Key result
ICD-based approach identifies ~86% more cancer cases than traditional registries, especially among rural Veterans.
Why the study?
Veterans in isolated and rural areas face barriers to timely oncology care, making improved cancer case identification essential for healthcare delivery and resource planning where access is limited.
Does an ICD-based analytic cohort identify more veterans with cancer compared to a registry-based oncology cohort?
Cohort (n=990,091)
Yes
Does an ICD-based analytic cohort identify more veterans with cancer compared to a registry-based oncology cohort?
Identifying veterans with cancer using ICD codes from clinical encounters captures a substantially larger population, particularly in rural areas, compared to using the cancer registry alone.
Answers are grounded in this paper and cite their sources.
ICD-based identification may better capture rural veterans with cancer than registries; leaves open effects on care delivery and research validity.
e13745 Background: Veterans living in isolated and rural geographies often face barriers to timely specialty oncology care services, like screening and treatment. Improved case identification of Veterans with cancer is essential to support health care delivery and resource planning in rural geographies where access is limited. Methods: Our team developed a SQL-based approach to create two distinct Veteran cohorts, an oncology cohort (OC) and analytic cohort (AC) using data from 2013-2023. The OC was created using VHA’s Corporate Data Warehouse (CDW) Oncology Raw Domain, which contains VA oncology patients with a confirmed cancer diagnosis and date of diagnosis as recorded in the Central Cancer Registry. The AC was created using a clinician validated list of ICD-9/10 diagnosis code groupings informed by the Surveillance, Epidemiology, and End Results (SEER) list of reportable neoplasms, to ensure alignment with standardized cancer classifications. ICD-9/10 codes were sorted into 12 classifications by prevalence. Diagnosis dates and ICD codes were derived from clinician-validated inpatient and outpatient encounter data. Results: The OC identified 488,273 Veterans with cancer. 908,060 Veteran encounters were included in the AC. A total of 406,242 Veterans were present in both the OC and AC. The AC identified substantially more Veterans not represented in the OC, suggesting a higher than expected upper-bound of cancer cases. Rural Veterans comprised a greater proportion of the AC compared to the OC (36.3% vs. 34.8%), indicating that case identification in rural areas is underestimated. Conclusions: Identifying Veteran’s with cancer using the registry alone may underestimate, especially in rural geographies. Given that coding clinical encounters in oncology has limitations, this study identified differences between traditional methods and our novel approach. By incorporating relevant diagnosis codes and encounter data, a more comprehensive estimate of oncology care demand can be used to determine the allocation of oncology resources. Veteran characteristics of defined cohorts (OC vs. AC). Oncology Cohort (N=488,273) Analytic Cohort (N=908,060) Characteristic n (%) Characteristic n (%) Sex Male 466,010 (95.4%) Male 877,454 (96.6%) Age 18-6465 & older 135,358 (27.7%)352,686 (72.2%) 18-6465 & older 164,708 (18.1%)742,290 (81.7%) Race WhiteBlack/African AmericanNHPIAsianOther 353,158 (72.3%)94,611 (19.4%)3,621 (0.7%)2,185 (0.4%)12,713 (2.6%) WhiteBlack/African AmericanNHPIAsianOther 679,842 (74.9%)129,261 (14.2%)7,190 (0.8%)5,300 (0.6%)49,620 (5.5%) Rurality Rural 169,903 (34.8%) Rural 329,888 (36.3%)
No takes yet. Share an insight, caveat, or question.
Phibbs et al. (2026) conducted a cohort in Cancer (n=990,091). ICD-9/10 diagnosis code groupings (Analytic Cohort) vs. Central Cancer Registry (Oncology Cohort) was evaluated on Identification of Veterans with cancer. An ICD-based analytic approach identified substantially more Veterans with cancer (N=908,060) compared to the traditional cancer registry (N=488,273), especially in rural geographies.
Synapse has enriched 10 related papers. Consider them for comparative context: