Identifying risk factors for healthcare-associated infections from electronic medical record home address data

Abstract

Abstract Background Residential address is a common element in patient electronic medical records. Guidelines from the U.S. Centers for Disease Control and Prevention specify that residence in a nursing home, skilled nursing facility, or hospice within a year prior to a positive culture date is among the criteria for differentiating healthcare-acquired from community-acquired methicillin-resistant Staphylococcus aureus (MRSA) infections. Residential addresses may be useful for identifying patients residing in healthcare-associated settings, but methods for categorizing residence type based on electronic medical records have not been widely documented. The aim of this study was to develop a process to assist in differentiating healthcare-associated from community-associated MRSA infections by analyzing patient addresses to determine if residence reported at the time of positive culture was associated with a healthcare facility or other institutional location. Results We identified 1,232 of the patients (8.24% of the sample) with positive cultures as probable cases of healthcare-associated MRSA based on residential addresses contained in electronic medical records. Combining manual review with linking to institutional address databases improved geocoding rates from 11,870 records (79.37%) to 12,549 records (83.91%). Standardization of patient home address through geocoding increased the number of matches to institutional facilities from 545 (3.64%) to 1,379 (9.22%). Conclusions Linking patient home address data from electronic medical records to institutional residential databases provides useful information for epidemiologic researchers, infection control practitioners, and clinicians. This information, coupled with other clinical and laboratory data, can be used to inform differentiation of healthcare-acquired from community-acquired infections. The process presented should be extensible with little or no added data costs.