Cost modelling rehabilitation in the home for reconditioning in the Australian context

Descriptive analysis of Delphi data was undertaken using SPSS V27. Each Delphi participant’s estimated potential utilisation of RITH (by AN-SNAP class) was calculated from their estimate of the percentage of reconditioning patients suitable for RITH, multiplied by their estimate of the percentage of suitable reconditioning patients who might wish to take part in RITH if available. The median estimates for each class were determined.

Cost modelling was developed in MS Excel (Office 365) and was restricted to reconditioning classes 4AR1 to 4AR4. RITH program costs per episode were determined (Tables 3 and 4) and compared to the known inpatient cost per episode (Table 5). The assumptions and information sources underlying the calculations are detailed within each Table. Weighted averages for each metric in Tables 3, 4 and 5 were also determined, based on the modelled number of substitutable episodes per annum. The modelled number of substitutable episodes per annum is the number of inpatient reconditioning rehabilitation episodes that might be able to be substituted by RITH, limited to capital cities, and is based on the AN-SNAP reconditioning episodes reported to AROC in the 2019 (the last pre-Covid) year [2], adjusted for the potential utilisation of RITH determined from the Delphi survey (Table 2). The choice to limit the estimate to capital cities is conservative, since it assumes the operational efficiency available to RITH providers which arises from greater population density, may not be available elsewhere. Details of the calculations to determine the modelled number of substitutable episodes are shown in detail in Additional File 1.

As previously reported, over 130 statements on aspects of the patient journey achieved consensus (that is, at least 70% of participants agreed/strongly agreed) during the Delphi survey [18]. The subset of these statements that were considered by the project team to have the most direct impact on the cost modelling for RITH are shown in Table 1. These statements relate to staff, program features and budgetary factors, and will be referred to in the text as T1.Item No.

Twenty-one participants (of 78 participating in the third Delphi survey round) indicated clinical familiarity with the AN-SNAP classification and responded to the questions relating to potential RITH utilisation. These were rehabilitation medicine physicians (n = 9, 42.9%), rehabilitation nurses (n = 6, 28.6%), physiotherapists (n = 3, 14.3%) and occupational therapists (n = 3, 14.3%). They estimated that around half of all patients in the higher functioning classes (AN-SNAP classes 4AR1 and 4AR2) may utilise RITH if it were available, but that few patients in the lowest functioning classes (AN-SNAP classes 4AR5 and 4AR6) would likely do so. See Table 2.

The underlying assumption was that RITH should provide as comprehensive a rehabilitation service as would be provided by inpatient rehabilitation and be no more costly than inpatient rehabilitation (T1.15; T1.17). Cost modelling was restricted to the provision of RITH for patients in AN-SNAP classes 4AR1 to 4AR4 (which comprise around 73% of all inpatient reconditioning episodes in Australia [4]), as these were groups for whom Delphi survey participants estimated the potential utilisation would be greatest (Table 2).

Table 5 estimates the potential cost savings per episode to the public hospital sector in Australia of the proposed RITH program compared to inpatient rehabilitation for reconditioning. The weighted average public inpatient cost per episode of A$15,820 has been estimated with reference to publicly available Independent Health and Aged Care Pricing Authority (IHACPA) data points. Each AN-SNAP class has an IHACPA associated cost weight (ranging from 2.3 for 4AR1 to 4.11 for 4AR4) ([22], p.84) and the IHACPA National Efficient Price per single cost weight of activity in 2021–22 was A$5,597 ([22], p.7). Thus, the derived price paid for an inpatient reconditioning episode in this period ranges from A$12,850 for 4AR1 to A$22,976 for 4AR4 (weighted average of A$15,820). The modelled RITH costs (calculations shown in Table 4 and reappearing in Table 5) indicate hypothetical savings per episode ranging from 23.6% for 4AR1, to 32.6% for 4AR4 (weighted average of 28.1%).

A conservative estimate of potential cost savings to the Australian public hospital sector was quantified, using the modelled number of substitutable episodes per annum. Details on the derivation of the number of substitutable episodes are provided in Additional File 1. In summary, they relate to the provision of RITH in the capital city only [28] of each state or territory, and use the estimated potential utilisation of RITH for reconditioning for the four AN-SNAP classes 4AR1 through 4AR4 (Table 2), and the number of reconditioning episodes (4AR1 to 4AR4) reported to AROC in the 2019 year ([2], p. 92). Based on the above, an estimated 2,459 public inpatient reconditioning rehabilitation episodes could be substituted with a RITH program annually. The project team restricted the estimate to capital cities because the greater population density in these cities would allow RITH service providers operational efficiencies from having critical mass. This is a conservative approach since other large Australian cities may offer similar efficiencies. RITH in regional and rural locations, however, might require a greater allowance for staff travel costs.

Should 2,459 public inpatient reconditioning rehabilitation episodes be substituted with RITH, this would translate into hypothetical cost savings to government of A$10.9 million dollars annually (Table 5).

We have attempted to present a fully costed RITH hospital substitution program. While the actual amount of therapy intensity for inpatient rehabilitation is not formally reported, the Australasian Faculty of Rehabilitation Medicine (AFRM) Standards recommend that therapy should consist of a minimum of three hours per day over five days per week for patients who can tolerate it ([19], p.10). Published data suggest that actual therapy levels may be less, especially for older patients who are not presenting with a predominantly neurological impairment [29]. Our modelled allied health therapy intensity is an average of 1.7 h/day over 7 days, which is equivalent to an average of 2.4 h per day over five days. This does not include nursing input or clinical case management time, both of which are likely to include a therapy component.

Our costing model provides a total of allied health therapy per RITH episode that ranges from 20.1 (4AR1) through to 31.7 (4AR4) hours, which increases to a total of 26.7 h (4AR1) to 42.3 h (4AR4) with the inclusion of other direct clinical input (i.e., therapy plus clinical case management, plus nursing, plus rehabilitation physician time). Data from the AROC Ambulatory Report [21] indicate the mean number of occasions of service per episode for patients with a reconditioning impairment code in Australia in 2021 was 26.4 ([21], p.32). If we assume that an OOS is one hour (since it is not defined by AROC), then this is supportive of our model. Direct comparison is difficult however, because the AROC ambulatory dataset includes patients whose rehabilitation may have only been in an ambulatory setting, as well as those for whom it may have been a continuation of the inpatient rehabilitation episode; and the extent to which AROC-reported episodes were ‘hospital substitution’ is unknown. Interestingly, the mean length of program for the reconditioning impairment code in the Report is 62.8 days ([21], p. 26), which suggests therapy in the ambulatory setting was spread over a long duration. This would be more akin to an alternative RITH service delivery model which Delphi survey participants also supported [18] on the assumption that it would be of lower or varied intensity, delivered over a duration of up to ten weeks, and cost-equivalent to the intensive, shorter duration model costed in this paper.

Our modelled cost per episode of RITH for reconditioning ranges from A$9,817 (4AR1) to A$15,491 (4AR4), with the estimated weighted average cost being A$11,371. While we were unable to locate any publicly available data on the cost of other RITH programs for reconditioning as hospital substitution, some comparative data are available for other programs delivered in the home. The Transition Care Program (TCP) and the Short-Term Restorative Care (STRC) Program are Australian government funded programs targeting community-dwelling older people who have experienced functional decline and who have (in the case of TCP) and have not (in the case of STRC) experienced recent hospitalisation [30]. These programs offer short term support (8 – 12 weeks) with low intensity therapy aimed at improving function and reducing premature admission to residential aged care. The duration of the TCP and STRC programs are more akin to the alternative RITH program described above (that is, lower or varied therapy intensity of up to 10 weeks duration) and the AROC reported average duration of ambulatory rehabilitation for reconditioning ([21], p. 26).

The cost to government of delivering an STRC program, which runs over a maximum of 56 days, was set at A$214.39 per day for 2021–2022 [31], equating to just over A$12,000 per 56-day episode, which is similar to our weighted average cost per RITH episode. The reported cost of delivering the TCP program (up to 12-weeks) varies between jurisdictions, ranging from A$247.34 per day in the ACT to A$348.68 per day in Victoria (2018 financial year) [32]. This would equate to between A$20,000 and A$30,000 per 12-week episode, thus considerably higher than our weighted average cost per RITH episode. This may be somewhat accounted for by the fact that the cost of TCP appears to increase as the proportion of episodes delivered in residential care increases, while our RITH model assumes delivery in the home setting only (Table 4). Further, TCP may provide additional nursing support and personal care to that which we have allowed within our costing of RITH.

We have shown hypothetical cost savings that range from 23.6% (4AR1) to 32.6% (4AR4) per episode (weighted average saving of 28.1%) compared to the cost of an inpatient rehabilitation program for reconditioning. Cost savings have been reported for existing RITH programs in other patient groups and for “non-rehabilitation” hospital-in-the home programs with partial or total episode substitution (e.g., [16, 17, 33‐36]). For example, savings were found to range from 4%—30% for various stroke early supported discharge programs versus usual care [16], while an average saving of 26.5% [34] was reported in a meta-analysis of ‘hospital in the home’ studies, which suggest our estimated savings are credible.

The majority of our hypothetical cost savings of RITH for reconditioning over inpatient rehabilitation is likely accounted for by the absence of costs associated with 24/7 inpatient nursing care and hotel services. Counter to this saving is the requirement for staff to travel, which we estimated to be 22.4% of the cost of RITH (based on an allowance of 40 min travel per OOS plus a staff travel allowance of 90 cents per kilometre for use of their own vehicle, Table 4). Travel cost is a variable that will depend on several factors, comprising the location serviced (including distances to be travelled, traffic congestion, road tolls and parking costs) and economies of scale arising with patient volume. For these reasons we modelled the potential costs (and cost savings) for capital cities only. Some allowance for the provision of in-home support for patients and carers has been made (6.0% of program cost), but this could be increased depending on assessed patient need, while still remaining below the cost of an inpatient rehabilitation episode.

With good patient selection, we expect re-admission rates to acute care from RITH to be similar to those that occur for inpatient rehabilitation (T1.10, T1.13). However, we expect some patients (less than 10%, T1.9, T1.12) who undergo RITH to subsequently require inpatient rehabilitation. While this might dilute the savings, it is likely that patients who failed to adequately progress in RITH would have required a longer than average inpatient rehabilitation episode had they remained in a hospital rehabilitation ward, but we have no data available to support this assertion.

When inpatient care is shifted to the home, informal carers are to a greater or lesser extent undertaking some of the care-work that would have previously been undertaken by paid staff [37]. As carers are integral to the success of RITH programs for many patients, carers must be fully informed, willing and supported [18, 37]. The provision of in-home support in our model was an attempt to reduce carer burden. There is a need for future research to measure costs associated with the carer role in order to understand fully the implications of RITH as hospital substitution [37]. Despite the potential cost and care-shift [37], patients and carers who had experienced RITH for neurological and orthopaedic conditions, found being at home beneficial for therapy and for the well-being of both the patient and their family [38]; thus for appropriate patients (and carers), RITH provides increased choice about care.

Our finding of potential aggregate public hospital system savings of A$10.9 million are based on a number of assumptions, including the estimated potential utilisation of RITH for reconditioning and the relative cost of inpatient rehabilitation versus our modelled RITH cost. Any such savings should be regarded as theoretical and imprecise for several reasons. First, estimated potential utilisation is based on the experience of a self-selected subset of survey participants and cannot be assumed to be generalisable (e.g., factors impacting patient selection may vary in different socio-demographic communities). Second, the uptake of RITH by individual health services may not be complete, and will vary depending on local factors (e.g., economies of scale that can be achieved, workforce availability, and competing service priorities). Third, ‘savings’ will only be achieved if inpatient rehabilitation bed utilisation is reduced, allowing actual savings to be realised.

In reality, ‘savings’ may better expressed as the delivery of greater capacity for the same overall system expenditure, leading to an alleviation of inpatient capacity constraints within the existing public hospital system. By extension, fewer additional inpatient beds may be required in the future, resulting in savings in both (a) capital expenditure (fewer beds needing to be built) and (b) recurring operational expenses (from the servicing of additional beds). Note that our modelling excludes any potential capital savings.

A detailed discussion of the limitations of the Delphi survey methodology used has been presented elsewhere [18]. In addition to the limitations listed in the preceding section, there are several other limitations associated with our cost modelling of RITH episodes which must be acknowledged. Of note, our estimates are hypothetical only, since an operating program was not costed. Some of the assumptions used in our costings have relied on the clinical and service expertise of the authors, due to the absence of other information sources. While this could present a source of bias, assumptions have been made explicit and transparent through the details provided in Tables 3, 4 and 5 and Additional File 1. It is possible that additional costs may be incurred while the patient is at home, that have not been accounted for in the model, such as medical investigations, pathology, and non-routine medical assessments (i.e., beyond the rehabilitation physician reviews). However, as per our admission criteria for RITH [18], patients should enter RITH medically stable (T1.10). Conversely, there may be opportunities for trimming and/or reallocating costs based on individual service configurations and overhead costs. Medical staffing configurations may include the use of junior medical staff to defray the amount of medical consultant time required. The transparency of our assumptions should enable such modifications to be made by readers. We must also acknowledge that the cost model is vulnerable to unpredictable costs, such as higher readmission rates and poorer clinical outcomes (e.g., falls at home). On the other hand, calculated public hospital sector ‘savings’ have been limited to RITH in capital cities only, thus are potentially conservative.

The number of inpatient reconditioning episodes relied on 2019 data, because of the impact of the COVID-19 pandemic on rehabilitation admissions during 2020—2022 [9]. Further, no adjustment to capital city numbers was made for the small proportion of inpatient episodes involving people in AN-SNAP classes 4AR1 to 4AR4 who were domiciled in residential care prior to the onset of their impairment (< 2%) [4]. While RITH for patients from a residential care facility was outside the scope of this project, the authors are of the view that viable RITH models could be developed for patients returning to residential care for rehabilitation following an acute hospital episode. In this paper we have only presented modelling for the cost of RITH that mirrors an inpatient rehabilitation episode in the public hospital sector. We did not model the cost of the second RITH program delivery model that was supported by Delphi survey participants (i.e. RITH of longer duration and of lesser or variable intensity) [18], but we assume it should be roughly cost equivalent.