Length of Stay, Discharge Destination, and Functional Improvement
Utility of the Australian National Subacute and Nonacute Patient Casemix Classification
Background and Purpose— Although implemented in 1998, no research has examined how well the Australian National Subacute and Nonacute Patient (AN-SNAP) Casemix Classification predicts length of stay (LOS), discharge destination, and functional improvement in public hospital stroke rehabilitation units in Australia.
Methods— 406 consecutive admissions to 3 stroke rehabilitation units in Queensland, Australia were studied. Sociodemographic, clinical, and functional data were collected. General linear modeling and logistic regression were used to assess the ability of AN-SNAP to predict outcomes.
Results— AN-SNAP significantly predicted each outcome. There were clear relationships between the outcomes of longer LOS, poorer functional improvement and discharge into care, and the AN-SNAP classes that reflected poorer functional ability and older age. Other predictors included living situation, acute LOS, comorbidity, and stroke type.
Conclusions— AN-SNAP is a consistent predictor of LOS, functional change and discharge destination, and has utility in assisting clinicians to set rehabilitation goals and plan discharge.
Australian National Subacute and Nonacute Patient Casemix (AN-SNAP) Classification, implemented in 1998, was developed because the Australian National Diagnostic-Related Groups casemix system did not adequately cover care episodes requiring extended hospital stays. AN-SNAP classifies patients requiring rehabilitation, palliative, psychogeriatric, geriatric, and maintenance care.1 In these care categories, diagnosis alone cannot adequately explain the need for and cost of services.2 Instead, AN-SNAP classifies rehabilitation patients on the basis of admission scores on the Functional Independence Measure3 (FIM) and occasionally age.
In the original AN-SNAP study, 4641 overnight rehabilitation cases recruited across Australia were classified into 32 classes, of which 6 specifically relate to stroke (classes 203 to 208).4 Table 1 outlines the FIM score and age definitions used to classify stroke patients into the 6 classes. In the AN-SNAP study, stroke comprised the largest impairment group (19%), with a mean length of stay (LOS) ranging from 18.2 to 42.3 days.4
Only 2 studies have investigated the utility of AN-SNAP to predict short-term outcomes of rehabilitation for patients with stroke. Lowthian5 showed that AN-SNAP accurately predicted 31% of the variation in LOS in 547 patients and recommended the further development of AN-SNAP as a casemix tool. However, their study used retrospectively collected data and was restricted to patients in a private rehabilitation setting who received a 7-day/week rehabilitation program of 5 to 6 hours per day. This frequency and intensity of rehabilitation is unlikely to occur in public settings. In the second study, we examined the ability of AN-SNAP to predict rehabilitation LOS, discharge destination, and functional improvement in 145 patients admitted to a public hospital stroke rehabilitation unit from 1993 to 1998.6 However, our study was limited by retrospectively collected data (with substantial missing data) and by having to convert Modified Barthel Index scores into FIM scores to compute AN-SNAP classes because the FIM was not routinely used in the hospital at the time. Consequently, we could not create cognitive FIM scores and, therefore, all the AN-SNAP classes for stroke. However, using the AN-SNAP motor FIM cutoffs (13 to 46, 47 to 62, 63 to 91) as a proxy, we showed that lower motor FIM scores were associated with longer LOS and changing living situation on discharge.
Although AN-SNAP was introduced as a casemix tool to predict costs of subacute and nonacute care, no study has prospectively evaluated its utility to predict outcomes after stroke rehabilitation in public hospital settings. We aimed to examine how well AN-SNAP predicts LOS, discharge destination, and functional improvement in public hospital patients undergoing rehabilitation after stroke.
Materials and Methods
Participants and Procedure
Participants were all patients prospectively discharged between September 1999 and December 2002 from inpatient stroke rehabilitation at 3 public hospitals in southeast Queensland, Australia. Approval was obtained from hospital and university ethics committees. During rehabilitation, the treating occupational therapists, in consultation with other rehabilitation staff, collected details about the patients’ sociodemographic and clinical characteristics, rehabilitation progress, and discharge. To reduce misclassification, clinical stroke details were confirmed against separate databases maintained by the directors of the rehabilitation units.
Standard sociodemographic and clinical information was collected (Table 2). Functional status was measured by the FIM, which was administered within 72 hours of admission and discharge by FIM-accredited occupational therapists. The FIM is a valid and reliable tool for measuring burden of care by assessing level of assistance required from another person and/or assistive device.3 The FIM can be separated into 2 subscales: motor (13 items; score range, 13 to 91) and cognitive (5 items; score range, 5 to 35), with higher scores reflecting more independent function.
Descriptive statistics (means, standard deviations [SDs], medians, and quartiles) were used to present sociodemographic and clinical characteristics and to show changes in functional status and living situation. AN-SNAP classes were created from FIM admission scores and age.
For multivariate prediction of rehabilitation, LOS, discharge destination, and change in FIM scores, some predictor variables needed recoding because of missing data or insufficient cell sizes, and others were excluded because of a lack of variability. The 6 patients in AN-SNAP class 203 (motor FIM=13) were either included in class 207 if they were 75 years or older (n=1) or in class 208 if they were 74 years or younger (n=5). In doing this, the mean LOS for classes 207 and 208 changed by −0.13 and −0.53 days, respectively. Pre-admission living situation was recoded into living alone/with others. Vocational status was recoded into employed/not working. Because of skewness, acute LOS was recoded into ≤7, 8 to 14, 15 to 31, and ≥32 days. Marital status was excluded because it was strongly related to pre-admission living situation (93% of married patients lived with others). Years of education was excluded because of 21% missing data. Because age is used to create 2 AN-SNAP classes, it was also excluded. Comorbidity was considered in 2 ways: the presence of a comorbid condition (yes/no) and a weighted Charlson Comorbidity Index7 score (recoded into 0, 1, 2, or ≥3). Because the Charlson Index score incorporated previous stroke, the variable number of previous strokes was excluded.
To predict rehabilitation LOS, the General Linear Model procedure was used. To reduce skewness, rehabilitation LOS was transformed using a natural logarithm. All predictor variables were categorical. For the analysis, the model screened main effects of predictor variables and their interactions. The final model was selected by progressive elimination of nonsignificant (P≤0.05) higher-order interactions until it contained only the significant main effects and their interactions.
Change in motor and cognitive FIM scores was calculated using the standard error of measurement (SEM),8 a responsiveness statistic. SEM accounts for the possibility that some observed change is random error. SEM was calculated as the SD multiplied by the square root of 1 minus the reliability (test re-test) of the FIM (SEM=SD√1− r). The SEM calculated for motor FIM scores was 10 and 3 for cognitive FIM scores. Motor and cognitive FIM change scores that were >1 SEM were recoded as improved; scores within ±1 SEM were recoded as no change; and scores <−1 SEM were recoded as deteriorated. Using SEM, 7 patients (1.8%) showed deterioration in motor FIM scores and 9 (2.3%) showed deterioration in cognitive FIM scores. Because of these small numbers, the motor and cognitive FIM change scores were recoded into 2 groups, improved scores or stayed the same/deteriorated. Prediction of improvement versus no change/deterioration in motor and cognitive FIM scores was analyzed using logistic regression.
Discharge destination was analyzed as follows: (1) whether patients were discharged into a care facility (nursing home/hostel/hospital) versus home; (2) whether patients changed their living situation, regardless of discharge destination; and (3) for those patients discharged back into the community (ie, not into care facilities, n=297), whether they changed their living situation. For these analyses, logistic regression with categorical predictor variables was used. In each, the full model was tested against the constant only model to determine if it was statistically different. Subsequent models, with nonsignificant (P≤0.05) interactions and main effects progressively removed, were also tested. The final model contained the significant main effects and interactions. All analyses were conducted using SPSS (version 11.5).
Patient Characteristics and Discharge Outcomes
The 3 rehabilitation units admitted 406 patients (Table 2). Most (44%) were in AN-SNAP class 204, followed by 206, 208, and 207. On discharge, 78% of patients were in class 204. During rehabilitation, the most improvement was for patients in classes 206, 207, and 208 (Figure 1).
Eleven patients (2.7%) died during hospitalization. For the 395 discharged alive, mean±SD motor FIM scores increased from 59.3±21.7 on admission to 75.5±18.1 at discharge. Using the SEM, 224 (57%), 161 (41%), and 7 (2%) patients were classified as having improved, showed no change, or deteriorated, respectively. The mean±SD motor FIM change for patients who improved was 25.9±12.3. From admission to discharge, mean±SD cognitive FIM scores increased from 25.9±7.7 to 28.4±6.3. Using the SEM, 110 (28%), 273 (69%), and 9 (2%) patients were classified as having improved, showed no change, or deteriorated in cognitive scores, respectively. The mean±SD cognitive FIM change for those who improved was 7.9±3.9.
Three-quarters of patients were discharged to community-based dwellings, with 3% going to another hospital and 22% requiring hostel or nursing home care. One was discharged back to prison. Figure 2 shows the admission AN-SNAP classes, discharge FIM score, and rehabilitation LOS for patients discharged to each destination. Proportionately, patients in AN-SNAP classes 203 and 207 were most likely to be discharged to a nursing home. On discharge, 30% changed either living setting or situation.
Prediction of Rehabilitation LOS
The overall best model was significant (F(13,365)=18.9; P≤0.001; Table 3). The 5 significant variables predicted 41% (39% adjusted) of the variance in LOS, with AN-SNAP accounting for 21% and the other variables uniquely accounting for <4% each. Longer LOS was predicted by poorer motor FIM scores, living alone, longer acute LOS, and comorbidity. Stroke type was also important; post hoc analyses of the Bamford clinical classification revealed patients with stroke types classified as other/SAH had significantly shorter LOS compared with partial anterior circulatory syndrome (PACS) or posterior circulatory syndrome (POCS) and those with lacunar syndrome (LACS) had significantly shorter LOS compared with POCS.
Prediction of Change in Functional Status
AN-SNAP was the only predictor of improved motor FIM scores (χ2(4)=112.9, P≤0.001; Table 4). Having lower motor FIM scores (13 to 62) and being younger (74 or younger) were associated with the greatest odds of improvement in discharge FIM scores.
Left-hemispheric stroke and AN-SNAP, in particular having poorer admission cognitive FIM scores or poorer admission motor FIM scores together with younger age, predicted improved cognitive FIM scores (χ2(5)=41.9, P≤0.001; Table 4).
Prediction of Change in Living Situation
The analysis of whether patients were discharged into a care facility (nursing home/hostel/hospital) versus home was significant (χ2(5)=58.9, P≤0.001; Table 4), with older age, poorer motor function, and living alone the important predictors.
The analysis of whether people changed their living situation, regardless of discharge destination, was significant (χ2(8)=16.7, P≤0.001; Table 4). Having a longer acute LOS, living alone, having poorer motor function, and being older were the strongest predictors. The prediction of change in living situation in patients who returned to noncare facilities (n=297) was significant (χ2(4)=47.1, P≤0.001; Table 4), with living alone and longer acute LOS being the important predictors.
We found AN-SNAP to be a consistent predictor of LOS, discharge destination, and functional improvement, and we support the utility of AN-SNAP as a clinical tool, assisting practitioners to set realistic rehabilitation goals and discharge plans for clients.5 Accurate outcome prediction based on patient classification on admission to rehabilitation is also an important management tool because it can indicate the need for funding.
Most patients were classified into AN-SNAP class 204 (motor FIM=63 to 91) or 206 (motor FIM=47 to 62) on rehabilitation admission, with only 30% having motor FIM scores <47 (classes 203/207/208). Compared with patients with the highest FIM scores, those in AN-SNAP classes 206 to 208 had longer LOS and were more likely to show motor and cognitive functional improvement and change their living situation on discharge, in particular, to require continued hospitalization or placement in a care facility. Similar to earlier research,5,6⇓ younger patients in AN-SNAP class 208 stayed longer in rehabilitation than their older counterparts (class 207), despite having the same admission FIM score range. There are possible reasons for this. Class 208 had the highest percentage of patients with TACS (23%). Younger patients, particularly those with more severe strokes, like TACS, tend to be more intensively rehabilitated.9 This is supported by the finding that patients in class 208 were more than twice as likely to improve in FIM score than patients in class 207. Second, patients in class 207 were more often discharged into care, a decision that may be made early in rehabilitation, especially if the patient lives alone before stroke. The impact of this on LOS would, however, depend on how quickly a care facility bed became available. Third, patients in class 208 were more likely to be discharged to a community-based dwelling, possibly because they are more likely to have competent informal carers or supports at home, and the preference is discharge to home rather than institutional care. They may stay longer in hospital, while environmental modifications are made to the home or alternative living arrangements organized.
AN-SNAP accounted for half the variance in LOS, which was also predicted by longer acute LOS, living alone before stroke, having a PACS, TACS, or POCS, and comorbidity, although the contribution of these was weaker. Stroke type has been previously linked with LOS,10 as has comorbidity.11 However, it was only when comorbidity was entered as a dichotomous variable (yes/no) that it was significant; the weighted Charlson Index score was unimportant. Investigation of other variables important in explaining variation in LOS is warranted.
Admission AN-SNAP class was the main predictor of improvement in both motor and cognitive function. Patients with lower admission FIM scores, especially those 74 years or younger, showed the greatest improvement because they have the greatest potential to recover.12 Brock13 argues that patients with higher initial function might actually show more improvement on a tool like the FIM because the improvement is more difficult to achieve and the FIM score levels are not interval. However, when considered in terms of burden of care, the improvement on the FIM from requiring 2 helpers to 1 helper arguably has more resource (burden) implications than the improvement from needing an assistive device to performing a task completely independently. Patients with left-hemispheric stroke also showed improved cognitive FIM scores, a finding supported by others.14,15⇓ This may reflect initial damage to, and subsequent recovery in, the major language areas and the precentral and postcentral gyri.
Because the patients in this study were those admitted to rehabilitation, the majority returned to community-based dwellings. Patients are selectively accepted for rehabilitation based on their suitability.5,16⇓ Between 19% and 46% of patients with acute stroke are referred for rehabilitation,17–19⇓⇓ with this varying by unit. We studied 3 geriatric and rehabilitation units, as opposed to specialized stoke units. Additionally, we did not capture those who died or were discharged home or into a care facility directly from acute care.
Whereas the discharge destination of patients after stroke rehabilitation in Australia has been documented,6,20⇓ the factors predictive of destination or changing living situation are less studied. The highest odds of requiring discharge into care facilities were for patients aged 75 years or older with poor functional status. These patients were also more likely to need to change their living situation, regardless of discharge destination. For patients not requiring care facility placement, living alone and acute LOS were associated with changing living situation, such as type of dwelling or who they lived with.
Though including all admissions at 3 hospitals, the results may not be generalizable to all patients undergoing stroke rehabilitation in Queensland. The 3 hospitals may differ on how acute LOS is determined as well as their guidelines for accepting patients for rehabilitation. As a general rule, patients are deemed to be ready for rehabilitation if they are medically stable and can participate in the intensive rehabilitation program, and if realistic rehabilitation goals are achievable in a reasonable time frame. Currently, although used administratively, AN-SNAP is not used in Queensland as a casemix tool for funding and there are no financial penalties for long LOS. In measuring functional change, we chose SEM to compensate for error and regression to the mean;9 we acknowledge there are other ways of measuring change. Finally, intensity of rehabilitation, although standard across the 3 units, was not measured and may have influenced the outcomes.9
We have shown a clear association between AN-SNAP, LOS, functional change, and discharge destination. This highlights the utility of AN-SNAP in assisting clinicians to set rehabilitation goals and discharge plans. Additionally, living situation, comorbidity, and stroke type should also be considered. Whether AN-SNAP predicts longer-term outcomes such as community reintegration, quality of life, and the sustainability of community placement is of interest.
We acknowledge the tremendous assistance of the medical and occupational therapy staff at the Princess Alexandra Hospital, Gold Coast Hospital, and Royal Brisbane Hospital Geriatric and Rehabilitation Units in the data collection process. Tooth was supported by a National Health and Medical Research Council of Australia Public Health Fellowship (#997032).
- Received December 16, 2004.
- Revision received February 13, 2005.
- Accepted April 6, 2005.
Eagar K. The Australian National Sub-Acute and Non-Acute Patient Casemix Classification. Australian Health Rev. 1999; 22: 180–196.
Eagar K. The Australian National Sub-Acute and Non-Acute Patient Classification (AN-SNAP): Report of the National Sub-Acute and Non-Acute Casemix Classification Study. Wollongong: Centre for Health Service Development, University of Wollongong; 1997.
Lowthian P, Disler P, Ma S, Eagar K, Green J, de Graaf S. The Australian National Sub-Acute and Non-Acute Patient Classification (AN-SNAP): Its application and value in a stroke rehabilitation programme. Clin Rehabil. 2000; 14: 532–537.
Wyrwich K, Tierney W, Wolinsky F. Further evidence supporting an SEM-based criterion for identifying meaningful intra-individual changes in health-related quality of life. J Clin Epidmiol. 1999; 52: 861–873.
Bode R, Heinemann A, Semik P, Mallinson T. Relative importance of rehabilitation therapy characteristics on functional outcomes for persons with stroke. Stroke. 2004; 35: 2537–2542.
Hakim E, Bakheit A. A study of the factors which influence the length of hospital stay of stroke patients. Clin Rehabil. 1998; 12: 51–156.
Roe C, Dodich N, Kulinskaya E, Adam W. Comorbidities and prediction of length of hospital stay. Aust N Z Med J. 1998; 28: 811–815.
Desmond D, Moroney J, Sano M, Stern Y. Recovery of cognitive function after stroke. Stroke. 1996; 27: 1798–1803.
Cadilhac D, Ibrahim J, Pearce D, Ogden K, McNeill J, Davis S, Donnan G. Multicentre comparison of processes of care between stroke units and conventional care wards in Australia. Stroke. 2004; 35: 1035–1040.