Predicting complicated appendicitis is possible without the use of sectional imaging—presenting the NoCtApp score

Abstract

Abstract Purpose Appendicitis is among the most common acute conditions treated by general surgery. While uncomplicated appendicitis (UA) can be treated delayed or even non-operatively, complicated appendicitis (CA) is a serious condition with possible long-term morbidity that should be managed with urgent appendectomy. Distinguishing both conditions is usually done with computed tomography. The goal of this study was to develop a model to reliably predict CA with widespread available clinical and laboratory parameters and without the use of sectional imaging. Methods Data from 1132 consecutive patients treated for appendicitis between 2014 and 2021 at a tertiary care hospital were used for analyses. Based on year of treatment, the data was divided into training (n = 696) and validation (n = 436) samples. Using the development sample, candidate predictors for CA—patient age, gender, body mass index (BMI), American Society of Anesthesiologist (ASA) score, duration of symptoms, white blood count (WBC), total bilirubin and C-reactive protein (CRP) on admission and free fluid on ultrasound—were first investigated using univariate logistic regression models and then included in a multivariate model. The final development model was tested on the validation sample. Results In the univariate analysis age, BMI, ASA score, symptom duration, WBC, bilirubin, CRP, and free fluid each were statistically significant predictors of CA (each p < 0.001) while gender was not (p = 0.199). In the multivariate analysis BMI and bilirubin were not predictive and therefore not included in the final development model which was built from 696 patients. The final development model was significant (x2 = 304.075, p < 0.001) with a sensitivity of 61.7% and a specificity of 92.1%. The positive predictive value (PPV) was 80.4% with a negative predictive value (NPV) of 82.0%. The receiver operator characteristic of the final model had an area under the curve of 0.861 (95% confidence interval 0.830–0.891, p < 0.001. We simplified this model to create the NoCtApp score. Patients with a point value of ≤ 2 had a NPV 95.8% for correctly ruling out CA. Conclusions Correctly identifying CA is helpful for optimizing patient treatment when they are diagnosed with appendicitis. Our logistic regression model can aid in correctly distinguishing UA and CA even without utilizing computed tomography.