Purpose: We constructed a clinical clue-based algorithm to identify the microbiology-positive post-cataract surgery endophthalmitis.
Methods: The Endophthalmitis Infectivity Measurement Algorithm (EIMA) was constructed using presenting Snellen vision (Letter score [LS]) and Inflammation Score (IS, from the cornea, anterior chamber, iris, and vitreous). Retrospective data (70% for training; 30% for testing) was fitted into CHAID (Chi-squared Automatic Interaction Detection). EIMA was validated with prospective data. EIMA-categorized disease severity was weighed against the symptom duration to detect infecting micro-organisms.
Results: EIMA was constructed from 1444 retrospective data. The average LS was 6.03 ± 12.11, median IS was 14 (8-24), and culture positivity was 38%. The accuracy and area under the curve of CHAID were 66.36% and 0.642, respectively. EIMA was validated with 175 prospectively collected data. Microbiology positivity (culture + sequencing) was 58.9%. EIMA sensitivity, specificity, and accuracy against microbiology-positive endophthalmitis were 73.7 (95% confidence interval [CI], 64.19-81.96), 81.9 (95% CI, 71.1-90.02), 77.1 (95% CI, 70.20-83.14), respectively. The positive and negative likelihood ratios were 4.08 (95% CI, 2.46-6.67) and 0.32 (95% CI, 0.22-0.45), respectively. There was higher microbial growth in two days or less than in three- to six-day symptom duration (69.9% vs. 28.2%; P = 0.018) endophthalmitis. Gram-negative infection was higher in two days or less (55.6% vs. 20.2%; P = 0.014), and gram-positive infection was higher in three- to six-day endophthalmitis (62.1% vs. 27.7%; P = 0.027).
Conclusions: EIMA identified microbiology-positive endophthalmitis three-quarters of the time.
Translational relevance: EIMA suggested infectivity and the class of microbial infection could help targeted management of endophthalmitis after cataract surgery.