The new dynamics of residential aged care in Australia: continuity and change

Funding of residential aged care in Australia has used ratings of dependency-based care needs since the Resident Classification Instrument (RCI) was introduced in mid-1988. The RCI applied only to nursing homes, and was replaced by the Resident Classification Scale (RCS) in 1997, with eight categories covering both nursing homes and hostels as part of a move towards integration of ‘high’ and ‘low’ care. This move was furthered with the introduction of the Aged Care Funding Instrument (ACFI) from 2008, and the distinction between high and low care was removed altogether from mid-2014. The ACFI comprises 64 categories of funding based on ratings at four levels across activities of daily living, cognition and behaviour, and complex health care.¹

Although the ACFI was intended to rebalance funding towards residents with high-care needs, marked increases in ratings saw Commonwealth recurrent funding rise 6.6% p.a., from $AUD6.5b in 2008–09 to $AUD13.0b in 2018–19,^2,3 far in excess of the 1.3% p.a. growth in permanent residents over that period. This divergence gives compelling grounds for investigating the extent to which funding growth was driven by increases in resident dependency and care needs vis-à-vis other factors at play.

Our analysis follows on from a study of dynamics associated with steady increases in dependency ratings over the 7 years 1998–99 to 2005–06, which found no association with increased turnover.⁴ Our analysis also provides baselines for monitoring future continuity and change; such information is especially timely as concerns over expenditure trends led to the development of the Australian National Aged Care Classification (AN-ACC), expected to be phased in during 2021.⁵

The specific objectives of this study are to:

1. develop a schema that aligns the eight RCS categories with the 64 ACFI levels and use it to summarise and track dependency ratings from 2008–09 to 2018–19;

2. use episode-level data in new analyses of turnover and length of stay (LOS), in addition to the person-level analyses of the earlier study;

3. test whether increasing dependency is associated with increasing turnover and reduced LOS; and

4. assess the implications for policy.

Data relating to all permanent residents in Commonwealth-subsidised residential aged care services in Australia for the 11 years from 1 July 2008 to 30 June 2019 were obtained from the National Aged Care Data Clearinghouse at the Australian Institute of Health and Welfare (AIHW). The dataset included dependency-level appraisal at admission and at separation, exact LOS, and client age at admission. Additional data published by the Data Clearinghouse were used to enumerate the number of residents at 30 June each year.

The earlier study was limited to published, aggregate data, and median LOS was interpolated from LOS distributions rather than calculated from data. Our dataset supports analysis at an episode level (a discrete period of care) and a person level (discrete periods aggregated for each person). Person-level analysis provides consistent comparison with the earlier study, whereas episode-level analysis better reflects changes in AIHW’s approach to reporting LOS, and allows for more in-depth investigation of system dynamics. The scope of the dataset and measures used are detailed in Box 1; multiple measures of LOS are used because of the highly skewed nature of the LOS distributions. Analysis was conducted using Tableau and Microsoft Excel.

To enable comparison of the RCS and ACFI, categories were mapped to a six-level schema of Common Care Groups (CCG). Primarily based on the dollar value [ranges] of the categories within the two scales at the time of transition,⁶ the schema also considered other properties noted in Box 2.

Box 2. Mapping of RCS and ACFI categories to 6 Common Care Groups, 2008

Following a summary of the mapping of the RCS and ACFI, three sets of results show the relationships between reported increases in dependency of residents and system dynamics.

The initial effect of introducing ACFI was a slight upwards shift in the care ratings of new admissions (Table 1), consistent with a policy objective of directing more funding to residents with higher-care needs.^7,8 The outcome was most pronounced in the higher-dependency categories, whereas the share of admissions in the two lowest CCGs fell marginally. The most noticeable change was the disappearance of the ‘waist’ at RCS 4 that reflected the requirement (under RCS) for resident dependency to increase by two categories for reclassification from low to high care.

Table 1.  Initial distribution of RCS and ACFI appraisals by Common Care Group, 2008

Over the 11 years, there was a strong trend towards higher-dependency ratings, whether viewed on admission, for residents at the end of each financial year, or at separation (Table 2). The largest increases were in CCG 1, and are most apparent in the ‘residents’ view, which shows more than an eight-fold increase to 2015–16, tapering in later years. A decline of the same order is seen for residents in CCGs 5–6, with progressive upwards shifts across other CCGs.

Table 2.  Trends in Common Care Group ratings for admissions, residents, final separations, and completed episodes, 2008–09 to 2018–19 (per cent)

Person-level turnover fluctuated slightly between 32.2 separations per 100 residents in 2008–09 and 34.3 per 100 in 2016–17 (Table 3). Recent person-level turnover is similar to that for 1998–99 to 2005–06, which ranged from 31.7 to 33.6 separations per 100 residents, with no evident trend.

Table 3.  Selected measures of residential care dynamics, by Common Care Group, 2008–09 to 2018–19
Note: Turnover uses CCG relevant to the numerator and denominator; LOS measures use CCG on admission, and are tabulated against separation year; n.a., not applicable

Episode-level turnover is higher in the earlier years at over 40 separations per 100 residents, but rather than increasing in line with higher-dependency ratings, this measure declines over the period. The very high turnover in CCG 1 in 2008–09 is due to the small number of residents in this group in the first year of ACFI, but this halved in 2009–10, and had halved again by 2016–17. The only exceptions to the gradual slowing were the steady turnover for CCG 3 and fluctuating turnover for CCG 6.

Average LOS conforms with the expected inverse relationship of shorter stays for higher dependency CCGs, but average LOS increased for each CCG and overall. Across all ratings, average LOS increased by 2 months to just on 2.5 years, but most notably, for CCG 1, this increased from 2.4 months to 16.6 months. The lower-dependency CCGs had longer average LOS at the start of the period, yet still almost doubled by the end.

Median LOS is shorter than average LOS, especially for the higher CCGs – a product of the skewed LOS distribution – but again increased for all CCGs over the 11 years. The median LOS overall of 17.3 months is not markedly shorter than the estimated median LOS over 1998–99 to 2005–06, which varied only slightly from 17.9 to 18.9 months.

The two final measures of the 90th percentile and proportion staying <6 months both show stays for all CCGs grew steadily, and considerably. In contrast to the expectation that increasing numbers in CCGs 1 and 2 would see more separations within 6 months, the proportions with such short episodes fell. At the same time, and despite the declining numbers in CCGs 5 and 6, the 90th percentile LOS for these groups increased.

Across the 11 years, there was a gradual upwards shift in age at admissions (Appendix Table 1). In 2008–09, 3.7% of admissions were for people aged <65 years, and 49.2% were for those aged ≥85 years; by 2018–19, the proportions were 3.0% and 52.9% respectively – most of this increase substituted for the drop in the proportion for people aged 65–84 years.

Length of stay was strongly inversely associated with age at admission. However, although LOS generally decreased as age increased, regardless of the measure used, it rose over time for all age groups, with one exception (65–74 years). Figure 1 demonstrates the highly skewed nature of LOS distributions; whereas outliers are more influential for the 0–49 years age group, by age 95 average and median LOS somewhat converge. For all age groups, these patterns persist over time.

Fig. 1.  Distribution of length of stay by age at admission, selected years.

This study has demonstrated striking changes in reported dependency ratings of residents in Australian aged care homes from 2008–09 to 2018–19, and highlighted how outcomes in system dynamics do not accord with the expected consequences of increasing resident dependency.

We have accounted for variations over time in conventions used for counting separations and derived measures, and such analysis was facilitated by the use of multiple measures. It was further aided by mapping the RCS and ACFI to six categories, enabling continuity across the transition years and simplifying comparison of dependency groups.

All measures showed a slowing of dynamics over the period: LOS increased and turnover fell, contrary to the expectation that more highly dependent residents would have shorter LOS leading to higher turnover across the system.

The more than four-fold increase in the highest ratings – and the converse trend among the lower ratings – cannot be accounted for by an aging–dependency effect. By 2018–19, 20% of admissions were in CCG 1 (rated high on all three ACFI domains) and another 43% were in CCG 2. The slight flattening in the second half of the decade likely reflects changes to the structure of the Complex Health Care domain of ACFI, which were introduced to curtail unexpected ACFI expenditure growth.⁷ Although it is not unreasonable that there is an upwards shift in aggregate ratings over time – given that residents can be reappraised when their care needs change, and dependency generally increases with advancing age – alternative explanations for the magnitude of the increases are discussed below.

Rather than showing change in system dynamics, the recent outcomes continue on from those following the introduction of the RCS,⁴ when the share of residents in the highest RCS category almost doubled from 12.2% to 23.4%. Both studies found higher-dependency residents had shorter LOS, but total median LOS changed little.

The results based on episode as the unit of analysis differ slightly from results published by the Australian Institute of Health and Welfare (AIHW) at various times,^11–13 and recently by Gibson¹⁴ using published AIHW data. The differences are largely due to changes in the treatment of transfers and readmissions and the lag effect of excluding episodes completed before 2008 in accumulating care days to person-level measures – this is particularly evident up to 2011–12 (Supplementary Table S3).

The first possible driver of higher dependency is the gradual increase in age at admission over the period. Until very recently, though, higher age was not associated with higher dependency rating on admission (Appendix Figure 1); for example, in 2008–09, 38% of people aged 50–64 years were rated in CCGs 1–2, compared with 29% of people aged 85–94 years. By 2018–19, the differences were less pronounced, and broadly pushed to higher ratings; 62% and 63%, respectively. Even so, LOS has been increasing within age groups, with the exception of the relatively small group aged 65–74 years.

A second possible driver is the effect of increasing use of residential respite, community-based and restorative programs (such as the Home Care Packages program) on delaying admission to permanent residential care. Several studies have shown varying associations between use of home-based care services and eventual take-up of residential care.^15–19 However, only Welberry et al., in a large-scale study of people with dementia,¹⁹ quantified the dependency level on admission to residential care against level of community-based care. They showed that the group who had received high-level home care were significantly more dependent on admission than those not using home care, but were not significantly more likely to die within 2 years than the comparison group.

A third possible driver is provider claiming behaviour. Auditing of more than 78 000 ACFI appraisals for the 5 years to June 2019 shows an average of 20.5% were downgraded each quarter.²⁰ A review of the ACFI in 2017 noted that consultancy and benchmarking services, and specialised ACFI coordinators, served to optimise ACFI income, and that some providers were using business models based on continuing ACFI revenue growth to build scale.⁷

Commonwealth action to control the growth of claims from 2016 – by way of changes to the ACFI and a freeze on indexation – saw the increase in the highest dependency ratings flatten, but also put providers under financial pressure, creating incentives to maximise claims. These tensions are summarised in the Aged Care Financing Authority’s submission to the Royal Commission into Aged Care Quality and Safety in 2019, observing that providers point to increasing frailty of residents (and consequently costs of care) and Government points to income-optimising behaviours of providers.²¹

A final driver is the small but persistent fall in occupancy over the period. The 22% expansion in operational places²² has considerably exceeded the 15% growth in permanent residents, and increases in admissions and separations have fluctuated from year to year, but neither has increased commensurately with capacity. Fewer residents for whom ACFI subsidies could be claimed has added to pressure on providers, but has not seen earlier entry of residents to fill ‘excess beds’, leading to subsequent longer episodes.

This paper underlines the value of monitoring dependency ratings, and has provided a means to follow changes across the transition from RCS to ACFI. With the likely adoption of AN-ACC, it will be critical to track the performance of the tool – and any associated perverse system dynamics – including changes in provider behaviour in response to policy changes, and vice versa. A technical mapping between ACFI and AN-ACC has been undertaken,²³ and further simplification for monitoring purposes would be valuable.

Second, the paper highlights the importance of monitoring respite, community and restorative programs (duration of use, mode of exit and use before admission to permanent care) to assess targeting in relation to risk of admission and impacts on dependency and subsequent LOS. An update of the AIHW analysis of pathways in aged care²⁴ would be valuable for this purpose.

Third, the Common Care Group schema presented here emphasises the need for a classification system with sufficient categories to differentiate residents with differing care needs and costs. The aggregation of nearly three-fifths of residents into two of the six CCGs – equivalent to 11 of the 64 ACFI levels – by mid-2019 masks substantial heterogeneity in care ‘streams’, such as palliative care and dementia care.

Finally, there are policy implications of the impacts of dynamics on access. The interplay between access, care ratings and LOS/turnover is most evident with regard to very long episodes. For example, in 2018–19, there were 838 separations of people admitted at CCG 6 who had LOS above the 90th percentile; these episodes accounted for 3.3 million care days, equivalent to some 12 000 episodes at the median LOS for CCG 1 in the same year. Hence, better understanding of these dynamics is important for informing the capacity of the system to meet the equity objectives of the Aged Care Act 1997, particularly as these pertain to certain geographies, provider types and special needs groups.

The major strength of this study was the use of a large administrative dataset with episode-level exact LOS – this avoids the bias inherent in other methods of estimating LOS. A further strength was the development of a schema for monitoring ratings across the RCS–ACFI transition.

A limitation was that it was not possible to analyse care ratings in the context of pre-admission eligibility assessment or use of home-based care programs. Joining up data sources to provide a comprehensive picture of assessed care needs and progressive care use would yield richer insights into the dynamics of the whole care system.^25,26

Using this study dataset, further work could include more in-depth analysis of particular subgroups of residents; system dynamics by provider type, location and service size; and relationships between LOS and mode of separation. The small group of residents with very long LOS especially warrants further investigation as they use a disproportionately large share of bed days. All of these topics would be relevant to monitoring the classification and funding scheme selected to replace ACFI.

The mapping of RCS and ACFI to a simplified, common rating enables the analysis of long-term trends in residential aged care dynamics. The analysis has shown a marked increase in dependency ratings over the decade to 2018–19, but found no evidence that the higher reported dependency levels have led to accelerated system activity. A comparison of these findings with those reported in an earlier study show more continuity than change.

This analysis provides a solid base for ongoing monitoring of care appraisals in the context of a possible new rating scheme. It highlights the potential interplay between policy changes and provider behaviour, and the need for robust validation of care appraisals. Analyses along these lines require that detailed data be designed into the system, and be routinely collected, interrogated, reported and available for research uses.

Finally, this study illuminates the relationships among service capacity, care patterns and turnover, in turn informing consideration of equity issues regarding access to residential aged care nationally, regionally and for particular client groups.

The authors declare that they have no competing interests.