Aim: To investigate movement of people from hospital into residential aged care.
Methods: An innovative record linkage method was implemented to create a national database to investigate transitions from hospital into aged care.
Results: In 2001–2002, 3.2% of hospitalisations for people aged 65+ ended with admission into residential aged care. A further 5.5% were for people already living permanently in care. Nationally, more people were admitted into permanent care from hospital than from the community. Factors important in predicting admission to aged care from hospital included length of hospital stay, diagnoses, region of usual residence and hospital jurisdiction.
Conclusion: Individually, national hospital and aged care datasets do not provide adequate information on movement between the sectors. Linking the data allowed the first national investigation into movement from hospital into aged care. Results indicate the importance of investigating interactions of service provision (both supply and demand driven) at the local level.
The interface between acute hospital care and residential aged care (RAC) has long been recognised as an important issue in aged care services research [1,2]. The policy significance of cross-sector movement was brought to the fore in 2001 with the establishment by the Australian Health Ministers' Advisory Council (AHMAC) of the Care of Older Australians Working Group (COAWG) , and subsequently by several COAWG-sponsored research reports [4–10].
Despite the importance of this interface, existing national data provide poor information on movement between the two sectors. Current national datasets are derived from routine data collections, and provide information on the specific program that they describe, not program interfaces. For example, in the hospital data residents returning to RAC and people returning to their home in the community cannot be distinguished . Results from previous linkage studies showed that often residents who were returning to permanent care in RAC were reported in the hospital data as newly going into RAC rather than going home to their usual residence, and vice versa [9,12]. Previous estimates based on preliminary linkage work suggested that around 3% of hospital discharges for people aged 70+ are associated with admission to RAC [13,14].
Other important gaps in understanding the interface between the two sectors include the absence of information on where people were prior to their admission into care, the proportion of people coming into RAC via hospital, whether people admitted from hospital have different characteristics from those coming directly from home and the causes of admissions to hospital for people already living in RAC.
Over recent years the Australian Institute of Health and Welfare (AIHW) has developed an event-based record linkage method to link national hospital morbidity data and RAC data; the effectiveness of this method has been established both through theoretical analysis and direct comparison with a well-established name-based linkage strategy [9,12,15–17]. In 2006–2007, AHMAC's Statistical Information and Management Committee (SIMC) provided funding support to establish a linked database and to produce a demonstration report containing the first ever national statistics on movement from hospital to RAC. National data were linked to identify unidirectional movement straight from hospital to RAC. The purpose of this article is to describe the movement of people from hospital into residential care at both the national and state and territory levels. Factors affecting length of stay in hospital are investigated, and predictors of admission into RAC on discharge from hospital are discussed.
Ethical and privacy issues
Approval to use administrative data for a particular project is required from all data custodians, and ethical clearance is needed. For the current project, permission to use the data was obtained from the relevant data custodians in nine jurisdictions (each state and territory and the Australian Government). In addition, ethics approval for the project was obtained from two ethics committees, as specified by the data custodians and the AIHW (all but one jurisdiction accepted approval from the AIHW Ethics Committee). To protect the privacy of individuals, linkage was carried out by the AIHW using the Institute's privacy protocol  and under the protection of the AIHW Act.
To identify transitions from hospital into RAC, hospital episode discharge data for 2001–2002 (held on the National Hospital Morbidity Database, or NHMD) and RAC entry data for the same period (recorded in the Aged and Community Care Management Information System) were linked using AIHW's event-based matching strategy. This method identified links between events using demographic variables in conjunction with available event date information and event descriptors, matching people's hospital discharges to their relevant entries into RAC . The event-based method was used as name data were not available for record linkage.
Direct comparisons with a name-based linkage method using Western Australian data showed that event-based linkage more often tends to miss matches than make false matches . As a consequence, event-based linkage underestimates the total number of transitions from hospital to RAC by about 10%. To allow for this undercount, national estimates of flow were calculated using adjustment factors based on the results from the comparative study . Because of the limited nature of data available from the comparative study, estimates of flow were derived only at the national level. For estimates at the finer level, we estimated ranges of the proportion of live hospital discharges relating to transition into RAC.
Within the above approach, several types of transitions were identified:
• Returns to care by people already living permanently in RAC
• Admissions into permanent RAC
• Admissions into respite RAC
Hospital episodes that lasted less than a day were excluded, as were hospital episodes ending with the person remaining in the hospital sector (e.g. hospital transfers).
A logistic regression model was fitted to identify factors associated with whether a person left hospital to be admitted to RAC rather than returning to their home in the community. Fitting such models allows simultaneous consideration of factors affecting entry into RAC from hospital . Variables used in the analysis were: age at hospital admission; sex; state/territory of hospital; remoteness of usual residence (prior to hospitalisation); marital status; English proficiency (EP) group (based on reported country of birth ); hospital sector; care type in hospital prior to discharge; hospital mode of admission; length of hospital episode; principal diagnosis of hospital episode; and presence or absence of specific diseases as additional diagnoses. An overall Type I error rate of 5% was maintained by using a Bonferroni adjustment for multiple comparisons . It should be noted that length of hospital stay is truncated when people have changed care type or transferred within the hospital system (Box 1).
Box 1: Technical note on estimates of length of stay in hospital
In two cases the patient remains within the hospital system at the end of a hospital episode: when a patient changes from one hospital episode care type to another and when a patient transfers from one hospital to another . Because a person identifier is not available on the National Hospital Morbidity Database it is not possible to string together a person's episodes of hospital care into a single stay. Consequently, measures of length of hospital episode understate the total length of stay in hospital to the extent that people change care type or transfer within the hospital system. The likelihood of the exit episode being the only one related to the period in hospital varies with transition group. Analysis has indicated that estimates of length of stay for people who were admitted into residential aged care were more affected by within-hospital changes than those who returned home; actual differences are likely to be greater than the estimated differences as presented in this paper.
Estimates of flow
The flow of people can be examined from two viewpoints: (i) looking at the destination of people who are leaving hospital (hospital discharges); and (ii) looking at where people were before entering permanent or respite RAC. During 2001–2002, there were nearly 948 200 hospital discharges for stays lasting at least one night for people aged 65+. Of these hospital discharges, an estimated 86% were for people returning to the community (including a small number going into non-RAC institutions) and an additional 5.4% ended with death (Table 1). The remaining 8.7% (82 500) were for discharges into RAC, either as returns to permanent RAC or as new admissions. Overall, 3.4% (30 400) of live discharges from hospital (or 3.2% of all hospital discharges) were for people then admitted into RAC and 5.8% (52 000) were for people returning to permanent RAC.
Table 1. Movement from hospital into RAC, people aged 65+, Australia 2001–2002 (hospital separations)
Type of movement
Adjusted estimates of flow
Live separations (%)
Separations into RAC (%)
Age is as at time of hospital admission. Adjusted numbers for movements from hospital are rounded to the nearest hundred. Totals may differ to the sum of their components due to rounding. †Based on linked hospital and RAC records. Same day and next day re-admissions in to permanent RAC are treated as transfers and so have been combined into a single period of care when identifying returns to RAC after hospital leave. ‡Based on unlinked records, excluding deaths as reported on the hospital data. ‘Other’ includes non-RAC institutions (such as prisons, shelters and group homes providing primarily welfare services). NA, not applicable; RAC, residential aged care.
From the perspective of RAC, there were just over 99 900 admissions into RAC, either from the community (again including a small number coming from non-hospital institutions), from hospital or through within-RAC transfers. More people made the transition to RAC on a permanent basis via hospital (57% of non-transfer permanent admissions) than from the community (43%) (Table 2). For respite care, however, admissions from the community accounted for almost four times as many respite admissions as did those from hospital.
Table 2. Admission into RAC, people aged 65+, Australia 2001–2002 (RAC admissions)
Type of movement
Adjusted estimates of flow
Age is as at time of RAC admission. Adjusted numbers for movements from hospital are rounded to the nearest hundred. †Based on linked records, estimates of transitions between hospital and RAC vary slightly depending on whether movements from hospital or into RAC are being examined due to transitions occurring across either the beginning or end of the financial year. ‡Based on unlinked records, excluding identified transfers between facilities. ‘Other’ includes non-hospital institutions (such as prisons, shelters and group homes providing primarily welfare services). §Transfers between RAC facilities (unadjusted). 41% of transfers into permanent RAC were from respite RAC and 87% of transfers into respite RAC were from respite RAC. NA, not applicable; RAC, residential aged care.
These direct estimates are similar to those based on reported post-hospital destination on the NHMD . This similarity may appear to suggest that the NHMD data item reporting admission to RAC as the post-hospital destination is suitable for use in analysing admissions from hospital into RAC. However, comparisons of reported destination and the transition as identified through data linkage showed that about 30% of people reported as being admitted into RAC were in fact returning to care in RAC [12, table A6.2]. In addition, a similar number of discharges, which were reported in the NHMD database as returning to their usual residence (in RAC for permanent RAC residents), were in fact people being newly admitted into RAC. Thus, while the aggregate estimates of flow from hospital to RAC admission were similar from both data sources, this is a serendipitous effect resulting from the cancelling out of data coding errors. Consequently, while the total number is comparable, the individual patients in each sub-population are substantially different.
There were differences between the states and territories in the movement patterns from hospital to RAC (Figure 1, Table 3). For example, the proportion of live hospital discharges corresponding to admission into RAC varied with state, ranging from just over 2% for Tasmania and the two territories to over 3.5% for New South Wales. There was also a substantial difference across the jurisdictions in the ratio of permanent to respite admissions after leaving hospital. Nevertheless, for all states and territories there were more moves from hospital into permanent RAC than into respite care.
Table 3. Key statistics on transitions into RAC, people aged 65+, state/territory, Australia 2001–2002
Because of the limited nature of data available from the comparative study, adjusted estimates of flow were derived only at the national level. For estimates at the finer level, estimated ranges of the proportion of live hospital discharges relating to transition into RAC are presented. Ratio of admissions from hospital to admissions from community/other is a minimum number based on unadjusted numbers. Transition data are based on linked and unlinked hospital and RAC records. See notes to Tables 1 and 2 for information on identification of transition groups. HACC usage figures are indicative only as participation in the HACC national minimum dataset varied with jurisdiction – see table 1 in . Age is as at time of RAC admission for admissions and as at hospital admission for hospital discharges. State/territory is state/territory of RAC facility for admissions and state/territory of hospital for hospital discharges. Sources: derived from data in ,  and . HACC, Home and Community Care; RAC, residential aged care.
Transitions from hospital
Median length of stay (days)
To permanent RAC
To respite RAC
As per cent of live discharges from hospital
All to RAC, includes return to permanent RAC (estimated range)
To RAC admissions (lower limit)
Number (lower limit)
All to RAC
RAC admissions from hospital
As per cent of transitions from hospital to RAC
Permanent to respite admissions (ratio)
Admissions into RAC
Admissions from hospital versus admissions from community (ratio)
Into permanent care
Into respite care
Provision of aged care
Places and packages per 1000 (70+ and Indigenous population aged 50–69) 30 June 2002
Average RAC occupancy rate (2001–2002)
HACC clients per 1000 (70+) 2001–2002
Results show that hospitals are an important entry point into RAC. Nationally, the most common pathway into permanent residential care was via hospital (Table 3). This pattern held at the state level for New South Wales, Victoria and South Australia. However, in Western Australia, Tasmania and the two territories the most common pathway into permanent RAC was from the community.
Length of hospital episode
The length of the hospital episode prior to discharge varied considerably with post-hospital destination (Figure 2). People who returned to their home (either in the community or in RAC) tended to have shorter stays than other people. On the other hand, people who were newly admitted into RAC tended to have relatively long stays. The 90th percentile for people admitted into permanent RAC on discharge from hospital was 73 days. This finding raised the question as to whether post-hospital destination (i.e. RAC or the community) was the key determinant in length of stay, or whether other factors such as diagnosis and type of care were also important.
People can receive a range of types of care while in hospital, depending on the main clinical intent of the hospital episode. These include acute care, rehabilitative care, palliative care, geriatric evaluation and management (GEM), psycho-geriatric care and maintenance care. In general, acute care episodes tended to be shorter than other types (median 4 days in 2001–2002), with palliative care also being relatively short (median 8 days) (Table 4).
Table 4. Hospital separations for people aged 65+: median and 90th percentile of length of stay, by hospital care type and movement type, Australia, 2001–2002
Hospital care type
Returning to permanent RAC
To permanent RAC
To respite RAC
To community/ other
Died in hospital
90th percentile (days)
Table is based on linked and unlinked hospital and RAC records (see notes to Table 1). Table excludes same-day hospital episodes, statistical discharges and transfers to other hospitals. Age is as at time of hospital admission. Care types Newborn care, Organ procurement – posthumous and Hospital boarder were excluded from the table (1 case). 6107 records with unknown care type were excluded from the table. All of these records related to separations from Tasmanian hospitals; 6095 were from private hospitals. GEM, geriatric evaluation and management; NA, not applicable; RAC, residential aged care.
Total hospital separations (number)
Length of stay varied both with hospital care type and with post-hospital destination within care type (Table 4). Within care type, among those who did not die in hospital, people who moved into permanent RAC on discharge were likely to have spent longer in hospital than others. For example, the median length of acute episodes prior to discharge was 21 days for those admitted into permanent RAC compared with 12 days for those admitted into respite RAC and 5 days for those returning to their home in the community.
Different hospital principal diagnoses were also associated with different lengths of stay (e.g. median of 4 days for circulatory disease and 12 days for mental and behavioural disorders ). However, once again – within health condition group – people who moved into permanent RAC on discharge were likely to have spent longer in hospital than others. Apart from episodes for people with a principal hospital diagnosis related to eye disease, the median length of the hospital episode prior to discharge was between 12 and 22 days longer for people who were admitted to permanent RAC on discharge compared with those who went home, and 5–9 days longer for people admitted into respite RAC, depending on the principal diagnosis (grouped into ICD-10-AM chapters).
While length of hospital stay for people returning to their home in the community varied little with jurisdiction, the time spent in hospital prior to admission into RAC did. The two states with the lowest relative use of respite care (Victoria and Tasmania) tended to have relatively short stays in hospital prior to admission into respite care (medians of 10 and 7 days, compared with 14 days nationally) (Table 3). Victoria, Western Australia, Tasmania and the Australian Capital Territory had relatively long stays prior to admission into permanent RAC (medians of 28 days or more, compared with 25 days nationally).
The cause of observed differences in length of hospital episode by post-hospital destination cannot be gauged from the datasets used for this analysis. However, people who were discharged to RAC tended to have more diagnoses recorded during their final hospital episode prior to discharge (Figure 3). This suggests that these patients had high levels of comorbidities affecting their hospital treatment. In addition, since people cannot enter RAC without first having an assessment by an Aged Care Assessment Team (ACAT), at least some of this additional time in hospital was due to the assessment process and the time needed to identify and arrange transfer to a suitable RAC facility. In 2003–2004, the median time taken for an ACAT to complete an assessment for people first seen in hospital varied between 3 and 10 days, depending on the jurisdiction, and the 90th percentile was between 16 and 45 days (compared with an overall median of 19 days and 90th percentile of 69 days) .
Propensity to be discharged to RAC
A total of 844 800 hospital discharges were for older people who either returned to the community (96.4%) or who left hospital to be newly admitted into RAC (3.6%); that is, they did not die in hospital and they were not already RAC residents. The fitted logistic regression model indicated that the most significant predictors of a higher rate of entry into RAC rather than a return to the community were:
• Principal and additional diagnoses such as ‘Awaiting admission elsewhere’ and ‘Dementia and related disorders’
• Longer stay in hospital
• Older age
• Usual residence in major cities
• More than one ‘episode of care’ during the period of hospitalisation
• State or territory of hospital admission (people in hospitals in New South Wales and South Australia were more likely than others)
• Care type prior to hospital discharge (acute and rehabilitative care were associated with lower probabilities)
Among lesser effects, being male reduced the probability of entering RAC while being widowed or single (that is, unlikely to have a co-resident carer) increased the probability, as did being born in a high English proficiency country. More detailed analyses and tables supporting these findings can be found in the technical report Movement from hospital to residential aged care.
Linking hospital and RAC data has proven to be even more technically challenging than the team envisaged; it also required extensive stakeholder consultation and a range of ethical clearances. The resulting database has demonstrated descriptive value and substantial analytic flexibility. In this article, we have presented previously unobtainable estimates of patient flows from acute hospitals to the residential care sector at the national level, and provided some indication of the enormous potential of linked datasets of this kind. Further analyses could explore and explain state differences in patterns of care and transitions in addition to changes over time, with a view to identifying causal links, such as the impact of access to health services or admission and transition rates, and the cost implications of particular policies and the associated patterns of service use.
The purposes of this article are both methodological and substantive. First, the findings reported here can be seen as indicative, alerting other researchers to the potential of linking routine program data for questions relating to the residential/acute care interface. Second, the findings are of substantive interest in their own right, providing key information on the number and proportion of older people who leave hospital and enter RAC (3.4% of live discharges) and their associated attributes (e.g. diagnosis, length of stay, age, sex). The analyses reported here also demonstrate that there are significant, and as yet unexplained, variations in patterns across the states and territories, suggesting that these flows between the sectors may be amenable to shifts in policy, changes in service mix and availability or patterns of practice.
The key considerations in health policy in the 21st century relate more to how incentives and disincentives work within the system to produce particular patterns of care and outcomes, rather than the effects of simple increases or decreases in the number of constituent elements in the system. The empirical questions, which inform these policy considerations, relate more to patient flows (and the characteristics of those patients) than to descriptive headcounts at any one time. While the number of beds in a hospital is indeed an absolute constraint on the number of people that can be cared for on a particular day, it is how those beds are used, which determines how many people are assisted over, for example, a one-year period. And for complementary systems such as hospitals and residential care, understanding the inter-dependencies of the two systems (expressed in the flows of patients between them) is a critical element in addressing how the health-care sector works to support and maintain the health and well-being of older people and their families.
That service provision practices are influencing hospital outcomes is indicated by the significant effects that the region of a patient's usual residence and the jurisdiction of the hospital had on the probability of admission into RAC from hospital. The jurisdictional differences in the use of respite (versus permanent) care following admission from hospital and the varying pre-eminence of hospital as a source for RAC admissions (Table 3) support this. However, there appears to be little relationship between the proportion of admissions into RAC that come from hospital and either the provision of RAC places or the average RAC occupancy rate for a state. Similarly, there is no evidence that either provision of community packages or relatively low average usage rates of Home and Community Care services across a jurisdiction are associated with a transition to RAC after a period in hospital. These latter findings suggest that the regional effects are most likely within jurisdiction.
Using linked data to investigate transition patterns from hospital to aged care allows the hospital–aged care interface to be examined in new ways. In particular, the results presented in this paper point to the importance of investigating interactions of service provision on post-hospital destination at the local level.
AHMAC's SIMC provided funding support to this work. The authors thank the Department of Health and Ageing for permission to use their RAC data for this project, and Peter Braun (AIHW), who prepared the RAC data for use in the project. Thanks also go to the jurisdictional data custodians for permission to use their hospital data, and to Katrina Burgess and Christina Barry (AIHW), who prepared the hospital data for use in this project.
• Linking hospital discharge and aged care entry data allowed direct investigation into national flows from hospital to aged care for the first time.
• An overwhelming majority of older people (96.6% of live discharges) returned to their usual residence following hospitalisation; 3.4% of live discharges were for people newly admitted to RAC.
• More people were admitted into permanent care from hospital than from the community.