An internally validated diagnostic tool for acute invasive fungal sinusitis

Background: Acute invasive fungal sinusitis (AIFS) is a potentially life-threatening diagnosis in immunocompromised patients. Identifying patients who could benefit from evaluation and intervention can be challenging for referring providers and otolaryngologists alike. We aimed to develop and validate an accessible diagnostic tool to estimate the probability of AIFS. Methods: Retrospective chart review from 1999 to 2017 identified all patients evaluated for possible AIFS at a tertiary care center. AIFS was diagnosed by pathologic confirmation of fungal tissue angioinvasion. Stepwise selection and univariate logistic regression were used to screen risk factors for a multivariable predictive model. Model performance was assessed using Tukey's goodness-of-fit test and the area under the receiver operator characteristic curve (AUC). Model coefficients were internally validated using bootstrapping with 1000 iterations. Results: A total of 283 patients (244 negative controls, 39 with AIFS) were included. Risk factors in our final diagnostic model included: fever ≥38°C (log-odds ratio [LOR] 1.72; 95% CI, 0.53 to 2.90), unilateral facial swelling, pain, or erythema (LOR 2.84; 95% CI, 1.46 to 4.23), involvement of the orbit or pterygopalatine fossa on imaging (LOR 3.02; 95% CI, 1.78 to 4.26), and mucosal necrosis seen on endoscopy (LOR 5.52; 95% CI, 3.81 to 7.24), with p < 0.01 for all factors. The model had adequate goodness of fit (p > 0.05) and discrimination (AUC = 0.96). Conclusion: We present an internally validated diagnostic tool to stratify the risk for AIFS. The estimated risk may help determine which patients can be observed with serial nasal endoscopy, which ones could be biopsied, and which ones would benefit from immediate surgical intervention.