Validation of ICD-10-CM Diagnosis Codes in Patients With Inflammatory Bowel Disease in the National Veterans Affairs Healthcare System

Introduction: We sought to determine if a previously validated algorithm for using International Classification of Diseases-9 (ICD-9) codes remained accurate when using ICD-10 codes to identify Inflammatory Bowel Diseases (IBD) [Ulcerative colitis (UC) and Crohn’s Disease (CD)] in the Veterans Health Administrative (VHA) database. Methods: We identified IBD patients by automated data collection (ADC) between Oct 1, 2015 and Nov 30, 2021 from the VHA National Corporate Data Warehouse (CDW) by a previously validated ICD-9 algorithm [at least 2 VA encounters with an ICD-10 code for IBD (K51 UC or K50 CD) with at least one code from an outpatient encounter]. Cases that did not have the same IBD code on both encounters were classified as unspecified IBD (UIBD). Manual chart review (MCR) was performed by 2 independent reviewers (PP and LS) by review of clinical notes, endoscopy, imaging and pathology reports. Cases were categorized as IBD, UC, CD or other [indeterminate colitis (IC), unable to determine if IBD or not IBD]. The 3 data sets (ADC and 2 MCR) were assessed for concordance by Cohen’s kappa. Results: We identified 57,823 IBD patients by ADC and randomly selected 250 patients, 134 UC (53.6%), 85 CD (34.0%) and 31 UIBD (12.4%) from VA facilities throughout the country for MCR. Reviewer 1 identified 104 (44.3%) as UC, 72 (30.6%) as CD, 3 (1.3%) as IC, 30 (12.8%) as unable to determine IBD diagnosis and 26 (11.1%) as not IBD. Reviewer 2 identified 99 (54.7%) as UC, 73 (40.3%) as CD, 1 (0.6%) as IC, 7 (3.9%) as unable to determine IBD diagnosis and 1 (0.6%) as not IBD. The algorithm had a positive predictive value (PPV) of 88.9 % to identify overall IBD diagnosis. The algorithm identified patients with CD and UC with high accuracy (concordance of 79.2% for CD and 94% for UC cases between reviewer 1 and ADC, 80.8% for CD and 95.6% for UC between reviewer 2 and ADC). After combining all IBD categories (UC, CD, other) the unweighted Cohen’s kappa between ADC and reviewer 1 (235 cases) was 0.501 (95% CI 0.418-0.584), between ADC and reviewer 2 (181 cases) was 0.735 (95% CI 0.650-0.820) and between reviewer 1 and reviewer 2 was 0.686 (95% CI 0.581-0.791). Conclusion: We validated an ICD-10 based algorithm to identify IBD cases in VHA datasets. It is highly accurate for UC (94%) and CD (80%) with an overall PPV of 88% for the diagnosis of IBD, similar to the ICD-9 based algorithm. Validation of ICD-10 IBD diagnosis codes is essential to study IBD in one of the largest national cohorts of IBD patients.