Development and validation of algorithms to identify acute diverticulitis

Purpose: The objectives of this study were to develop and validate algorithms to accurately identify patients with diverticulitis using electronic medical records (EMRs).

Methods: Using Kaiser Permanente Southern California's EMRs of adults (≥18 years) with International Classification of Diseases, Clinical Modifications, Ninth Revision diagnosis codes of diverticulitis (562.11, 562.13) between 1 January 2008 and 31 August 2009, we generated random samples for pilot (N = 692) and validation (N = 1502) respectively. Both samples were stratified by inpatient (IP), emergency department (ED), and outpatient (OP) care settings. We developed and validated several algorithms using EMR data on diverticulitis diagnosis code, antibiotics, computed tomography, diverticulosis history, pain medication and/or pain diagnosis, and excluding patients with infections and/or conditions that could mimic diverticulitis. Evidence of diverticulitis was confirmed through manual chart review. Agreement between EMR algorithm and manual chart confirmation was evaluated using sensitivity and positive predictive value (PPV).

Results: Both samples were similar in socio-demographics and clinical symptoms. An algorithm based on diverticulitis diagnosis code with antibiotic prescription dispensed within 7 days of diagnosis date, performed well overall. In the validation sample, sensitivity and PPV were (84.6, 98.2%), (95.8, 98.1%), and (91.8, 82.6%) for OP, ED, and IP, respectively.

Conclusion: Using antibiotic prescriptions to supplement diagnostic codes improved the accuracy of case identification for diverticulitis, but results varied by care setting.