Prevalence and patterns of multimorbidity in Australia

The health care system in Australia, as in many other countries, focuses on single diseases. Payment structures increasingly support single chronic diseases,1,2 and guidelines for care usually take a single-morbidity approach.3 This is a challenge for primary care physicians, who care for the “whole” patient, as the presence of multimorbidity has implications for managing chronic illness. It is argued that patients are more than the sum of their parts, and that, using a single-morbidity model, we fail to grasp the pattern of disease, leading to inadequate management.4 Calls are growing for a more holistic consideration of the patient.2,5-7

Research into multiple morbidity first concentrated on measurement of comorbidity, defined as “the existence or occurrence of any distinct additional disease entity in a patient who has the index disease under study”.8 More recent interest is in the mix of morbidities (multimorbidity) — the “co-occurrence of two or more diseases within one person without defining an index-disease”.7 Studies have investigated the relationship of multimorbidity (often with severity scores) with quality of life,9-11 utilisation of medical services,12 hospitalisations and mortality.13 Some studies use self-report surveys, and others use administrative data12 or data from primary care records.14

However, the study of multimorbidity is still in its infancy. A 2005 literature review identified few articles on multimorbidity, compared with the vast literature on specific diseases.15 This is despite the fact that multimorbidity has an estimated prevalence of 60% among 65–74-year-olds, far higher than the prevalence of each individual disease.15

As in any new field, there are problems in measuring multimorbidity. In particular, the number and types of disease selected and their grouping greatly affect prevalence estimates. An international literature review of published data found widely varying methods and results. Although multimorbidity was always defined as the presence of two or more diseases, the morbidities included varied widely, as did study populations, age ranges and subgroups, and denominators for prevalence estimates.16

There are currently no published Australian data on population prevalence of multimorbidity. There are a few reports on comorbidities of selected target diseases drawn from the National Health Survey17 but, as discussed elsewhere,18 this is based on patient self-report, which is of questionable reliability.

A Western Australian study in the early 1990s used hospital chart review to describe “comorbidity”, using the Charlson Comorbidity Index. For a subsample of patients, the patient’s general practitioner at the time of admission was also asked to report (by tick box) the diseases present at that time. The authors rejected the GP survey as unreliable because GPs reported only 58% of morbidities found in hospital charts, but 20% more morbidities not matched in hospital charts.19 It is curious that the additional GP diagnoses were judged erroneous. The project was geographically limited, involved a small selected sample, and must have relied on retrospective review of GP records.

In this study, we take a first step in the prospective measurement of the prevalence of multimorbidity in the population attending general practice in Australia and extrapolate these results to provide estimates for the Australian population.

The data collection methods have been described elsewhere.18 In summary, this is a secondary analysis of data from a substudy of the BEACH (Bettering the Evaluation And Care of Health) program, a continuous national study of general practice activity.20,21 The substudy investigated prevalence of selected conditions among 9156 patients attending 305 GPs in July–November 2005. GPs were asked which of a list of morbidities were currently being managed for each patient, using their knowledge of the patient, the patient’s self-report, and the medical record.

The BEACH study uses random samples of GPs, each providing data about a cluster of encounters. SAS version 9.1.3 (SAS Institute Inc, Cary, NC, USA) was used to adjust for clustering at individual GP level. The sample survey results were adjusted to provide prevalence estimates for the population attending a GP in 2005–2006, and the population of Australia, as described elsewhere.18

Three methods have been validated to some degree for measuring multimorbidity: the Johns Hopkins Adjusted Clinical Group case-mix system,12 the Charlson Comorbidity Index,22 and the Cumulative Illness Rating Scale (CIRS).10,23,24 We chose the last as it has been validated25 and applied in family practice.9,10,16,24 It uses the presence of illness in each of 14 organ domains plus a measure of severity of each domain to provide an index of total chronic medical illness burden. We allocated the surveyed morbidities into these synergistic domains (ie, management of one problem is usually beneficial for management of another within the group). They fell into eight of the 14 domains. Our category “malignant neoplasms” was too broad to be allocated to a single CIRS domain, so we added it as an independent morbidity type, giving a total of nine morbidity domains (Box 1).

Respondents with more than one surveyed morbidity within a single CIRS domain were counted only once for that domain. Prevalence of illness in each CIRS morbidity domain among the sample was calculated. Multimorbidity was defined as presence of illness in two or more morbidity domains. Results are presented as estimated prevalence of each domain and of combinations of domains in the sample; estimated prevalence among patients who attended a primary care practitioner at least once in 2005–2006, after adjustment for attendance rates by age–sex category; and estimated population prevalence.18 Estimated numbers of national encounters with patients having each multimorbidity combination were made through simple extrapolation from the crude prevalence estimate to the number of services by GPs in April 2005 – March 2006.26

The most common combinations of domains were first investigated for all patients with at least two morbidity domains, followed by patterns among those with three or more (a subgroup of the first analytical group), and then patterns among those with four or more (a subgroup of both previous groups). Only the 12 most common combinations of two or more morbidity types are presented here. The full results for all combinations are available at the website of the Family Medicine Research Centre (http://www.fmrc.org.au/publications/appendices/).

This study has given a first insight into the prevalence of multimorbidity in Australia. It suggests that about three in 10 people who attend general practice, and one in four Australians overall, have multimorbidity as defined in this study. This aligns with overseas estimates of 30% in Quebec, Canada, and the Netherlands.14 Extrapolation suggests that over 5 million Australians are currently being managed for diagnosed diseases from at least two of the morbidity domains.

Our finding that prevalence and complexity do not differ by sex but are highly age-related are consistent with the results of other studies.14 However, our prevalence estimates for older patients (75% for 65–74-year-olds and 83% for those aged 75 years or over) are higher than those from Quebec in 1999 (63.4%),15 the Netherlands in 1990 (> 60%) and the United States in 1999 (ranging from 54% among 65–70-year-olds to 73% in those aged 80 years or more).16 All studies used different methods to measure multimorbidity, raising further questions rather than providing answers.

Our study has the limitations reported elsewhere in the adjustment processes used to account for higher chance of selection associated with higher morbidity independent of age and sex of patient.18 In addition, it includes only diagnosed morbidities currently under management and cannot consider morbidities not yet diagnosed. Further, it includes a limited number of morbidities, selected because they are National Health Priorities or because of their high frequency of management in general practice.18 However, together these morbidities account for about 80% of all occasions of chronic disease27 management in general practice in Australia (unpublished BEACH data), and this is the only approach possible given the cross-sectional patient sampling processes. Surveyed patients with none of the listed morbidities may have other, unlisted, chronic diseases, and patients who had one of the morbidity domains plus other chronic diseases not studied would be classed as not having multimorbidity. We therefore cannot conclude that 29% of the population has multimorbidity, but we can conclude that this proportion has multimorbidity as defined in this study.

Obesity was not included in the study, as it did not fulfil the inclusion criteria at the time. However, in 2005–2006, 22% of patients attending a GP were classified as obese (through self-reported height and weight),28 about the same prevalence as, for example, arthritis (22.8%).18 As obesity has now been added to the National Health Priorities, it should be included in future studies. Other morbidities that could be considered for inclusion are renal impairment, osteoporosis and other musculoskeletal conditions, and their inclusion would certainly raise prevalence estimates for multimorbidity. We believe the next study should be structured on the basis of the CIRS domains and include tick boxes for each of the more commonly managed morbidities in general practice, as well as other individual morbidities within each domain relevant to government health policy. Ideally, severity should be assessed with multimorbidity, but would require a separate study as this would be too demanding for BEACH participants in the confines of the GP–patient encounter.

Underlying these limitations is the lack of an internationally accepted standard for measuring multimorbidity. A clear conceptual framework must be developed that includes a consistent approach to measurement of multimorbidity and age categorisation in the study population to allow comparison between studies and populations.29

Some suggest that, if we are to deal with an increasing prevalence of multimorbidity in an ageing population, we need to know about disease combinations so we can design best practice guidelines for clinicians.5 They also note that continuing to combine care plans will increase polypharmacy and the chance of adverse drug events.5 Some multimorbidity combinations are particularly problematic for both patient and clinician. For example, combinations of cardiac and respiratory system morbidity have a particularly strong synergistic negative effect.10 The common combination of musculoskeletal morbidity with vascular and upper gastrointestinal problems may result in a non-steroidal anti-inflammatory agent increasing the mortality of the other two morbidities, particularly in older patients.30 These issues are often not considered in guidelines for individual conditions.

Governments are currently considering the structure and function of the Australian health care system. Our study, although only a first step in measuring prevalence of multimorbidity, may help inform planning of health services of the future.