ABSTRACT
Background
International Classification of Diseases, Tenth Revision (ICD-10) codes are commonly used for identifying myocardial infarction (MI) in clinical research and increasingly used to capture events in clinical trials; however, their accuracy for distinguishing MI subtypes is uncertain.
Methods
In this secondary analysis of a single-center prospective cohort study, consecutive individuals who underwent highsensitivity cardiac troponin testing between October 2023 and February 2025 were adjudicated by physicians as either an MI subtype, myocardial injury, or non-elevated troponin. Clinician adjudicated diagnoses were compared to ICD-10 codes attributable to type 1 MI, type 2 MI, and type 3-5 MI.
Results
In a study of 24,524 individuals undergoing troponin testing, type 1 MI ICD-10 codes showed moderate sensitivity (53%) but high specificity (99%), whereas the ICD-10 codes for type 2 MI and types 3-5 MI demonstrated extremely low sensitivities (3% and 0%, respectively) despite high sepificitiy (99%).
Conclusions
These data suggest the ICD-10 codes for each individual MI subtype have such low sensitivity that they may not reliably and accurately identify these MI subtypes in clinical research or in quality metrics and value-based programs.
Background
Beyond facilitating institutional diagnosis capture and billing claims, International Classification of Diseases, Tenth Revision (ICD‑10) codes are frequently used to examine the epidemiology of myocardial infarction (MI). More recently, ICD-10 codes have also been utilized to capture events in clinical trials as an alternative to adjudication. , Consequently, their use has become integral to contemporary clinical research.
While identification of individuals with MI is important, differentiation of the subtypes (type 1-5) is also imperative as evidence-based treatments differ across MI subtypes. Additionally, subtype differentiation affects the inclusion of individuals with MI in value-based programs and quality metrics. For example, the Hospital Readmissions Reduction Program intends to only include individuals with type 1 MI. , Furthermore, therapies may vary in their efficacy for preventing and treating these MI subtypes. ,
In response to the growing need to differentiate MI subtypes, ICD-10 codes for type 2 MI (I21.A1) and the other MI subtypes, i.e., type 3, 4, or 5 MI (I21.A9) were introduced in 2017. Few studies have evaluated the accuracy of these MI subtype codes, and those undertaken have primarily examined it among individuals assigned an ICD‑10 code for type 1 or type 2 MI. , These studies likely miss individuals with MI who were not coded as such, preventing a comprehensive understanding of MI code accuracy. Accordingly, we assessed MI code accuracy in a prospective cohort undergoing routine cardiac troponin testing.
Methods
This research was supported by a National Heart, Lung, and Blood Institute Career Development Award (K23HL167659). The authors alone are responsible for the design and conduct of this study; the analysis of the data; and the drafting, editing, and finalization of the manuscript. This content does not necessarily represents the official views of the National Institutes of Health.
This secondary analysis used data from a single‑center prospective cohort study at Massachusetts General Hospital (MGH). Consecutive patients undergoing high‑sensitivity cardiac troponin testing in the Emergency Department or an inpatient setting between October 2023 and February 2025 were identified during pre‑screening for a prospective health‑status study. The study was approved by the Mass General Brigham Institutional Review Board, with a waiver of informed consent for pre‑screened individuals. Individuals aged less than 18 years, those with two or more distinct troponin encounters, and those who could not be matched to a clinical dataset capturing ICD‑10 coding were excluded.
Physician investigators (CMC, SM, AR, EL) reviewed the electronic medical record to adjudicate each encounter as an MI subtype, myocardial injury, or non-elevated troponin using all the available clinical data. Uncertain cases were reviewed by at least one additional investigator (CMC, JJ, JW, or PN) to form consensus. MI was defined according to the Universal Definition. The criteria to diagnose each MI subtype are provided in Supplementary Table I.
Participants were matched to the MGH clinical data registry, Research Patient Data Repository (RPDR), which captures demographics and ICD-10 coded diagnoses from our electronic health record. Individuals with ICD-10 codes I21.0, I21.1, I21.2, I21.3, I21.4, or I21.9 were considered to be coded as type 1 MI; those with the code I21.A1 as type 2 MI; and those with the code I21.A9 as type 3, 4, or 5 MI. Classification into an MI subtype required the relevant ICD-10 code to be recorded as the primary diagnosis or within diagnosis positions 1–10 of the associated emergency department or inpatient encounter, or to be coded within 30 days following discharge from the corresponding encounter.
Descriptive statistics were used to present the participant characteristics. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for the MI subtype codes using adjudicated clinician diagnosis as the gold-standard comparator. Statistical analyses were performed using R software, version 4.2.2 (R Foundation for Statistical Computing).
Results
Over 16 months, we identified 41,236 encounters with one or more high-sensitivity cardiac troponin measurements. After removing individuals with two or more encounters, 27,293 remained. This was reduced to 24,839 after excluding individuals who could not be matched with RPDR. After additionally excluding individuals under 18 years of age, the final cohort comprised 24,524 individuals. The mean age of the study population was 60.1 (±18.7) years; 11,666 (47.5%) were female, and 16,814 (68.6%) were White. Among the cohort, 7,515 (30.6%) had hypercholesterolemia, 8,767 (35.7%) had hypertension, and 6,371 (26.0%) had known coronary artery disease and/or a prior MI ( Table 1 ).
Table 1
Baseline characteristics of study population.
| Study population | N = 24,524 |
|---|---|
| Demographics | |
| Age, mean (standard deviation) | 60.11 (18.70) |
| Gender (%) | |
| Female | 11,666 (47.5) |
| Male | 12,852 (52.4) |
| Unknown | 6 (0.1) |
| Race (%) | |
| White | 16,814 (68.6) |
| Black | 2,453 (10.0) |
| Asian | 1,240 (5.1) |
| American Indian or Alaska Native | 27 (0.1) |
| Native Hawaiian or Other Pacific Islander | 14 (0.1) |
| Other | 2,801 (11.4) |
| Two or more races | 260 (1.1) |
| Declined | 268 (1.1) |
| Unknown/Missing | 647 (2.6) |
| Ethnicity (%) | |
| Hispanic | 806 (3.3) |
| Non-Hispanic | 20,997 (85.6) |
| Declined | 1,331 (5.4) |
| Unknown/Missing | 1,390 (5.7) |
| Past medical history (%) | |
| Cerebrovascular disease, n (%) | 1,383 (5.6) |
| Peripheral vascular disease, n (%) | 3,806 (15.5) |
| Chronic lung disease, n (%) | 3,757 (15.3) |
| Congestive Heart Failure, n (%) | 2,589 (10.6) |
| Hypercholesterolemia, n (%) | 7,515 (30.6) |
| Hypertension, n (%) | 8,767 (35.7) |
| Coronary artery disease (including prior myocardial infarction), n (%) | 6,371 (26.0) |
| Diabetes Mellitus, n (%) | 3,710 (15.1) |
| Chronic kidney disease, n (%) | 2,727 (11.1) |
| Current smoker, n (%) | 2,404 (9.8) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
