348. Predicting Carbapenem-Resistant Enterobacteriaceae (CRE) Carriage at the Time of Admission Using a State-Wide Hospital Discharge Database
Session: Poster Abstract Session: HAI: Multi Drug Resistant Gram Negatives
Thursday, October 27, 2016
Room: Poster Hall

Background: Hospitals need efficient ways to identify patients at high risk of CRE carriage at the time of admission. Prior healthcare exposure and antibiotic use are commonly cited risk factors for carriage of antibiotic-resistant organisms, but such data are difficult to obtain in practice. We assessed whether information from a state hospital discharge database could predict CRE carriage on admission.

Methods: We performed a case-cohort study using healthcare exposure data from the Illinois Hospital Discharge Database. We defined cases as patient encounters with a positive CRE culture collected from the first 3 hospital days (as reported to the Illinois XDRO registry; only the first qualifying encounter per patient was retained). We defined matched controls as all CRE-negative patient admissions from the same hospital and month time period. We randomly split the sample into training (60%) and validation (40%) sets. Fitting a multivariable logistic regression on the derivation set, we assessed the following predictors in the prior year: age, number of short term and long-term acute care hospital (STACH and LTACH) visits, average STACH and LTACH length of stay, receipt of endoscopic retrograde cholangiopancreatography (ERCP; based on procedure codes), prior admission with infection diagnosis (based on ICD-9 code associated with bacterial infection, as a marker of antibacterial receipt). We also included whether the patient was currently admitted to an LTACH versus STACH. We used the validation set to generate an ROC curve.

Results: We identified 486 case (CRE+) and 340,055 control (CRE-) patient hospital encounters. Using the training set, we identified 7 independent predictors of being CRE-positive on admission (Table). We tested both a full prediction model that used healthcare exposure + billing codes (B) and a model with only healthcare exposure (A) on the validation set (Model A shown in Figure). Model A performed as well as Model B, and had a c-statistic of 0.86.

Conclusion: Detailed healthcare exposure history using regional hospital discharge databases can identify patients who have CRE carriage at the time of admission. Automating such models could provide alerts to healthcare facilities, improving identification of high-risk patients.

Michael Y. Lin, MD, MPH1, Serena Rezny, MS2, Michael J. Ray, MPH2, Dejan Jovanov, BSc2, Robert A. Weinstein, MD1,3, William E. Trick, MD1,3 and For the CDC Prevention Epicenters Program, (1)Rush University Medical Center, Chicago, IL, (2)Illinois Department of Public Health, Chicago, IL, (3)Cook County Health and Hospitals System, Chicago, IL

Disclosures:

M. Y. Lin, None

S. Rezny, None

M. J. Ray, None

D. Jovanov, None

R. A. Weinstein, None

W. E. Trick, None

Findings in the abstracts are embargoed until 12:01 a.m. CDT, Wednesday Oct. 26th with the exception of research findings presented at the IDWeek press conferences.