Attributing Cause of Death for Patients with Clostridium difficile Infection

et al. Isolation of Rickettsia parkeri and identifi cation of a novel spotted fever group Rickettsia sp. from Gulf Coast ticks (Amblyomma maculatum) in the rates of Rickettsia parkeri infection in Gulf Coast ticks (Amblyomma maculatum) and identifi cation of " Candidatus Rickettsia andeanae " from Fairfax County, Virginia. A ten-year study of tick biting in Mississippi: implications for human disease transmission. To the Editor: Hota et al. report that for deceased patients who had Clostridium diffi cile infection (CDI), agreement is poor between causes of death reported on death certifi cates and those categorized by a review panel (1). Our data support the diffi culty of attributing cause of death for patients with CDI. In 2004 in Quebec, Canada, a mandatory CDI surveillance program was implemented. Deaths that occurred within 30 days after CDI diagnosis were classifi ed as 1) directly attributable to CDI (e.g., toxic megacolon, septic shock), 2) having a CDI contribution (e.g., acute decompensation of chronic heart failure), or 3) unrelated to CDI (e.g., terminal cancer) (2). To determine accuracy of the surveillance classifi cations, we compared cause-of-death classifi cation of 22 deceased CDI patients reported to surveillance by 1 hospital in 2007 with causes of death reported by 13 external reviewers who examined summaries of medical fi les of the deceased patients. Reviewers were 11 infectious disease and 2 public health physicians involved with CDI surveillance at their respective hospitals but not this hospital. The median (minimal, maximal) κ statistics for comparison of external reviews with surveillance classifi cation were 0.495 (0.252, 0.607) for directly attributable, 0.182 (−0.091, 0.182) for contributed, and 0.321 (0.124, 0.614) for unrelated. Comparison within external reviewers yielded 0.697 Complete agreement was found for only 6 cases (4 directly attributable and 2 unrelated) (Figure). Variation among reviewers suggested that categorizations reported to surveillance were inaccurate. Number of deaths among patients with CDI, regardless of the cause of death, seemed to better indicate CDI severity. Since 2008, only the crude numbers of deaths, not subjected to individual interpretation, have been reported to surveillance. A questionnaire addressing concurrent medical conditions, prognosis, level of care, and circumstances of death is being implemented in Quebec hospitals participating in CDI surveillance and should help determine the role of CDI in deaths. Figure. Classifi cation of cause of death among 22 patients with Clostridium diffi cile infection (CDI), by 13 external reviewers, Quebec, Canada, 2007. Bars indicate …