OBJECTIVE To investigate the distribution of Candida and predict the detection trend in the southern and northern campuses of Affiliated Hospital of Jiangnan University, Wuxi, between 2021 and 2024.

METHODS A total of 27 056 patients with common Candida infections from the southern and northern campuses of Affiliated Hospital of Jiangnan University between 2021 and 2024 were selected to analyze the distribution of Candida species and predict the detection trend.

RESULTS Among the 27 056 patients, there were 11 061 males and 15 995 females, aged from 1 to 101 years, with a median age of 68 years.Over the past four years, the top five most commonly detected Candida species in the hospital were Candida albicans, Candida glabrata, Candida tropicalis, Candida parapsilosis and Candida krusei.Statistically significant differences were found in infection characteristics among patients with C.albicans and C.glabrata in terms of gender, age, specimen source and related diseases (P < 0.05).From 2021 to 2024, the number of detected cases declined in 2022 and then rebounded (P < 0.001).Among the detected patients, those aged 70 and above accounted for the highest proportion.Regarding the distribution of specific diseases, the top three were vaginitis (4 176 cases, 15.43%), bacterial pneumonia (1 842 cases, 6.81%) and cancer (1 279 cases, 4.73%).Patients with vaginitis were mainly infected with C.albicans, while patients with bacterial pneumonia were predominantly infected with C.albicans and C.glabrata. The LSTM model showed a good fit to the training set, with an root-mean-square error (RMSE) of 145.03 and an mean absolute error (MAE) of 131.19.Model predictions indicated that the number of patients with Candida infections in the hospital would remain low from Jan.to May 2025, which was basically consistent with actual clinical observations (RMSE=94.71, MAE=84.00).

CONCLUSIONS The common diseases associated with Candida infections in the hospital include vaginitis, bacterial pneumonia and cancer.C.albicans and C.glabrata are the main pathogenic species, and the infection situation is relatively severe.The LSTM model performs well in short-term prediction and dynamic analysis of Candida detection trends.