Methods: A panel of toxigenic C. difficile assays was performed on 1000 unformed stool samples by our research laboratory. CDI was defined by the IDSA definition. Clinicians were only aware of the EIA results. 127/1000 (12.7%) patients had true-positive CDI and were divided into Group 1 (EIA +/PCR +) = 56 and Group 2 (EIA -/PCR +) = 71. Demographic, clinical, and outcome data were collected through retrospective chart review. Uni- and multi- variate analyses controlling for patient demographic and health status were conducted to assess the differences between the two groups on outcomes. A subgroup analysis was performed to examine the mediating effects of treatment on outcomes stratified by the two groups (i.e. whether the cases were detected by PCR only).
Results: Patients in Group 2 were younger (60 vs. 71 years, p=0.003), more commonly had CKD (31.3% vs. 14.6%, p=0.03), less commonly had a history of CDI (29.9% vs. 47.2%, p=0.05), received antibiotics less often in the last 90 days (63.6% vs. 80%, p=0.52), and were less often treated for CDI (35% vs. 79%, p < 0.001) when compared to Group 1. Patients in Group 2 were less likely to have recurrent CDI (12% vs. 33%, p=0.005) or a readmission related to CDI within 90 days (10% vs. 23%, p=0.04). Multivariable analysis demonstrated that patients in Group 2 were less likely to have severe CDI, less likely to receive treatment, and had shorter LOS. The subgroup analysis revealed a significant effect of receiving treatment on reducing 90 day readmission for any reason, inpatient LOS, total LOS (inpatient + readmission LOS), and total charges in Group 2 rather than in Group 1.
Conclusion: Patients with CDI detected by both EIA and PCR are sicker and have worse outcomes than those only PCR positive, but outcomes in Group 2 are significantly improved by CDI treatment suggesting that those patients had CDI and not just colonization. Hospitals may make the business case for using the more expensive PCR assay to identify all patients with CDI based on a strong beneficial treatment effect on multiple relevant clinical outcomes.
E. Liu, Cepheid: Employee , Salary
M. Schoonmaker, Cepheid: Employee , Salary
J. Ridgway, None
A. Robicsek, None
L. Peterson, None
See more of: Poster Abstract Session