Background: Discharge codes for acute myocardial infarction (AMI) in the primary position have declined since 2011 but increased in any secondary position among Medicare beneficiaries. The objective of this study was to assess validity of AMI hospital discharge codes in primary vs. secondary positions compared to adjudicated AMI criteria from the Atherosclerosis Risk in Communities (ARIC) Study.
Methods: From ARIC, a continuous surveillance of hospital discharges for AMI among residents aged 35-74 years in 4 US communities between 2005-2013 was evaluated. Based on annual hospital discharges, a probability sample of AMI (ICD9-410) codes was captured, further sub-categorized into AMI in primary vs. secondary positions. The accuracy of ICD-9-CM codes for AMI in primary vs. secondary positions was compared to the gold standard, i.e. diagnosis of AMI by ARIC criteria (based on cardiac pain, electrocardiographic criteria and cardiac biomarkers).
Results: The study population was 53% men, 72% white. After weighting for sampling probabilities, 15,097 AMI discharge codes were identified. When discharge diagnosis of AMI was coded in the primary position (n=9,968) 79% were identified as definite/probable AMI and 21% identified as suspected/no AMI using ARIC criteria (table). In contrast, when discharge diagnosis of AMI was coded in any secondary position (n=5,129), 43% were identified as definite/probable AMI and 57% as suspected/no AMI using ARIC criteria. The positive predictive value for ICD-410 in the primary position was 1.8 times higher than in any secondary position.
Conclusions: The accuracy of a discharge code for AMI in a secondary position was low compared to AMI classification by ARIC criteria, indicating that this entity has different clinical, economic and epidemiological implications compared to AMI in a primary position. Researchers using claims data should be cautious about the use of ICD-9 code 410 in a secondary position to identify patients with AMI.