Abstract

Objective

Patients with abdominal diagnoses constitute 5 to 10% of all ED presentations. The goal of this study is to identify which of these patients will have a non-concerning diagnosis based on demographic, physical exam and basic laboratory testing.

Methods

Consecutive patients from 7/2013 – 3/2014 discharged with a GI diagnosis who presented to an urban, university affiliated ED were identified. The cohort was split into a derivation set and a validation set. Using univariate and multivariable logistic regression analysis, a risk score was created based on the deviation data, and then tested on the validation data.

Results

There were 8,852 patients with a GI diagnosis during the study period. 7,747 (87.5%) of them had a non-concerning diagnosis. The logistic regression model identified 13 variables that predict a concerning GI diagnosis and created a scoring system ranging 0–20. The area under the ROC was 0.81. When dichotomized at ≥7 vs. <7, the risk score has a sensitivity of 91% (95CI 88–94), specificity 46% (95CI 44–48), PPV 17% (95CI 15–19) and NPV 98% (95CI 97–99).

Conclusion

One can determine with a high degree of certainty, based only on an initial evaluation and screening lab work (excluding radiology) whether a patient who presents with a gastrointestinal related complaint has a non-concerning diagnosis. This model could be utilized as a tool to aid in quality assurance when reviewing patients discharged with GI complaints and with future study, could as a secondary triage instrument in a crowded Emergency Department environment and aid in resource allocation.