Abstract 826: Heart Failure Diagnostic Codes: Concordance of Medical Records and Medicare Administrative Claims. The ARIC Study
The identification of heart failure (HF) during review of medical data is made complex by limitations of data sources. Medical records provide detailed information but require costly abstraction. Administrative claims data are more accessible but their diagnostic codes may be influenced by nonclinical factors. Claims are increasingly used in research but their susceptibility to errors and bias are not well understood. We compared HF ICD-9-CM diagnosis codes in hospital medical records to linked Medicare hospital claims in the Atherosclerosis Risk in Communities (ARIC) cohort, years 2003–2005. We included >7000 cohort members ≥65 years of age in Medicare during this time, encompassing 21,741 participant-years of observation. 853 (3.9%) had at least one HF-coded hospitalization claim. Of these, 80.0% (683/853) had at least one HF-coded medical record hospitalization. 757 (3.5%) had at least one HF-coded medical record hospitalization. Of these, 90.2% (683/757) had at least one HF-coded hospitalization claim. Probability of agreement=98.9%, kappa=0.84. Cohort Participant-years in 2003–5 Medicare Data, cross-classified by presence of HF diagnosis codes in claims vs. hospital medical record.
Disagreement was noted where hospital record HF diagnosis codes had no matching claim (n=74). Reasons for this included date mismatch, failure to capture diagnostic codes in claims, diagnosis code disagreements and errors. Disagreement where HF-coded claims were found without a matching HF hospital medical record (n=170) were mostly due to incomplete ascertainment of information from hospitals outside the ARIC surveillance areas. There is generally good agreement of HF-related diagnosis codes comparing hospital medical records and Medicare claims. Areas of disagreement suggest approaches to calibrate claims-based definitions of HF for refinement of hospital surveillance systems.