The ageing of the population is expected to lead to increases in the prevalence of chronic conditions, multimorbidity, and raised demand for primary care services. To enable health systems to respond to these increases, the prevalence of chronic conditions and multimorbidity need to be measured in an accurate and timely manner. However, prevalence estimates of multimorbidity vary widely due to inconsistent definitions and measurement methods used in research. The aim of this thesis is to develop a reliable and practical method of measuring multimorbidity in Australia.
The research reported in this thesis is based on two sets of sub-studies of the Bettering the Evaluation and Care of Health (BEACH) program, a continuous national survey of Australian general practice activity.
The first survey was conducted between August 2008 and May 2009, and involved 290 randomly selected general practitioners (GPs) who recorded all diagnosed chronic conditions in 8,707 patients at their encounters.
Having GPs record patients’ diagnosed chronic conditions avoids the limitations of self-reported data used in most large population prevalence studies. However, patients sampled at GP encounters are not representative of the population as only about 87% of people visit a GP in any year and because older people are more likely to attend and to attend more often. To estimate population prevalence, I weighted each age-sex group to match the distribution of the population. I then weighted the outcome by the proportion in each age-sex group who visited a GP at least once in the survey year, assuming those who did not see a GP did not have a diagnosed chronic condition.
I estimated that two-thirds (66.3%) of patients at GP encounters had at least one diagnosed chronic condition as did half (50.8%) of the Australian population. Hypertension was the most prevalent condition, 26.6% of patients at GP encounters and 17.4% of the population having this diagnosed condition.
While multimorbidity has been most often defined as 2+ chronic conditions, there have been recent moves towards using 3+. There have been calls for standardisation of multimorbidity research, inconsistent definitions and methods having led to large variance in estimated prevalence between studies. I examined the independent effects on prevalence estimates of:
1. how ‘morbidity’ is defined either as a single chronic condition or a ‘group’ of conditions using the chapter/domain structure of the International Classification of Primary Care (Version 2) (ICPC-2), the International Classification of Disease (10th revision)(ICD-10), or the Cumulative Illness Rating Scale (CIRS);
2. the number of ‘morbidities’ required in the definition of multimorbidity;
3. the number of diagnosed chronic conditions included in the study.
I found that data grouped by ICPC-2 chapters, ICD-10 chapters or CIRS domains produced similar multimorbidity prevalence estimates. Multimorbidity defined as 2+ morbidities provided similar estimates whether individual conditions or groups of conditions were counted and whether as few as 12 prevalent chronic conditions were studied or all chronic conditions, but it lacked the specificity to be useful, especially among older people. Multimorbidity, defined as 3+ morbidities, required more measurement conformity and inclusion of all chronic conditions, but provided greater specificity than the 2+ definition.
These results led to a set of guidelines for multimorbidity researchers, which if followed, will produce results that can be compared with results from other studies adhering to the same guidelines. I also proposed the concept of ‘complex multimorbidity’, the co-occurrence of three or more chronic conditions classified in three or more different body systems within one person, without defining an index chronic condition. Using ‘complex multimorbidity’ may identify high-need individuals.
I estimated that: 47.4% of patients at GP encounters and one-third (32.6%) of the population had multimorbidity (2+); further, that 27.4% of patients at GP encounters and 17.0% of the Australian population had complex multimorbidity. The most prevalent pattern of three conditions was hypertension + hyperlipidaemia + osteoarthritis (5.5% of patient at encounters and 3.3% of the population).
In my second, larger, survey, conducted between November 2012 and March 2016, 1,449 randomly selected GPs recorded all diagnosed chronic conditions for 43,501 patients. They also recorded the number of times each patient had seen a GP in the previous 12 months. Data collected in Survey 1 had not allowed adjustment for high and low attenders within each age-sex group. The individual attendance data in survey 2 allowed me to adjust for each patient’s chance of being in the survey sample.
My prevalence estimates for patients at encounters were similar to those from Survey 1, with 26.5% of patients at encounters having diagnosed hypertension, 51.6% multimorbidity and 30.4% having complex multimorbidity. However, the population prevalence estimates produced with the new method were significantly lower than those from the previous method, an estimated 12.4% of the population having diagnosed hypertension, 25.7% multimorbidity and 12.1% complex multimorbidity. This suggests that patients with more chronic conditions attend more often than others in their age-sex group. Adjusting for individual patient attendance is therefore required to produce reliable population estimates from data collected from patients sampled at GP encounters.
My final task was to develop a parsimonious model to predict patient GP-visit rate, testing the assumption that the number of chronic conditions is driving GP service use. In Survey 2, the number of diagnosed chronic conditions alone accounted for a significant proportion of the variance (25.5%) in patient GP-visit rate. The number of body systems involved also explained a significant proportion of variance (23.9%). Including patient age, sex and Commonwealth concession health care card status only marginally increased the predictive value of the model to 27.9%.
In summary, this thesis demonstrates a practical method of measuring multimorbidity in Australia, using GPs as expert interviewers and adjusting for each patient’s individual attendance. I have shown that to produce robust results that can be compared with other studies, multimorbidity researchers should ideally define multimorbidity as 3+ conditions and include as many chronic conditions as possible in their study. Finally the measure has practical application as the number of diagnosed chronic conditions in an individual is the most significant driver of general practice service use. The results of this research will help inform health policy makers in their response to the challenges posed by continued growth in the prevalence of multimorbidity.