The authors reply:

We thank Dr. Oud (1) for his thoughtful and important points about our recently published study (2) in Critical Care Medicine. Oud (1) is correct that using the National Inpatient Sample precludes linkage of inpatient data to other post-hospital data sources, and that, as noted in our Discussion, “earlier hospital transfer to long-term acute care hospitals and/or increased discharge to hospice over time (3, 4) may shift deaths out of hospital without meaningfully changing overall mortality.”

However, the premise of performing stratified analysis with 48-hour mortality, 3–14-day mortality, and greater than 14-day mortality subgroups was recent evidence suggesting that “early” deaths are more directly attributable to sepsis itself (and thereby, modifiable by sepsis-specific initiatives), while late deaths are associated with pre-ICU comorbidities. Therefore, in contrast to prior studies investigating unstratified (e.g., 30-d or in-hospital) mortality, we expected that “early” (i.e., 48-hr) mortality would be a more specific way to capture the effect of sepsis initiatives in the past 20 years. Our finding that septic shock 48-hour mortality declined markedly over 2 decades is unlikely to be explained by discharge to hospice or long-term acute care hospitals, which generally occur later in a patient’s course. Furthermore, the opposite direction of 48-hour mortality trends between mechanically ventilated patients with and without septic shock is also unlikely to be explained by increased discharge to long-term acute care hospitals or hospice.

We agree that evaluation of trends in 3–14-day mortality and greater than 14-day mortality, like prior studies of 30-day or in-hospital mortality, remains vulnerable to changing trends in patient disposition. We view our findings as hypothesis-generating, and further studies linking post-hospital data (including date of death) may help explain whether our observed decreases in 3–14-day in-hospital mortality or greater than 14-day in-hospital mortality are meaningful decreases in mortality (perhaps due general ICU quality improvements, rather than sepsis-specific initiatives) or artifacts of changing trends in disposition. In the meantime, our finding of decreased early, 48-hour septic shock mortality (contrasted with increasing early mortality in mechanical ventilation) may provide context to understanding the impact of sepsis management strategies over time.