Predicting factors and prediction model for discriminating between fungal infection and bacterial infection in severe microbial keratitis

Abstract

© 2019 Jongkhajornpong et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. A retrospective medical record review including 344 patients who were admitted with severe microbial keratitis at Ramathibodi Hospital, Bangkok, Thailand, from January 2010 to December 2016 was conducted. Causative organisms were identified in 136 patients based on positive culture results, pathological reports and confocal microscopy findings. Eighty-six eyes (63.24%) were bacterial keratitis, while 50 eyes (36.76%) were fungal keratitis. Demographics, clinical history, and clinical findings from slit-lamp examinations were collected. We found statistically significant differences between fungal and bacterial infections in terms of age, occupation, contact lens use, underlying ocular surface diseases, previous ocular surgery, referral status, and duration since onset (p < 0.05). For clinical features, depth of lesions, feathery edge, satellite lesions and presence of endothelial plaque were significantly higher in fungal infection compared to bacterial infection with odds ratios of 2.97 (95%CI 1.43–6.15), 3.92 (95%CI 1.62–9.45), 6.27 (95%CI 2.26–17.41) and 8.00 (95%CI 3.45–18.59), respectively. After multivariate analysis of all factors, there were 7 factors including occupation, history of trauma, duration since onset, depth of lesion, satellite lesions, endothelial plaque and stromal melting that showed statistical significance at p < 0.05. We constructed the prediction model based on these 7 identified factors. The model demonstrated a favorable receiver operating characteristic curve (ROC = 0.79, 95%CI 0.72–0.86) with correct classification, sensitivity and specificity of 81.48%, 70% and 88.24%, respectively at the optimal cut-off point. In conclusion, we propose potential prediction factors and prediction model as an adjunctive tool for clinicians to rapidly differentiate fungal infection from bacterial infection in severe microbial keratitis patients.