Abstract: SA-PO015
Comparative Accuracy of Self-Reported vs. Electronic Health Record (EHR)-Derived Charlson Comorbidity Index among Veterans with Advanced CKD
Session Information
- Augmented Intelligence, Large Language Models, and Digital Health
October 26, 2024 | Location: Exhibit Hall, Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Augmented Intelligence, Digital Health, and Data Science
- 300 Augmented Intelligence, Digital Health, and Data Science
Authors
- Senthil, Nandita, University of Florida, Gainesville, Florida, United States
- Scheiffele, Grant D., VA North Florida South Georgia Veterans Health System, Gainesville, Florida, United States
- Parmar, Cydney El Cid, University of Florida, Gainesville, Florida, United States
- Shell, Popy, VA North Florida South Georgia Veterans Health System, Gainesville, Florida, United States
- Jia, Huanguang, VA North Florida South Georgia Veterans Health System, Gainesville, Florida, United States
- Shukla, Ashutosh M., University of Florida, Gainesville, Florida, United States
Background
Inaccurate or missing diagnosis codes limit the prognostic power of EHR-derived comorbidity indices, such as, Charlson Comorbidity Index (EHR-CCI). CCI calculated by patients self-reporting individual comorbidities (SR-CCI) can address these errors but are subject to patient awareness and recollection bias. The relative accuracy of SR-CCI vs. EHR-CCI has not been examined, especially among the US Veterans with CKD.
Methods
Using the baseline data of the ongoing randomized Trial to Evaluate and Assess the impact of Comprehensive pre-kidney failure education on Home dialysis among Veterans (TEACH-VET) on variety of patient outcomes, this study aimed to assess the concordance and accuracy of SR-CCI and EHR-CCI among veterans with advanced CKD. In addition to the detailed data for demographic and socioeconomic variables, all enrollees were self-queried for the individual comorbidities by trained study staff to collect their CCI at the enrollment. SR-CCI was compared against EHR-CCI, extracted from VA database, anchoring for each patient’s enrollment date.
Results
Of the 551 Veterans with advanced CKD enrolled in TEACH-VET, 338 participants with complete baseline data, including SR-CCI were included for this analysis. The primary outcome, the aggregate CCI score did not differ significantly between the SR-CCI and EHR-CCI (p=0.1),however, there were significant differences in individual comorbidities composing CCI. Of the 17 comorbidities comprising CCI, 13 (76.5%), notably, CKD-related comorbidities, e.g., moderate to severe CKD and liver disease, diabetes with or without complications, and myocardial infarction, etc., had substantial disagreements (p<0.0001) between the two methods, whereas peripheral vascular disease, mild liver disease, hemiplegia, and AIDS had no significant difference between the two methods. There were no significant differences in the demographic characteristics across the score patterns.
Conclusion
Most comorbidities used for CCI are either over or under accounted for by EHR-CCI, raising concerns for its routine use in clinical and epidemiological research. Future qualitative studies comparing SR-, and EHR-CCI comparison are needed to determine optimal method for reporting CCI and similar comorbidity indices.
Funding
- Veterans Affairs Support