Abstract P372: Evaluation of International Classification of Disease-Ninth Edition (ICD-9) codes for Determining Subclasses and Anatomic Location of Myocardial Infarction in the Community

Abstract

Background: Epidemiology studies of acute myocardial infarction (MI) often rely on hospital discharge codes or claims data to identify events. The fourth digit of ICD-9 code 410 is meant to identify anatomic location of an MI. Although the validity of ICD-9 410 codes to identify the general category of MI has been studied, far less is known about the validity of ICD codes to identify ST segment MI (STEMI) and non-STEMI (NSTEMI) and to identify anatomic location of STEMI infarcts. Methods: From 1987 to 2010 we evaluated random samples of hospitalizations with ICD-9 discharge codes 410-414, 402, 427, 428, or 518.4 among men and women age 35-74 years from hospitals serving the 4 communities of the Atherosclerosis Risk in Communities (ARIC) Study (400,000 base population in 2010). Trained staff abstracted medical records and copied up to three 12-lead electrocardiograms (ECG) that were coded by Minnesota Code. A standardized algorithm was applied to data on chest pain, cardiac biomarkers, and ECG evidence to determine MI diagnosis. Validated MI events with abnormal biomarkers were further classified by ECGs as STEMI or NSTEMI. ICD-9 codes 410.0-410.6 and 410.8 were used to define STEMI while codes 410.7 or 410.9 defined NSTEMI. STEMI infarct location was assessed by ECG and categorized as anterior, inferior, lateral, or multi-location. We determined the validity of code-based definitions using the ARIC algorithm and ECG evidence as referent standards. All analyses were weighted to account for sampling. Results: Between 1987 and 2010, 208,920 (weighted) hospitalizations with discharge codes suggestive of MI occurred in the 4 ARIC communities. Of these, 19% (38,729/208,920) were validated as MI. The positive predictive value (PPV) of an ICD-9 410 code to identify a validated MI was 72% (22218/30652). This PPV declined slightly from 78% (862/1111) in 1987 to 71% (1031/1462) in 2010. Center differences by community were seen (range 63% (197/315) to 78% (173/222) in 2010). Sensitivity of a 410 code to identify validated MIs remained stable from 1987 to 2010 at about 57% (22218/38,729). The PPV of the ICD-9 code-based STEMI definition improved after 2005 but remained moderate at 41% (175/430) in 2010. The PPV of the ICD-9 code based definition of NSTEMI was 63% (599/945) in 2010 and was stable over time. The PPV of codes to identify anterior and inferior infarctions were high (66% (1145/1741) and 78% (1956/2518), respectively). However codes for lateral and multiple site infarctions had lower PPV (53% (327/619) and 21% (153/727), respectively). Conclusions: The PPV of an ICD-9 code 410 to identify MI remained stable over the past 2 decades, but geographic differences persist. ICD-9 codes are better at correctly identifying NSTEMI than STEMI and better at identifying inferior infarcts than other anatomic locations. These data suggest caution in interpreting studies of MI trends based solely on ICD-9 codes.