Table 2 Approaches and their properties
From: Estimands to quantify prolonged hospital stay associated with nosocomial infections
Approach | Properties / pros and cons |
|---|---|
A 1 | - is a real quantity |
- undesired properties related to hazard ratio \(\frac {\lambda _{12}}{\lambda _{02}}\) | |
- yields positive values even if NI patients are discharged faster, i.e., \(\frac {\lambda _{12}}{\lambda _{02}}>1\) | |
- does not distinguish between pre- and post-infection time | |
- does not allow causal interpretation about attributable length of stay associated with NI | |
- not appropriate to quantify the burden of NI | |
A 2 | - is a hypothetical quantity |
- desired properties related to hazard ratio \(\frac {\lambda _{12}}{\lambda _{02}}\) | |
- considers only the post-infection time for NI patients | |
- contributes pre-infection time to patients without NI | |
- allows a causal interpretation about attributable length of stay associated with NI | |
- appropriate to quantify the burden of NI at patient-level | |
A 3 | - is a hypothetical quantity |
- desired properties related to hazard ratio \(\frac {\lambda _{12}}{\lambda _{02}}\) | |
- considers only the post-infection time for NI patients | |
- contributes pre-infection time to patients without NI | |
- allows a causal interpretation about attributable length of stay associated with NI | |
- appropriate to quantify the burden of NI at population-level | |
A 4 | - is a real quantity |
- desired properties related to hazard ratio \(\frac {\lambda _{12}}{\lambda _{02}}\) | |
- distinguishes between pre- and post-infection time | |
- appropriate to quantify the burden of NI at patient-level |