Time to Next APE
In the PS-matched sample, all 48 of the encounters with follow-up data
required new antimicrobial treatment and the median event time was 82
days. Figure 3 displays Kaplan-Meier plots for each group and we can see
that steroid administration had a time-varying effect on next APE. To
overcome this violation of the proportional hazard assumption, we
created a Cox model censoring patients still at risk after 82 days, and
another model including only patients still at risk beyond 82 days. The
effect of steroid administration reduced the risk of APE for the first
model (fully-adjusted HR=0.534, 95% CI: 0.208, 1.372, p-value=0.193;
Table 3), while steroid use appeared to increase risk in the second
model (HR=1.716, 95% CI: 0.431, 6.838, p-value=0.444). However, the
results were not significant.