2006 HSR&D National Meeting Abstract
1057 — Is There Overlap of Coding of Diagnoses for Veterans Who Use Both VA and Medicare Services?
Petersen LA (Houston Center for Quality of Care and Utilization Studies)
Pietz K (Houston Center for Quality of Care and Utilization Studies)
Kuebeler M (Houston Center for Quality of Care and Utilization Studies)
Byrne M (University of Miami, Miami, FL)
Many elderly veterans use both VA and Medicare services (dual users). It is not clear whether VA data reflect diagnostic codes for conditions treated in Medicare and vice versa. We determined whether all major diagnostic codes and total illness burden of dual users can be obtained from examination of administrative data from only one source.
Cohorts included all age-eligible fee-for-service Medicare dual users in fiscal years 2000-2002. We calculated and compared relative risk scores (RRS) (a measure of illness burden developed by DxCG, Inc.) using VA, Medicare, and the union of all diagnoses from both data sources. We explored whether differences in number of diagnoses (truncated at the 3-digit level) and RRS between VA and Medicare data sources were due to pharmacy-only users or to an underweighting of mental health diagnoses in the DxCG software.
On average for a dual user, more diagnoses were recorded in Medicare (13-15) than in VA data sources (8), and only approximately 2 of these diagnoses overlapped at the 3-digit ICD-9-CM level. Medicare data captured 80% of individuals’ total illness burden; VA data only captured one-third. The ratio of RRS derived from Medicare vs. VA sources was 2.4. After excluding veterans who primarily use the VA for pharmacy purposes, the ratio of Medicare to VA RRS dropped approximately 20-25%. Removing mental health diagnoses did not significantly affect the ratios.
The majority of major diagnostic codes for dual users were present in Medicare data sources, but falsely low estimates of the total illness burden and disease prevalence would result if calculated solely using either data source. The differences in RRS when calculated in VA vs. Medicare data were not driven solely by pharmacy-only VA users or by undervaluing of mental health diagnoses by the DxCG software.
There does not appear to be a phenomenon of coding but not treating chronic illnesses in the VA. However, using data from just Medicare or VA sources when conducting research or calculating performance measures on dually eligible veterans may seriously underestimate comorbidity burden and fail to identify individuals who suffer from common chronic diseases.