Cases of candidemia were prospectively identified by blood culture surveillance at 50 of 52 public and private microbiology laboratories in Australia (population ≈20.1 million) from August 2001 to July 2004. One nonparticipating laboratory provided services to a university pediatric hospital where candidemia rarely occurred (A. Daley, pers. comm.) and the other provided services to an adult university hospital (36 cases identified; H. Sheorey, pers. comm.). Approval for the study was obtained from the human ethics review committees of institutions providing clinical data.

Adults, children, and neonates with <1 blood culture yielding Candida species were eligible for enrollment. State or territory coordinators were responsible for case identification. Information was collected on a standard form regarding age, sex, patient location at time of candidemia diagnosis, healthcare setting, risk factors within the preceding 30 days (including surgery, vascular access devices [VADs], hyperalimentation, and use of antimicrobial and systemic antifungal agents), major concomitant conditions (International Statistical Classification of Diseases and Related Health Problems, 10th revision, Australian modification) (15), portal of entry, clinical signs of sepsis (16), complications of candidemia, results of diagnostic studies, antifungal therapy, and clinical outcome <30 days after diagnosis. All data were collected prospectively, and forms were completed on days 5 and 30 after the date of the initial positive blood culture or at death if it occurred earlier. Data were collected and analyzed at a central site. Periodic audits of laboratory records ensured that all cases of candidemia were reported. For the study period, number of hospital beds and annual different day patient separations (defined as completed admissions) were obtained from 40 hospitals.

A case was defined as incident isolation of Candida species from blood during the study period. For patients with >1 episode of candidemia, the second episode was defined as a new case if it occurred >30 days after the previous episode. A total of >2 episodes fulfilling the case definition and occurring in different patients who were epidemiologically linked (e.g., with regard to risk factors, location, and time) constituted a case cluster.

Episodes were classified as inpatient healthcare-associated (IHCA) if they occurred >48 hours after hospital admission and had not clinically manifested on admission (17). Cases occurring <48 hours after hospital admission were considered outpatient-acquired. Among outpatient-acquired candidemia, episodes associated with an indwelling medical device, surgical procedure, or chemotherapy-related neutropenia (<1 × 10⁹ cells/L, adjusted for children) were classified as outpatient healthcare associated (OHCA), and CA infections were classified as those occurring in patients with no healthcare-related risk factors (17). Adult intensive care unit (ICU) acquisition was when candidemia developed >48 hours after ICU admission for nonneutropenic patients. Source of candidemia was considered VAD related if culture of the device tip grew the same species isolated from blood. Endocarditis was classified by modified Duke criteria (18).

All laboratories cultured blood specimens in BACTEC (Becton Dickinson, Sparks, MD, USA) or BacT/Alert 3D (bioMérieux, Marcy l'Etoile, France) automated systems. Candida organisms were speciated by using standard phenotypic methods (19). Isolates were forwarded to a reference laboratory (Women's and Children's Hospital, Adelaide) for susceptibility testing and species confirmation by using conventional methods (20). Candida dubliniensis was distinguished from C. albicans by PCR fingerprinting (21). Where identification was discordant, species determination at the reference laboratory was used.

Population and age-specific incidences were calculated by using denominator data from the 2004 Australian census (22) and hospital-specific incidences by using day of separation denominator data from individual hospitals for the study period. Overall incidences were expressed as pooled mean rates calculated by aggregating numerators and dividing by the sum of denominators from all sites. Data were analyzed with SPSS version 10.0.7 (SPSS Inc., Chicago, IL, USA). Continuous variables were compared with Student t test, and categorical variables were compared with χ² or Fisher exact tests. Incidence data from university hospitals were pooled and compared with pooled data from other hospital types with χ² test. A p value <0.05 was considered significant. Univariate analyses were performed to identify risk factors associated with overall illness. Candidate variables with a univariate p<0.15, or previously published risk factors, were analyzed by multiple logistic regression. The same independent predictors of death were identified with forward, backward, and stepwise variable selection methods.

Permission was denied to include 36 candidemia episodes identified at the nonparticipating adult hospital. Thus, the number of incident episodes was 1,095 (in 1,095 patients), with 337, 352, and 406 during the first, second, and third years, respectively, of the study. Data on hospital characteristics were available for 1,095 patients, demographic and clinical data for 1,005 (91.7%) episodes, and outcome data for 857 (78.3%). Species were identified for 1,068 (97.5%) isolates.

The average population-based incidence of candidemia was 1.81/100,000 (1.87/100,000 if inclusive of cases at the nonparticipant hospital) per year (Table 1) and was highest in infants (24.8/100,000) and persons >65 years of age (13.7/100,000; Figure 1). Most cases (36.4%) occurred in the most populous state, New South Wales, but incidence was highest in Queensland. Age composition of the population was similar by jurisdiction and similar to the national average (data not shown) except for the 2 least populated areas, Northern Territory and Australian Capital Territory, where the percentage of persons >65 years of age was 4.4% and 9.3%, respectively, compared with the national average of 13%.

The pooled mean annual hospital-specific incidence of candidemia (data from 40 institutions) was 0.21/1,000 separations (range by jurisdiction 0.05–0.36) (Table 1). No case clusters occurred. The estimated mean incidence in university (n = 28) and university-affiliated hospitals (n = 9) was similar (0.22/1,000 separations, range by jurisdiction 0.05–0.90) and greater than in private hospitals (n = 3, 0.1/1,000 separations, range by jurisdiction 0.04–0.16).

Neonates <1 month of age accounted for 33 (3.3%) of 1,005 cases, children 2 months to 14 years of age for 95 (9.5%) cases, adults 15–64 years of age for 527 (52.4%) cases, and patients >65 years of age for 350 (34.8%) cases. The median age was 56 years (range 0–98); 537 (53.4%) episodes affected male patients.

Most cases occurred in medical wards (including cancer and hemopoietic stem cell transplant [HSCT] units; 390 [35.6%] of 1,095 cases), critical care units (273 [24.9%] cases), and surgical wards (193 [17.3%] cases). Emergency (4.3%) and obstetrics and gynecology (0.5%) services reported small percentages of episodes. Pediatric cases occurred in pediatric critical care units (16 cases, 1.5%) and pediatric medical or surgical wards (24 cases, 2.2%). Neonatal infections occurred in premature or low birthweight infants in neonatal critical care units (35 cases, 3.2%).

Concomitant conditions and risk factor data were available for 1,005 episodes. Cancer was the most common underlying condition (323 cases, 32.1%), with 164 episodes in solid tumor (50.8% cancers) cancer patients. Of 159 cases diagnosed in patients with hematologic malignancy, 91 occurred in patients with leukemia and 58 in those with lymphoma. Only 52 episodes occurred in transplant recipients (17 allogeneic and HSCTs, 11 autologous HSCTs, and 24 solid organ transplants). Gastrointestinal disorders (19%), chronic cardiovascular disease (13.8%), and diabetes (13.6%) were common, but pancreatitis was infrequent (2.5%), and coincident HIV infection was rare (0.6%). Common iatrogenic risk factors included indwelling VADs (72.6%), antimicrobial agents (77%), major surgery (37.1%), and hyperalimentation (33%, Table 2).

Of 919 episodes for which information on the healthcare setting was available, 749 (81.5%) were IHCA, 107 (11.6%, or 62.9% of outpatient-acquired cases) were OHCA, and 63 (6.9%) were CA. The proportion of disease acquired outside a hospital by jurisdiction ranged from 9.8% (4 of 41 cases in South Australia) to 50% (but only 1 of 2 cases in Northern Territory). Of the major states, Western Australia reported the lowest percentage of OHCA cases (1 [2.5%] of 40 episodes, Table 1). Comparison of major concomitant conditions and risk factors according to healthcare setting is summarized in Table 2; >4 risk factors were evident in 195 (21.2%) episodes. A total of 183 (20%) episodes were associated with use of an adult ICU. Twenty-nine (93%) neonates received hyperalimentation compared with 35.4% of the adults. Systemic antifungal agents were administered as prophylaxis in 67 (55.4%) of 121 instances; 48 (68.6%) of these prescriptions were for hematology/HSCT patients.

Most (90%–100%) episodes associated with prematurity, organ transplantation, and burn or trauma were IHCA (Table 2). Compared with all cases acquired outside a hospital, patients with IHCA infection were significantly more likely to have had recent surgery (49% vs. 12.5%, p<0.001), neutropenia (19.7% vs. 12.6%, p = 0.04), indwelling VADs (90.1% vs. 48.4%, p<0.001), corticosteroid therapy (33% vs. 19%, p<0.001), or parenteral nutrition (44.1% vs. 8.9%, p<0.001) or to have died within 30 days of infection (31.1% vs. 1.8%, p<0.001). These risk factors, with the exception of surgery, were more common in cases with OHCA candidemia than in those with CA candidemia (Table 2).

Recent surgery and VADs, hyperalimentation, and antimicrobial agents were more common in IHCA patients than in OHCA patients (Table 2). Conversely, coincident malignancy (44% vs. 34% episodes, p = 0.05) and cancer chemotherapy (25.2% vs. 14.3%, p = 0.01) were associated with OHCA infection. OHCA patients also included hemodialysis recipients and patients with diabetes. In CA candidemia, common concomitant conditions included intravenous drug use (IVDU, 23.8%), gastrointestinal disorders (19%), diabetes (19%), behavioral disorders such as chronic alcohol abuse (31.8%), and infectious diseases (23.4%); 15 (50%) of 30 episodes in IVDU patients were CA (Table 2). Differences in clinical manifestations were identified in the 3 groups. Patients with IHCA candidemia were more likely to have sepsis, die within 30 days, remain in hospital longer, and have concomitant healthcare-associated bacteremia (Table 2). Of 34 patients with Candida endocarditis (13 definite and 21 possible), 7 were OHCA with an increasing trend when compared with those with IHCA candidemia (Table 2).

Species of Candida was correctly identified in 96.7% instances in which identity was established. C. albicans was the most common species (505 [47.3%] of 1,068 episodes) followed by C. parapsilosis (19.9%) and C. glabrata (15.4%). C. tropicalis, C. krusei, and C. dubliniensis were identified in 5.1%, 4.3%, and 1.9%, respectively, of the patients. The remaining 42 (4.0%) episodes were caused by C. guilliermondi (n = 11), C. lusitaniae (n = 7), C. kefyr (n = 5), C. famata (n = 3), C. rugosa (n = 3), and C. pelliculosa (n = 3). There were 24 (2.2%) episodes of polycandidal infection and an increase in the proportion of C. glabrata candidemia from 10.1% episodes the first year to 19.5% episodes in the third year of the study (p = 0.02, data not shown).

C. albicans was the predominant species in Tasmania (60%), Australian Capital Territory (77%), and Western Australia (54.1%). However, this species caused only 37.8% of episodes in South Australia, whereas 31.1% were caused by C. parapsilosis (7.7%–21.9% of cases elsewhere). C. krusei and C. dubliniensis candidemia were rare in the least populous jurisdictions where complex procedures such as HSCT are not performed. Age distribution, including proportion of neonates and children in the population, was comparable for all jurisdictions (data not shown).

Species other than C. albicans were isolated more often from patients with candidemia acquired outside a hospital (60.5% vs. 49.9% for IHCA candidemia, p = 0.02). C. parapsilosis was recovered from 30.9% of these episodes (compared with 16.7% of IHCA episodes, p<0.001), although C. krusei and C. dubliniensis candidemia was rare. Relative proportions of infections caused by C. albicans and non–C. albicans species in cases of OHCA candidemia and CA candidemia were similar.

In neonates and children, C. parapsilosis accounted for a similar proportion of episodes as did C. albicans; other species were rare. C. albicans was the most common causative pathogen in adults (48.6% of cases) and C. glabrata, C. parapsilosis, and other Candida species were approximately equally distributed (Figure 2). Eighty (53.3%) of 150 episodes in patients >65 years of age were caused by C. glabrata.

C. albicans was the most common pathogen in ICUs (61.7% of episodes) and surgical (54.4% of episodes) patients but caused only 32% of episodes in hematology patients, who had the highest proportion (24 [66.7%] of 36 episodes) of infection with C. krusei. C. parapsilosis accounted for 42.9% of episodes in neonatal ICUs and for 42.4% of cases in patients without hematologic cancer. C. glabrata was present in all patient groups. Twelve (63.2%) of 19 episodes of C. dubliniensis candidemia were in surgical and nonhematology patients.

The all-cause 30-day death rate was 27.7% (236 of 853 episodes) and similar in all age groups. Clinician-reported attributable deaths were lower (93 episodes, 10.9%). The death rate was higher in IHCA episodes than in either OHCA or CA episodes (Table 2). Univariate analysis showed that numerous variables were associated with increased death (Table 3). Multiple logistic regression analysis of 16 variables showed that an age >65 years, ICU stay, sepsis at diagnosis, and corticosteroid therapy (p<0.001, Table 4) were associated with the greatest risk for death. C. glabrata infection was also an independent predictor of death. Treatment with an antifungal agent (752 [85.8%] of 876 episodes) was independently associated with lower odds of death but removal of a VAD (76.4% of 878 instances where VADs were in situ at diagnosis) was not (Table 4). However, in the subset of VAD-related candidemia episodes (n = 409), removal of the VAD (n = 339) was associated with a better outcome (odds ratio 0.48, 95% confidence interval 0.26–0.88, p = 0.02).