BACKGROUND: Carbapenem-resistant Enterobacteriaceae (CRE) are a serious public health concern in New York State (NYS). The NYS Department of Health (DOH) mandated that all regulated hospitals report CRE infections using CDC’s National Healthcare Safety Network (NHSN) laboratory-identified (LabID) infection surveillance protocol beginning in July 2013. A risk adjustment model is needed for publicly reporting hospital onset (HO) CRE rates so that hospital performance can be fairly compared.
METHODS: NHSN CRE Klebsiella and E. coli events were probabilistically matched onto the 2014 all payer hospital discharge database (APHDD) to obtain additional information on patient demographics and discharge diagnoses. Multivariable Poisson regression was used to identify patient risk factors for HO CRE, adjusting for length of stay. The first three days of each stay were excluded from the analysis because by definition patients are not at risk for HO infections on these days. Hospital-level variables such as admission prevalence CRE rates were not included in this model to avoid adjusting away the hospital-level effects of interest.
RESULTS: A total of 1,353 HO CRE events were reported by 178 hospitals among 8,458,892 days at risk. The majority of specimens were obtained from the urinary tract (48%), followed by 23% respiratory, 12% skin/soft tissue, and 11% blood. Significant predicators of HO CRE included older age, hospitalization in previous 90 days, direct transfer from nursing home, and present on admission conditions including infections, respiratory failure, and kidney failure. Significant protective factors included admission for mental disorder, substance abuse, pregnancy, birth, and bone/joint issues. These factors were combined into a risk index for each hospital, which was stable between calendar years. Hospital-level CRE rates adjusted using the risk index were consistent with the individual-level analysis.
CONCLUSIONS: Information available in the APHDD was useful for creating a risk index that characterizes differences in hospital populations that impact risk for HO CRE. The previous year’s risk index can be used for preliminary analyses before the APHDD is available. NYS will provide assistance to hospitals with significantly high HO CRE rates and require them to submit quality improvement plans.