Use of Hospital Discharge Data to Assign Epidemiologic Infection Categories to MRSA and Clostridium difficile Infections

Background: Public reporting using administrative data of methicillin-resistant Staphylococcus aureus (MRSA) and Clostridium difficile infections is mandatory for Illinois hospitals.  MRSA infections must also be reported as present on admission (POA) or occurring during the hospitalization.  No published studies have evaluated the validity of the POA variable in hospital discharge data with respect to healthcare-associated infections such as C. difficile and MRSA.

Objective: (1) Summarize infection rates and POA classification and assign epidemiologic infection categories based on administrative data and (2) compare proportions in each category to expected proportions, based on other data sources.

Methods: We searched the first 25 diagnostic codes for all 2008 hospital discharges in Illinois (IL) for codes indicating infection due to MRSA (ICD-9 code V09.0) and C. difficile (ICD-9 code 008.45), and categorized these infections using the POA code.  We assessed evidence of previous healthcare exposure using the source of admission and previous hospitalizations recorded for each discharge and other diagnostic codes, including those for POA surgical site infection and end-stage renal disease.  Infections coded as not POA were classified as hospital-onset.  Infections coded as POA with evidence of healthcare exposure in the previous year for MRSA or 4 weeks for C. difficile were classified as healthcare-associated infections.  Infections coded as POA with no evidence of previous healthcare exposure were classified as community-associated.  We compared the proportions in each category to other recent studies that used clinical or laboratory data.

Results: Out of a total of 1,699,853 discharges, we identified 20,714 (1.2 %) coded for MRSA and 17,369 (1.0 %) coded for C. difficile.  93% of MRSA infections and 75% of C. difficile infections were coded as POA.  Among MRSA infections, we classified 5% as hospital-onset, 25% as healthcare-associated, and 68% as community-associated.  Among C. difficile infections, we classified 24% as hospital-onset, 20% as healthcare-associated, and 56% as community-associated.  These distributions differed significantly from those obtained in studies using other data sources.  A CDC study of MRSA conducted at a single site in IL using clinical data found 19% of infections to be hospital-onset and 57% to be healthcare-associated (Chi-square = 333.19, p<0.0001) (Schaefer 2007).  A study of C. difficile conducted using laboratory and administrative data found 51% of infections to be hospital-onset and 13% to be healthcare-associated (Lambert 2009).

Conclusions: These discrepancies underscore the need for a case-based comparison of hospital discharge data and clinical data in order to assess the accuracy of POA coding.  Until the validity of POA coding is clarified, POA classifications for MRSA and C. difficile should be interpreted with caution.