A Predictive Model for Diagnosis of Acute Invasive Fungal Rhinosinusitis Among High-Risk Patients

Abstract

Background

Acute invasive fungal rhinosinusitis (AIFR) is a life-threatening disease mainly affecting immunocompromised patients. Early detection is therefore key to improving patient survival. To date, there are still no standard clinical criteria for AIFR diagnosis.

Objective

This study develops a predictive model that utilizes clinical presentation and computed tomography (CT) findings to diagnose AIFR.

Methods

A retrospective cohort study was conducted on patients with high risk for AIFR at King Chulalongkorn Memorial Hospital over the past 15 years (2008-2022). We constructed several multivariate logistic regression models for AIFR diagnosis based on different subsets of variables from 3 categories: signs/symptoms, endoscopy, and CT imaging.

Results

There were 67 AIFR-positive patients and 68 AIFR-negative patients. Combining variables from 3 categories, a 6-variable model (fever, visual loss, mucosal discoloration, crusting, mucosal loss of contrast, retroantral fat stranding) achieved the highest area under the receiver operating characteristic curve of 0.8900 (74.63% sensitivity, 89.71% specificity).

Conclusions

We proposed predictive models for AIFR diagnosis in high-risk patients using clinical variables. The models can be used to guide the decision for further management such as biopsy or surgical intervention.