Fungal infections are an expanding public health challenge. This single-center retrospective study evaluated the clinical and microbiological epidemiology of Candida spp. infections at a high-complexity referral hospital in Ibagué, Tolima, Colombia, over a ten-year period (2014–2024). The primary objectives were to describe recorded clinical candidiasis episodes, species-level isolation patterns, antifungal susceptibility where available, and exploratory associations with candidemia and invasive candidiasis.
The study used two institutional data sources from the same hospital: (A) a clinical/administrative database of recorded candidiasis episodes and (B) a microbiology laboratory database (WHONET-format exports). Data were collected retrospectively from routine care records spanning 2014–2024. The databases originated from separate institutional workflows and were analyzed independently; they were not linked record by record.
Eligible records included patients with a clinical diagnosis of candidiasis and/or laboratory identification of Candida spp. The unit of analysis was an infection episode in the clinical/administrative database and an isolate record in the laboratory database. Records lacking clinical or microbiological corroboration, contaminated samples, or missing key variables were excluded. Clinical cases were categorized operationally as mucocutaneous/localized non-invasive candidiasis, candidemia, deep-seated non-candidemic invasive candidiasis (organ-specific without documented candidemia), or other/unspecified candidiasis. EORTC/MSG criteria were applied for suspected invasive fungal disease when appropriate.
Microbiology processing followed routine laboratory practice. Identification methods included conventional culture and BD Phoenix (2014–2021) and VITEK 2 Compact (2022–2024) with the YST card. Antifungal susceptibility testing during the VITEK period used the AST-YS08 card and WHONET MIC exports for amphotericin B, fluconazole, voriconazole, caspofungin, micafungin, and 5-flucytosine (5-FC). MICs were interpreted using CLSI breakpoints where available; 5-FC was assessed using EUCAST ECOFFs and reported as wild-type (WT) or non-wild-type (NWT). Isolates lacking MICs for a given agent were excluded from that agent-specific analysis.
Data cleaning and preparation were performed in Excel; analyses used R 4.4.2. Categorical variables are reported as counts and percentages; continuous variables as medians and interquartile ranges (IQR). Group comparisons used χ² and nonparametric tests. Multiple correspondence analysis (MCA) summarized co-occurrence among categorical variables. Logistic regression estimated odds ratios (ORs) with 95% confidence intervals (CIs); Firth-penalized regression was applied for candidemia due to low event frequency. Age and length-of-stay categories were predefined. Regression models used complete-case analysis and were exploratory and data-source specific. The institutional ethics committee approved the research and waived informed consent for use of de-identified retrospective data.
Across 2014–2024 the hospital administrative denominator included 243,013 institutional care records. The clinical database identified 987 episodes compatible with candidiasis, corresponding to 0.41% of institutional records (4.06 per 1,000). Annual recorded rates ranged from 0.22% to 0.61%, peaking in 2018 (n = 147). In the microbiology database there were 776 fungal/yeast isolate records; 314 were confirmed Candida isolates (40.46% of fungal isolates). Cumulative confirmed isolates rose over time.
Mucocutaneous disease predominated among the 987 clinically recorded episodes: vulvovaginal candidiasis accounted for 50.66% (n = 500) and oropharyngeal candidiasis 24.32% (n = 240). Unspecified candidiasis represented 11.96% (n = 118), deep-seated non-candidemic invasive candidiasis 8.21% (n = 81), and candidemia 2.03% (n = 20). Women comprised 77.81% of episodes (female:male ratio 3.5:1), largely reflecting vulvovaginal disease. Median age was 29 years (IQR 22–53); invasive forms had higher median ages (candidemia median 38 years, deep-seated invasive 46 years). Length of stay was longer for candidemia and deep-seated invasive infections and was substantially associated with care setting, with longer stays in ICU and general wards.
Among 314 microbiologically confirmed Candida isolates, Candida albicans was the most frequent species (58.92%, n = 185). Non-albicans species comprised 41.08% (n = 129), including C. parapsilosis (16.56%, n = 52), C. tropicalis (12.10%, n = 38), and Nakaseomyces glabratus (6.37%, n = 20). Less common species included C. dubliniensis, Wickerhamomyces anomalus, and a single C. auris isolate. Specimen sources were concentrated in superficial tissues/cutaneous sites (36.62%, n = 115) and urine (33.44%, n = 105). Blood isolates constituted 7.32% (n = 23); bloodstream etiology was heterogeneous (C. albicans 39.13%, C. parapsilosis 34.78%, W. anomalus 13.04%, N. glabratus 8.70%, C. auris 4.35%). Non-albicans proportions increased over time, with approximately 41.2% in 2023 and 43.0% in 2024. By service, C. albicans predominated in outpatient and lower-complexity settings, while non-albicans species were more prevalent in critical care areas; C. parapsilosis was the only species reported from the neonatal ICU in this dataset.
MIC-based susceptibility data were available for 196 of 314 isolates (62.4%) during 2022–2024 when structured VITEK MIC exports were available. Overall, echinocandin activity remained high across tested isolates, while azole susceptibility showed greater variability. Results for 5-flucytosine were analyzed and reported using EUCAST ECOFF methodology and presented separately to avoid overinterpretation; isolates lacking MICs for a given agent were excluded from agent-specific analyses.
Multiple correspondence analysis (MCA) of selected categorical variables showed separation between invasive and mucocutaneous presentations: candidemia clustered with higher-complexity care, invasive procedures/devices, and blood samples, whereas mucocutaneous presentations clustered with lower-complexity profiles. Penalized multivariable logistic regression (clinical/administrative data) identified recorded candidemia associations with invasive devices (OR 5.54, 95% CI 2.01–15.64), recent surgery (OR 7.11, 95% CI 1.20–30.88), and presence of tumor (OR 19.88, 95% CI 3.14–97.51). Models were exploratory, source-specific, and not interpretable as linked patient-level clinical–microbiological risk models because databases were not record-linked.
This single-center retrospective analysis documents a sustained recorded burden of candidiasis over 2014–2024 with substantial non-albicans diversity and heterogeneous antifungal susceptibility patterns. Findings support continued local surveillance, species-level identification, and isolate-level susceptibility testing to inform empiric therapy and detect emerging pathogens such as C. auris. The authors deposited de-identified datasets and analysis code in Zenodo and GitHub as reported; funding and competing-interest statements indicate no specific funding and no declared competing interests.