Article Text
PDF
Abstract
Objectives As life expectancy increases, older people are living longer with multimorbidity (MM, co-occurrence of ≥2 chronic health conditions) and complex multimorbidity (CMM, ≥3 chronic conditions affecting ≥3 different body systems). We assessed the impacts of MM and CMM on healthcare service use in Australia, as little was known about this.
Design Population-based cross-sectional data linkage study.
Setting New South Wales, Australia.
Participants 248 496 people aged ≥45 years who completed the Sax Institute’s 45 and Up Study baseline questionnaire.
Primary outcome High average annual healthcare service use (≥2 hospital admissions, ≥11 general practice visits and ≥2 emergency department (ED) visits) during the 3-year baseline period (year before, year of and year after recruitment).
Methods Baseline questionnaire data were linked with hospital, Medicare claims and ED datasets. Poisson regression models were used to estimate adjusted and unadjusted prevalence ratios for high service use with 95% CIs. Using a count of chronic conditions (disease count) as an alternative morbidity metric was requested during peer review.
Results Prevalence of MM and CMM was 43.8% and 15.5%, respectively, and prevalence increased with age. Across three healthcare settings, MM was associated with a 2.02-fold to 2.26-fold, and CMM was associated with a 1.83-fold to 2.08-fold, increased risk of high service use. The association was higher in the youngest group (45–59 years) versus the oldest group (≥75 years), which was confirmed when disease count was used as the morbidity metric in sensitivity analysis.
When comparing impact using three categories with no overlap (no MM/CMM, MM with no CMM, and CMM), CMM had greater impact than MM across all settings.
Conclusion Increased healthcare service use among older adults with MM and CMM impacts on the demand for primary care and hospital services. Which of MM or CMM has greater impact on risk of high healthcare service use depends on the analytic method used. Ageing populations living longer with increasing burdens of MM and CMM will require increased Medicare funding and provision of integrated care across the healthcare system to meet their complex needs.
- Chronic Disease
- GERIATRIC MEDICINE
- Aging
- EPIDEMIOLOGY
- PUBLIC HEALTH
- Primary Care
Data availability statement
Data may be obtained from a third party and are not publicly available. Data that support the findings of this study are available from the Sax Institute, but restrictions apply to the availability of these data, which were used under licence for the current study and so are not publicly available. The data, however, are available from the authors upon reasonable request and with permission from the Sax Institute.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
A strength was that the analytic population included nearly quarter of a million community-dwelling adults aged from 45 to 90 years old, not just those engaged with health services.
A limitation was that the 45 and Up Study was not designed to be representative of the broader New South Wales population; however, observed cross-sectional exposure–outcome relationships have validity beyond the study.
A limitation was that multimorbidity (MM) and complex multimorbidity (CMM) were defined using the 11 chronic health conditions self-reported in the baseline questionnaire only.
A strength was that healthcare service utilisation was defined using records of hospitalisations and visits to emergency and primary care within linked healthcare administrative datasets.
A strength was that the effects of MM and CMM on high average annual healthcare service use were assessed using two alternative approaches to multiple regression analysis; with chronic disease count used as an alternative morbidity metric in sensitivity analysis.
Introduction
Multimorbidity (MM), co-occurrence of two or more chronic health conditions in an individual, becomes more common as people age. With advances in healthcare, increasing life expectancy and declining birth rates, proportions of older people in populations are increasing globally.1 Recent estimates indicate the proportion of the world’s population aged ≥65 years is 10.1%, while in Australia it is 16.7%.2 The population of older adults aged ≥65 years in Australia is projected to increase by 54% from 4.31 million in 2021 to 6.66 million in 2041.3 In 2017–2018, the estimated prevalence of MM in Australia by age group was: 11.6% for 15–44 years, 30.1% for 45–64 years and 50.5% for ≥65 years4; and increasing prevalence of MM with age has been shown internationally.5 6
Management of MM is complex; often requiring more intensive treatment and monitoring by general practitioners (GPs), specialists, nurses and other healthcare providers.7 8 People with MM are more likely to have higher healthcare service utilisation than people with one or no chronic condition.9–12 In English adults aged ≥18 years, an estimated 52.9% of GP consultations, 78.7% of prescriptions and 56.1% of hospital admissions are attributable to MM.10 About 90% of all Australians visit a GP at least once annually13; and the proportion of GP consultations attributable to MM and CMM are 52% and 27%, respectively.14 15
Previous international9 16 17 and Australian studies18–23 exploring the impact of MM on health service use showed greater health service utilisation among those with MM versus those without MM. However, limitations of Australian studies include that they have varied in how they defined MM, and most were restricted to subpopulations defined by sex,19 22 location (small urban areas or rural)18 20 or employment status.21 Some restricted their analysis to one modality of service use (emergency admissions or GP visits only)18 22 or to older adults ≥60 years old20; while others assessed the impact of specific comorbidities of interest on service use (cancer, psychological distress, or anxiety and depression).20 22 23
Complex multimorbidity (CMM; co‐occurrence of three or more chronic conditions affecting three or more different body systems) is an alternative concept intended as a more specific metric for assessing complexity of individual healthcare needs.15 While this metric provides lower prevalence estimates than MM and enables greater differentiation among older adults,24 it is unclear whether it provides greater insights into targeting patient care and health resource planning. As reported by the Primary Health Care Advisory Group in 2015,25 and noted by Ng et al,23 the Australian healthcare system is not optimally equipped to provide care for people with MM. Therefore, it is important to evaluate and understand the patterns of healthcare utilisation by people with MM and CMM in order to plan and develop models of integrated care aligning with the Australian Strategic Framework for Chronic Conditions.23 26 Internationally, CMM has been found to be associated with increased risk of mortality27 28 and healthcare utilisation.29 30 Though there has been limited evaluation of the impact of CMM on mortality in Australia,31 to our knowledge, no studies have evaluated the impact of CMM on healthcare utilisation in Australia or evaluated whether MM or CMM has greater impact on service use. Hence, the objective of our study was to measure and compare the impact of MM and CMM on healthcare utilisation in a large community-dwelling population of Australian adults aged ≥45 years.
Methods
Study design and population
We conducted a cross-sectional study among 267 357 older adults aged ≥45 years residing in New South Wales (NSW), Australia and enrolled in the Sax Institute’s 45 and Up Study which has been described elsewhere.32 Briefly, potential study participants aged ≥45 years in NSW were randomly sampled from the Services Australia Medicare enrolment database and invited to participate between 2005 and 2009. Services Australia maintains this database, which includes all those who have applied for Medicare enrolment and are accessing publicly funded health services. About 19% of those invited participated, accounting for ≈11% of the NSW population aged ≥45 years because not everyone ≥45 years living in NSW is listed in the Medicare enrolment database, and not everyone ≥45 years in the database was invited—it was a random sample. Participants consented to self-completing the baseline questionnaire and to long-term follow-up with linkage of survey data to administrative health records. Data collected via baseline questionnaire included sociodemographic and lifestyle characteristics, and self-reported health status and chronic conditions.33 We used a 3-year baseline period (the year before, the year of and the year after recruitment) to assess healthcare service utilisation to obtain a more accurate picture of participant’s service use for our analysis.34 We excluded from our analytic population those who: (1) withdrew from the 45 and Up Study (n=414); (2) had a date of birth in the 45 and Up Study dataset indicating they were <45 years old (n=7); (3) had a linkage error, multiple death records or an unusual date of death (n=20); (4) had completed their baseline survey before 20 February 2006, as death records from the NSW Register of Births, Deaths and Marriages (RBDM) were only available from that date (n=15 227); or (5) had died within the 3-year baseline period during which healthcare service utilisation was assessed (n=3193) (figure 1). This left a population of 248 496 participants for analysis.
Flow diagram—analytic population.
Data linkage and outcome ascertainment
Deterministic linkage of 45 and Up Study questionnaire data to Medicare claims data was facilitated by the Sax Institute using a unique identifier provided by Services Australia.
The Medicare claims data contained all claims from participants for benefits from Australian Government publicly funded Medicare health insurance between 2001 and 2017. We identified Medicare Benefits Schedule item numbers to ascertain claims for GP consultations (online supplemental table 1). We defined ‘high GP use’ as an annual average of ≥11 GP consultations during the baseline period.35
Supplemental material
The NSW Centre for Health Record Linkage probabilistically linked 45 and Up Study questionnaire data to: deaths data from NSW RBDM and hospital records from NSW Admitted Patient Data Collection (APDC) and NSW Emergency Department Data Collection (EDDC). The current estimated false positive (linkage error) rate for this probabilistic linkage is 0.05%.36
The RBDM contained all deaths data for study participants from 20 February 2006 onwards. We used these data to identify participants who died within the 3-year baseline period during which healthcare service utilisation was assessed so they could be excluded (see above).
The APDC contained all hospitalisation records for study participants between 2004 and 2018. We used these data to calculate the average annual number of hospitalisations for each participant during the 3-year baseline period. We defined ‘high hospital use’ as an annual average of ≥2 overnight hospital stays during the baseline period.34
The EDDC contained records for all emergency department (ED) presentations for participants between 2006 and 2018. We defined ‘high ED use’ as an annual average of ≥2 ED presentations during the baseline period.34 Given ED data were unavailable before 2006, we excluded people who completed the baseline questionnaire on or before 2006 for this outcome only.
Multimorbidity and complex multimorbidity ascertainment
We used self-reported chronic conditions in the 45 and Up Study baseline questionnaire to define MM and CMM.33 Study participants were categorised as having a chronic condition if they answered ‘yes’ to either of the following questions: “Has a doctor ever told you that you have (name of condition)?” and “In the last month have you been treated for (name of condition)?” The 11 conditions included in the baseline questionnaire (cancer (all types), heart disease, diabetes, stroke, Parkinson’s disease, depression, anxiety, asthma, allergic rhinitis, hypertension, thrombosis and musculoskeletal conditions) were classified into nine categories based on the body systems they affected, as per the International Classification of Primary Care, 2nd edition (ICPC-2) (online supplemental table 2).37 As per Harrison et al, MM was defined as the co-occurrence of two or more chronic health conditions in an individual, and CMM was defined as co-occurrence of three or more chronic conditions affecting three or more different body systems within an individual.24
Statistical analysis
We categorised continuous variables, calculated values for health status variables (eg, body mass index and psychological distress38) and included one additional category as ‘missing’ for variables with ≥5% missing values to maintain power in our analysis. We generated a stacked area plot to display the count of chronic conditions in individuals with increasing age, using 5-year age groups between 45 years and 90 years and one group ≥90 years. We calculated frequencies and proportions for participant characteristics and compared participant characteristics by MM and CMM status using χ2 tests. We calculated the frequencies and proportions of study outcomes (high hospital use, high GP use, high ED use) for participants by each chronic condition and body system affected, and by MM and CMM status. We used simple and multiple Poisson regression modelling to estimate crude and adjusted prevalence ratios (PRs) with their 95% CIs to assess the impact of MM and CMM on health service use across the three settings (hospital, GP and ED).39 We applied a two-step process to select covariates for adjustments. First, we selected variables found to be associated with MM or CMM at p<0.20 using χ2 tests.40 Second, we applied the change-in-effect estimate method using the ‘chest’ package in R to select the final model for adjustment.41 We checked for effect modification for age groups by adding an interaction term into the Poisson model. We conducted subgroup analysis stratified by three age categories (youngest group (45–59 years), middle group (60–74 years) and oldest group (≥75 years)) to assess any differential effect of MM or CMM according to age as was observed in our previous study.31 We set 0.05 as the significance level for all statistical tests and used R V.3.6.3 software (R Foundation, Vienna, Austria) for data analysis and SAS V.9.4 (SAS Institute, Cary, North Carolina, USA) for data management.
Sensitivity analysis
If the effect of CMM on healthcare utilisation is assessed using binary categories with a comparison group for ‘no CMM’ which includes participants with MM, this may underestimate the true impact of CMM. Also, the effect of MM on healthcare utilisation may be overestimated by including participants with CMM in the group of participants with MM. To address this, we conducted a sensitivity analysis by categorising our morbidity exposure variable into three different categories with no overlap in morbidity status: participants with neither CMM nor MM were the reference category, participants with MM but without CMM were the second category, and participants with CMM only were the third category. We then followed the same steps as we did in the primary analysis. In a further sensitivity analysis that was requested during peer review, we assessed the impact of morbidity status on healthcare utilisation by means of a count of chronic health conditions (also called a disease count).30 Instead of having categories defined by MM and CMM status, we assessed the impact of having the following numbers of chronic health conditions on healthcare utilisation: zero, one, two, three, four and five or more conditions. We then estimated crude and adjusted PRs with 95% CIs using the same modelling approach across three settings as per the primary analysis.
Patient and public involvement
Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans for this research.
Results
The analytic population comprised 248 496 participants (figure 1), of which 46.4% were male and 46.6% were aged 45–59 years at baseline. Among the analytic population, 75.5% had at least one chronic condition, 43.8% had two or more, 20.7% had three or more, and 8.3% had four or more: with the number of chronic conditions increasing with age (figure 2). Overall prevalence of MM was 43.8% versus 15.5% for CMM (table 1). The prevalence of MM and CMM increased as age increased: 32.1% and 9.2% in the youngest group (45–59 years), 50.1% and 18.8% in the middle group (60–74 years) and 63.0% and 26.0% in the oldest group (≥75 years), respectively. Males had higher prevalence of MM than females (44.7% vs 43.1%), while females had higher prevalence of CMM than males (16.2% vs 14.7%). Prevalence of MM and CMM was much higher among participants not working (55.3% and 22.3%) compared with those working part-time (36.0% and 11.0%) or full-time (30.5% and 7.4%). Participants with lower household income, those without current partners, current smokers, those not doing adequate physical exercise, those without self-reported good quality of life, those with high psychological distress and those requiring help with daily activity were more likely to have higher prevalence of MM and CMM compared with their counterparts.
Multimorbidity (MM) and complex multimorbidity (CMM) by participant characteristics
Number of chronic conditions and proportion for each by participant age group.
The most frequent chronic conditions among participants were cancer (34.5%), hypertension (30.9%) and depression or anxiety (18.4%) (online supplemental table 2). The most frequent body systems affected by at least one chronic condition were the cardiovascular system (50.7%), the respiratory system (22.0%) and the psychological system (18.4%). Regarding chronic conditions, participants with stroke had the greatest proportion of high healthcare service utilisation across three settings: 4.3% for hospital, 47.9% for GP and 5.7% for ED. Regarding body systems, participants with morbidity affecting the neurological system had the greatest proportion of high healthcare service utilisation across three settings: 2.9% for hospital, 42.5% for GP and 4.3% for ED.
Impact of MM and CMM on high hospital use
Mean annual hospital admissions was 2.25-fold higher for those with MM than without MM (0.27 vs 0.12 admissions), and crude prevalence of high hospital use was 4.23-fold higher with MM than without MM (1.7% vs 0.4%) (table 2). After adjusting for covariates, MM was associated with a 2.26-fold (95% CI 2.04 to 2.49) increased risk of high hospital use. Stratified analysis by age group showed that the impact of MM on high hospital use was highest in the middle group (60–74 years) and lowest among the oldest group (≥75 years).
Impact of multimorbidity (MM) and complex multimorbidity (CMM) on high hospital use (≥2 average annual admissions within −/+1 year of recruitment)
Mean annual hospital admissions was 2.27-fold higher with CMM than without CMM (0.34 vs 0.15 admissions), and crude prevalence of high hospital use was 4.12-fold higher with CMM than without CMM (2.7% vs 0.7%) (table 2). After adjusting for covariates, CMM was associated with a 2.08-fold (95% CI 1.91 to 2.26) increased risk of high hospital use, and stratified analysis by age group showed a similar pattern to that observed for MM (highest in the middle group (60–74 years) and lowest in the oldest group (≥75 years)). Impact of CMM on high hospital use was similar to that for MM (crude PR 4.12 vs 4.23; adjusted PR 2.08 vs 2.26). However, in our sensitivity analysis when we compared impact using three categories with no overlap in morbidity status, the impact of CMM on high hospital use overall was greater than the impact of MM (crude PR 6.82 vs 2.93) (online supplemental table 3). After adjusting for covariates, participants with CMM had a 2.97-fold (95% CI 2.66 to 3.33) increased risk of high hospital use, compared with those without CMM or MM.
Impact of MM and CMM on high GP use
Mean annual number of GP visits was 1.68-fold higher with MM than without MM (9.44 vs 5.62 visits), and crude prevalence of high GP use was 2.85-fold higher with MM than without MM (31.6% vs 11.1%) (table 3). After adjusting for covariates, MM was associated with a 2.02-fold (95% CI 1.98 to 2.06) increased risk of high GP use. When stratified by age group, the risk of high GP use associated with MM decreased linearly as age increased.
Impact of multimorbidity (MM) and complex multimorbidity (CM) on high general practitioner (GP) use (≥11 average annual visits within −/+1 year of recruitment)
Mean annual GP visits were 1.73-fold higher with CMM than without CMM (11.33 vs 6.55 visits), and crude prevalence of high GP use was 2.70-fold higher with CMM than without CMM (42.9% vs 15.9%) (table 3). After adjusting for covariates, CMM was associated with a 1.83-fold (95% CI 1.79 to 1.86) increased risk of high GP use, and stratified analysis by age group showed a decreasing trend with increased age as for that with MM. Impact of CMM on high GP use was slightly lower than that for MM (crude PR 2.70 vs 2.85; adjusted PR 1.83 vs 2.02). However, in our sensitivity analysis when we compared impact using three categories with no overlap in morbidity status, the impact of CMM on high GP use overall was greater than the impact of MM (crude PR 3.90 vs 2.31) (online supplemental table 4). After adjusting for covariates, participants with CMM had a 2.47-fold (95% CI 2.42 to 2.53) increased risk of high GP use compared with those without CMM or MM.
Impact of MM and CMM on high ED use
Mean annual number of ED visits was 1.81-fold higher with MM than without MM (0.29 vs 0.16 visits), and crude prevalence of high ED use was 2.99-fold higher with MM than without MM (2.6% vs 0.9%) (table 4). After adjusting for covariates, MM was associated with a 2.04-fold (95% CI 1.89 to 2.20) increased risk of high ED use. When stratified by age group, the risk of high ED use associated with MM decreased with increasing age.
Impact of multimorbidity (MM) and complex multimorbidity (CM) on high emergency department (ED) use (≥2 average annual visits within −/+1 year of recruitment)
Mean annual ED visits were 1.95-fold higher with CMM than without CMM (0.37 vs 0.19 visits), and crude prevalence of high ED use was 3.04-fold higher with CMM than without CMM (3.8% vs 1.2%) (table 4). After adjusting for covariates, CMM was associated with a 1.90-fold (95% CI 1.76 to 2.04) increased risk of high ED use, and stratified analysis by age group demonstrated a decreasing trend with increasing age as for that with MM. Impact of CMM on high ED use was similar to that for MM (crude PR 3.04 vs 2.99; adjusted PR 1.90 vs 2.04). However, as with other outcomes, when we compared impact using three categories with no overlap in morbidity status, the impact of CMM on high ED use overall was greater than the impact of MM (crude PR 4.37 vs 2.28) (online supplemental table 5). After adjusting for covariates, participants with CMM had a 2.58-fold (95% CI 2.36 to 2.81) increased risk of high ED use compared with those without CMM or MM.
Impact of disease count on high healthcare service use
If disease count was used as the morbidity metric instead of MM or CMM, mean annual hospital admissions increased with increasing disease count, with 0.30% of participants with zero conditions (overall) versus 5.39% of those with ≥5 conditions having high hospital use (online supplemental table 6). Crude prevalence of high hospital use demonstrated a dose–response relationship with increasing disease count and was 18.16-fold higher in participants with ≥5 conditions (overall) versus those with none. After adjusting for covariates, having ≥5 conditions was associated with a 5.34-fold (95% CI 4.42 to 6.46) increased risk of high hospital use overall. Stratified analysis by age group showed that the impact of disease count on high hospital use was greatest in the youngest group (45–59 years) and smallest in the oldest group (≥75 years). The adjusted PR for high hospital use in those with ≥5 conditions was 7.69 (95% CI 5.13 to 11.53) in the youngest group versus 3.17 (95% CI 2.40 to 4.21) in the oldest group. The outcomes of the analyses for high GP use (online supplemental table 7) and high ED use (online supplemental table 8) followed the same pattern established in the high hospital use analysis just described when disease count was used as the morbidity metric.
Discussion
This is the largest population-based study in Australia evaluating the impact of MM on healthcare utilisation, and the first Australian study evaluating impact of CMM on healthcare utilisation, comparing that with the impact of MM. Both MM and CMM were found to be common among people aged ≥45 years in NSW and their prevalence increased with age. In this population, 75.5% had at least one chronic condition, and overall prevalence of MM and CMM was 43.8% and 15.5%, respectively. Stroke and the neurological system, respectively, were the chronic condition and body system affected by morbidity with greatest impact on service utilisation. Across the three settings (hospital, GP and ED), MM was significantly associated with an approximately two-fold increased risk of high service use. And while CMM was also significantly associated with an increased risk of high service use, the point estimates were slightly lower than that of MM. Hence, when we assessed the impacts of MM and CMM on service use using binary categories in our primary analysis, MM was found to have greater impact than CMM. However, when we assessed the impacts of MM and CMM using three groups with no overlap in morbidity status (no MM/CMM, MM with no CMM, and CMM) in our sensitivity analysis, we found CMM had greater impact than MM. Notably, impact of both MM and CMM on high service use was greater for age groups ˂75 years than for the oldest group (≥75 years): being highest in the middle group (60–74 years) for hospital use, and highest in youngest group (45–59 years) for GP and ED use. When disease count was used as the morbidity metric instead of MM or CMM, findings across the three settings were similar: the mean number of service events increased with increasing disease count, and the PRs for high service use demonstrated a dose–response relationship with increasing disease count.
Our findings that MM and CMM were common, and their prevalence increased with age are consistent with Australian and international literature.4 10 14–16 Though not all chronic health conditions are degenerative, the onset and development of multiple degenerative conditions in individuals suggests acceleration of ageing processes.42 Mechanisms driving these conditions include chronic oxidative stress, immuno- and cellular senescence, mitochondrial dysfunction, defective DNA repair causing accumulating mutations, impaired autophagy, epigenetic changes and chronic low grade inflammation.42 43 Chronic conditions and failure of compensatory mechanisms lead to physical and cognitive impairment with functional decline and frailty, which impacts on the severity and burden of MM and CMM.43 Additionally, there appears to be increased risk of MM developing if cumulative lifestyle factors (inadequate physical activity, smoking, increased alcohol intake, obesity and unhealthy diet) are present.44
While no prior Australian evaluation was conducted on the impact of CMM on healthcare utilisation, several studies reported that older people with MM used more healthcare services than those without MM.9 16–19 However, they are not directly comparable due to the variation in the number and type of chronic conditions involved. A cross-sectional study, evaluating the impact of MM on healthcare utilisation and health status in people aged ≥50 years from 16 European countries, reported the mean number of annual visits to physicians by people with MM was twofold higher than people with no chronic disease.16 It also reported that increasing the number of chronic conditions by one was associated with a 49% increased risk of hospitalisation.16 A meta-analysis of 17 British studies measuring the impact of MM on healthcare costs and utilisation reported that people with MM had 2.56-fold higher odds (95% CI 1.88 to 3.47) of using healthcare services (including any GP visits, any hospital admissions and any ED visits) than people without MM.9 Similarly, in our study, MM was associated with 2.26-fold, 2.02-fold and 2.04-fold increased risk of high hospital, GP and ED use, respectively. A systematic review of 35 international studies (including 1 Australian and 23 American studies) also found that MM was associated with increased physician, hospital and ED service use.17 One cross-sectional study evaluating the association of psychological distress and MM with self-reported healthcare utilisation among adults aged ≥60 years in rural South Australia reported that compared with people without any chronic conditions, people with two conditions had 3.8-fold increased odds (95% CI 3.2 to 4.5) of GP consultation in the last year, 2.5-fold increased odds (95% CI 1.5 to 4.0) of hospitalisation and 2.0-fold increased odds (95% CI 1.2 to 3.4) of ED presentation.20 Our findings regarding increasing adjusted PRs for high healthcare service use across three settings with increasing disease count reflect those of a recent analysis in a Scottish ED where the ORs for ED admission and ED reattendance also increased with increasing disease count.30
Our study showed a trend indicating a greater impact of MM, CMM and disease count on high healthcare utilisation among younger versus older participants. While the absolute difference in the prevalence of high service use between MM and CMM groups versus groups without MM and CMM increased with age, the adjusted association between MM and CMM and high service use showed a decreasing trend with increasing age in two of the three settings (GP and ED use) and was lowest in the oldest group (≥75 years) in all three settings. For the adjusted association between disease count and high service use, this showed a decreasing trend with increasing age in all three settings. So, while the youngest group (45–59 years) had much lower prevalence of MM and CMM than the oldest group, the adjusted association with high service use was greater in the youngest group compared with the oldest group. Potentially, people in this youngest group may be more mobile, independent, capable of managing their own health and able to access care than older Australians.45 Also, a larger proportion of younger participants had no chronic conditions (among those without MM or CMM) than older participants (see figure 2), adding to the contrast in outcomes between those with MM or CMM and those without when outcomes were stratified by age. Notably, and similar to our findings in this and our previous study,31 Meng et al have shown greater impact of MM (or ‘high need, high cost’ in their analysis) on service use and mortality among middle-aged (40–64 years) versus older (≥65 years) adults (adjusted HR: 4.41 vs 2.44 for potentially preventable hospitalisations; 2.09 vs 1.75 for any ED visit; 7.20 vs 3.35 for mortality).46 We recommend targeting interventions towards people with MM and/or CMM in the youngest group (45–59 years) as that is where mortality risk is greatest (from our previous paper),31 to try to prevent acquisition of further comorbidities and to try to reduce the impact of MM and CMM on healthcare utilisation in the longer term. In older age groups, broader population-based strategies could be considered instead of targeted interventions designed for people with MM and CMM.
In our primary analysis, we found the impact of MM on high healthcare utilisation was higher than that for CMM. The impact of MM was increased by the group of participants with MM including those with CMM, and the relative impact of CMM on service use was decreased by the group of participants without CMM including those with MM. When we adjusted for this in our sensitivity analysis using three groups with no overlap in morbidity status, we found the impact of CMM on high service use to be greater than that of MM across all three settings. Hence, which of MM or CMM has greater impact on risk of high healthcare service use depends on the analytic method used and our sensitivity analysis approach lends support to the hypothesis that CMM may be a better metric for identifying patients with higher healthcare resource needs.24
One of the key challenges currently facing the Australian health system is increasing demand on health services due to the increasing prevalence and impact of MM and CMM within the ageing population.47 Life expectancy is increasing while the birth rate is falling and has been insufficient to sustain the population without immigration since 1976.48 The health system and workforce must be adequately resourced and better prepared to meet this increasing need,49 50 and health service provision should be further adapted to integrate care and better manage chronic health conditions and MM.51 52 There should be increased efforts in prevention to enable older adults to remain healthier for longer; and additional Medicare funding is needed to shift the health system focus from single disease to holistic management, and from reactive to proactive healthcare. Our recommendations align with key action areas identified in the United Nations Decade of Healthy Ageing (2021–2030) Action Plan: creating integrated and responsive healthcare systems and services; and ensuring access to long-term care for older people who need it.53
Strengths and limitations
The major strength of this study was our use of a large community-dwelling population of older adults which was not restricted only to those engaged with health services, thus providing a more realistic denominator. The 45 and Up Study was set up to investigate ageing, morbidity, mortality and patterns of health service use, and is the largest longitudinal cohort study of its kind in the southern hemisphere.32 Recruitment of individuals across the age spectrum from 45 to 90 years at baseline in the 45 and Up Study enabled us to assess the impacts of MM and CMM on health service use across that age range. Also, to our knowledge, this is the first Australian study to compare the effect of MM versus CMM on health service utilisation and our sensitivity analysis indicates an alternative analytic approach that can be used to assess the impact of CMM on high health service use versus that of MM without CMM.
Our study had several limitations. While non-response at baseline may mean the study population varies from the broader population, the 45 and Up Study was not designed to be representative of the general population and therefore does not necessarily provide population-level measures of MM and CMM.32 Though the 45 and Up Study baseline response rate was modest (≈19%), representativeness is not essential in cohort studies and observed cross-sectional exposure–outcome relationships have been similar to state-based surveillance systems that reported higher response rates.32 54 We used self-reported data on chronic health conditions from the baseline questionnaire to define MM and CMM without any clinical diagnosis, so non-differential misclassification might occur. However, no method or data source (eg, self-reported, hospital administrative or medication data) used to define morbidity status is perfect, and each method or data source has its strengths and weaknesses.55–58 Collecting data via self-report in surveys can identify chronic health conditions after they are diagnosed, but before people have been hospitalised with them; plus surveys enable collection of socioeconomic, functional and other participant characteristics of interest that are useful for analysis.32 58 Self-report can potentially be affected by recall bias, while medication data only capture those conditions that are treated with drugs and there are issues with drugs that can be used to treat multiple conditions.58 Hospital administrative datasets may only capture conditions that are relevant to specific hospital admissions and do not capture data for those who have chronic conditions but have not yet been hospitalised.58 Internationally, systematic reviews have found that counting chronic conditions, most often via self-reported data, are the most common method of defining MM in published studies and the use of MM defined this way to determine outcomes is valid.55–57 Self-reported data have been used to define MM in the majority of published Australian research, and in the 45 and Up Study self-report has been used as the reference standard when evaluating other methods of defining MM via hospitalisation and medication data.58 For most chronic conditions assessed by Lujic et al, there was greater than 70% agreement between self-reported data and hospital administrative data in the 45 and Up Study,58 notwithstanding the well-known issues of under-reporting in hospital administrative data.55 Yurkovich et al found in their review that while agreement can vary between self-reported versus health administrative data used to define morbidity status, both data sources had similar abilities to predict various health outcomes.56 Notably, our analysis only included the 11 chronic conditions listed in the baseline survey, but some other important chronic conditions which also increase the risk of high service use (eg, chronic liver disease, chronic renal disease, etc) were not included. Hence, the effect of MM or CMM might be underestimated due to non-differential misclassification bias. Finally, to define CMM, we classified self-reported chronic conditions into body systems using ICPC-2, rather than the clinically coded International Classification of Diseases, 10th edition (ICD-10). Given the chronic health condition data used to define CMM in our study were self-reported by participants in a community setting, and not coded by healthcare clinical and clerical staff in a hospital setting, we believe that ICPC-2 was more suitable than ICD-10 to use for classification in our study. When Harrison et al defined the concept of CMM, they evaluated different classification systems and found near perfect concordance between ICPC-2 and ICD-10 for defining and measuring CMM from chronic health condition data.24
Conclusion
Increased healthcare service use among older adults with MM and CMM demonstrates the impact of MM and CMM on demand for primary care and hospital services. Notably, the impact of both MM and CMM on high health service use was higher in the youngest group (45–59 years) compared with the oldest group (≥75 years), despite the prevalence of MM and CMM increasing with age. These findings were confirmed when chronic disease count was used as the morbidity metric in our sensitivity analysis. Which of MM or CMM has greater impact on risk of high healthcare service use depends on analytic methods used. An ageing population living longer with an increasing burden of MM and CMM will require increased Medicare funding and integrated care service provision across the healthcare system to meet their complex needs.
Data availability statement
Data may be obtained from a third party and are not publicly available. Data that support the findings of this study are available from the Sax Institute, but restrictions apply to the availability of these data, which were used under licence for the current study and so are not publicly available. The data, however, are available from the authors upon reasonable request and with permission from the Sax Institute.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by NSW Population and Health Services Research Ethics Committee (reference number 2016/06/642). The 45 and Up Study was approved by the University of New South Wales Human Research Ethics Committee. Participants gave informed consent to participate in the study before taking part.
Acknowledgments
This research was completed using data collected through the 45 and Up Study (https://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 Heart Foundation and the NSW Ministry of Health. We thank the many thousands of people participating in the 45 and Up Study. We acknowledge the NSW Centre for Health Record Linkage for data linkage and provision of the death data (https://www.cherel.org.au/). We acknowledge the Sax Institute’s Secure Unified Research Environment (SURE) for the provision of secure data access.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors All authors substantially contributed to the manuscript and met the authorship criteria. AK, AT, SA and MB conceived the study. AK, AT, SA, DPC, JJR and MB contributed to the design, analysis and interpreting the results. AK drafted the manuscript, AK and DPC coordinated its revision, and all authors critically reviewed the manuscript. All authors read and approved the final version of the manuscript. AK acts as the guarantor for the overall content.
Funding This research was funded by Sydney Local Health District, South Eastern Sydney Local Health District and the Central and Eastern Sydney Primary Health Network. Grant numbers N/A.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.