The health loss from ischemic stroke and intracerebral hemorrhage: evidence from the North East Melbourne Stroke Incidence Study (NEMESIS)
© Cadilhac et al; licensee BioMed Central Ltd. 2010
Received: 4 September 2009
Accepted: 14 May 2010
Published: 14 May 2010
People suffering different types of stroke have differing demographic characteristics and survival. However, current estimates of disease burden are based on the same underlying assumptions irrespective of stroke type. We hypothesized that average Quality Adjusted Life Years (QALYs) lost from stroke would be different for ischemic stroke and intracerebral hemorrhage (ICH).
We used 1 and 5-year data collected from patients with first-ever stroke participating in the North East Melbourne Stroke Incidence Study (NEMESIS). We calculated case fatality rates, health-adjusted life expectancy, and quality-of-life (QoL) weights specific to each age and gender category. Lifetime 'health loss' for first-ever ischemic stroke and ICH surviving 28-days for the 2004 Australian population cohort was then estimated. Multivariable uncertainty analyses and sensitivity analyses (SA) were used to assess the impact of varying input parameters e.g. case fatality and QoL weights.
Paired QoL data at 1 and 5 years were available for 237 NEMESIS participants. Extrapolating NEMESIS rates, 31,539 first-ever strokes were expected for Australia in 2004. Average discounted (3%) QALYs lost per first-ever stroke were estimated to be 5.09 (SD 0.20; SA 5.49) for ischemic stroke (n = 27,660) and 6.17 (SD 0.26; SA 6.45) for ICH (n = 4,291; p < 0.001). QALYs lost also differed according to gender for both subtypes (ischemic stroke: males 4.69 SD 0.38, females 5.51 SD 0.46; ICH: males 5.82 SD 0.67, females 6.50 SD 0.40).
People with ICH incurred greater loss of health over a lifetime than people with ischemic stroke. This is explained by greater stroke related case fatality at a younger age, but longer life expectancy with disability after the first 12 months for people with ICH. Thus, studies of disease burden in stroke should account for these differences between subtype and gender. Otherwise, in countries where ICH is more common, health loss for stroke may be underestimated. Similar to other studies of this type, the generalisability of the results may be limited. Sensitivity and uncertainty analyses were used to provide a plausible range of variation for Australia. In countries with demographic and life expectancy characteristics comparable to Australia, our QoL weights may be reasonably applicable.
Worldwide, stroke is a significant contributor to disease burden. In Australia, stroke is the second leading cause of death . Of those with first-ever stroke, about 35% die within 12 months of their stroke  and about half of survivors at 12 months are dependent on others . The cost of stroke is high, with the present value of lifetime costs of first-ever stroke in 2004 estimated to cost more than AUD2 billion (~2% of total annual health expenditure ) to the Australian community .
In the NEMESIS study, 72.5% of strokes were ischemic stroke while 14.5% were intracerebral hemorrhage (ICH); 4.3% were subarachnoid hemorrhage and in 8.7% the subtype was undetermined . Importantly, these different types of stroke have different risk factors, treatments and outcome. Case-fatality at one year is lower for first-ever ischemic stroke (reported range from Australian incidence studies 23% to 31%) than for ICH (39% to 50%) and undetermined stroke (67% to 89%) [6, 7]. When assessing the effect of stroke on society it is important to include measures of both mortality and morbidity, as stroke affects both of these outcomes.
Summary measures of population health include Health Adjusted Life Expectancy (HALE), Disability Adjusted Life Years (DALYs) and Quality Adjusted Life Years (QALYs). The HALE value represents the number of expected years of life equivalent to years lived in full health adjusted for time spent in poor health based on current rates of ill-health (e.g. chronic disease) and mortality in a community. The DALY is a health gap measure and captures the years of life lost (YLL) due to premature mortality (in this case from stroke) and the years of life lived with disability (YLD) as a consequence of having had a stroke. QALYs are based on a similar conceptual framework (life expectancy plus quality of life [QoL]), but are often based on different assumptions and methods [8, 9]. DALYS have gained prominence in recent years with the extensive systematic review of burden of disease .
One of the main differences for DALYs and QALYS is the way health states (i.e. the physical, social and emotional functioning of individuals) , such as stroke, are weighted. Traditionally, QALYs are based on a heath related QoL weight that is directly derived from patients or the general population while DALY weights have more commonly been elicited from expert panels[8, 12]. There are both advantages and disadvantages for each approach. Regardless, the preference weights derived for both DALYs and QALYs reflect departures from good health [8, 13]. It has been found that stroke survivors usually assign themselves higher utilities (i.e. a value that represents the strength of an individual's preference for a particular health outcome) than do the general community or caregivers of stroke survivors . Current estimates of DALYs for stroke in Australia have used preference weights directly elicited from patients .
QALYs are usually reported as something society will want to gain, while DALYs are to be avoided. To keep the concept similar to a DALY, one can instead report the QALYs lost. This provides a measure of the health gap experienced by stroke survivors compared to the normal population and provides an estimate of the health gain that could be achieved if a stroke was prevented. Such information is important when making assessments of the value of various interventions, such as in cost-effectiveness analysis.
Health-related QoL of stroke patients has been well documented as part of the North East Melbourne Stroke Incidence Study (NEMESIS). This provided an opportunity to estimate the health loss attributable to stroke using prospectively elicited QoL data from individual stroke patients. In NEMESIS a health state classification questionnaire, the Assessment of Quality of Life (AQoL) tool , was used whereby a utility score (an index of the strength of a person's preference for a health state where death is represented as 0.0 and normal health as 1.0) can be derived. This utility score can be used to provide the preference weight in estimating health loss. Because of the recognized differences between the main stroke subtypes, we hypothesized that average QALYs lost would be different for ischemic stroke and ICH.
We developed a 'Lifetable' model, created in Excel (Microsoft Corporation, 2003), to accommodate a 'lifetime' perspective for an incident cohort of people with stroke.
Incidence and case-fatality data obtained from NEMESIS were applied to Australian population data to estimate the size of the cohort. The NEMESIS data constitute 'best available' information. Authors of the 2003 Australian BoD study have used NEMESIS data and have compared representative Australian hospital admission rates from 1996/1997 with those in NEMESIS mapped to the relevant statistical local areas. The ratios of hospital admissions in NEMESIS catchment areas to those in all of Australia in 1996/97 ranged from 0.9 to 1.3 by age and sex with an overall ratio close to unity (personal communication, T. Vos March 2010). Nonetheless, we included point estimates used from NEMESIS in the sensitivity and uncertainty analyses as outlined in the Analysis section below. In addition, because early mortality in ICH is much greater than in ischemic stroke and would dominate comparisons with ischemic stroke, estimates of health loss were based on 28 day survivors. Twenty-eight day survivors were categorized into age-groups according to their starting age (mid-point ages for < 55, 55 to 64, 65 to 74, 75 to 84 and 85+ years) up to age 100. Deaths attributable to stroke (not 'all cause' mortality) were calculated for day 28, 12 months and then for each year up to 5 years after stroke using NEMESIS data. Between 5 and 10 years the probability of dying each year attributable to stroke was maintained at the same year 5 probability since no other data were available. After 10 years, remaining survivors were assumed to have the same probability of death as the general population for that age band and gender. Although one may argue that mortality risk may not return to that of the general population, since risk factors in the general population will have increased over time, and may be at similar levels to those of the stroke population, we therefore felt this was a plausible approach given the current limitations of data.
The QoL weights used to estimate QALYs lost were derived from the AQoL Mark 1 instrument which has previously been validated for use in stroke using NEMESIS data . We used published 'normal' population values derived from the AQoL instrument  by age band as a measure of pre-stroke QoL. The QoL (preference) weight was then calculated as the net difference between these two scores, and could be described as the utility loss attributable to stroke. Because it was important that change in QALYs over time be reported using longitudinal data we used the average of paired utilities obtained from the same cases at 12 months and 5 years. In applying these utility values in the 'Lifetable' model a linear relationship between the 12 month and 5 year values were assumed. This is because the direction and magnitude of change between the 12 month and 5 year time points for each age band varied and was usually small (min 0.006 in < 55 year olds and max 0.12 in 75-84 year olds). Therefore, a more elaborate approach was considered unnecessary. After 5 years, we assumed that survivors would have the same utility loss as the 5 year utility loss in each relevant age band.
To avoid overestimating life expectancy, we used HALE values calculated for the Australian 2003 BoD study  for life expectancy in our Lifetable model according to age group. That is, the net difference in health loss due to stroke was estimated by subtracting the estimated HALE value for someone without stroke from that estimated for a person of the same age and gender with stroke to provide the final QALY loss result. We assumed HALE values for the 2003 population were applicable to the 2004 population used in this analysis. Thus, results from this research could be used for economic evaluations using our redeveloped stroke economic model . Standard population life expectancies were not used as they were considered insufficient to address the impact on QoL, since people with stroke may also have non stroke-related disability as they get older.
T-tests for continuous variables were applied using Intercooled STATA version 8 (Stata Corporation, 2003). The level of significance was set at p < 0.05 (two-sided).
We used a 3% discount rate to accord with methods recommended in the Australian BoD study . Since the issue of discounting QALYs is still debated , undiscounted QALYs were also estimated. We also provide estimates for different age groups and gender using a 5% discount rate to accord with other studies. Multivariable probabilistic uncertainty analyses were undertaken using @Risk software version 4.5 (Palisade Corporation, 2005). The sampling variations incorporated for point estimates were based on triangular distributions (minimum, most likely and maximum) that approximate a normal distribution. The minimum and maximum values were obtained from literature reviews or best available data and applied as a proportion greater or less than the point estimate. Variables included incidence (range used -5% to +1%) [20, 21]; survival (-2% to +1%) [22, 23]; and average QALYs lost (ischemic stroke 4%; ICH 2% based on the plausible variation found using NEMESIS data). Three thousand 'Monte Carlo' simulations were undertaken to ensure convergence. Convergence was defined as less than 1.5% variation in primary outcome statistics, such as numbers of strokes. The 3,000 individually simulated point estimates were used to estimate a mean, median and 95% uncertainty interval for the results.
To test the sensitivity of the QALY loss estimates we substituted the NEMESIS QoL weights and case fatality rates with estimates used in the 2003 BoD study for stroke. To distinguish between these estimates, we define these results as DALYs. The results were compared to describe the potential variation that might occur when different input parameters (e.g. QoL weights and case-fatality rates) for Australia are used.
Average and net difference in utility scores between Australian population and first-ever stroke survivors.
NEMESIS utility scores*
Net difference (QoL weights)
The estimated 'lifetime' health loss attributable to first-ever strokes
Quality Adjusted Life Years lost according to stroke subtype and gender
N= 14 139
N= 13 521
N = 2 087
N = 2 204
QALY lost per case
Age Group years
QALYs with 3% discounting (5% discounting, 0% discounting)
5.5 (3.7, 11.1)
9.8 (4.7, 20.2)
10.7 (7.3, 21.7)
12.5 (8.6, 25.6)
6.3 (5.0, 9.7)
7.6 (5.8, 12.1)
7.4 (5.9, 11.0)
10.9 (8.6, 16.3)
6.6 (5.7, 8.5)
7.6 (6.5, 10.2)
6.6 (5.7, 8.4)
7.6 (6.4, 10.1)
3.7 (3.3, 4.3)
4.9 (4.4, 5.9)
3.9 (3.5, 4.5)
5.1 (4.6, 6.1)
2.3 (2.1, 2.5)
2.5 (2.4, 2.8)
2.2 (2.1, 2.4)
2.4 (2.2, 2.6)
Weighted average (SD)†
Total weighted average (SD)†
Undiscounted total weighted average
Modelled estimates of stroke subtype and QALYs lost: Results of uncertainty analyses.
Modeled point estimates†
95% Uncertainty interval†
Lower bound (2.5%)
Higher bound (97.5%)
Number of strokes
Total number of QALYs lost (incident cases) *
Number of strokes
Total number of QALYs lost (incident cases)‡
Difference in DALYs and QALYs: Sensitivity analyses for average lifetime health loss*.
Intracerebral hemorrhage (n = 4 291)
Average per first-ever stroke event
Total health loss for cohort
Ischemic stroke (n = 27 660)
Average per first-ever stroke event
Total health loss for cohort (years lost)
We found that people with ICH incurred greater health losses over a lifetime than those having an ischemic stroke and that this is explained by greater stroke-related case fatality in cases with ICH (average over 5 years was 40% greater for ICH). Moreover, since ICH's experience stroke a younger age, those that survive also have a greater duration of disability since these survivors will have a greater life expectancy. We also found that QALY losses were greater for females than males in both stroke subtypes. Previous research on QALYs lost from stroke have been provided for generic stroke or by stroke severity [8, 19]. It is reported that only about half of the variance in QoL is explained by stroke severity . Because the QoL weights used were the same for both stroke subtypes and gender, the differences that we observed in QALYs between the stroke types reflect differences in the age of stroke onset and case fatality between these subtypes.
There has been considerable discussion about the best method for deriving the QoL weight. It is important that this is measured carefully because it comprises a major component of the QALY and inaccuracies will lead to inaccurate estimates. The estimates of QoL weights reported in the literature vary considerably for generic stroke, ranging from 0.29 to 0.903 in a recent meta-analysis . These large differences have been attributed to different elicitation methods  (e.g. from people with and without stroke or health experts ) including the type and range of tool used; timing of elicitation between the event and the assessment; differences in age and gender; and the variance for weights obtained . In our study, the QoL utility loss varied by age and time since stroke and ranged between 0.025 and 0.514. Our summary QoL weight estimate at 12 months was 0.19 and at 5 years 0.25 (adjusted for age and population size). These preference weights are consistent with the lower bound of those reported in the literature.
There are a number of strengths in this study. First, we used comprehensive data obtained from a large community-based stroke incidence study (NEMESIS). Importantly, the QoL weights were directly elicited from the same patients at 12 months and 5 years using the AQoL instrument and are appropriate for the reference population . Use of QoL weights for survivors for up to 5 years is an added advantage to previous studies that have been based on 12-month estimates . This provides a longer-term perspective rarely included in the estimation of health loss for stroke. Furthermore, because health loss was estimated separately for ICH and ischemic stroke, this will enable investigators to evaluate interventions specific to ICH and ischemic stroke.
Similar to other studies of this type, the generalisability of the results may be limited. Although NEMESIS estimates are unbiased at a community level as we have obtained almost every case in a specific population (i.e. community-based rather than hospital-based) and evidence for the 2003 Australian BoD study provide good evidence that these data are fairly representative for Australia as a whole, potential differences across Australia in ethnicity and socioeconomic status may be important. Therefore, it was necessary to provide detailed sensitivity and uncertainty analyses for these data which provide a plausible range of variation for Australia. In countries with demographic and life expectancy characteristics comparable to Australia, our QoL weights should, therefore, be reasonably applicable.
Other limitations include the assumption that NEMESIS incidence rates are applicable for Australia; that use of the normal population utility values to estimate pre-stroke utility are applicable to people who experience stroke when their risk profile may mean they have greater pre-morbid disability; and longer-term (5 year) disutility in people with stroke may reflect co-morbidity from other diseases. Another important limitation was the assumption that in those surviving beyond 5 years the QoL weight values that were applied were those of the next older age band. Because the available longitudinal data re quality of life in the long-term are limited in stroke cohorts we adopted this approach as the most feasible. It also ensured that these data were consistent with those of other studies. Methods used to address this limitation included (1) describing the health loss for stroke according to stroke subtype since this is one of the major factors likely to influence health status over time; (2) use of the 5 year QoL weight values; and (3) multivariable probabilistic uncertainty analysis to assess the impact of a range of important variables including survival rates where evidence is limited but a plausible range of values could be considered. Lastly, the use of triangular distributions rather than use of normal distributions for estimates derived from NEMESIS, such as the QoL weight may have underestimated the uncertainty. It was the considered view of the author group, that triangular distributions would enable ranges to be selected that would best reflect the data we had, as samples size were small when QoL data were disaggregated by the 5 age groups nominated.
These identified limitations may provide sources of over and under estimation of health loss. Inaccurate estimation of health-related QoL can have major implications when used to undertake cost-effectiveness analysis. This is because, as a summary measure for population health which attributes a perceived social value to different health states, variance in QALYs may result in inappropriate resource allocation decisions . In other words, use of different QoL weights may over- or under-estimate the QALY gain for an intervention therefore producing contradictory cost-effectiveness results. Use of multivariable probabilistic uncertainty and sensitivity analyses in this study were used to address such potential limitations and, overall, the estimates were fairly robust.
Estimates of health loss measured as QALYs were presented for Australia. Notably, these estimates were based on patient-level data obtained up to 5 years following stroke. We provide evidence that the health loss attributable to ICH and ischemic stroke are different; health loss also varies between males and females. Therefore, when undertaking studies of disease burden in stroke, investigators should account for these differences between subtype and gender. Otherwise, in countries where ICH is more common, the disease burden for stroke may be underestimated. Lastly, the estimates provide quantifiable measures of the average health loss over a lifetime per first-ever case of stroke that may be applied in economic evaluations to determine the cost-effectiveness of prevention interventions.
All authors have a PhD. DC is the Head of the Public Health Division of the National Stroke Research Institute and has a clinical background in nursing. HD is an Associate Professor for the Department of Medicine (The University of Melbourne) and is the Deputy Director of Neurology at Austin Health (Vic, Australia). TV is the Director, Centre for Burden of Disease and Cost-Effectiveness, School of Population Health at The University of Queensland (Australia) as is also qualified as a medical practitioner. RC is a Professor and the inaugural Chair of Health Economics at Deakin University (Burwood, Australia). AT is an NHMRC Senior Research Fellow, the Head of Stroke Epidemiology at the Baker IDI Heart and Diabetes Institute, as well as an Associate Professor for Monash University (Australia).
We wish to acknowledge Judith Katzellenbogen for providing details of her methods used to estimate the stroke burden in the 2003 Australian Burden of Disease study. We acknowledge the important contribution of the research nurses who have directly collected the outcome data for NEMESIS.
Helen Dewey, Rob Carter and Amanda Thrift were supported by research fellowships from the National Health and Medical Research Council (NHMRC). Dominique Cadilhac was supported by an NHMRC postgraduate public health scholarship and a public health grant from the Victorian Health Promotion Foundation (VicHealth). The North East Melbourne Stroke Incidence Study was supported by grants from the NHMRC, VicHealth, the Foundation for High Blood Pressure Research, and the National Stroke Foundation. None of the funders contributed to the study design; in the collection, analysis, and interpretation of data; in the writing of the manuscript; and in the decision to submit the manuscript for publication.
- Australian Bureau of Statistics: 3303.0 Causes of death 2206: Australia. Canberra: ABS; 2008:88.Google Scholar
- Thrift AG, Dewey HM, Sturm JW, Srikanth VK, Gilligan AK, Gall SL, Macdonell RA, McNeil JJ, Donnan GA: Incidence of stroke subtypes in the North East Melbourne Stroke Incidence Study (NEMESIS): differences between men and women. Neuroepidemiology 2009,32(1):11–18. 10.1159/000170086PubMedView ArticleGoogle Scholar
- Sturm JW, Dewey HM, Donnan GA, Macdonell RA, McNeil JJ, Thrift AG: Handicap after stroke: how does it relate to disability, perception of recovery, and stroke subtype?: the North East Melbourne Stroke Incidence Study (NEMESIS). Stroke 2002,33(3):762–768. 10.1161/hs0302.103815PubMedView ArticleGoogle Scholar
- Australian Institute of Health and Welfare: Health expenditure Australia 2006–07. Health and Welfare Expenditure Series no. 35. Cat. no. HWE 42. Canberra: AIHW; 2008:186.Google Scholar
- Cadilhac DA, Carter R, Thrift AG, Dewey HM: Estimating the long-term costs of ischemic (IS) and hemorrhagic (ICH) stroke for Australia: new evidence derived from the North East Melbourne Stroke Incidence Study (NEMESIS). Stroke 2009, 40: 915–921. 10.1161/STROKEAHA.108.526905PubMedView ArticleGoogle Scholar
- 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(8):1732–1738.PubMedView ArticleGoogle Scholar
- Sudlow CL, Warlow CP: Comparable studies of the incidence of stroke and its pathological types: results from an international collaboration. International Stroke Incidence Collaboration. Stroke 1997,28(3):491–499.PubMedView ArticleGoogle Scholar
- Post PN, Stiggelbout AM, Wakker PP: The utility of health states after stroke: a systematic review of the literature. Stroke 2001,32(6):1425–1429.PubMedView ArticleGoogle Scholar
- Sassi F: Calculating QALYs, comparing QALY and DALY calculations. Health Policy Plan 2006,21(5):402–408. 10.1093/heapol/czl018PubMedView ArticleGoogle Scholar
- Mathers C, Bernard C, Iburg K, Inoue M, Ma Fat D, Shibuya K, Stein C, Tomijima N: The Global Burden of Disease in 2002: data sources, methods and results. Organization WH. Geneva: World Health Organization; 2003.Google Scholar
- Drummond M, O'Brien B, Stoddart GL, Torrance GW, eds: Methods for the Economic Evaluation of Health Care Programmes. Second edition. New York Oxford University Press; 1997.Google Scholar
- Gold MR, Stevenson D, Fryback DG: HALYS and QALYS and DALYS, Oh My: similarities and differences in summary measures of population Health. Annu Rev Public Health 2002, 23: 115–134. 10.1146/annurev.publhealth.23.100901.140513PubMedView ArticleGoogle Scholar
- Tengs TO, Yu M, Luistro E: Health-related quality of life after stroke a comprehensive review. Stroke 2001,32(4):964–972.PubMedView ArticleGoogle Scholar
- Begg SJ, Vos ET, Barker B, Stevenson CE, Stanley L, Lopez AD: The burden of disease and injury in Australia 2003. PHE 82. Canberra: Australian Institute of Health and Welfare; 2007.Google Scholar
- Hawthorne G, Richardson J, Osborne R, McNeil H: The Australian Quality of Life Instrument: Construction, Initial Validation & Utility Scaling. Melbourne: Centre for Health Program Evaluation; 1997.Google Scholar
- Sturm JW, Osborne RH, Dewey HM, Donnan GA, Macdonell RA, Thrift AG: Brief comprehensive quality of life assessment after stroke: the assessment of quality of life instrument in the North East Melbourne Stroke Incidence Study (NEMESIS). Stroke 2002,33(12):2888–2894. 10.1161/01.STR.0000040407.44712.C7PubMedView ArticleGoogle Scholar
- Hawthorne G, Osborne R: Population norms and meaningful differences for the Assessment of Quality of Life (AQoL) measure. Aust N Z J Public Health 2005,29(2):136–142. 10.1111/j.1467-842X.2005.tb00063.xPubMedView ArticleGoogle Scholar
- Begg SJ, Vos T, Barker B, Stanley L, Lopez AD: Burden of disease and injury in Australia in the new millennium: measuring health loss from diseases, injuries and risk factors. Med J Aust 2008,188(1):36–40.PubMedGoogle Scholar
- Bosworth HB: Health-related quality of life after stroke a comprehensive review: Editorial comment. Stroke 2001,32(4):972.Google Scholar
- Islam MS, Anderson CS, Hankey GJ, Hardie K, Carter K, Broadhurst R, Jamrozik K: Trends in Incidence and Outcome of Stroke in Perth, Western Australia During 1989 to 2001. The Perth Community Stroke Study. Stroke 2008,39(3):776–82. 10.1161/STROKEAHA.107.493643PubMedView ArticleGoogle Scholar
- Anderson CS, Carter KN, Hackett ML, Feigin V, Barber PA, Broad JB, Bonita R: Trends in stroke incidence in Auckland, New Zealand, during 1981 to 2003. Stroke 2005,36(10):2087–2093. 10.1161/01.STR.0000181079.42690.bfPubMedView ArticleGoogle Scholar
- Carter KN, Anderson CS, Hackett ML, Barber PA, Bonita R: Improved survival after stroke: is admission to hospital the major explanation? Trend analyses of the Auckland regional community stroke studies. Cerebrovasc Dis 2007,23(2–3):162–168. 10.1159/000097054PubMedView ArticleGoogle Scholar
- Dunn C, Sadkowsky K, Jelfs P: Trends in deaths: Analysis of Australian data 1987–1998 with updates to 2000. Canberra: Australian Institute of Health and Welfare; 2002:198.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.