Development and Validation of a Risk Equation for Appendicitis in Children Presenting With Abdominal Pain

Authors

Gabriela Vazquez-Benitez, HealthPartners Institute, Minneapolis, MNFollow
Elyse Kharbanda, HealthPartners Institute, Minneapolis, MN
Dustin Ballard, Kaiser Permanente Northern California, Oakland, CA
David Vinson, Kaiser Permanente Northern California, Oakland, CA
Richard Bachur, Boston Children's Hospital, Boston, MA
Uli Chettipally, Kaiser Permanente Northern California, Oakland, CA
Mamata Kene, Kaiser Permanente Northern California, Oakland, CA
Patrick O'Connor, HealthPartners Institute, Minneapolis, MN
Steven Dehmer, HealthPartners Institute, Minneapolis, MNFollow
Heidi Ekstrom, HealthPartners Institute, Minneapolis, MN
Peter Dayan, Columbia University, New York, NY
Nathan Kuppermann, University of California, Davis, Sacramento, CA
Anupam Kharbanda, Children's Hospital and Clinics Minnesota, Minneapolis, MN

Publication Date

8-10-2017

Keywords

child and adolescent health, clinical decision-making

Abstract

Background: Appendicitis is a common surgical emergency in children, yet the diagnosis remains challenging. A widely used risk score, the Pediatric Appendicitis Score, is not sufficiently sensitive or specific to be used alone, with many patients classified as “intermediate risk.” Goals of this study were to develop and validate an improved appendicitis risk calculator for children with acute abdominal pain to aid in clinical decision-making.

Methods: We developed our risk calculator using data from a multicenter cohort of children 5 to 18 years old presenting to the emergency department (ED) with acute abdominal pain. We validated the risk calculator in two independent cohorts with similar enrollment criteria. Patient history, physical examination and laboratory data were prospectively recorded by clinicians during ED visits. Appendicitis was confirmed by pathology reports and follow-up telephone survey. Variables evaluated for inclusion in the risk calculator were: age, sex, pain duration, pain with walking, migration of pain, temperature, heart rate, guarding, maximal tenderness in right lower quadrant, white blood cell count, and absolute neutrophil count. A step-wise regression approach was followed to select the best model, using Akaike information criteria and the C-statistic. We forced inclusion of age and sex, including first-order interaction terms. Laboratory values were evaluated for nonlinear associations with appendicitis, and a two-step linear association was included. Validation included calibration and discrimination analyses.

Results: The development sample included 2,423 children, of whom 40% had appendicitis; the validation sample included 1,426, and 35% had appendicitis. Our final risk calculator included sex, age, duration of pain, guarding, migration of pain and absolute neutrophil count. In the validation sample, calibration plot and Hosmer and Lemeshow test (P < 0.0001) showed high calibration and a high discrimination (C-statistic = 0.86). Among 248 (17%) patients in the validation sample at < 5% predicted risk, we observed 4% had appendicitis. Of an additional 318 (22%) patients with predicted risk 5% to < 15%, appendicitis occurred in 8%. Of 48 (3.4%) patients in the validation sample at predicted risk of > 90%, 96% had appendicitis.

Conclusion: Our validated pediatric appendicitis risk calculator can accurately quantify risk for appendicitis and can identify children with acute abdominal pain at high or low risk for appendicitis.

Recommended Citation

Vazquez-Benitez G, Kharbanda E, Ballard D, Vinson D, Bachur R, Chettipally U, Kene M, O'Connor P, Dehmer S, Ekstrom H, Dayan P, Kuppermann N, Kharbanda A. development and validation of a risk equation for appendicitis in children presenting with abdominal pain. J Patient Cent Res Rev. 2017;4:160.