French intensive care physicians’ practices for the management of high-risk patients with invasive candidiasis: results from the multicenter INSPIRE survey

Abstract

Background

Candidiasis remains a major global health challenge, with rising incidence and antifungal resistance complicating management. This multicenter survey aimed to describe diagnostic and therapeutic management practices of ICU physicians in France for high-risk invasive candidiasis cases, using randomized clinical case simulations distributed to French ICU physicians via email and academic congresses.

Results

A total of 75 physicians completed 283 cases. Overall, clinical suspicion justified diagnostic investigation in 72.1% of cases. Key factors associated with the decision to test for invasive candidiasis included post–abdominal surgery (OR = 3.92; p = 0.002), central venous catheter (OR = 2.28; p = 0.04), parenteral nutrition (OR = 2; p = 0.027), prolonged antibiotic exposure (OR = 1.90; p = 0.041), and physician experience >10 years (OR = 2.79; p = 0.014). Echinocandins were the preferred first-line treatment (88.2%; n = 120/136). Their use was guided by clinical guidelines (78.3%), pharmacokinetic/pharmacodynamic profile (38.3%), SOFA score (29.2%), patient status (25.8%), and immunosuppression (20%). De-escalation to azoles upon documented mycological susceptibility occurred in 87.5% of cases initially treated with an echinocandin and 60.0% of cases initiated on amphotericin B, primarily after 5 days of treatment (51.2%; n = 64).

Conclusions

Patient severity, immunosuppression, and catheter presence were the main factors guiding antifungal treatment decisions Management of confirmed invasive candidiasis was largely consensual among respondents, whereas empirical treatment decisions were heterogeneous. While conclusions are primarily reflective of practice patterns within specialized university hospital centers, the INSPIRE survey highlights the need for enhanced training, broader access to rapid diagnostics, and improved decision-support systems to better harmonize clinical practices.

Background

Invasive candidiasis is a serious and prevalent infection in critically ill patients admitted to intensive care units (ICUs), with Candida albicans being the predominant agent [1]. The incidence of invasive candidiasis is tenfold higher in ICUs compared to medical and surgery wards and represents 70%–90% of all ICU fungal infections [2], with a crude mortality of 40–55% [[3], [4], [5]]. In France, it affects approximately 4%–5% of ICU patients [3].

The management of invasive candidiasis is particularly complex due to diagnostic and therapeutic challenges [6]. Although culture-based approaches remain the gold standard, their prolonged turnaround time combined with their low sensitivity and specificity, can delay confirmation of invasive candidiasis, especially during the early stages of infection [7]. Therefore, it is essential that the ICU physicians keep a high level of suspicion for this infection in their differential diagnosis [1], and select the most appropriate, prompt, and targeted antifungal therapy to decrease the mortality rate [8].

Nonetheless, the incidence of invasive candidiasis is relatively low in the ICU, which limits the direct clinical experience of ICU physicians and may contribute to variability in management. Over the past few years, antifungal therapy has evolved and despite the availability of clinical global guidelines for the management of candidiasis [9], real-world decision-making can differ significantly from standardized recommendations, especially in high-risk ICU patients. This discrepancy appears to be driven less by clinical preference and more by practical constraints. On the diagnostic level, although several microbiological tests are available, surveys, including the 2024 nationwide survey [10], have shown that access to these tools remains limited in routine ICU settings. Therapeutically, while treatment durations for candidemia are generally well adhered to, practices for other deep-seated candidiasis, particularly intra-abdominal infections, remain heterogeneous.

Given the gap between recommendations and everyday practice, there is a pressing need to assess how ICU physicians in France manage such high-risk patients.

Methods

Aim of the survey

The aim of the INSPIRE survey was to describe common management practices from suspicion to confirmed diagnosis and clinical evolution of high-risk patients with invasive candidiasis in the ICU based on virtual cases. Secondary objectives were to describe diagnostic approaches to identify key criteria influencing therapeutic decisions, and to explore the use and decision-making around antifungals.

Survey design

This survey of practices was based on the validated case-vignette method [[11], [12], [13], [14], [15]]. Fictitious clinical cases were developed with a French expert committee involving ICU physicians and a mycologist, based on combinations of the following parameter variations: patient clinical situation, immunosuppression status, presence of a central venous catheter, use of parenteral nutrition, length of ICU stay, severity of illness in the ICU, prior antibiotic exposure, and confirmation of invasive candidiasis (Supplementary Table S1). To ensure clinical coherence, certain incompatible combinations were excluded, specifically, cases combining absence of a central venous catheter with either lung cancer or a SOFA score >8.

Each simulated clinical case included four main components: (i) a sociodemographic description of the patient (age, gender, profession); (ii) a clinical examination; (iii) diagnostic test results; and (iv) treatment modalities, defined according to the selected parameters. By randomizing identified parameters and addressing their incompatibilities, a pool of cases was created, each featuring a unique combination of modalities. Notably, the clinical scenario was not limited to the four broad patient categories alone. Each clinical vignette was generated through the random combination of modalities from eight predefined clinical parameters (one modality randomly assigned per parameter), which represent established high-risk factors for invasive candidiasis.

To maintain strict clinical realism and directly address the clinical suspicion of severe infection, all vignettes began with an identical baseline description: the patient was introduced as receiving empirical broad-spectrum antibiotic therapy for a suspected septic shock, with no bacteriological confirmation yet available. These variables reflected four different clinical situations detailed with their standard parameters in Supplementary Table S1: (i) patients with abdominal surgery and upper mesocolic perforation; (ii) patients with lung cancer; (iii) patients with decompensated cirrhosis; and (iv) obese patients with diabetes and decompensated renal and/or cardiac failure.

Diagnostic testing was left to physician discretion, and testing was often prompted by high‑risk profiles, prolonged ICU stays, and prior antibiotic exposure, even in the absence of overt sepsis. Risk factors included in the scenarios were deliberately selected to reflect established high‑risk profiles for invasive candidiasis (e.g., central venous catheter, immunosuppression, parenteral nutrition). Unrelated factors, such as hypertension, were not incorporated, as the primary objective was to evaluate physician practices in populations where invasive candidiasis is most frequently suspected.

Importantly, the generated scenarios could involve either suspected or mycologically confirmed invasive candidiasis. The diagnostic tests for mycological confirmation were not explicitly defined within the simulation questionnaire. The confirmed/non-confirmed status of invasive candidiasis was assigned through randomization and subsequently inferred from the diagnostic tests selected by the responding physician (Supplementary Fig. S1). To support a more robust analysis of clinical decision-making, the “confirmed candidiasis” status was intentionally overrepresented within the randomized pool.

Data collection

The survey was conducted between November 2024 and February 2025. Each voluntary ICU physician received four clinical cases presented via a chatbot on a limited-access platform. This “conversational interface” is a computer program that allows a conversation between the ICU physician and a virtual interlocutor. To prevent multiple individual responses from the same department, the survey explicitly targeted unique ICUs, and the formal invitation specified that only one response should be submitted per ICU service. Prior to completing the cases, the ICU physicians provided their sociodemographic information (demographics, practice setting, ICU type and capacity, experience, number of annual ICU admissions, and estimated number of annual admissions for invasive candidiasis). Then, they completed a questionnaire on diagnostic tools and treatment choices across multiple stages (diagnosis and initial management, confirmed diagnosis, treatment monitoring and adaptation) (Supplementary Table S2 and Supplementary Fig. S1). Regarding the systemic workup for confirmed candidemia, the option to perform cardiac imaging was presented broadly as “echocardiography” without mandatory differentiation between transthoracic (TTE) and transesophageal (TEE) modalities, focusing primarily on the global decision to initiate cardiac screening.

Study outcomes

Study outcomes were frequencies and patterns in diagnostic and therapeutic practices, timing of antifungal initiation and discontinuation, first-line antifungal preferences, justifications for treatment decisions (clinical factors, guidelines such as the global guideline for the diagnosis and management of candidiasis [9], local Candida epidemiology), and management of non-responsive cases (diagnostic reassessment, treatment de-escalation).

Statistical methodology

Continuous variables are reported as means, standard deviations, minima, maxima, medians and interquartile ranges. Categorical variables are reported as frequencies and percentages.

To identify factors associated with the therapeutic management of confirmed cases of invasive candidiasis, two types of analysis were conducted: the first involved describing decision-making parameters as directly reported by physicians through the questionnaire; the second involved testing the influence of randomized parameters using a mixed-effects multinomial logistic regression model (echinocandins vs. amphotericin B vs. azoles), accounting for the non-independence of responses from the same physician (random physician effect).

An additional analysis was included to investigate factors associated with the decision to test for invasive candidiasis. This was prompted by the observation that in many scenarios, the search for candidiasis was not initiated, contrary to expectations at the time of survey planning. This analysis used a mixed-effects binary logistic regression model. Due to a collinearity between the SOFA score and central venous catheterization, these two parameters could not be entered into the multivariate model simultaneously. Consequently, two separate multivariate analyses were constructed: a primary analysis model incorporating the SOFA score, and a sensitivity analysis model substituting the SOFA score with the presence of a central venous catheter. The model incorporating the SOFA score was defined a priori as the primary analysis because a validated organ dysfunction score serves as a more accurate substitute for overall patient acuity, systemic physiological severity, and clinical prognosis. Conversely, the presence of a central venous catheter represents a context-dependent exposure factor, making it more suitable for a secondary sensitivity analysis to verify the stability of the remaining covariate estimates across different adjustment strategies. Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were presented, statistical significance level was set at 5%. To ensure the validity of multivariate logistic models, a minimum of one explanatory variable per 10 events in the smallest group was required. If the number of events observed was too low relatively to the number of potential explanatory variables (the eight randomized factors), only descriptive analyses were performed. In addition, a stratified analysis was performed according to four predefined patients’ clinical situations: (1) patients with abdominal surgery and upper mesocolic perforation; (2) patients with lung cancer; (3) patients with decompensated cirrhosis; and (4) obese patients with diabetes and decompensated renal and/or cardiac failure. Statistical analysis was carried out using SAS (Statistical Analysis System, SAS Institute Inc., version 9.4) software.

Results

Characteristics of the respondents

A total of 160 invitations to participate in the survey were extended to ICU physicians, including 145 sent via email and 15 during an academic congress. Of these, 75 ICU physicians responded (response rate: 46.9%) completing a total of 283 cases.

The majority (78.7%; n = 59) were practicing in university hospitals, and 23 (30.7%) physicians had more than 20 years of experience. Thirty-eight (50.7%) physicians reported working in ICUs with a capacity of more than 20 beds, with a median number of 850 annual admissions [IQR: 500.0 to 1,200.0] across participating centers. The majority of physicians (64%, n = 48) reported managing fewer than 10 invasive candidiasis cases in the ICU per year (Supplementary Table S3).

Characteristics of the clinical cases

The 283 clinical situations were distributed in approximately equal proportions across the four predefined patients’ clinical situations: (i) 70 (24.7%) cases were patients with post–upper abdominal surgery and mesocolic perforation; (ii) 69 (24.4%) cases had lung cancer; (iii) 74 (26.1%) were hospitalized for decompensated cirrhosis; and (iv) 70 (24.7%) were obese patients with diabetes and decompensated renal and/or cardiac failure.

By study design, the remaining randomized baseline clinical parameters were well-balanced and distributed across approximately half of the generated vignettes. Invasive candidiasis was mycologically confirmed in 195 (68.9%) cases, while 88 (31.1%) cases were managed without mycological confirmation (Supplementary Table S4).

Diagnostic testing for invasive candidiasis: frequency and associated factors

Overall, the majority of the respondents (72.1%; n = 204) reported that they would likely perform diagnostic testing for invasive candidiasis. Specifically, when presented with the initial clinical vignettes, respondents were asked to determine whether the scenario justified a clinical suspicion and subsequent investigation for an invasive fungal infection. Candidiasis was subsequently confirmed in 136/204 (66.7%) of these investigated cases, with a relatively balanced distribution across the four clinical situations: (i) 38 (64.4%) cases were patients with abdominal surgery and upper mesocolic perforation; (ii) 36 (64.3%) cases were patients with lung cancer; (iii) 32 (71.1%) were hospitalized for decompensated cirrhosis; and (iv) 30 (68.2%) were obese patients with diabetes and decompensated renal and/or cardiac failure.

Based on the multivariate analysis adjusted for clinical severity via the SOFA score, patients with abdominal surgery with upper mesocolic perforation, compared to those with decompensated cirrhosis, were 4 times more likely to undergo diagnostic testing for invasive candidiasis (OR = 3.92; 95% CI: 1.66–9.28, p = 0.002). The use of parenteral nutrition was also significantly associated with increased likelihood of diagnostic testing (OR = 2; 95% CI: 1.08–3.70, p = 0.027). Also, a SOFA score >8 showed a strong trend toward an increased likelihood of testing (OR = 2.04; 95% CI: 0.97–4.30, p = 0.061). In addition, an antibiotic treatment duration longer than 48 h was associated with greater odds of testing (OR = 1.90; 95% CI: 1.05–3.52, p = 0.041). Finally, physicians with more than 10 years of intensive care experience were more likely to perform diagnostic testing for invasive candidiasis (OR = 2.79; 95% CI: 1.23–6.29, p = 0.014; Table 1 – ‘Without catheter – With SOFA score’ column). The presence of a central venous catheter was itself a significant independent driver of testing (OR = 2.28; 95% CI: 1.06–4.89, p = 0.04; evaluated in a separate multivariable model iteration due to design-driven collinearity with SOFA score; Table 1 – ‘With catheter – Without SOFA score’ column). Importantly, the ORs, 95% CIs, and significance levels for all other variables remained stable between the two models, confirming that the adjustment choice did not alter the study’s main conclusions.

Table 1.

Factors associated with a diagnostic testing for invasive candidiasis (univariate and multivariate analysis).

				Without catheter - With SOFA score		With catheter - Without SOFA score
		Univariate analysis OR (CI 95%)	Univariate analysis p-value	Multivariate analysis OR (CI 95%)	Multivariate analysis p-value	Multivariate analysis OR (CI 95%)	Multivariate analysis p-value
Patient’s clinical situation	Patient with post–upper abdominal surgery with mesocolic perforation (vs patient with decompensated cirrhosis)	3.53 (1.56; 7.89)	0.002*	3.92 (1.66; 9.28)	0.002*	4.04 (1.70;9.61)	0.002*
	Patient with lung cancer (vs patient with decompensated cirrhosis)	2.82 (1.30; 6.12)	0.009*	2.09 (0.80; 5.45)	0.13	1.94 (0.74;5.12)	0.18
	Obese patient with diabetes and decompensated renal and/or cardiac failure (vs patient with decompensated cirrhosis)	1.09 (0.55; 2.16)	0.80	0.82 (0.39; 1.74)	0.60	0.82 (0.34;1.74)	0.60
Immunocompromised status	Yes (vs No)	2.04 (1.14; 3.51)	0.001*	1.75 (0.98; 3.14)	0.061	1.77 (0.99; 3.18)	0.056
Central venous catheter	Yes (vs No)	1.94 (1.13; 3.34)	0.017*	–	–	2.28 (1.06; 4.89)	0.035*
Parenteral nutrition	Yes (vs No)	1.84 (0.83; 4.08)	0.035*	2.00 (1.08; 3.70)	0.027*	2.01 (1.09; 3.72)	0.026*
Length of ICU stay	>15−20 days (vs <15−20 days)	1.15 (0.68; 1.96)	0.60	1.16 (0.64; 2.08)	0.63	1.15 (0.64; 2.07)	0.65
SOFA score	>8 (vs ≤8)	1.87 (1.08; 3.22)	0.025*	2.04 (0.97; 4.30)	0.061	–	–
Antibiotic treatment duration	>48 h (versus ≤48 h)	1.34 (0.82; 2.33)	0.25	1.90 (1.03; 3.52)	0.041*	1.95 (1.05; 3.62)	0.035*
ICU experience	<10 years (vs >20 years)	1.85 (0.97; 3.51)	0.061	1.99 (0.96; 4.11)	0.063	2.00 (0.97; 4.14)	0.062
	10−20 years (vs >20 years)	2.42 (1.16; 5.05)	0.019*	2.79 (1.23; 6.29)	0.014*	2.70 (1.19; 6.10)	0.018

Abbreviations: CI, confidence interval; OR, odds ratio; SOFA, sequential organ failure assessment.

*Statistical significance level set at 5%.

Catheter and SOFA are linearly dependent by design and cannot be included in the model simultaneously: 100% of patients with a central catheter have a SOFA score >8, and 100% of patients without a central catheter have a SOFA score ≤8 (arbitrary decision to retain catheter instead of SOFA). Two separate multivariate models are presented side-by-side: Model 1 (Primary Analysis) controls for baseline patient severity using the SOFA score (omitting the catheter), while Model 2 (Sensitivity Analysis) substitutes the SOFA score with the presence of a central venous catheter.

Therapeutic approaches according to the level of mycological confirmation of the invasive candidiasis

Among the 204 cases tested for invasive candidiasis, empirical treatment was the predominant approach overall while awaiting test results, used in 129 (63.2%) cases, with similar distribution across clinical situations. In contrast, for the remaining 31.4% of cases, clinicians deferred antifungal therapy until mycological documentation was available. This targeted approach was executed in two distinct stages depending on laboratory reporting timelines: 33 (16.2%) cases initiated targeted curative treatment immediately upon microbiological confirmation (defined as the initial laboratory detection of yeast in blood or deep sampling), while 31 (15.2%) cases received curative treatment after identification of the Candida species (species‑level determination and susceptibility testing). Treatment based on antifungal susceptibility profiles remained infrequent at 5.4% (n = 11).

Among the 68 cases without confirmed candidiasis, the most common therapeutic approach was also empirical treatment, used in 47 (69.1%) cases prior to receiving diagnostic test results. Once invasive candidiasis was ruled out, empirical treatment was discontinued in most cases (74.5%; n = 35/47), particularly in patients with decompensated cirrhosis (100%; n = 10) and obese patients with diabetes and decompensated renal and/or cardiac failure (83.3%; n = 5/6). A smaller subset discontinued therapy after 8 days (10.6%; n = 5/47) or after 14 days (6.4%; n = 3/47), with a few reporting other reasons (8.5%; n = 4/47) for discontinuation (Table 2).

Table 2.

Discontinuation of empirical treatment if candidiasis is not documented (N = 47).

	Clinical situation 1 (N = 18)	Clinical situation 2 (N = 13)	Clinical situation 3 (N = 10)	Clinical situation 4 (N = 6)	Total (N = 47)
After 8 days of treatment	4 (22.2%)	1 (7.7%)	0 (0.0%)	0 (0.0%)	5 (10.6%)
After 14 days of treatment	2 (11.1%)	1 (7.7%)	0 (0.0%)	0 (0.0%)	3 (6.4%)
Discontinuation as soon as negative results are obtained	11 (61.1%)	9 (69.2%)	10 (100.0%)	5 (83.3%)	35 (74.5%)
Other	1 (5.6%)	2 (15.4%)	0 (0.0%)	1 (16.7%)	4 (8.5%)

Clinical situation 1: Patient with post–upper abdominal surgery with mesocolic perforation; Clinical situation 2: Patient with lung cancer; Clinical situation 3: Patient with decompensated cirrhosis; Clinical situation 4: Obese patient with diabetes and decompensated renal and/or cardiac failure.

First-line treatment and reasons for treatment choice in confirmed candidiasis

Once the diagnosis of invasive candidiasis was mycologically confirmed, echinocandins were the first-line treatment choice (88.2%; n = 120/136), followed by azoles (8.1%; n = 11/136), and amphotericin B (3.7%; n = 5). Among cases treated with echinocandins, caspofungin was widely preferred (93.3%; n = 112/120), with micafungin (5.8%; n = 7/120) and anidulafungin (0.8%; n = 1/120) used less frequently. For azoles, fluconazole was the most selected agent (63.6%; n = 7/11), followed by voriconazole (27.3%; n = 3/11) and itraconazole (9.1%; n = 1/11).

The decision to initiate treatment with echinocandins was primarily guided by guidelines (78.3%; n = 94), the pharmacokinetic/pharmacodynamic (PK/PD) profile (38.3%; n = 46), SOFA score (29.2%; n = 35), patient status (25.8%; n = 31), and presence of immunosuppression (20%; n = 24). Other factors influencing this treatment choice included the length of ICU stay (15.0%; n = 18), presence of a central catheter (14.2%; n = 17), the use of parenteral nutrition (11.7%; n = 14), and the duration of prior antibiotic treatment (10.0%; n = 12). Dosage adjustment in case of hepatic impairment was reported in 12 (40%) cases treated with echinocandin. Also, dosage adjustments were most reported in case of obesity (BMI of 34 Kg/m²; 64.3%; n = 18), followed by both obesity and renal failure (21.4%; n = 6), and renal failure alone (14.3%; n = 4).

When azoles were chosen as first-line therapy, patient status (72.7%; n = 8), local Candida epidemiology (54.5%; n = 6) and the PK/PD profile (36.4%; n = 4) were key considerations for treatment initiation. For amphotericin B, the patient’s initial status (60.0%; n = 3) and immunosuppression (60.0%; n = 3) were the main drivers of the decision.

Due to the predominance of echinocandins as the first-line treatment in this survey (88.2%; n = 120/136), variability across antifungal groups was insufficient to support statistical analysis. As a result, no comparative assessment of factors associated with initial treatment selection could be conducted.

Secondary objectives

Diagnostic approaches and use of scoring systems

Among the 204 cases tested for invasive candidiasis, blood culture was performed in almost all cases (96.6%; n = 197). Serum β-D-glucan testing was the most frequently used biomarker assay, performed in 164 (80.4%) cases. Deep sampling was chosen for 102 (50%) cases, predominantly in patients with abdominal surgery with upper mesocolic perforation (62.7%; n = 37/54), and patients with decompensated cirrhosis (57.8%; n = 26/46) compared with 37.5%–40.9% in the other clinical situations. Clinical scoring systems were used to guide diagnosis or management in half of the cases (51%; n = 104/204). The Candida score was the most used tool (63.5%; n = 66), especially in non-surgical clinical situations (patients with lung cancer, patients with decompensated cirrhosis, and obese patient with diabetes and decompensated renal and/or cardiac failure). The Candida colonization index was used in 39 (37.5%) out of 204 cases. The peritonitis score, relevant mainly to surgical cases, was used in 20 (60.6%) patients with post–upper abdominal surgery with mesocolic perforation, but it was rarely or not used in other clinical situations. Additional tests included antigen–antibody assays for Candida or mannan detection in 79 (38.7%) cases and multiplex PCR on blood cultures in 34 (16.7%) cases (Supplementary Table S5).

Treatment monitoring

Among the 136 simulated cases with confirmed invasive candidiasis, adaptation or de-escalation of antifungal therapy was reported in approximately half of the cases (51.7%; n = 66) when the clinical and/or microbiological evolution was favorable. When isolate susceptibility to azoles was confirmed, stepdown to an azole was considered in 87.5% (n = 105/120) of cases initially started on an echinocandin and in 60% (n = 3/5) of cases initially started on amphotericin B formulations (Fig. 1). In total, when the patient was clinically stable without severity criteria, de-escalation was considered in 38 (30.4%) cases. De-escalation in accordance with guidelines was reported in 34 (27.2%) cases. The timing of de-escalation most commonly occurred after 5 days of treatment (51.2%; n = 64), followed by 3 days (28%; n = 35) and other timings (20.8%; n = 26).

Fig. 1.

Therapeutic adaptation and de-escalation pathways across initial antifungal classes (Echinocandins and Amphotericin B formulations).

The de-escalation patterns were collected via a multiple-choice question. Data are expressed as percentages within each initial treatment group. Stepdown therapy to azoles was defined as de-escalation or class transition guided by confirmed microbiological susceptibility. Note: Given the small sample sizes in the azole and amphotericin B cohorts, proportions for these groups should be interpreted descriptively.

As part of the additional diagnostic workup for all cases with confirmed candidemia (N = 136), a fundus examination was performed for 126 (92.6%) cases, and echocardiography for 121 (89%) cases. Removal of the central venous catheter for microbiological testing was performed for 67 (93.1%) cases when applicable. Other procedures related to patient management were reported in 30 (22.1%) cases. These included therapeutic drug monitoring, computed tomography, Doppler ultrasound, or, no additional tests. Among cases treated with echinocandins (N = 120), fundus examination was performed in 113 (94.2%) cases, and echocardiography in 110 (91.7%) cases. Removal of the central venous catheter for microbiological testing was performed for 60 (96.8%) cases when applicable. Additional management-related procedures were documented in 28 (23.3%) cases.

Strategies in case of lack of clinical improvement

Among high-risk ICU patients with confirmed invasive candidiasis (N = 136), lack of clinical improvement prompted further diagnostic reassessment and therapeutic modification. Antifungal susceptibility testing was performed in 107 (78.7%) cases, most frequently among cases treated with echinocandin (82.5%; n = 99/120), followed by azoles (54.5%; n = 6/11), and amphotericin B (40.0%; n = 2/5). Blood sampling was done in 101 (74.3%) out of 136 cases, with slightly higher frequency among patients treated with echinocandins (75.8%; n = 91/120) compared to those receiving amphotericin B (60.0%; n = 3/5) or azoles (63.6%; n = 7/11). Deep sampling was performed in 67 (49.3%) out of 136 cases, with similar rates across antifungal groups. Serum β-D-glucan testing was used in 56 (41.2%) cases (Supplementary Table S6).

A new antifungal was prescribed in 46 (33.8%) out of 136 cases, with higher rates among cases initially treated with azoles (45.5%; n = 5/11) compared to those receiving echinocandins (33.3%; n = 40/120) or amphotericin B (20.0%; n = 1/5). When a new antifungal was prescribed, a different antifungal was initiated without waiting for test results in 28 (60.9%) out of 46 cases, while the initial antifungal therapy was continued pending diagnostic confirmation in 18 (39.1%) cases. Among the 40 echinocandin-treated cases where a new antifungal was initiated, 25 (62.5%) cases involved the immediate prescription of another antifungal without waiting for the results of diagnostic tests. In contrast, among the 5 cases initially treated with azoles and who received a new antifungal, 3 (60.0%) cases continued the initial azole therapy while awaiting test results (Supplementary Table S6).

Discussion

The INSPIRE survey shed light on the current practices of French ICU physicians in managing patients at high risk of invasive candidiasis, based on simulated clinical scenarios. The results highlight key diagnostic and therapeutic decision-making processes, and the criteria influencing antifungal choices.

The distribution of the 283 clinical situations was balanced across the four patient types, underscoring the heterogeneity of the cases managed. In general, the diagnosis and management of invasive fungal infections (IFIs) are similar regardless of the patient’s clinical situation, with only minor adaptations, primarily related to the diagnostic process and therapeutic approach.

Diagnostic strategies and scoring systems

The high rate of blood culture requests (96.6%) across all clinical situations reflects strong awareness among respondents of its role as the gold standard in the diagnosis of invasive candidiasis, despite its limited sensitivity (around 50%) [16]. Serum β-D-glucan was the most frequently used biomarker assay (80.4%), aligning with the finding of the French national survey conducted in 2024 among senior ICU physicians from 221 French ICUs, where 75% reported using this test [10]. Serum β-D-glucan appears to be a widely favored diagnostic tool, despite its low specificity and modest level of supporting evidence in the literature [17]. This common use is likely due to its perceived utility as a non-invasive adjunct in high-risk patients [18]. By contrast, multiplex PCR was modestly used (16.7%) despite its good diagnostic performance [16]. This suggests that molecular diagnostics have yet to be integrated into routine ICU workflows, likely due to availability, cost, or turnaround time constraints. This finding was somewhat unexpected, given that most respondents were practicing in university hospitals (CHUs), where such platforms should generally be accessible. In practice, the diagnostic workflow is largely determined at the laboratory level, where every first positive blood culture should be followed by multiplex PCR testing. The low rate of reported use may indicate a gap in communication since ICU physicians may benefit from these assays without necessarily labeling them as “multiplex PCR,” which could explain why they did not mention them explicitly in the survey. Overall, the integration of multiplex PCR into ICU practice remains limited, indicating potential for improvement in both hospital availability and physician awareness. This challenge in translating advanced diagnostic and management algorithms into uniform clinical practice mirrors current European data; for example, the multicenter German CAMEO study [19] similarly highlighted notable variations in the implementation of standard care bundles and diagnostic tracking at the institutional level.

Only half of respondents reported using clinical scoring systems, most commonly the Candida score (63.5%), especially in non-surgical clinical situations. So, despite their utility in stratifying risk [20], clinical tools are selectively used, and clinical judgment remains predominant in antifungal decision-making. Particularly, deep sampling was performed more often in surgical (62.7%) and cirrhotic (57.8%) clinical situations compared with others, highlighting the influence of underlying pathology on diagnostic aggressiveness.

Therapeutic management and antifungal selection

Echinocandins were the most prescribed first-line antifungal therapy (88.2%), reflecting both guideline adherence and their perceived safety and efficacy in critically ill or unstable patients [9]. Amphotericin B (<5%) and azoles (8.1%) were less frequently used since they are mainly reserved to cases with intolerance or confirmed susceptibility [9].

Antifungal therapy was adapted once microbiological results or clinical evolution became available. De-escalation was reported in half of the cases (52.8%), primarily when clinical or mycological improvement was documented. This proportion seems higher than antifungal de-escalation rates of 20% described in previous European studies in critically ill patients admitted to ICUs [21,22], emphasizing growing attention to antifungal stewardship. Among cases treated with echinocandin, 87.5% of de-escalations to azoles were guided by susceptibility results, confirming the essential role of mycological confirmation in driving antifungal optimization. De-escalation was typically implemented after 5 days of treatment, consistent with the timeframe needed for clinical response evaluation and culture results [22].

Patient clinical situation and context-specific practices

Variations in diagnostic and therapeutic approaches to invasive candidiasis in the ICU setting across patients’ clinical situations were particularly informative. In patients with abdominal surgery and upper mesocolic perforation, deep sampling and peritonitis scoring were more frequent, reflecting a higher level of diagnostic intervention due to surgical complexity and intra-abdominal contamination risk. Empirical antifungal therapy was more frequently initiated in this clinical situation, aligning with a high pre-test probability of invasive candidiasis and diagnostic urgency. In contrast, non-surgical (cancer and obese/diabetic) clinical situations showed greater reliance on non-invasive tools such as the Candida score, colonization index, and antigen/antibody assays. This practice reflects both a lower use of invasive procedures and an emphasis on risk stratification through validated scoring systems. Also, therapeutic decisions in these clinical situations were more conservative, with lower rates of empirical antifungal initiation, especially in obese patients with renal failure, where PK/PD considerations and the need for dose adjustment influence drug selection. Among patients with decompensated cirrhosis, diagnostic caution was evident, with moderate use of deep sampling and high reliance on colonization indices and antigen/antibody testing. Despite diagnostic uncertainty, empirical antifungal treatment was initiated less frequently compared to surgical cases, potentially due to concerns about antifungal toxicity and hepatic metabolism in this clinical situation.

These trends demonstrate that antifungal management in the ICU remains highly contextual. Although guideline adherence appears strong overall, decisions are influenced by patient-specific factors, perceived infection risk, and diagnostic confidence rather than rigid algorithmic adherence.

Reassessment and follow-up investigations

In the subgroup of 136 cases with confirmed invasive candidiasis, additional diagnostic workup and treatment adjustment were common, reflecting appropriate clinical vigilance. Additional diagnostic workup such as fundus examination and echocardiography were used in over 90% of cases, reflecting excellent alignment with recommended testing for invasive candidiasis [9].

Removal of the central venous catheter, performed in 93.1% of applicable cases, further demonstrates adherence to guidelines [9]. However, approximately one-third of the respondents reported prescribing a new antifungal in the absence of microbiological or clinical improvement, and in 60.9% of these cases, therapy was modified before diagnostic confirmation. This gap underscores the ongoing challenge between the need for rapid therapeutic response and the principles of antifungal stewardship.

Strengths and limitations

Certain strength points of the INSPIRE survey should be acknowledged. First, the use of standardized case vignettes allowed for the simulation of realistic, clinically relevant ICU scenarios. Second, the generation of medical and surgical clinical situations, and both documented and probabilistic infections, enhanced the external validity of findings by reflecting the spectrum of invasive candidiasis presentations encountered in the ICU setting. Finally, the vignette-based survey design provided unique insights into the physicians’ diagnostic and therapeutic decisions, extending beyond mere guideline adherence.

However, the survey has some limitations. It was subject to non-response bias, as participation was voluntary and the number of respondents may not fully represent the entire ICU physician population in France. The response rate was moderate (46.9%), and the sample was not fully representative, with an overrepresentation of university hospitals and possible clustering of respondents, which may limit external validity and the generalizability of findings to smaller or non-academic centers. Also, there was a selection bias because respondents were frequently affiliated with university or referral hospitals, where invasive fungal infections are more commonly managed. While our recruitment strategy explicitly targeted all hospital categories to capture a broad mix of community and regional settings, the voluntary nature of the survey ultimately favored these highly specialized academic centers. The recruitment strategy (email invitations and congress distribution) may have further introduced self‑selection bias, favoring more engaged or academically inclined clinicians. A desirability bias cannot be excluded, as physicians may have provided answers aligned with perceived best practices rather than their actual real-life practice.

Responses to questionnaires do not necessarily reflect bedside practice. Although the vignette design was validated, simulated scenarios cannot fully replicate the complexity of real‑life ICU decision‑making. In our simulation, diagnostic testing decisions were strongly driven by the strategic accumulation of predefined fungal risk factors under an overarching clinical baseline of suspected septic shock. However, because the study design intentionally overrepresented “confirmed candidiasis” cases to support a robust multivariate regression, and because participants were naturally aware of the survey’s main focus on mycology, an inherent salience bias likely shifted physician thresholds toward heightened diagnostic vigilance. This may have inflated suspicion of fungal infection compared to routine practice. Respondents’ prior knowledge that the questionnaire focused on fungal infections and antifungal therapy may also have biased responses, leading to an overestimation of both diagnostic testing and empirical treatment. The reported rate of mycological diagnostic testing (72%) and the high frequency of empirical antifungal therapy across scenarios appear higher than what is typically observed in clinical practice.

The restriction to four selected high‑risk profiles further limits scope. These scenarios (post‑upper abdominal surgery with mesocolic perforation, lung cancer, decompensated cirrhosis, and obese diabetic patients with renal/cardiac failure) were chosen based on published evidence and expert consensus to represent patient groups at high risk for invasive fungal infections. Importantly, patient populations with active hematological aplasia (neutropenia) or severe necrotizing pancreatitis were intentionally excluded during the study design phase. In daily ICU workflows, these specific cohorts carry an exceptionally high, well-established baseline risk where fungal investigations and empirical treatments are nearly automatic and highly institutionalized. Including them would have created an artificial ceiling effect, thereby obscuring the nuanced risk-factor thresholds and practice clinical reasoning we aimed to capture in more ambiguous clinical settings. Nevertheless, results might have differed had these other ICU populations with more algorithmic management pathways been included.

Another limitation is that the predominance of echinocandin use restricted comparative statistical analysis across antifungal groups. This high frequency of echinocandin selection (88.2%) and diagnostic vigilance observed among respondents directly mirrors current international clinical guidelines for the management of high-risk ICU patients. Thus, these patterns reflect robust adherence to established global recommendations and clinical evidence rather than external sponsor influence. Furthermore, the sequential unveiling of the mycological confirmation status during the chatbot interaction meant that early diagnostic choices were strictly probabilistic, whereas subsequent therapeutic adaptations were guided by definitive results. Predefined time points for de‑escalation (e.g., after 3 or 5 days) were provided to ensure comparability across responses, but the use of these suggested time points may have biased answers and does not fully reflect bedside practice. Similarly, the phrasing of the treatment choice question (“select up to 3 reasons and rank them”) and the inclusion of the option “Guidelines (if echinocandin chosen)” may have oriented respondents toward selecting guidelines when echinocandins were chosen, thereby reinforcing guideline adherence as a reported driver of therapy choice. Furthermore, the questionnaire design may have oriented physicians toward specific responses, as most risk factors presented were classical fungal risk factors. The exclusion of unrelated factors (e.g., hypertension) limited the ability to assess whether physicians selectively responded to true fungal risk factors versus non‑specific comorbidities.

Finally, some topics were not specifically addressed in the survey, such as the detailed use of imaging modalities which could have provided additional insights into diagnostic pathways. Additionally, the clinical vignettes did not specify the exact mycological diagnostic modalities (such as standard blood cultures versus non-culture-based assays like β-D-glucan or PCR) that physicians were expected to consider. This was a deliberate design choice intended to prevent over-complexifying the simulation matrix and preserve the global response rate; however, it limits our ability to evaluate how specific diagnostic technologies with varying performance profiles and laboratory turnaround times dynamically influence real-world empirical treatment and de-escalation timelines at the bedside. Similarly, when evaluating the extension workup for documented candidemia, the survey questionnaire did not differentiate between TTE and TEE, leaving the specific modality to the interpretation of the respondents. While our primary goal was to evaluate general clinician awareness regarding the necessity of systemic screening, we acknowledge that the lack of granularity prevents us from capturing practice variations or structural barriers related to the deployment of TEE versus TTE in daily ICU workflows.

Conclusions

The INSPIRE survey identified patient severity (SOFA score), immunosuppression status, and the presence of a central catheter, as the main factors guiding antifungal treatment decisions. These parameters significantly influence physicians' choices between empirical initiation, specific antifungal classes, and eventual treatment adaptation.

The presence of high-risk profiles, particularly those with poor clinical status, immunodeficiency, or requiring dosage adjustments due to obesity or organ dysfunction, underscores ongoing challenges and unmet needs in the management of invasive candidiasis in the ICU setting. These complex scenarios are often the most influential in motivating physicians to explore alternative antifungal strategies. Also, this survey emphasizes the complexity of antifungal management in the ICU setting, where early therapeutic intervention must be balanced; once diagnosis of invasive candidiasis is mycologically confirmed, de-escalation needs to be done cautiously. The case-vignette design revealed that, even in virtual scenarios, clinical reasoning remains individualized and influenced by the patient’s underlying condition.

Given the rarity of invasive candidiasis, there is a clear need to strengthen physician training to ensure familiarity with evolving diagnostic methods and antifungal stewardship principle. Continued education and decision-support tools may further promote harmonization of diagnostic and therapeutic strategies. Moreover, broader access to rapid diagnostics could facilitate earlier targeted therapy and reduce reliance on empirical treatment. By identifying the drivers of clinical decision-making, this survey provides valuable insights on optimizing the care of critically ill patients at risk of IFIs.

Authors' contributions

All authors (including SG-S and KG) met all ICMJE criteria for authorship. All authors contributed substantially to the conception, design, protocol elaboration, and critical interpretation of the data; participated in drafting and revising the paper for intellectual content; gave final approval of the version to be published; and agreed to be accountable for all aspects of the work.

Consent for publication

Not applicable.

Funding

This study was funded by MUNDIPHARMA. MUNDIPHARMA employees (SG-S, KG) participated in the study design, protocol coordination, and data interpretation in their capacity as co-investigators utilizing independent scientific judgment. The sponsor did not influence their scientific opinions or decisions. FAST4 (France) provided editorial and medical writing assistance in accordance with Good Publication Practice (GPP 2022) and the ICMJE requirements, this assistance was funded by MUNDIPHARMA. The authors retained full editorial control over the content, final interpretation of the data, and decision to submit for publication. There was no financial reward associated with writing the paper. The views and opinions expressed are those of the authors.

Ethics approval and consent to participate

Given the absence of real patient data and the non-interventional nature of the survey, ethics committee approval was not needed under European law. ICU physicians who responded to the survey were informed of its purpose and of their ability to withdraw from the questionnaire at any time without consequences, thus ensuring their autonomy and comfort throughout the process. Neither this information nor any other confidential information that could identify a responding ICU physician was retained for the analysis of the results. The protection of individual data was rigorously ensured in accordance with the guidelines of the Information Security and Confidentiality Policy (PSI).

Availability of data and material

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Data availability

The data that has been used is confidential.

Data will be made available on request.

Declaration of competing interest

Hervé Dupont reports consultancies from Mundipharma, Pfizer, and Shionogi; lectures from Pfizer, Mundipharma, and Shionogi. He declares no other conflicts of interest.

Jean-Paul Mira reports consultancies from Mundipharma, Viatris, and Pfizer; lectures from Mundipharma and Viatris. He declares no other conflicts of interest.

Muriel Cornet reports consultancies from Mundipharma and BioMérieux. She has received travel funding from Mundipharma, Gilead Sciences, and Pfizer.

Sandrine Girardot-Seguin and Karine Gouhier are employed at MUNDIPHARMA.

Nour El Khatib is employed at Fast4.

Emmanuel Novy reports consultancies from Mundipharma; lectures from Mundipharma and Gilead; participation in clinical studies with Mundipharma; and research grants from Mundipharma and MSD. He declares no other conflicts of interest.

HD reports consultancies from Mundipharma, Pfizer, and Shionogi; lectures from Pfizer, Mundipharma, and Shionogi. He declares no other conflicts of interest.

J-PM reports consultancies from Mundipharma, Viatris, and Pfizer; lectures from Mundipharma and Viatris. He declares no other conflicts of interest.

MC reports consultancies from Mundipharma and BioMérieux. She has received travel funding from Mundipharma, Gilead Sciences, and Pfizer.

SG-S and KG are employed at MUNDIPHARMA.

NEK is employed at Fast4.

EN reports consultancies from Mundipharma; lectures from Mundipharma and Gilead; participation in clinical studies with Mundipharma; and research grants from Mundipharma and MSD. He declares no other conflicts of interest.

While the study was funded by Mundipharma, the survey content was developed collaboratively with independent academic experts. The choice of scenarios and risk factors was guided by established literature and expert consensus, rather than sponsor input, to ensure scientific validity and relevance. Final interpretation of the results remains the sole responsibility of the authors.

Appendix A. Supplementary data

The following are Supplementary data to this article: