OBJECTIVES: Explore the relevance of using a national administrative hospital discharge database (French acronym PMSI) coding in a large tertiary hospital to identify candidemia and quantify its burden.

METHODS: A retrospective, observational, single-center study used two independent data extraction approaches to collect PMSI data reported for the ICD-10 code B37.7 (candidal sepsis) and chart review data for clinically relevant candidemia cases (confirmed, unconfirmed and possible) during 2018-2020. Analyses compared patient demographics, antifungal treatment and care resources, including length of stay (LoS) in the in the intensive care unit (ICU) and total LoS in the hospital, as well as cost per stay based on (DRG).

RESULTS: The analysis included 128 relevant hospital admissions (123 patients); 101 PMSI-reported B37.7 admissions (97 patients) and a further 27 admissions (26 patients) from chart review (that is without B37.7 coding). Mean (standard deviation [SD]) age was 58.1 (16.7) years and 39.0% were female. Candida albicans was the most prevalent species detected (65.0%), followed by C. glabrata (19.0%). Antifungal therapies included echinocandins plus azoles (34.4%), azoles (23.4%) and echinocandins (21.1%), and 5.5% of admissions received no antifungal treatment. For PMSI-coded admissions median cost per stay was 23,386 Euro for PMSI-reported admissions and 41,716 Euro for admissions identified by chart review.

CONCLUSIONS: Analysis of PMSI-reported and chart review data from a large tertiary hospital identified coding inconsistencies that may cause the cost of hospital admissions due to candidal sepsis to be underestimated. Multi-center research is required to fully evaluate how missing data or coding errors/omissions may impact understanding regarding the incidence, costs and trends in candidal sepsis as well in other invasive Candida infections.