LAUSR

Critical Care Medicine www.ccmjournal.org e1033 Although we did find that some conditions unexpectedly suggested a “protective” odds ratio (OR) (e.g., congestive heart failure and previous history of myocardial infarction/percutaneous coronary intervention/coronary artery bypass grafting), the former did not meet the prespecified α value for significance, and the latter is statistically significant but not clinically meaningful (0.972). Hypertension is not an identified risk factor for mortality in sepsis and has been found to be possibly protective of mortality in sepsis in prior studies (4). Although it is difficult to surmise a plausible explanation for the protective OR of an International Classification of Diseases (ICD) diagnosis of diabetes or chronic lung disorder, our findings are consistent with a large retrospective cohort study of over 100,000 patients (5). When we examined the predictors for in-hospital mortality in encounters for sepsis and AF, we adjusted for the following hospital-level characteristics: bed size, geographic region, and teaching status. The calculated OR for the factors associated with mortality in sepsis with AF was all adjusted for these hospital-level characteristics. Adjusting for these factors is standard in analyses of National Inpatient Sample (NIS) data. Although the prognostic association of hospital-level covariates with hospital mortality is interesting, it was not a goal of the present study. Finally, we reported the odds of hospital death among septic hospitalizations decreased over time on adjusted analysis, suggesting that this improvement is likely secondary to improved healthcare measures. We agree that using this as a surrogate for short-term mortality may be less accurate than a combination of mortality and discharge to hospice, but this information was not available for the years examined. More specifically, the variable disposition of patient, UB92 coding (DISPUB92) is only available in NIS data for the years 1998–2006. Disposition of patient, UB04 standard coding (DISPUB04) also contains these data but is only available on data from 2007 to 2011. The present study examined data from 2010 to 2014, for which only the variable disposition of patient, uniform coding (“DISPUNIFORM”) could be used, which does not contain data regarding discharge to hospice (6). The cited article was published in the same year as an article by her team suggesting a rise in prevalence and decrease in mortality of sepsis (as assessed by a clinical surveillance definition), albeit to a lesser extent than would have been estimated on the basis of ICD, 9th Edition, codes alone (7). Dr. Kumar received funding from Teleflex and Osprey Medical. The remaining authors have disclosed that they do not have any potential conflicts of interest.