Development and validation of a diagnostic prediction model distinguishing complicated from uncomplicated diverticulitis.

OBJECTIVES: Most diverticulitis patients (80%) who are referred to secondary care have uncomplicated diverticulitis (UD) which is a self-limiting disease and can be treated at home. The aim of this study is to develop a diagnostic model that can safely rule out complicated diverticulitis (CD) based on clinical and laboratory parameters to reduce unnecessary referrals.

METHODS: A retrospective cross-sectional study was performed including all patients who presented at the emergency department with CT-proven diverticulitis. Patient characteristics, clinical signs and laboratory parameters were collected. CD was defined as > Hinchey 1A. Multivariable logistic regression analyses were used to quantify which (combination of) variables were independently related to the presence or absence of CD. A diagnostic prediction model was developed and validated to rule out CD.

RESULTS: A total of 943 patients were included of whom 172 (18%) had CD. The dataset was randomly split into a derivation and validation set. The derivation dataset contained 475 patients of whom 82 (18%) patients had CD. Age, vomiting, generalized abdominal pain, change in bowel habit, abdominal guarding, C-reactive protein and leucocytosis were univariably related to CD. The final validated diagnostic model included abdominal guarding, C-reactive protein and leucocytosis (AUC 0.79 (95% CI 0.73-0.84)). At a CD risk threshold of ≤7.5% this model had a negative predictive value of 96%.

CONCLUSION: This proposed prediction model can safely rule out complicated diverticulitis. Clinical practitioners could cautiously use this model to aid them in the decision whether or not to subject patients to further secondary care diagnostics or treatment.