Candidemia is Costly--Plain and Simple

As physicians, we are notoriously ignorant when it comes to money matters, both in personal and professional matters. As advocates for our patients, our decisions regarding interventions or therapies are more often based on issues of availability, safety, and efficacy than on cost. It is only in our roles as members of the hospital epidemiology team or pharmacy and therapeutics committee that issues of costs play a pivotal role in decision making, especially regarding the use of therapeutics or technical devices that may prevent health carerelated infections, such as bloodstream infections. The clear and obvious costs of products or programs aimed at preventing health carerelated bloodstream infections often overshadow estimates of attributable mortality or attributable costs of these infections. In the end, hospital committees often choose not to invest in these prevention efforts or products (e.g., antimicrobial-impregnated catheters, improving catheter care, and improved diagnostic capabilities). What is disturbing is that much of the information that we use in making these assessments is flawed. Both the mortality and the costs attributable to a health carerelated infection are very closely tied to the severity of underlying illness in these patients; the disease-specific attributable outcomes of these infections can rarely be teased out. It follows that decisions regarding investments in health carerelated bloodstream infection prevention measures are often made with flawed data. Such is the case with Candida bloodstream infections. In this issue of Clinical Infectious Diseases, Zaoutis et al. [1] attempt to fill a knowledge gap regarding the attributable mortality and cost of invasive candidiasis through the analysis of 2 large national data sets. During the past decade, candidemia has become a significant problem throughout the health care system, where roughly one-third of candidemias occur in outpatients and another third occur in inpatients outside of an intensive care unit setting [2]. The incidence of candidemia was estimated in Baltimore, Maryland, and Connecticut, with use of the traditional gold standard approach of active laboratory-based surveillance, to be 8 infections per 10,000 discharges from the hospital (10 infections per 100,000 persons per year) [2], and in comparable European studies, the incidence of candidemia was estimated to be 25 infections per 10,000 discharges from the hospital [3–5]. Zaoutis and colleagues utilized a less sensitive methodology, relying on an existing International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), code database, to estimate the incidence in a more generalizable sample of US hospital admissions. They report an incidence of 34 infections per 10,000 discharges from the hospital, which is approximately one-half of the population-based incidence reported from Baltimore and Connecticut. As pointed out by Zaoutis et al. [1], this lower rate is likely to be an underestimation because of miscoding of candidemia using codes not captured in the analysis (e.g., candidemia of unspecified site) or not coding the candidemia at all (i.e., low sensitivity). Regardless, both approaches (i.e., ICD-9-CM code data and active laboratory-based surveillance) likely underestimate the true incidence of candidemia because of deficient diagnostic testing. Blood cultures for candidemia may only be 50%60% sensitive; the true incidence may be double the published rates. This underestimation should be considered by decision makers when considering the implementation of candidemia prevention and treatment strategies. Prevention efforts may have an enormous impact if applied broadly, because conventional diagnostic techniques fail to identify 30%50% of cases of true disease. Candidemia is frequently associated with a sepsis syndrome and often involves disease outside of the bloodstream [6, 7]. This component of disseminated candidiasis may help to explain why candidemia appears to be such a devastating disease. The attributable mortality rate associated with candidemia in a well-matched case-control study at the University of Iowa was recently estimated to be 49%, which is similar to the rate observed more than a decade earlier, despite the availability of several newer agents to treat this disease [8]. Data from a population-based, matched case-control study in Baltimore and Connecticut estimated the attributable mortality of candidemia to be 19%24%, depending on the patient's age [9]. When extrapolated to the United States, the annual number of excess deaths due to candidemia was estimated to be 42565376 [9]. These estimates are applicable to patients overall, because they include both inpatients and outpatients. In the study by Zaoutis et al. [1] reported here, the attributable mortality is much lower and is estimated to be 10% among nonnewborn hospitalized pediatric patients and 14.5% among hospitalized adult patients. These data are the most representative of the US patient population published to date. Moreover, the authors used a matching procedure based on a logistic regression model of the entire Nationwide Inpatient Survey 2000 (a 20% sample of hospital admissions from 28 states) and the 2000 Kid's Inpatient Database (an 80% sample of pediatric admissions from 27 states). This propensity-matched analysis identified a subset of patients who would be as likely to have developed candidemia as those who actually did, based on the initially developed predictive model. This approach differs from previous studies that relied on a matching algorithm based on a smaller number of parameters. However, the authors point out that, because of misclassification bias inherent in the process of coding patients' medical records (i.e., some patients with invasive candidiasis were classified as noninfected patients), the calculated attributable mortality is likely to be smaller than the true attributable mortality. Given these limitations, it is safe to assign the attributable mortality of candidemia determined by Zaoutis and colleagues as a very conservative estimate of the outcomes of candidemia among adults and nonnewborn children. This brings us back to dollars. Zaoutis et al. [1] present an eloquent cost analysis that is easy to follow and intuitively makes sense. Total hospital charges attributable to invasive candidiasis for nonnewborn pediatric patients are estimated to be $65,058$119,474 per episode; those for adults are estimated to be $33,604$45,602 per episode. These excess hospital charges are higher than those reported from Baltimore and Connecticut ($6214$29,094 per episode) using different methodology that adjusted for several confounders, including age group, death, hospital, and, most importantly, treatment status. Although the reported charges were lower in the study by Zaoutis et al. [1], all attributable charges in the Baltimore and Connecticut study [9] were driven by costs of antifungal treatment. Unfortunately, the Zaoutis study [1] does not determine which patients were treated for candidemia. However, other investigators have described an alarming proportion of patients with candidemia who received inadequate antifungal therapy: 10% received no therapy [10] and 30%39% received <7 days of therapy [9]. It is unclear why some patients with candidemia are not treated. However, as more treatment options become available, the approach to treatment becomes more standardized, the agents become safer, and concerns of fluconazole resistance (whether rational or irrational) increase, treatment-related costs to the hospital will continue to increase even as attributable mortality decreases. Considering this moving target, treating patients with candidemia will continue to be costly to hospitals. Regarding estimating costs of candidemia, Oscar Wilde was right: the truth is rarely pure and never simple. However, substantial data do exist that describe effective measures for the prevention of Candida bloodstream infections, which are currently focused on catheter care. Prevention of candidemia through evidence-based recommendations (i.e., improved hand hygiene, optimal catheter care, and prudent antimicrobial use) [11] can serve as a cornerstone to reducing candidemia-related hospital costs by reducing the need to treat central linerelated candidemia. This study by Zaoutis et al. [1] points us in the right direction. Potential conflicts of interest. S.K.F.: no conflicts.