Length of hospitalization in clinical trials

Hello all,

Is there any reference on how to analyze and interpret the length of hospitalization in superiority clinical trials?

I always saw this outcome as a useless surrogate endpoint: if a treatment is associated with substantial more deaths, wouldn’t the average hospitalization length in that group be lower?

I’ve seen several Covid-19 studies reporting the length of symptoms and length of hospitalization as sole outcomes, and I’m not sure how much we can conclude from this.

Thank you in advance,

1 Like

You’re right. Many analyses of length of stay reward for early death. It’s better to think of time until successful hospital discharge, with censoring on death. But you might think more generally by modeling daily patient status on an ordinal scale as described here.

1 Like

or how about days-alive-and-out-of-hospital composite? used in cvd: https://www.ahajournals.org/doi/10.1161/CIRCOUTCOMES.118.004755

That endpoint has many problems. First some researchers are using parametric models for the bizarre distribution this outcome variable has. That can be easily fixed by using a proportional odds or proportional hazards semiparametric regression model. t-test, no.

But the harder to fix problems are (1) showing that death receives the proper weight, (2) handling loss to follow-up properly, and (3) figuring out the correct maximum follow-up duration.

1 Like

Thank you very much for the reference @f2harrell.

There is an additional worrisome point, I guess. Effective treatments often increase the risk of adverse events. Some patients with serious adverse events probably will have a much more extended hospitalization period, and this can skew the data substantially.

Even if you analyze this data correctly, I see a very problematic interpretation for the clinical practice, especially if this is the primary outcome of the study.

1 Like