BACKGROUND: Clostridium difficile infections (CDIs) are an important cause of morbidity and mortality in New York State (NYS) with approximately 10,000 community onset (CO) CDIs reported to CDC’s National Healthcare Safety Network (NHSN) by hospitals each year. To efficiently reduce CO CDIs, NYS sought to identify pairs of skilled nursing facilities (SNFs) and hospitals that frequently share CDI patients for participation in a CDI prevention project and then to validate the methods.
METHODS: The NYS all payer hospital discharge database (APHDD) contains patient demographics and discharge diagnoses for all inpatient stays. The Health Facility Information System (HFIS) lists the names and addresses of regulated SNFs. CO CDI cases from NHSN were matched to the 2013 APHDD to obtain an address, and APHDD addresses were matched with SNF addresses from HFIS to identify SNF-associated CDIs. To validate whether the APHDD address is truly the admission address, we used 2015 data and compared the SNF status (yes/no) of APHDD address to a “gold standard” categorical variable derived from SNF status indicators from NHSN and APHDD: NHSN ‘last overnight location immediately prior to arriving into the hospital’ (new optional field in 2015) and APHDD variable ‘point of origin’. The derived variable values were: true SNF if both yes, possible SNF if discrepant, or not SNF if both no.
RESULTS: Based on the APHDD address, in 2013 16% of CO CDI cases were directly admitted from SNFs. Forty-seven SNFs (8%) that transferred more than five CDI patients were targeted to join the prevention project, and nine of those (19%) joined. There were 5,473 NHSN CO CDI cases successfully matched with the APHDD in 2015; 3,701 were missing last overnight location in NHSN, leaving 1,472 matched cases with complete information from 115 (65%) hospitals. APHDD addresses identified SNFs in 69% of 137 true SNFs, 49% of 260 possible SNFs, and 3% of 1,075 non-SNFs.
CONCLUSIONS: SNF-associated CDI rates may be underestimated by roughly 30% based on APHDD addresses. However, these addresses are useful for identifying SNFs transferring patients with a specific type of infection. The validation was hampered by missing location data in NHSN and will be repeated using a better gold standard (i.e. direct information from the SNFs that participated in the prevention project). Data linkage using readily available databases may improve efficiency by focusing infection prevention efforts on facilities with the largest numbers of patients who would directly benefit from the intervention.