Parsimonious clinical prediction model for the diagnosis of complicated appendicitis

Abstract

Objective: To develop a logistic regression model that combines clinical and radiological parameters for prediction of complicated appendicitis. Methods: 248 patients with histologically proven uncomplicated (n = 214) and complicated (n = 34) acute appendicitis were analyzed retrospectively. All patients had undergone a presurgical abdominal and/or pelvic computed tomography (CT) scan, assessed by two radiologists. A model using univariate and multivariate logistic regression analyses was developed, and the strength of association between independent predictors and complicated acute appendicitis was evaluated by adjusted odds radio. Clinical parameters were gender, age, anorexia, vomiting, duration of symptoms, right lower abdominal quadrant (RLQ) tenderness, rebound tenderness, body temperature, white blood cell (WBC) count, and neutrophil ratio. Radiological parameters were appendix diameter, appendicolith, caecal wall thickening, mesenteric lymphadenopathy, extraluminal air, abscess, fat stranding, and periappendicular fluid. Results: Four features (body temperature＞37.2 °C, vomiting, appendicolith, and periappendiceal fluid) were included in the logistic regression model, and yielded an area under the curve (AUC) of 0.87 (95% confidence interval (CI), 0.80–0.93), sensitive of 88%, and specificity of 74%. Conclusion: The logistic regression model makes an accurate and simple prediction of complicated appendicitis possible.

Keywords

• Complicated appendicitis
• Predictive factors
• Logistic regression model
• Computed tomography
• Abdominal imaging