Surveillance of Hospital Onset Clostridium difficile Infection; International Classification of Disease Versus Microbiological Database

Abstract

Purpose: Currently, surveillance of Clostridium difficile Infection (CDI) is done using billing database where discharge diagnoses are maintained using International Classification of Disease, 9th revision - Clinical Modification (ICD9-CM). The pitfall of using ICD9-CM is that it does not differentiate between community onset and hospital onset CDI. Estimation of hospital onset CDI is crucial in determining effect of allocated resources and compliance to infection prevention and isolation policies. Primary discharge diagnosis is the reason for admission and present at the time of diagnosis. Hence forth, ICD9-CM of 008.45 as primary diagnosis suggest against hospital onset CDI. We hypothesize that secondary diagnosis code (any sequence other than primary) is a better estimate of hospital onset CDI. In this study, we aimed to compare accuracy of microbiological database of Clostridium difficile toxin and ICD9-CM. Methods: A retrospective cohort of all the patients discharged from the medical service over a six year period from October 2005 to September 2011 was studied from a 360-bed community hospital located in a suburb of Baltimore, MD. Diagnosis of hospital onset CDI was established when new onset unformed stool after 48 hours of hospital admission was tested positive for Clostridium difficile toxin on enzymeimmune assay or colonoscopy evidence of pseudomembranus colitis. Accuracy of ICD9-CM coding of 008.45 as secondary diagnosis and microbiological database in detecting hospital onset CDI was calculated by constructing 2x2 contingency Table. Results: A total of 52,065 patients were admitted during the study period and 1,102 of them were coded as 008.45, 413 (37.5%) primary diagnosis and 689 had secondary diagnosis. Over this period; total of 241 cases of hospital onset CDI were identified, 22 of them were not coded and two were coded as primary diagnosis but no evidence of diarrhea at admission and no mention of suspicion of CDI in admission note and assumed to be coding error. The ICD9-CM coding was 90.0% (95% CI; 85.4-93.4) sensitive and 99.1% (95% CI; 99.0-99.2) specific where as microbiology database was 100% (95% CI; 98.0-100) sensitive and 98.9% (95% CI; 98.9-99.1) specific. When choosing subset of toxin positive after 48 hours of hospital admission, sensitivity remained 100% and specificity improved to close to 100% (95% CI; 99.9-100) with only 45 (15.7%) false positive cases. Conclusion: Surveillance using microbiological database is more accurate in estimating true incidence of hospital onset CDI when compared to ICD9-CM coding. Collecting time of toxin positive can improve the specificity and provided most accurate estimation for CDI surveillance until a system is developed to report hospital onset CDI separately.