Previous Research Using Administrative Data to Ascertain Cases of Coronary Heart Disease

Date: December, 2006

Table 1 : Summary of previous research on methods to identify coronary heart disease cases from administrative data


Data Source


Diagnosis/Treatment Codes and Algorithms

Study Cohort

Validation Methodology


Rawson et al. (1995)

Country: Canada


Source: Hospital discharge abstracts


Year: 1986



Codes: ICD-9 410 – 414


Algorithms: Hospital discharge data with a relevant diagnosis code in the primary or secondary discharge diagnosis field.

Age of cohort was not specified.


Validation was performed with medical chart review for AMI only. Physician billing claims were used to validate AMI, angina, and chronic IHD.


Concordance between hospital discharge abstracts and medical chart review for AMI was 96.9%.


Concordance between hospital discharge abstracts and physician claims was 69.3% for AMI, 70.0% for angina, and 55.6% for chronic IHD when the primary diagnosis was used.



“When diagnostic agreement was broadened to include any physician diagnosis in the same ICD sub-chapter, the concordances all increased ranging from 79% to 94% in the IHD groups…” p.2639

Mahonen (1997)

Country: Finland


Source: Finnish national hospital discharge register


Year: 1983-1990

Codes: ICD-9 410 – 414


Algorithm: Diagnosis of CHD in hospital discharge abstract

Ages 25-65 years

FINMONICA acute myocardial infarction register


Max. Sensitivity: 87.1%

Max. PPV: 95.7%

“There are no medical record abstractors in Finland ; clinicians taking care of the patients assign both the diagnoses and ICD codes for hospitalizations.” p.405

O’Connor (1998)

Country: USA


Source: Clinic (i.e., physician) data


Year: 1992-1994


Codes: ICD-9 428.0 (congestive heart failure), 412 (history of myocardial infarction), 429.2 (arteriosclerotic cardiovascular disease), 413.9 (angina)


Algorithm: At least one visit with at least on of the codes in a defined 1-year period

HMO members. Age is not specified

Survey data. The question was: “Have you ever been told by a doctor that you had a heart attack?”

When survey data and database data conflicted, a chart review was done.


Max. Sensitivity: 89%

Max. Specificity: 99%

Max. PPV: 85%


Shah et al. (2000)

Country: Canada


Source: CIHI hospital discharge abstracts for Ontario


Years: 1981-1997



ICD-9 410, 411, 413, 414.

CCP codes for angioplasty and bypass: 48.0, 48.1


Algorithm: An event of heart disease was defined as a hospital discharge that has a selected ICD-9 code as one of the first two discharge diagnoses, or a selected CCP code as one of the first 2 procedures listed.

Residents of Ontario communities that had regular census participation and at least 95% of their population claiming Native origins (N=16875 in 1991)

No validation was conducted.




Pajunen et al. (2005)

Country: Finland


Source: Finnish national hospital discharge register


Years: 1988 – 2002

ICD-9 410 (AMI)

ICD-10 I21-I22 (AMI)


ICD-9 411.0 (unstable angina)

ICD-10 I20.0 (unstable angina) – these were investigated in a secondary component of the analysis


Non-fatal events were identified where one of the selected codes was either the main diagnosis or an additional diagnosis.

35 years of age and older.

Validation was conducted using the myocardial infarction (MI) register.


For both fatal and non-fatal events:

Max Sens = 85%

Max PPV = 90%


For non-fatal events:

Max Sens = 94%

However, there were wide geographic variations in sensitivity, from 52% to 94%.


©2006 Manitoba Centre for Health Policy (MCHP)