Categorising major cardiovascular disease hospitalisations from routinely collected data July 2015, Volume 25, Issue 3

Introduction

Routine administrative hospital data are an important source of information for population health, health services and research. They have been particularly important for cardiovascular disease (CVD) statistics and research because there are few disease registers for CVD, and reporting of CVD is not mandatory, unlike most cancer types and certain infectious diseases.

Hospital data can be coded using a variety of systems. In practice, the International Classification of Diseases (ICD) has become the international standard diagnostic classification for many general epidemiological and health management purposes. ¹ These include analysing the general health status of population groups, and monitoring the incidence and prevalence of diseases and other health problems in relation to other variables such as the characteristics and circumstances of the individuals affected. Generally, the ICD orients diagnoses around the organ system affected and the pathological or clinical diagnosis within each organ system, with additional broader and less specific codes where this is not possible or appropriate. The ICD allows ascertainment of specific CVD endpoints (e.g. I21: acute myocardial infarction), homogeneous ‘blocks’ of three-character categories (e.g. I20−I25: ischaemic heart diseases), through to an entire ‘chapter’ (Chapter IX, ‘Diseases of the circulatory system’: ICD-10-AM codes I00−I99), although hospital data vary in their validity for this purpose. ^{2, 6} However, methods of ascertaining and categorising CVD using hospital data vary considerably^7-10 , often with little explicit justification for why a particular method has been applied. In particular, it is common for all codes at the chapter level to be used to identify ‘All CVD’ or ‘All circulatory disease’ hospitalisations.^10-12 However, this results in a heterogeneous category that combines common severe atherosclerotic and thrombotic CVD (caused by plaque formation or blood clots in blood vessels), such as myocardial infarction and pulmonary embolism, with common, less severe and pathologically dissimilar conditions such as varicose veins and haemorrhoids. In addition, hospital data collections contain a range of data fields, including those relating to primary and additional diagnoses, and those relating to procedures. All of these have the potential to contribute valuable information on CVD. Depending on how CVD is defined, the reporting of the total burden of CVD in the population can vary considerably – for example, if CVD is defined as ‘heart, stroke and vascular diseases’, 1 million Australians were affected in 2011−12, but if the definition includes all circulatory diseases, 3.7 million Australians were affected.^{10, 13}

This paper proposes a pragmatic approach to using primary and additional ICD diagnosis codes and procedure codes to capture major atherosclerotic, arteriovenous thromboembolic and related CVD. The implications of differences in CVD ascertainment methods are then explored using a large-scale cohort study and its linked datasets.

Methods

Classifying and categorising major CVD diagnoses using hospital data

Context

Australia uses the International Statistical Classification of Diseases and Related Health Conditions, to code and classify a range of health data, including hospital diagnoses and causes of death, with the Tenth Revision Australian Modification (ICD-10-AM) in use since 1997-98.¹⁴ Procedures are coded using the Australian Classification of Health Interventions, which is used in conjunction with ICD-10-AM.¹⁵

The New South Wales (NSW) Admitted Patient Data Collection (APDC) is a complete census of all public and private hospital admissions in NSW. In this dataset, a ‘hospitalisation’ refers to an episode of admitted care, which can be a total hospital stay (from admission to discharge, transfer or death) or a portion of a hospital stay beginning or ending in a change of type of care (e.g. from acute care to rehabilitation). A person can have multiple ‘separations’ within the same hospitalisation episode (e.g. transfers between hospitals or wards). Data linkage – either internal or to an external cohort – is required to track individuals to allow the number and proportion of people admitted to hospital with CVD to be estimated.

Diagnoses are recorded as principal and additional in the APDC.¹⁰ The principal diagnosis is the diagnosis established after an episode of admitted patient care that is considered to be mainly responsible for the care episode. The principal diagnosis is usually a disease, but can also be injury or poisoning, a specific care or service provided for a current condition (e.g. dialysis for kidney disease), or other reasons for hospitalisation.¹⁰ An additional diagnosis is a condition or complaint that affects patient management and either coexists with the principal diagnosis or arises during the episode of care. Up to 54 additional diagnoses may be recorded.

Outcome and grouping definitions

The aim of this study was to develop a pragmatic and inclusive definition of major atherosclerotic and arteriovenous thromboembolic CVD for use in reporting and research. We were not aiming to identify clinically or pathologically ‘pure’ atherosclerosis or arteriovenous thromboembolic disease (such a grouping is not necessarily feasible using routine data), but to focus on serious disease that has a significant atherosclerotic or thrombotic component.

The authors (GJ, EB, RK, KS) reviewed the ICD-10-AM three-character codes (one letter and two digits) for all circulatory diseases, including ‘Diseases of the circulatory system’ (I00−99), as well as G45 (‘Transient cerebral ischaemic attacks and related syndromes’) and G46 (‘Vascular syndromes of brain in cerebrovascular diseases’). From these, we selected conditions that were considered to have a significant atherosclerotic or arteriovenous thromboembolic contribution, and placed them in a group that we termed ‘major CVD’. Where aetiology was likely to be mixed, we erred on the side of inclusiveness. We excluded conditions that chiefly and primarily affect the lymphatic system or are caused by other factors, such as CVD caused by infections; nonthrombotic diseases of veins, lymphatic vessels and lymph nodes; and other or unspecified disorders of the circulatory system. We excluded essential hypertension on the grounds that it was a risk factor rather than a disease outcome. The excluded conditions were grouped as ‘other circulatory diseases’ for the purposes of this paper. We also considered procedure codes related to CVD interventions¹⁰ and identified those likely to relate to those conditions considered ‘major CVD’. Groupings were refined further in consultation with an expert cardiologist (author WA).

Applying the approach to a cohort study with linked hospitalisation data

Data sources

The Sax Institute’s 45 and Up Study is a large-scale Australian cohort study of 267 153 men and women aged 45 years and over who were randomly sampled from the general population of NSW, Australia. Individuals joined the study by completing a baseline postal questionnaire (distributed from 1 January 2006 to 31 December 2008) and giving informed consent for follow-up through repeated data collection and linkage of their data to population health databases. The study methods are described in detail elsewhere.¹⁶

Questionnaire data from study participants have been linked probabilistically to the NSW APDC by the Centre for Health Record Linkage. The linked data contain details of admissions in participants from 1 July 2000 to 31 December 2011, including dates of admission and discharge, the primary reason for admission using ICD-10-AM, up to 54 additional clinical diagnoses and up to 50 operations or procedures, coded using the Australian Classification of Health Interventions.¹⁵ Dates of death were ascertained from the date of recruitment up to 31 December 2011, using linkage to the NSW Register of Births, Deaths and Marriages. Death registrations capture all deaths in NSW.

Ethical approval for the study was obtained from the NSW Population and Health Services Research Ethics Committee and the Australian National University Human Research Ethics Committee.

Analysis

The 45 and Up Study linked data were used to follow participants from baseline until 31 December 2011 or the date of death, whichever occurred first. We ascertained the total number of hospitalisations attributable to the CVD codes under investigation and the total number of participants hospitalised with these codes over the follow-up period. Participants with invalid age or date of recruitment and those with linkage errors (n = 428, 0.16%) were excluded.

The numbers of CVD hospitalisations were identified using two different definitions: the proposed definition of major CVD and the definition using codes for all circulatory diseases (I00−99, plus G45, G46 and relevant procedures). Using both definitions, the numbers of cases were ascertained using primary diagnosis codes only, as well as using both primary and additional diagnosis codes. We also investigated the number of additional cases ascertained when using procedure codes with and without paired diagnosis codes (specific to each procedure).

Results

The Australian Classification of Health Interventions procedure codes and ICD-10-AM diagnosis codes categorised as major CVD are outlined in Tables 1 and 2, respectively, and the other circulatory disease codes excluded by the review process (categorised as ‘other circulatory disease’) are shown in Table 3.

Table 1.     Australian Classification of Health Interventions codes for CVD-related intervention procedures, used to define
major CVD

CVD = cardiovascular disease

Notes: Codes in italics are additional codes that we have included to reflect coding changes during the follow-up period.

Computerised tomographic brain scans (56001-00, 56007-00, 56010-02, 56010-03 and diagnosis codes G45, I60–I64), echocardiography (55112-00, 55118-00, 55130-00 and diagnosis codes I00–I52), magnetic resonance imaging brain scans (90901-00 and diagnosis codes G45, I60–I64) and coronary angiography (38215-00, 38218-00, 38218-01, 38218-02, and diagnosis codes I20–I25, I30–I52) are CVD-related diagnostic procedures and do not amount to diagnosis of CVD. These procedure codes are not used to define major CVD. Source for the full list of procedure codes: Australian Institute of Health and Welfare.¹⁰

Table 2.     ‘Major CVD’ codes, number of hospitalisations and number of participants hospitalised during follow-up

ICD-10-AM = International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification.

* Procedure codes in any of the 50 code fields; see Table 1 for the full list of procedure codes.

Notes:  Episodes of hospitalisation with changes in type of care (e.g. transfer from acute to rehabilitation) could contribute to more than one ICD-10-AM three-character code; as a result, numbers may not add up precisely.

Codes I70−I74 are often grouped together to identify hospitalisations for peripheral vascular disease.

Codes I60−I64 are often grouped together to identify hospitalisations for stroke. Although I60 (subarachnoid haemorrhage) is a major illness, it has a different aetiology from ischaemic stroke and is neither atherosclerotic nor artherothrombotic. We have excluded I60 from the definition of major CVD.

Table 3.     Codes not included in ‘major CVD’, number of hospitalisations and number of participants hospitalised during follow-up

ICD-10-AM = International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification

^a Categories for ICD-10-AM Chapter IX (Diseases of circulatory system), which implies that they are optional additional codes for the manifestation.

Study participants (n = 266 725) were followed for a median of 3.42 years (range 0.005–6.33 years). Overall, there were 39 459 hospital admissions and 24 543 participants admitted for major CVD as a primary diagnosis; the corresponding figures for all diagnosis codes were 65 340 and 33 242 (Table 2). Considering primary diagnoses only, the greatest contributions to this group were from angina pectoris (I20: 6725 hospitalisations), atrial fibrillation and flutter (I48: 6270 hospitalisations) and chronic ischaemic heart disease (I25: 4845 hospitalisations). Including secondary diagnoses substantially increased the number of hospitalisations and the number of participants hospitalised (by 66% and 35%, respectively), but the proportions of hospitalisations attributed to the individual codes were broadly similar to those for primary diagnosis codes alone. Procedures (coded with a relevant diagnosis code) contributed an additional 1.1% of major CVD admissions (increasing from 39 459 to 39 899); dropping the requirement for a procedure code to include a relevant diagnosis code made little difference to the numbers ascertained (1.6% increase, from 39 459 to 40 100).

Of the other circulatory diseases, the largest contributors based on primary diagnosis codes were haemorrhoids (I84: 3810 hospitalisations), varicose veins (I83: 1281 hospitalisations) and hypotension (I95: 1002 hospitalisations) (Table 3). When both primary and additional diagnosis codes were included, 59% of all hospital admissions with these codes were for essential hypertension (I10: 39 820 hospitalisations). The grouping of ‘all circulatory disease’ (I00−99, G45 and G46) included 47 678 hospitalisations and 30 389 participants using primary diagnosis codes, increasing to 105 759 and 55 036, respectively, when all diagnosis codes were used.

Discussion

To maximise the usefulness of routine administrative hospital data for CVD reporting and research, such data must be classified and categorised judiciously. The findings presented here demonstrate large differences in the composition and numbers of events in grouped CVD outcomes depending on the methods used. In particular, use of the ICD-10-AM ‘all circulatory disease’ (I00−I99, G45, G46) category results in a large, heterogeneous group; inclusion of both primary and secondary diagnoses in this category leads to a substantial contribution from common comorbidities such as hypertension.

Where the reporting and research interest relates to incident disease, it may be appropriate to prioritise specific disease categories and pathological homogeneity. Where grouping is necessary, it is important that the goals of and underlying rationale for such a grouping is clear. In defining the category for ‘major CVD’, we included conditions with a substantial contribution from atherosclerotic and arteriovenous thromboembolic disease. Along with its specific subcomponents, this category is particularly suited for research aimed at understanding conditions with this aetiology. We excluded conditions known to be secondary to nonatherosclerotic causes, such as those caused by infectious and infiltrative disease (e.g. rheumatic heart disease, infectious endocarditis and portal vein thrombosis). We aimed for overall pathological homogeneity in outcome ascertainment while maintaining inclusiveness. We also wanted to minimise assumptions about underlying relationships. For example, excluding venous thrombosis from the definition of major CVD could be seen as assuming that risks for arterial and venous thrombosis differ fundamentally for all exposures.

How the proposed groupings are implemented depends on the outcome and research questions of interest. For example, where the interest is in incident hospitalisation attributable to major atherosclerotic and related CVD, then the primary diagnosis codes for these conditions may be most relevant. When evidence of a history of CVD is sought, the use of ‘major CVD’ diagnoses, including primary and secondary diagnoses codes and CVD-related intervention procedures may be more useful. Because procedure codes introduce only a marginal gain in case identification, these may be dropped from the algorithm. In addition, because linked hospitalisation data may cover only a limited time period before baseline, self-reported survey data may be used as a complementary source of information about history of CVD. Where there is interest in comorbidity such as hypertension, and other elements of health status and health services use, broader groupings and those that include primary and secondary diagnoses may be more appropriate, but the validity of any included diagnoses should be borne in mind. ^2-6 Investigations involving CVD mortality are also likely to benefit from the use of secondary diagnosis codes. For example, a significant number of fatal cases of myocardial infarction are coded in secondary diagnosis fields, with the principal diagnosis being a code for arrhythmia.

The classification method and grouping outlined here acknowledges that hospital administrative diagnoses are imperfect and variable⁶, and we chose not to go beyond the three-character ICD-10-AM codes. Reporting patterns are also likely to change with the recent introduction of activity-based funding in many Australian states.¹⁷ The pragmatic method outlined here is open to being refined and may benefit from adaptation to address specific research questions, or according to the preferences of other research groups.

Conclusion

In investigations of CVD outcomes using routine administrative hospital data, where classification and grouping is necessary, it is important to clearly specify the goal and rationale for the grouping method employed. Prioritisation of specific disease categories and pathological homogeneity may be appropriate, particularly where the reporting and research interest relates to incident disease.

Acknowledgements

This research was completed using data collected through the 45 and Up Study (www.saxinstitute.org.au). The 45 and Up Study is managed by the Sax Institute in collaboration with major partner Cancer Council NSW; and partners: the National Heart Foundation of Australia (NSW Division); NSW Ministry of Health; NSW Government Family and Community Services – Carers, Ageing and Disability Inclusion; and the Australian Red Cross Blood Service. We thank the many thousands of people participating in the 45 and Up Study.

This specific study was supported by a National Health and Medical Research Council (NHMRC) of Australia project grant (reference: 1024450), a NHMRC partnership grant (reference: 1092674), a NHMRC Centre for Research Excellence grant and a Cardiovascular Disease Network Development Grant from the Heart Foundation of Australia. Emily Banks is supported by an NHMRC Fellowship (no.  042717).

Copyright:

© Joshy et al. This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Licence, which allows others to redistribute, adapt and share this work non-commercially provided they attribute the work and any adapted version of it is distributed under the same Creative Commons licence terms.