Inpatient care of community-acquired pneumonia: The effect of antimicrobial guidelines on clinical outcomes and drug costs in Canadian teaching hospitals
TK Marras | L Jamieson | CK Chan
BACKGROUND: Evidence supporting antibiotic treatment guidelines and respiratory quinolones (RQs) in community-acquired pneumonia (CAP) is limited.
OBJECTIVE: To study associations among guideline adherence, specific antibiotics, clinical outcomes and antibiotic costs.
METHODS: A retrospective cohort study in three tertiary care university teaching hospitals in Toronto, Ontario, studying CAP inpatients between November 1997 and June 2000. The period encompassed 12 months when an early version of empirical antibiotic guidelines was used (early cohort) and 18 months when recent guidelines (including RQs) were used (recent cohort).
RESULTS: Six hundred ninety-eight cases of CAP were reviewed, and 91% were guideline adherent. In multivariable analyses, no association was observed between guideline adherence and mortality or duration of hospitalization. Guideline-adherent cases received fewer antibiotics in both cohorts and 0.9 days less of intravenous antibiotics (P=0.04) in the recent cohort. There was no significant difference in antibiotic cost according to guideline adherence, but recent cohort guideline-adherent cases had lower drug costs than early cohort guideline-adherent cases. Antibiotic selection was associated with illness severity and was mirrored by clinical outcomes, despite controlling for the pneumonia severity index (PSI). Treatment with anaerobic agents (odds ratio 2.7, P=0.001) or cephalosporin plus macrolide (odds ratio 2.7, P=0.02) was associated with higher mortality. Treatment with RQ monotherapy was associated with a 2.3 day shorter duration of intravenous therapy (P<0.0001) and a $19.19 lower total antibiotic cost (P<0.0001).
CONCLUSION: Findings support empirical treatment guidelines for CAP and their recommendations regarding RQs. The association between mortality and anaerobic coverage or combination therapy may reflect prognostic information available at presentation but not captured by the PSI.