Attributable mortality of candidemia at a German tertiary hospital from 1997 to 2001 before the introduction of echinocandins

The relevance of candidemia has increased over the last decades due to higher incidence rates in an ageing society. Studies on amphotericin B and fluconazole have shown high attributable mortality rates of 38% and 49% in the United States. Incidence rates and locational factors might have an impact on the mortality rates at the University Hospital of Cologne (UHC), Germany.


| INTRODUC TI ON
Candidemia often results in a prolonged hospital stay and high morbidity and mortality. [1][2][3] Candida spp. are one of the most common causes of nosocomial bloodstream infections (BSI). A study performed in 2011 in Germany, including data from 586 intensive care units (ICUs), observed Candida albicans as the fourth leading pathogen of central venous catheter (CVC) associated BSI. 4 In 2004, Candida spp. ranked seventh of the pathogens of nosocomial BSI in Switzerland. 5 Studies performed in the United States (US) reported that Candida spp. are among the four most common pathogens causing nosocomial BSI. [6][7][8] Well-established risk factors for candidemia are prior and prolonged antibiotic treatment, parenteral nutrition, haemodialysis, presence of CVC, prolonged hospitalisation and abdominal surgery. Immunocompromised patients and patients on ICU are at high risk for candidemia. 7,9,10 In 1988, a monocentric, retrospective, case-control study performed in Virginia, US, by Wey et al observed an attributable mortality of candidemia of 38%. 3 11 and emergence of triazole-resistant species 12 influence the attributable mortality. In this second study, an increase of the attributable mortality to 49% was reported. 2 In 2020, a similar study was performed at the University Hospital of Cologne (UHC), Germany, analysing the attributable mortality of candidemia after the introduction of echinocandins. Cornely et al observed a significantly lower attributable mortality of 26%. 13 Until now, little data on epidemiology, risk factors and mortality rates of candidemia have been published in Germany at the turn of the millennium. To study medical progress and to analyse trends and developments, historical comparators are needed. We therefore performed a retrospective, matched case-control study analysing the attributable mortality of candidemia at the UHC from 1997 to 2001.

| PATIENTS AND ME THODS
We performed a retrospective, monocentric, case-control study analysing all hospitalised patients with nosocomial candidemia at the UHC, between 1 July 1997 and 30 June 2001. The UHC is a tertiary care teaching hospital, which had an average annual number of 44 863 admissions during that period. Nosocomial candidemia was defined as at least one positive blood culture for any Candida species occurring >24 hours after admission. Control patients were matched to candidemia cases, which did not have a sign of invasive Candida infection.
To ascertain the best matching to a control patient, we used a stepwise procedure. In the first step, we used SAP databases (SAP Graphical User Interface (SAP GUI) 750 for Windows, SAP SE) to generate a list of possible control patients matched by age ±6 years (children: age ±1 year), sex, admission ±18 months, treatment on ICU ±14 days, days at risk and underlying disease. In a second step, we took surgical procedures and the Charlson Comorbidity Index (CCI) into consideration. 14 During the matching process, YB and PK were blinded to the outcome of each matched patient chosen by the predefined algorithm.
For reasons of comparability, we defined the day of candidemia diagnosis as day d0. This was the day at which a preliminary identification of a yeast was communicated to the treating physician, which triggered antifungal therapy and further measures (eg removal of indwelling central lines). The timespan from the initial identification of a yeast in blood culture to the subsequent identification to species level was zero to five days. For controls, d0 was defined as the last day of the timespan t0 after admission, which was equivalent to the timespan t0 between admission and first positive blood culture of case patients. If no suitable control patient matched by days at risk could be identified, the patient with the longest time of hospitalisation was chosen. In those cases, d0 was defined as the last day of hospitalisation for non-survivors and patients lost to follow-up, whereas for survivors the definition of d0 did not differ from other control patients and occurred beyond the date of discharge.
For both case and control patients, epidemiologic and demographic characteristics were documented. We captured data of the underlying disease, Charlson Comorbidity Index, and risk factors for candidemia. ICU stay and indwelling CVC were only considered as a risk factor if they were present at least two days before d0. Both groups were compared by time of hospitalisation before and after d0, the latter, respectively, for survivors and non-survivors.

| Baseline characteristics
Characteristics serving as matching criteria such as sex, median age, CCI and major surgery were similar across both groups (P = n.s.;
Susceptibility results were available for only seven isolates at the time of hospitalisation. At a later stage, 55 isolates were retrospectively tested in 2007. 16 All analysed C albicans isolates (39) were susceptible to amphotericin B and fluconazole. All C glabrata isolates (8)

(2.5-4.5)
Underlying conditions Hem./Onc. disease   (1.8%) was treated with CVC removal only. In three cases (5.3%), no information regarding treatment could be retrieved due to incomplete data on archived files (Table 4).

| Guideline adherence
Diagnostic procedures were provided to more survivors than to nonsurvivors, in particular susceptibility testing (17.1% vs 4.5%, P =.404) and ophthalmoscopy (28.6% vs 0%, P =.005). 45.7% of survivors were treated for 14 days after first negative follow-up blood culture whereas only 13.6% of non-survivors were treated in the same manner (P =.003). CVCs were removed in 72.4% of survivors and 64.7% of non-survivors (P = n.s.), and follow-up blood cultures were performed in 82.8% and 63.6% of surviving and non-surviving patients (P =.003; Table 5).

| Mortality
During our study period, we observed an in-hospital mortality of 33.3% for candidemia cases (17/51) and 11.8% for control patients (6/51). Day 30 mortality rates for cases and controls were 23.5%

| DISCUSS ION
We performed a retrospective, single-centre, matched-pair study analysing the attributable mortality in patients with nosocomial candidemia from 1997 to 2001.
Our study observed a lower attributable mortality than studies previously performed in Iowa during the same period (21.5% vs 49%) and from 1983 to 1986 (38%). 2,3 Overall mortality in candidemia patients was significantly lower in our study (33% vs 61%) whereas mortality in control patients was similar (11.8% vs 12%). 2 Our prior single-centre study showed higher case / control inhospital-mortality rates in comparison with our more recent study (cases 33.3% vs 43%, controls 11.8% vs 17%; attributable mortality 21.5% vs 26%). 13 Furthermore, comorbidity rates and overall candidemia incidence were lower in our study (3.5/10 000 admissions vs 6/10 000 admissions). 13 Higher morbidity and mortality rates in more recent publications could be explained by the demo- An improved disease management, including aspects recommended by the IDSA Guideline from 2000 like follow-up blood cultures, treatment for 14 days after first negative follow-up blood culture and removal of CVCs, 11 might also improve morbidity rates.
Improved diagnostic and prophylactic procedures as well as a better disease management by adherence to current guidelines contribute to a better outcome of candidemia.

ACK N OWLED G EM ENTS
We thank Roman Voskoboynik for SAP data retrieval. We thank Irmgard Henseler and Susanna Proske for assistance in health record collection. Open access funding enabled and organized by ProjektDEAL.

E TH I C A L A PPROVA L
The authors confirm that the ethical policies of the journal, as noted in the author's guideline page, have been adhered to.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.