now.
There is an assumption that the marginal admission is mostly constant. The model assumes that there is always a bed available. The model also assumes that some patients should always be admitted/re-admitted to the hospital and some patients have a health status that dictates that they should never be re-admitted to the hospital. However there are some patients who could benefit from a hospitalization but who could also be just fine so the decision to send someone upstairs to an inpatient bed is a coin flip. These are the marginal patients.
In non-COVID times, the marginal patient might be pretty damn close to a constant over time and space once we take into account hospital and physician fixed effects and any regional trends. If that is the case, then there are other issues with readmission rates but a basic measurement problem can be waived away. However I don’t think the assumption that the marginal patient is a constant is a good assumption in COVID times. We know that the marginal patient admitted to a New York City hospital in April 2020 is very different than the marginal patient admitted to a New York City hospital in August 2020 much less April 2022. We know that entire regions were down to a handful of staffed beds available at different times. We know, in discussions with clinicians, that patients in pre-COVID times that would have been NO DOUBT ABOUT IT, ADMIT THEM, were being sent home with instructions to call if things got really bad.
More simply, someone who normatively should be re-admitted won’t be re-admitted if there is no staffed bed available for them to be readmitted to. Pulling this speculation a step further back, we could plausibly expect to see readmission rates decrease in regions/hospitals with COVID surges more than regions where COVID was not as severe at a given time just because the marginal patient who was discharged does not have a bed available to be re-admitted.
There are other factors at play here. We could plausibly believe that the severity of people who get initially admitted in high COVID region/times may be higher than those who would have been admitted to that same hospital in a no-COVID counterfactual or admitted to hospitals in low COVID regions/times. This higher hypothetical severity could lead to more re-admissions balancing or swamping the marginal patient problem. It is also conceivable that patients admitted to a hospital in high COVID region/time get less intensive care or less bed days due to a need of the hospital to clear a bed faster and thus leading to more re-admissions.
I think all of those stories are plausible right now. As a reviewer, I’m having a hard time figuring out what I should learn when a manuscript shows differences in re-admission rates across regions and time when the time period includes 2020-2022, and especially 2020-2021.