Lifetime Cost of Stroke Subtypes in Australia
Findings From the North East Melbourne Stroke Incidence Study (NEMESIS)
Background and Purpose— Little is known about any variations in resource use and costs of care between stroke subtypes, especially nonhospital costs. The purpose of this study was to describe the patterns of resource use and to estimate the first-year and lifetime costs for stroke subtypes.
Methods— A cost-of-illness model was used to estimate the total first-year costs and lifetime costs of stroke subtypes for all strokes (subarachnoid hemorrhages excluded) that occurred in Australia during 1997. For each subtype, average cost per case during the first year and the present value of average cost per case over a lifetime were calculated. Resource use data obtained in the North East Melbourne Stroke Incidence Study (NEMESIS) were used.
Results— The present value of total lifetime costs for all strokes was Aus $1.3 billion (US $985 million). Total lifetime costs were greatest for ischemic stroke (72%; Aus $936.8 million; US $709.7 million), followed by intracerebral hemorrhage (26%; Aus $334.5 million; US $253.4 million) and unclassified stroke (2%; Aus $30 million; US $22.7 million). The average cost per case during the first year was greatest for total anterior circulation infarction (Aus $28 266). Over a lifetime, the present value of average costs was greatest for intracerebral hemorrhage (Aus $73 542), followed by total anterior circulation infarction (Aus $53 020), partial anterior circulation infarction (Aus $50 692), posterior circulation infarction (Aus $37 270), lacunar infarction (Aus $34 470), and unclassified stroke (Aus $12 031).
Conclusions— First-year and lifetime costs vary considerably between stroke subtypes. Variation in average length of total hospital stay is the main explanation for differences in first-year costs.
There is little doubt that stroke is a costly disorder.1 However, few investigators have detailed costs for stroke subtypes.2–7 Costs of the major stroke subtypes are likely to differ, given that they have different risk factors, average age of onset, incidence rates, treatments, and outcomes. There is also some evidence that the cost of care varies with stroke subtype.3,4,7–10 In the United States, the average total lifetime cost per person is estimated to be highest for subarachnoid hemorrhage, followed by intracerebral hemorrhage (ICH) and cerebral infarction.4 The in-hospital cost of cerebral infarction may also vary between subtypes.10 Apart from these data, little is known about any variations in resource use and costs of care between stroke subtypes, particularly if nonhospital costs are considered.
The Oxfordshire Community Stroke Project (OCSP) classification was developed for use in community studies and defines 4 distinct syndromes of ischemic stroke (IS) with consistent patterns of mortality, disability, and risk of recurrence.11 The classification is simple and does not require sophisticated investigations.12 Furthermore, it succinctly describes how patients with IS present to clinicians at the bedside.
The purpose of the present study was to describe the patterns of resource use and to estimate the first-year and lifetime costs for the major stroke subtypes and OCSP subtypes of IS in Australia.
This study builds on an existing model of the lifetime economic costs of stroke in Australia. The methodology used in this model has been previously reported1 and forms part of a larger modeling project, Model of Resource Utilization, Costs and Outcomes for Stroke (MORUCOS).13 Briefly, the model was constructed with an incidence-based, “bottom-up” costing approach from a societal perspective using linked spreadsheets. The costs of subarachnoid hemorrhages and transient ischemic attacks were not assessed. Stroke incidence and resource use data obtained in the North East Melbourne Stroke Incidence Study (NEMESIS) were used to estimate first-year costs.1,14–16 “Rest of life” costs (ie, costs beyond the first year) were modeled from published long-term survival and recurrence data from the OCSP (see Reference 1 for further details).17 There was comprehensive inclusion of all stroke-related resource use and costs. Included costs comprised all health sector and community costs, additional out-of-pocket costs to patients, time and out-of-pocket costs to informal caregivers, and costs of lost production for those patients participating in both the paid workforce and unpaid productive home-based activity before stroke.1 Detailed information about resource use was obtained by interview at 3, 6, and 12 months after stroke as previously described.1,15,16 The value of time lost from productive activity (indirect costs) was estimated with an approximation of the friction cost method with mortality costs confined to the first 12 months after stroke.1,18,19 For each use of healthcare and community resource during each follow-up period, we calculated the proportion of cases of each stroke subtype using that particular resource and the average stroke-related frequency of use. Because we were interested in all possible frequency values and our use of frequencies was for further calculation in a multiplicative model to estimate total costs, average rather than median values were used.1 Unit costs for this study have been published previously.1 Costs and earnings were discounted with a 5% discount rate. Purchasing power parity (see the Organisation for Economic Co-operation and Development Web site at http://www.oecd.org/ dataoecd/61/56/1876133.xls) was used to convert estimates in Australian dollars to the equivalent value in 1997 US dollars. The 1997 Australian dollar estimates can be converted to 1997 US dollars, 1997 Canadian dollars, and 1997 British pounds by multiplying by 1/1.32, 0.9, and 0.493 respectively. Purchasing power is the amount of prespecified real goods and services each unit of currency will buy; purchasing power parity exists when the equivalent amounts of 2 currencies have identical purchasing power in their respective countries.
The main assumption underlying the model was that the incidence rates and resource use data obtained in NEMESIS were representative of the situation in Australia generally. Estimates of the aggregate total cost and average cost per person during the first year after stroke and the estimated present value of the aggregate lifetime costs and average lifetime costs per person for all first-ever-in-a-lifetime strokes that occurred in Australia in 1997 have previously been reported.1 Cost-of-illness models for the lifetime economic costs of subtypes of IS have now been constructed with identical methodology. Incidence and resource use data for these subtypes of IS were obtained from NEMESIS.14 A range of 1-way sensitivity tests have been performed to assess the robustness of the cost estimates.1 This procedure included variation of the discount rate (0%, 3%, and 7%), the costs of hospitalization, and the value of caregiver time. Tests were also performed by substituting the lower and upper bounds of the 95% confidence intervals (CIs) for the proportions of patients resident in a nursing home before stroke, patients hospitalized for acute care, those admitted for inpatient rehabilitation, and patients newly admitted to a nursing home after stroke.
Stroke was defined according to the World Health Organization definition of stroke.20 Stroke cases were categorized as IS or ICH on the basis of brain imaging (CT or MRI) or autopsy findings. Cases of stroke that did not undergo brain imaging or autopsy were categorized as unclassified. Cases of IS were subdivided according to the OCSP classification as total anterior circulation infarction (TACI), partial anterior circulation infarction (PACI), posterior circulation infarction (POCI), and lacunar infarction (LACI).11 OCSP subtype was determined on clinical grounds alone without knowledge of specific findings on brain imaging and was based on the symptoms and signs present at the time of maximal deficit after stroke.
Acute hospitalization costs were determined by use of individual patient-specific costs for the subgroup of stroke patients registered in NEMESIS who were admitted to the Austin & Repatriation Medical Centre (ARMC). These costs were determined from the in-house hospital financial costing system (transition 2) and are untrimmed (ie, patients with particularly long or expensive admissions are not excluded). Stroke patients admitted to hospital for rehabilitation were subclassified according to classes within the Australian National Sub-Acute and Non-Acute Patient Classification (AN-SNAP), a case-mix classification.21 Inpatient rehabilitation costs were determined with the mixed per-diem/per-episode funding model recommended by the authors of the AN-SNAP.21 This method accounts for both the length of stay and relative intensity of resource use for different categories of stroke patients according to age, disability level, and cognitive ability.
Two-tailed t tests were used to compare mean values between groups. Differences in the proportion of first-ever and recurrent cases receiving inpatient rehabilitation and CIs for the differences between these proportions were calculated with the methods outlined by Gardner and Altman22 for calculating CIs for proportions and their differences for unpaired samples.
NEMESIS was approved by the ethics committee at each participating institution. Informed consent was obtained before any interview was conducted.
Using incidence rates of stroke subtypes from NEMESIS14 and Australian population data provided by the Australian Bureau of Statistics, we estimated the total number of first-ever-in-a-lifetime IS, ICH, and unclassified cases that occurred in Australia in 1997 to be 22 246, 4548, and 2495, respectively. Of the IS cases, 3735 were calculated to be TACI, 7992 as PACI, 5063 as POCI, and 5449 as LACI.
The proportion of patients admitted to hospital after stroke and the average length of stay varied with stroke subtype (Table 1). The mean length of stay for TACI was approximately twice as long as that for other first-ever stroke subtypes (P=0.006). Unclassified strokes had the shortest length of stay of all first-ever subtypes, although the number of cases was very small and this difference was not statistically significant (P=0.46). Among the subtypes of IS, LACI had the shortest mean length of stay. On average, recurrent stroke cases stayed 1 extra day in hospital compared with first-ever cases, although this difference was not statistically significant (P=0.48).
The costs of acute hospitalization according to stroke subtype are also shown in Table 1. When untrimmed individual patient-specific costs for the ARMC cohort are used, recurrent cases cost 30% more on average than first-ever-in-a-lifetime cases (an additional Aus $2,000 (US $1515) per admission), although this was not statistically different (P=0.28). Among first-ever cases, TACI had the greatest average cost per admission, and this was more than twice the average for all other cases of first-ever-in-a-lifetime stroke (P=0.004). ICH was the next most expensive, followed by POCI, PACI, and LACI. In contrast, when trimmed average Australian National–Diagnosis Related Group (AN-DRG) costs (a case-mix payment system) were substituted, ICH cases were the most expensive, and PACI was the least expensive.
The proportion of first-ever and recurrent 3-month survivors admitted for rehabilitation was similar (45% versus 50%; difference in proportions, −5%; 95%CI, −23 to 13), and there was no significant difference in mean length of inpatient stay for these 2 groups (P=0.95; Table 2). The longest mean lengths of stay occurred for ICH and TACI; however, they were not significantly higher than for the rest of first-ever strokes (P=0.12 for both comparisons).
The cost of inpatient rehabilitation also varies according to stroke subtype (Table 2). Cases of ICH are most expensive; cases of LACI are least expensive. Furthermore, on average, the cost of rehabilitation for cases of recurrent stroke is greater than for first-ever-in-a-lifetime cases.
Other Resource Use During the First Year After Stroke
Estimated average costs per case for other categories of resource use during the first year after stroke are shown in Table 3. Of note, out-of-pocket costs incurred by stroke cases constituted an important contribution to first-year costs for all stroke subtypes.
Total First-Year Costs
There are major differences in average first-year costs between the major stroke subtypes. On average, unclassified cases are the least expensive subtype, whereas the costs of ICH are similar to those of IS. There are, however, large differences between the cost per case for subtypes of IS. Cases of TACI are most expensive, followed by PACI, POCI, and LACI. On average, cases of TACI are 1.9 times as expensive as cases of LACI. This is explained primarily by the fact that the costs of acute hospitalization for TACI are more than 3 times those for LACI. These differences disappear to a large degree when the average AN-DRG costs for hospital admissions for stroke (ie, trimmed of outlying values) are substituted. PACI then becomes the most expensive infarct subtype, followed by TACI, POCI, and LACI. Given the clear differences in length of hospital stay between the subtypes of IS among the NEMESIS cohort, the cost estimates shown in Table 3 (incorporating ARMC costs of hospitalization) are likely to be most indicative of the true differences between subtypes.
Total Lifetime Costs
A summary of the present value of average lifetime cost per case and the present value of total lifetime costs for all first-ever-in-a-lifetime stroke in Australia in 1997 according to stroke subtype is presented in Table 4. As for first-year costs, lifetime cost per case varies considerably between subtypes of first-ever-in-a-lifetime stroke. Compared with the situation when only first-year costs are considered, ICH is the most expensive stroke subtype, costing on average 1.7 times the lifetime cost per case of IS.
With sensitivity testing, the average lifetime costs for ICH range from 1.6 to 2 times the average lifetime costs for IS cases. Among cases of IS, the lifetime cost per case is greatest for TACI, followed by PACI, POCI, and LACI. On an average-cost-per-case basis, LACI costs approximately one third less than TACI. The cost hierarchy for subtypes of IS is largely maintained throughout a range of sensitivity testing. Only 2 scenarios result in another subtype (PACI) being the most expensive infarct subtype over a lifetime: (1) the use of AN-DRG hospital costs, which explicitly exclude cases with very long lengths of stay, and 2) the use of the high point of the 95% CI for the proportion of cases resident in a nursing home at the time of stroke.
In this study, both the costs during the first year and costs over a lifetime after first-ever-in-a-lifetime stroke varied according to major stroke subtype and between subtypes of IS. The relationship between subtype and cost was robust as it was maintained across a range of sensitivity tests. Over a lifetime, on average, cases of ICH were 1.7 times as expensive as cases of IS. This is explained largely by the fact that, on average after ICH, ongoing stroke-related costs are incurred for a greater number of years. Among cases of IS, lifetime costs also varied with stroke subtype. TACI was the most expensive subtype, followed by PACI, POCI, and LACI. The differences in costs between subtypes of IS were, however, less for lifetime costs than for first-year costs. This is not surprising because, on average, TACI patients are older than patients with other infarct subtypes and thus costs continue for fewer years.
During the first year after stroke, average costs were similar for IS and ICH cases. However, there were large differences in cost between subtypes of IS. TACI was again the most expensive infarct subtype, costing ≈1.9 times the cost per case of LACI. The major determinant of these differences was the total cost of hospitalization (acute care and rehabilitation). When AN-DRG trimmed average costs are substituted in the sensitivity analysis, cases of ICH are most expensive. Because the AN-DRG code for any admission is strongly influenced by any surgical procedures, it is likely that the costs of surgery and postoperative care for ICH cases are responsible for the greater costs of this subtype. In contrast, the high costs of hospitalization for TACI reflect the relatively long average length of stay for these patients.
The differences in first-year and lifetime costs seen between the stroke subtypes reflect clear differences in resource use. For TACI, >40% of first-year costs relate to the costs of acute hospitalization, 30% to inpatient rehabilitation, and ≈6% to nursing home care. Measures that result in a reduced length of acute and rehabilitation hospital stay with no adverse effect on outcome have the potential to realize substantial savings for this subgroup. PACI patients have the highest recurrent stroke costs, so targeted interventions to optimize secondary prevention among this subgroup may result in substantial savings and health gains. Potential strategies, however, need to be assessed through formal economic evaluation.
A note of caution is appropriate with regard to our total cost estimates for stroke subtypes in Australia. These are based on the findings among a cohort of stroke patients residing in inner urban Melbourne, in close proximity to the ARMC, a center for stroke care and research. It is very likely that patterns of care and costs vary considerably across Australia, particularly in nonurban and remote regions. Additional resource use data from other regions of Australia, should they become available, would improve the accuracy of our overall cost estimates for Australia.
At the bedside, clinicians are presented with clinical stroke subtypes with easily recognized characteristics. The patterns of incidence, outcome, and risk of recurrence for the major stroke subtypes and the OCSP subtypes are now well recognized.11,12,14 We have demonstrated that patterns of resource use and costs also vary with stroke subtype. Although other identifiable factors clearly influence stroke costs (eg, age and comorbidities), subtype-specific resource use and cost information is immediately understood by the clinician and healthcare provider. This is the first investigation of the differences in resource use and cost for subtypes of IS beyond the initial period of acute hospitalization. Our findings provide a detailed overview of current stroke care and its costs in Australia and provide input to our economic model, MORUCOS.13 These data provide the necessary starting point (or base case) for future economic evaluations of various aspects of the prevention, treatment, and care of stroke patients. This economic evaluation work is now in progress.
This work was supported by grants from the Victorian Health Promotion Foundation, National Health and Medical Research Council, Foundation for High Blood Pressure Research, and National Stroke Foundation. While chief executive officer of the National Stroke Foundation, Franca Smarelli commissioned the Centre for Health Program Evaluation to develop MORUCOS in several stages as a collaborative initiative. The model is fully owned by the National Stroke Foundation. Lichun Quang provided assistance with database management and analysis. The following research nurses completed most interviews, and their contribution is gratefully acknowledged: Stephen Cross, Barbara Dowell, Elspeth Freeman, Meg Hooton, Catherine Sharples, Mary Staios, and Dennis Young.
- Received April 28, 2003.
- Revision received June 16, 2003.
- Accepted June 24, 2003.
Dewey HM, Thrift AG, Mihalopoulos C, Carter R, Macdonell RA, McNeil JJ, Donnan GA. Cost of stroke in Australia from a societal perspective: results from the North East Melbourne Stroke Incidence Study (NEMESIS). Stroke. 2001; 32: 2409–2416.
Taylor TN, Davis PH, Torner JC, Holmes J, Meyer JW, Jacobson MF. Lifetime cost of stroke in the United States. Stroke. 1996; 27: 1459–1466.
Bergman L, van der Meulen JH, Limburg M, Habbema JDF. Costs of medical care after first-ever stroke in the Netherlands. Stroke. 1995; 26: 1830–1836.
Weimar C, Weber C, Wagner M, Busse O, Haberl RL, Laterbach KW, Diener HC, for the German Stroke Data Bank Collaborators. Management patterns and health care use after intracerebral hemorrhage: a cost-of-illness study from a societal perspective in Germany. Cerebrovasc Dis. 2003; 15: 29–36.
Mamoli A, Censori B, Casto L, Sileo C, Cesana B, Camerlingo M. An analysis of the costs of ischemic stroke in an Italian stroke unit. Neurology. 1999; 53: 112–116.
Mihalopoulos C, Carter R, Dewey H, Smarrelli F. Modelling the Costs of Stroke: Development and Results from the Model of Resource Utilisation, Costs and Outcomes of Stroke (MORUCOS). Melbourne, Australia: Centre for Health Program Evaluation; 2003. Research report. In press.
Thrift AG, Dewey HM, Macdonell RA, McNeil JJ, Donnan GA. Incidence of the major stroke subtypes: initial findings from the North East Melbourne Stroke Incidence Study (NEMESIS). Stroke. 2001; 32: 1732–1738.
Dewey H, Thrift A, Mihalopoulos C, Carter R, Macdonell R, McNeil J, Donnan G. Informal care for stroke survivors: results from the North East Melbourne Stroke Incidence Study (NEMESIS). Stroke. 2002; 33: 1028–1033.
Dewey HM, Thrift AG, Mihalopoulos C, Carter R, Macdonell RA, McNeil JJ, Donnan GA. “Out of pocket” costs to stroke patients during the first year after stroke: results from the North East Melbourne Stroke Incidence Study. J Clin Neurosci. 2003. In press.
Dennis MS, Burn JP, Sandercock PA, Bamford JM, Wade DT, Warlow CP. Long-term survival after first-ever stroke: the Oxfordshire Community Stroke Project. Stroke. 1993; 24: 796–800.
Eagar K, Gordon R, Hodkinson A, Green J, Eagar L, Erven J, Eckstein G, Spooner G, Kennedy C, Owen A, et al. 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, Australia: Centre for Health Service Development, University of Wollongong; 1997.
Gardner MJ, Altman DG. Confidence intervals rather than P values: estimation rather than hypothesis testing. BMJ. 1986; 292: 746–750.