The association of prior hospitalization with clinical outcomes among patients admitted with pneumonia: a propensity score matching study

Abstract

Background Although prior hospitalization (PH) has been considered as a risk factor for infection with potentially drug-resistant (PDR) pathogens in patients admitted with pneumonia, the evidence is limited. We aimed to elucidate the clinical impact of PH on these patients. Methods PH was defined as hospitalization for two or more days in the preceding 90 days. Patients with PH-associated pneumonia (PHAP) or community-acquired pneumonia (CAP) were matched using the propensity score matching method, and the clinical outcomes were compared. We also conducted subgroup analyses based on intravenous antibiotic use during PH, duration of PH, and time to re-admission. Results A total of 704 patients were identified; the PHAP group included 97 patients (13.7%). After matching according to propensity scores, the baseline characteristics of the PHAP group were similar to those of the CAP group. The isolation rate of PDR pathogens as well as the 30-day and total in-hospital mortality did not differ between propensity score-matched PHAP and CAP patients (13.6% vs. 10.2%, P = 0.485; 10.2% vs. 14.8%, P = 0.362; and 13.6% vs. 15.9%, P = 0.671, respectively). In subgroup analyses, only intravenous antibiotic use during PH was associated with the isolation rate of PDR pathogens (adjusted OR: 5.066; 95% CI: 1.231–20.845). Conclusions PH itself might not be related with higher isolation rates of PDR pathogens or mortality in patients admitted with pneumonia. Therefore, it seems reasonable that broad spectrum antibiotic therapy for PDR pathogens should be selectively applied to PHAP patients with intravenous antibiotic use during PH.

Hospitalizations for Community-Acquired and Non-Ventilator-Associated Hospital-Acquired Pneumonia in Spain: Influence of the Presence of Bronchiectasis. A Retrospective Database Study

“…Therefore, propensity scores can be thought of as an advanced matching technique that makes possible the comparison of populations with very different sizes and distributions according to possible confounding variables [26]. This approach, considering exposed and non-exposed patients or those without a condition and not a treatment or an intervention, has been previously used by other authors [27][28][29][30][31].…”