Methods

We studied a cohort of patients $ \ge \!$ 65 years of age with advanced cancer who had developed NV-HAP after receipt of palliative chemotherapy at Yale New Haven Hospital, a 1,541-bed tertiary care center in New Haven, Connecticut, from January 1, 2016, through September 30, 2017. Advanced cancer was defined as stage III–IV solid tumors and acute, refractory, and relapsed or active liquid tumors requiring chemotherapy or targeted therapies including lymphoma, leukemia, and multiple myeloma. Advanced cancers were identified by the International Classification of Diseases (ICD), Tenth Revision codes and confirmed via pathology or medical record review. We defined cases of NV-HAP using Centers for Disease Control and Prevention National Healthcare Safety Network criteria for clinically defined pneumonia, restricted to the index event following receipt of palliative chemotherapy. ⁵ We excluded patients presenting with complicated pneumonias associated with abscess, bacteremia, or fungal, necrotizing, or organizing pneumonia due to the expected extended duration of antibiotic therapy. We also excluded patients readmitted or transferred from another institution for any reason due to potential confounding with all-cause readmissions and mortality within 90 days of discharge. The Yale Human Investigation Committee approved this study.

For each patient, we collected demographics, comorbidities, history of multidrug-resistant organisms, and hospitalization characteristics including intensive care unit admission for any period of time. We also evaluated inpatient microbiological tests. All antibiotics administered orally, intravenously, or intramuscularly for NV-HAP were recorded from pharmacy data and confirmed by medical record review. For antibiotics, we determined the total days of therapy, including both inpatient and postdischarge days of therapy. We also evaluated the antibiotic spectrum index, a metric designed to compare a spectrum of activity ranging from the most narrow at 1 (eg, dicloxacillin) to the broadest spectrum at 13 (eg, tigecycline). We recorded the total antibiotic spectrum index per patient per day. ^{Reference Gerber, Hersh, Kronman, Newland, Ross and Metjian6}

We evaluated the distribution of days of therapy and determined all-cause readmissions, all-cause mortality, detection of Clostridioides difficile, and detection of a multidrug-resistant organism within 90 days of discharge for all patients. A multidrug-resistant organism was defined as methicillin-resistant Staphylococcus aureus, vancomycin-resistant enterococci, extended-spectrum β-lactamase-producing Enterobacteriaceae, or carbapenem-resistant Enterobacteriaceae. To evaluate the association between antibiotic exposure and readmission or death, we used a multivariable logistic regression model adjusted for duration of therapy (≤7 d vs 8–14 d and ≥15 d) and intensive care unit admission. We calculated adjusted odds ratios and 95% CI. We used R version 3.6.0 software (Vienna, Austria) for analyses.