Burden of Candidemia in the United States, 2017

Abstract

Background

Candidemia is a common healthcare-associated bloodstream infection with high morbidity and mortality. There are no current estimates of candidemia burden in the United States (US).

Methods

In 2017, the Centers for Disease Control and Prevention conducted active population-based surveillance for candidemia through the Emerging Infections Program in 45 counties in 9 states encompassing approximately 17 million persons (5% of the national population). Laboratories serving the catchment area population reported all blood cultures with Candida, and a standard case definition was applied to identify cases that occurred in surveillance area residents. Burden of cases and mortality were estimated by extrapolating surveillance area cases to national numbers using 2017 national census data.

Results

We identified 1226 candidemia cases across 9 surveillance sites in 2017. Based on this, we estimated that 22 660 (95% confidence interval [CI], 20 210–25 110) cases of candidemia occurred in the US in 2017. Overall estimated incidence was 7.0 cases per 100 000 persons, with highest rates in adults aged ≥ 65 years (20.1/100 000), males (7.9/100 000), and those of black race (12.3/100 000). An estimated 3380 (95% CI, 1318–5442) deaths occurred within 7 days of a positive Candida blood culture, and 5628 (95% CI, 2465–8791) deaths occurred during the hospitalization with candidemia.

Conclusions

Our analysis highlights the substantial burden of candidemia in the US. Because candidemia is only one form of invasive candidiasis, the true burden of invasive infections due to Candida is higher. Ongoing surveillance can support future burden estimates and help assess the impact of prevention interventions.

Candida, candidemia, healthcare-associated infections, communicable diseases

Candidemia is one of the most common healthcare-associated bloodstream infections in the United States (US) [1, 2]. It is associated with significant morbidity and 25%–40% all-cause in-hospital mortality [3]. Risk factors for candidemia include critical illness, hematologic and solid organ malignancy, hematopoietic cell and solid organ transplantation, recent abdominal surgery, hemodialysis, presence of a central venous catheter, receipt of total parenteral nutrition and broad-spectrum antibiotics, preterm birth in neonates, and injection drug use [4, 5]. Although candidemia is associated with severe illness and death, the burden of this condition has not been estimated for the US. Understanding the magnitude and scope of candidemia nationally is important for designing, implementing, and measuring the impact of large-scale prevention efforts and informing policy decisions. The US does not have mandatory national surveillance for candidemia, which complicates the estimation of its burden.

The Centers for Disease Control and Prevention (CDC) has conducted active, population-based surveillance for candidemia through the Emerging Infections Program (EIP), starting with 2 sites in 2008 and increasing to 9 sites in 2017, covering approximately 17 million people (5% of the national population) [6]. We used these surveillance data to estimate the national burden of candidemia during 2017.

METHODS

The EIP is a partnership between CDC, state health departments, and academic collaborators that serves as a platform to conduct surveillance for many types of infections, including influenza and those caused by foodborne and healthcare-associated pathogens [7]. In 2017, surveillance for candidemia was conducted in 45 counties in 9 states (California, Colorado, Georgia, Maryland, Minnesota, New Mexico, New York, Oregon, and Tennessee) with a total catchment area population of 17 million persons. Laboratories and healthcare facilities that serve the catchment counties reported all bloodstream infections with Candida species in residents of catchment areas to local EIP surveillance staff. Periodic audits were conducted at each laboratory to ensure completeness of reporting.

A case of candidemia was defined as a blood culture positive for Candida species in a surveillance area resident. Subsequent cultures from the same individual within 30 days were considered part of the same case, even if the individual was discharged and readmitted to a healthcare facility. If a blood culture yielded Candida in the same patient beyond 30 days from the initial positive culture, it was considered a new case in the same person. Surveillance staff applied the case definition criteria to each laboratory report and determined whether it met the case definition. If so, they reviewed the patient’s medical records and abstracted demographic and clinical data using a standardized case report form. Data were entered into an electronic database and securely transmitted to CDC. Available Candida isolates from incident blood cultures were sent to CDC for species confirmation and antifungal susceptibility testing using Clinical and Laboratory Standards Institute guidelines [8].

We calculated sampling weights by dividing the Census Bureau 2017 US population estimate by estimated EIP site populations, stratified by age, sex, and race. We then weighted 2017 case numbers, stratified by demographic characteristics, to estimate the national burden of candidemia and generate 95% confidence intervals (CIs). We estimated incidence rates per 100 000 persons, by age group, sex, race, and census division. Using similar methods as for incidence estimates, we estimated national candidemia-associated mortality by extrapolating surveillance data on all-cause mortality within 7 days of candidemia and all-cause mortality during candidemia-associated hospitalization. To address missing patient race information, we performed multiple imputations based on the distribution of known values for 2017 cases, stratified by age and sex. We performed analyses in SAS 9.4 software (Cary, North Carolina) using multiple imputed data and combined results to account for imputation errors.

Institutional review board review was not required as EIP candidemia surveillance is a public health surveillance system.

RESULTS

In 2017, we identified 1226 cases of candidemia in 1140 patients in 9 EIP sites, with 7% of cases being recurrent in the same patient. Forty-two percent of cases occurred in patients aged ≥ 65 years, and 3% were among children (≤ 18 years). Fifty-five percent of cases occurred in males. Overall, 64% of cases occurred in white patients, and 32% of cases occurred in black patients (Table 1). All-cause in-hospital mortality within 7 days of candidemia was 15%, and mortality during the entire hospitalization with candidemia was 25%. In-hospital mortality was highest among those aged ≥ 65 years (31%) and lowest among children aged ≤ 18 years (9%).

Table 1.

Estimated Cases and Incidence Rates of Candidemia by Demographic Characteristic, United States, 2017

Characteristic	Cases, No. (%)	Estimated Cases, No. (95% CI)	Estimated Incidence Rate per 100 000
All	1226 (100)	22 660 (20 210–25 110)	7.0
Age, y
< 1	18 (1.5)	303 (238–368)	7.7
1–18	14 (1.1)	242 (195–290)	0.3
19–44	271 (22.1)	4629 (3816–5441)	4.1
45–64	414 (33.8)	7281 (5966–8596)	8.6
≥ 65	509 (41.5)	10 205 (8305–12 104)	20.1
Sex
Male	676 (55.1)	12 625 (10 674–14 577)	7.9
Female	550 (44.9)	10 035 (8554–11 516)	6.1
Race
Black	388 (31.6)	5366 (4055–6677)	12.3
White	785 (64.0)	16 515 (14 455–18 574)	6.6
Othera	53 (4.3)	779 (490–1069)	2.4

Abbreviation: CI, confidence interval.

^aAsian, Native Hawaiian/Pacific Islander, and American Indian/Alaska Native.

In 2017, 81 laboratories forwarded to CDC 1122 Candida species isolates. The most frequent Candida species isolated from culture was Candida albicans (38%), followed by Candida glabrata (30%), Candida parapsilosis (14%), and Candida tropicalis (7%). Six percent of all Candida isolates were resistant to fluconazole, with rates ranging from 0.5% in C. albicans to 7% in C. glabrata and 9% in C. parapsilosis. Two percent of Candida isolates were resistant to an echinocandin antifungal, of which most were C. glabrata.

We estimated that 22 660 candidemia cases (95% CI, 20 210–25 110) occurred in the US in 2017. The overall estimated incidence of candidemia was 7.0 cases per 100 000 persons. Incidence was highest among adults aged ≥ 65 years (20.1 per 100 000), accounting for nearly half of all candidemia cases (10 205 [95% CI, 8305–12 104]). Among racial groups, incidence was highest among black patients (12.3 per 100 000), accounting for nearly a quarter of all cases (5366 [95% CI, 4055–6677]).

Candidemia rates were highest in the South Atlantic (7.9 per 100 000 [95% CI, 3.2–12.7]) and East South Central (7.7 per 100 000 [95% CI, 3.1–12.4]) census divisions and lowest in the Pacific (5.9 per 100 000 [95% CI, 2.1–9.7]), although these rates were not statistically different from each other.

An estimated 3380 (95% CI, 1318–5442) deaths occurred within 7 days of positive culture for candidemia and 5628 (95% CI, 2465–8791) deaths occurred during the hospitalization with candidemia.

Discussion

Using data from population-based surveillance for candidemia, we estimated that US candidemia incidence for 2017 was 7 cases per 100 000 persons and that the national annual burden was nearly 23 000 cases. Candidemia represents only a portion of the burden of invasive candidiasis; other forms of invasive disease include intra-abdominal candidiasis, hepatosplenic candidiasis, endophthalmitis, osteomyelitis, and meningitis, which may not manifest with positive blood cultures, but also have severe sequelae and contribute to morbidity and mortality. Previous studies of autopsy-proven invasive candidiasis suggest that blood cultures are positive in 21%–71% of cases [9], which suggests that the estimated burden of invasive candidiasis would be even higher, although further evaluation of the relative burden of candidemia in overall invasive candidiasis is needed.

The national burden of candidemia mortality was substantial, with annual deaths numbering in the thousands. However, determining attributable mortality for a condition such as candidemia is extremely challenging, as patients frequently have multiple concurrent complex and morbid conditions, making it difficult to determine candidemia’s contribution. Estimates of mortality attributable to candidemia range from 5% to 71%, varying based on setting, age group, and study design [10]. Therefore, we used a conservative estimate of death within 7 days of positive blood culture for Candida, assuming that the bloodstream infection with Candida species contributed, at least in part, to death in these cases. In EIP surveillance, 15% of patients died of any cause within 1 week of positive culture for candidemia, resulting in an estimated nearly 3400 deaths in 2017. All-cause in-hospital mortality was 25% (97% of cases were hospitalized), which corresponds to an estimate of 5600 deaths nationally in 2017. Assuming deaths associated with candidemia account for only part of invasive candidiasis–associated deaths, the US burden of all-cause in-hospital mortality from invasive candidiasis is likely higher than what we have reported.

We expect that this burden estimate will allow assessment of trends in candidemia over time. Although candidemia incidence in EIP surveillance has been relatively stable for the past 5 years [6], recent events suggest that substantial changes in burden and epidemiology of candidemia may occur in the near future. First, multidrug-resistant Candida auris emerged in the US in mid-2015 and has now been identified in > 300 bloodstream infections reported by early 2019 [11]. In some countries where C. auris emerged approximately a decade ago, it is now a leading cause of candidemia, including being the most common Candida species in some hospitals [12]. In hospitals experiencing C. auris outbreaks, this species adds to the overall candidemia burden rather than displacing other species [13]. Second, injection drug use, rising in prevalence in many areas in the United States, is increasingly recognized as a risk factor for candidemia. For example, in the Tennessee EIP site, candidemia incidence rose in tandem with the proportion of patients with candidemia who injected drugs in the prior year, accounting for 17% of cases in 2017 [5]. These changes have the potential to offset gains made by preventive interventions such as improved catheter care, antibiotic stewardship, and antifungal prophylaxis policies.

Candida infections have shown increasing resistance to antifungal drugs in recent years. Drug-resistant Candida was highlighted as a serious threat and C. auris an urgent threat in CDC’s 2019 “Antibiotic Resistance Threats in the United States” [14]. Although we did not include an extrapolation of drug-resistant infections in our national estimates, resistance rates found in 2017, 6% to fluconazole and 2% to echinocandins, correspond to > 1400 fluconazole-resistant and 340 echinocandin-resistant bloodstream infections nationally in 2017. The consequences of drug-resistant infections can be severe. Candida is already associated with high mortality rates, and drug-resistant infections may result in even higher mortality, especially when infections are deep-seated or when diagnosis and adequate treatment are delayed [15]. The proportion of Candida infections that are drug resistant is not uniformly distributed. Fluconazole and echinocandin resistance have been reported to be as high as 20%–50% in some facilities, primarily large tertiary care hospitals that provide care for patients with numerous comorbidities and previous antifungal treatment [16, 17]. Prospective monitoring of the burden of drug-resistant infections is important, as echinocandin use may be increasing following the release of 2016 Infectious Diseases Society of America guidelines recommending echinocandins as first-line treatment [18], as well as ongoing spread of multidrug-resistant C. auris infections.

In this analysis, we found that the burden of candidemia among black individuals was disproportionately high. Thirteen percent of the US population was black in 2017 Census Bureau estimates [19], whereas nearly a quarter of all estimated candidemia cases occurred among black individuals. This racial disparity may result from underlying patient-level factors (eg, differences in underlying health conditions, socioeconomic status, access to medical care, and health-seeking behavior) or due to disparities in the healthcare system itself [20]. Racial disparities have been documented with other healthcare-associated pathogens, including methicillin-resistant Staphylococcus aureus (MRSA) [21]. Investigating and addressing the cause of these disparities will be important in reducing the burden of candidemia.

With regard to geography, previous studies summarizing antimicrobial prescribing consistently show higher rates in the Southeast US [22, 23]. In our study, rates of candidemia were similarly higher in eastern and southern parts of the country, although not statistically different possibly due to small sample size. These findings may be related to underlying risk factors for candidemia not accounted for here (eg, diabetes), broader factors (eg, socioeconomic status, health behaviors), or differences in antibiotic use by region, since broad-spectrum antibiotic use is an independent risk factor for candidemia [24–26]. Our ability to evaluate geographic variation in rates was limited, but this would be an area of interest in future analysis and work.

Because candidemia is not a nationally notifiable condition, we explored multiple data sources from which to measure candidemia burden, including the National Healthcare Surveillance Network (NHSN) and other data sources that utilize International Classification of Diseases (ICD) billing codes. The NHSN is CDC’s national healthcare-associated infection surveillance system that captures central line–associated bloodstream infections (CLABSIs), including those caused by Candida species. Although NHSN captures data from nearly all acute care hospitals in the US, candidemia cases would only be captured from CLABSIs, missing cases occurring in patients without central lines [27]. The Healthcare Utilization Project (HCUP) [28] is a large publicly available database with billing codes for hospital discharges that includes approximately 20% of hospital discharges in the US each year, allowing for national extrapolation and has been used for burden estimation [29, 30]. However, using ICD billing codes to identify people with a condition can be problematic, as these codes are assigned primarily for billing purposes, and discrepancies have been identified between billing codes and actual diagnoses for another fungal pathogen [31]. Furthermore, no single ICD code exists for candidemia. A study of pediatric candidemia determined that the sensitivity of ICD-9 codes for candidemia was 60% [32], making ICD coding data a poor source for burden estimation of candidemia.

Although EIP surveillance does not have the breadth of national coverage as other sources such as NHSN or HCUP, it has substantial depth and is population based. Furthermore, because the populations under surveillance in the 9 sites are diverse (by age distribution, race, urban/rural, and other factors), EIP findings can and have been used to extrapolate to national estimates of other infections [33, 34]. Burden estimates for other organisms derived through EIP surveillance data have provided a benchmark in the epidemiology of these infections and allowed for future evaluation and comparison. For example, burden estimates performed for invasive MRSA in 2005 (111 000 infections) and 2011 (80 000 infections, 80% of which were bloodstream infections) revealed a 31% decrease in overall number of cases and changing trends in hospital vs community-onset illness [33, 35]. The national burden of Clostridioides difficile infection, also using EIP data, was estimated to be 453 000 infections based on 2011 data [34]. The estimated case burden of candidemia is similar to that of certain invasive bacterial infections, despite the fact that candidemia accounts for only a portion of invasive candidiasis (Table 2).

Table 2.

Comparison of Burden Estimates Derived From Emerging Infections Program Surveillance for Selected Invasive Infections, United States

Infection	Year	Estimated Cases	Incidence Rate per 100 000	All-Cause Mortality	Estimated Deaths
Candidemiaa	2017	22 660	7.0	25%	5628
Clostridioides difficileb [34]	2011	453 000	147.2	1.3%–9.3%	29 000
Invasive MRSAc [33]	2011	80 461	25.8	13%	11 285
Invasive Streptococcus pneumoniaec [36]	2017	31 000	9.5	11.4%	3590
Invasive group A Streptococcusc [37]	2017	23 650	7.6	8.2%	1980

Abbreviation: MRSA, methicillin-resistant Staphylococcus aureus.

^aIncludes bloodstream infections only.

^bIncludes stool specimens.

^cIncludes bloodstream infections plus other invasive infections.

Our estimates are limited in that they extrapolate active surveillance data from 5% of the US population to the national level using demographic data and do not fully account for complexities in underlying medical conditions and access to care. Because we identified cases based on positive blood cultures, we likely missed cases for which cultures were not obtained, not available, or for which cultures were falsely negative. Molecular diagnostics, such as polymerase chain reaction, are increasingly used as an adjunct to culture, and our surveillance does not capture culture-negative cases that might be detected by these methods, although use of molecular diagnostics for Candida in blood was likely not widespread in 2017. Additionally, autopsy studies have revealed invasive candidiasis as an often-missed diagnosis [9]. For these reasons, we have likely underestimated the overall burden of candidemia in the US. Although candidemia surveillance started in 2008, the number of sites participating was limited before 2017, thereby limiting our ability to conduct national-level trend estimation over the last decade. Additionally, we may have underestimated mortality as our surveillance only captures mortality that occurs in patients who die during a hospitalization for candidemia, and more robust methods of estimating candidemia attributable mortality are needed.

Despite these limitations, this analysis represents the first population-based estimate of candidemia burden in the United States and will serve as a starting point for evaluating future rates and epidemiology, in addition to the impact of interventions aimed at preventing candidemia.

Notes

Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Potential conflicts of interest. The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.