Summary

The purpose of this project is to recommend methods that can be used to monitor the incidence of cardiovascular disease using routinely collected data. Various special data collections were used to examine the validity and reliability of routinely collected data, during the period when International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) was used for classification of disease. With the recent introduction of ICD-10-AM (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification) additional validation studies will be needed to examine the concordance between the two classifications and to reestimate the adjustment factors. This report outlines the methodology that can be used.

This report focuses on the components of cardiovascular disease described below.

CHD is a generic term describing disease that results from insufficient blood flow to the heart caused by the narrowing of the coronary arteries due to atherosclerosis. AMI is the most severe form of CHD and occurs when the heart muscle is damaged as a result of a sustained blockage in a coronary artery. Unstable angina or preinfarction angina is part of the same biological process but may not progress to AMI if the blockage is cleared before the lack of oxygen causes permanent damage to the heart (Crea et al. 1997; Kristensen et al. 1997).

Cerebrovascular disease (stroke) comprises several disorders which results from a deficient blood supply to the brain due to the formation of a blood clot (most common type) or where an artery leaks blood into the brain. Congestive cardiac failure occurs when the heart is unable to pump enough blood to meet the needs of the body’s other organs, which often leads to a build up of fluid, either in the lungs or in other parts of the body.

Acute myocardial infarction

For monitoring incidence of coronary heart disease (CHD) it is recommended that:

  1. The rate of coronary events should be calculated as the sum of the rate of coronary deaths estimated from death certificates and the rate of non-fatal acute myocardial infarctions (AMIs) estimated from hospital separations.
  2. For fatal coronary events, deaths with (ICD-9) codes 410–414 should be used with adjustment factors to account for underestimation. The single ICD-9-CM code of 410 is not adequate.
  3. For non-fatal AMI, hospital separations should be used where the patient is discharged alive, the primary diagnosis is coded 410 using ICD-9-CM and the length of stay is greater than two days. Adjustment factors should be used to account for overestimation due to hospital transfers, readmissions and other effects.
  4. Further studies are needed on the use of more detailed hospital information, such as additional ICD coding and whether a hospital admission was unplanned to improve the validity of data on non-fatal AMIs. Also the occurring of routinely collected data for people aged over 65 years requires further investigation.
  5. Separate validation studies are needed for fatal and non-fatal events as the data sources and diagnostic criteria differ.

Investigations and procedures

For monitoring numbers of coronary investigations and procedures, booked admissions coded 413 to 414 should be subdivided into:

  1. admissions with percutaneous transluminal coronary angioplasty (PTCA) and coronary artery bypass grafts (CABG); and
  2. admissions without PTCA or CABG.

The admission rates for each of these categories (separately) should be monitored, without any adjustments, as they are essentially indicators of health services rather than disease incidence.

Angina pectoris

For monitoring incidence of angina pectoris it is recommended that:

  1. The rate of angina pectoris can be obtained by counting all patients who had an unbooked (emergency) admission to hospital and who were given a primary discharge diagnosis coded 411 or 413 according to ICD-9-CM.
  2. Primary discharge codes of 411 and 413 should be considered together for validation studies of angina pectoris. There is insufficient information in medical records to distinguish between cases of unstable angina pectoris and stable angina pectoris.

Stroke

For monitoring incidence of stroke it is recommended that:

  1. At present, hospital morbidity data (HMD) should not be used to measure past trends in attack rates for acute stroke because of rapidly changing proportions of non-fatal cases admitted to hospital. If the relatively large proportion of cases admitted to hospital in Perth in 1995–96 (particularly in patients under 75 years of age) is confirmed in further studies, it should be possible to use HMD to monitor trends in hospitalised cases of stroke from 1995 onwards.
  2. HMD may be used for obtaining improved estimates of rates of admission to hospital in a particular year for acute stroke and in hospital case fatality using the following selection algorithms. Total acute stroke is the sum of non-fatal stroke and fatal stroke where non-fatal stroke is defined as:
    • main diagnosis coded as acute stroke (430, 431, 434 or 436), OR acute stroke is coded in another diagnostic field for an admission of at least three days’ duration that is unbooked,

    and fatal stroke is defined as:

    • fatal cases where length of stay <29 days AND EITHER the main diagnosis was coded as acute stroke (430, 431, 434, 436), OR acute stroke was coded in another diagnostic field and the admission was unbooked.
  3. Validation studies are needed for the coding of deaths from stroke which occur out of hospital. Further studies should be undertaken in different geographical areas or health regions (or at least involving more than one major hospital catchment area) to test the algorithms described above and to determine the proportions of non-fatal cases admitted to hospital.
  4. The present study should be repeated using HMD and death records that have first been linked to provide episodes of fixed length to remove the effects of multiple admissions relating to the same person. Similar methods should be used to determine ‘first’ events, defined in terms of no previous admission because of acute stroke within a defined period (for example, five years).

Congestive cardiac failure

There are severe limitations to monitoring trends in congestive cardiac failure using hospital admissions. This is because:

  • signs and symptoms are poorly recorded in medical records
  • diagnostic criteria vary and are not used uniformly
  • large changes in rates can be caused by changes in coding practice.

As the incidence of congestive cardiac failure is believed to be increasing due to changes in the treatment of cardiovascular disease it is necessary to improve data quality. At present little credence can be given to available data.

Validation of hospital data on cardiac conditions

Based on a pilot study of validation methodology for cardiac conditions (but not stroke) using hospital data it is suggested that:

  1. Diagnosis of AMI can be validated through retrospective review of hospital records as the necessary information is usually available
  2. Information in hospital records is insufficient to distinguish between unstable angina; angina pectoris and chest pain, but if a broader category of angina is used then validation is possible
  3. For congestive cardiac failure, lack of universally accepted diagnostic criteria or evidence from a definitive test and inadequacies in hospital records make validation from retrospective review of records unfeasible. Only prospective data collection for patients admitted for a broad range of conditions could produce adequate information.