Lifetime Cost of Stroke in the United States
Background and Purpose Stroke imposes a substantial economic burden on individuals and society. This study estimates the lifetime direct and indirect costs associated with the three major types of stroke: subarachnoid hemorrhage (SAH), intracerebral hemorrhage (ICH), and ischemic stroke (ISC).
Methods We developed a model of the lifetime cost of incident strokes occurring in 1990. An epidemiological model of stroke incidence, survival, and recurrence was developed based on a review of the literature. Data on direct cost of treating stroke were obtained from Medicare claims data, the 1987 National Medical Expenditure Survey (NMES), and insurance claims data representing a group of large, self-insured employers. Indirect costs (the value of foregone market and nonmarket production) associated with premature morbidity and mortality were estimated based on data from the US Bureau of Economic Analysis and the 1987 NMES.
Results The lifetime cost per person of first strokes occurring in 1990 is estimated to be $228 030 for SAH, $123 565 for ICH, $90 981 for ISC, and $103 576 averaged across all stroke subtypes. Indirect costs accounted for 58.0% of lifetime costs. Aggregate lifetime cost associated with an estimated 392 344 first strokes in 1990 was $40.6 billion: $5.6 billion for SAH, $6.0 billion for ICH, and $29.0 billion for ISC. Acute-care costs incurred in the 2 years following a first stroke accounted for 45.0%, long-term ambulatory care accounted for 35.0%, and nursing home costs accounted for 17.5% of aggregate lifetime costs of stroke.
Conclusions The lifetime cost of stroke varies considerably by type of stroke and entails considerable costs beyond the first 2 years after a stroke.
Stroke is a leading cause of morbidity and mortality in the United States, imposing an enormous economic burden on individuals and society overall. Data compiled by the Stroke Prevention Patient Outcomes Research Team1 indicate that annually there are as many as 550 000 hospitalizations and 150 000 deaths attributable to stroke in the United States.
The economic burden of stroke can be defined in terms of the direct costs of providing medical care to patients and the indirect costs associated with lost productivity. Studies of the economic burden of stroke typically use a prevalence-based approach to estimating the cost of stroke in a given year and often focus solely on the direct costs of care. Prevalence-based studies have been used to estimate the cost of treating stroke in Canada,2 Sweden,3 4 the United Kingdom,5 6 and the United States.1 7 8 9 10 The most recent prevalence-based study of the cost of stroke in the United States, conducted by the Stroke Prevention Patient Outcomes Research Team,1 estimated the economic burden of stroke to be $30 billion in 1993: $17 billion in direct medical costs and $13 billion in indirect costs associated with lost earnings. Although prevalence-based studies are helpful in identifying the costs of stroke at a given point in time, they provide little insight into the lifetime costs associated with incident cases.
Incidence-based, or lifetime, cost-of-illness studies are more difficult to perform than prevalence-based studies because of the need for detailed information on the natural history of the disease and because they require that assumptions be made about future costs of treatment.11 Lifetime cost-of-illness models are useful for evaluating the economic implications of alternative prevention and treatments options and public policies affecting stroke treatment. Incidence-based cost-of-illness models have been developed for estimating the lifetime costs of injury,12 cigarette smoking,13 and epilepsy.14
Two studies estimate lifetime costs associated with stroke. Hartunian et al15 developed an incidence-based, lifetime cost-of-illness model for ischemic strokes using data collected during the 1970s. They estimated the lifetime cost of stroke to be $1.7 billion for hemorrhagic strokes and $4.6 billion for infarctions occurring in 1975. Indirect costs accounted for 86% of the total costs of hemorrhagic stroke and 56% of total costs of ischemic stroke. Bergman et al16 estimated the lifetime direct cost of medical care for patients in The Netherlands in 1991. In their model, nursing home costs accounted for 50% of lifetime costs.
This study presents estimates of the lifetime cost of stroke in the United States by type of stroke for individuals experiencing a first stroke in 1990. We measure cost from the societal perspective, including both direct and indirect costs. We build on previous research in this area by estimating costs for three stroke subtypes and by incorporating more recent and more comprehensive measures of the epidemiology and costs of stroke.
We developed a computer-simulation model of the lifetime economic burden of stroke using standard incidence-based, cost-of-illness methodology.11 12 13 14 15 We estimated lifetime cost per person and aggregate lifetime costs for all first strokes occurring in 1990. Lifetime cost of stroke per person was defined as the sum of direct medical costs and indirect costs associated with lost productivity. Separate estimates of the per-person lifetime cost of stroke were developed for the three main subtypes: SAH, ICH, and ISC. Estimates of the aggregate lifetime cost of stroke were obtained by multiplying per-person lifetime cost estimates times the estimated incidence of first strokes in 1990. Details of the mathematical model used are provided in the “Appendix.” All costs and earnings were estimated in terms of 1990 dollars. Future costs and earnings were discounted at the 5% rate commonly used in economic evaluations in health care.17 18
Stroke cases were classified by subtype according to the ICD-9-CM: SAH, code 430; ICH, codes 431 and 432; and ISC, codes 433, 434, and 436.
The epidemiological component of the model was designed to reflect the community history of the disease in terms of three key variables: incidence, survival, and recurrence rates. The epidemiological data used in this model were based on a review of the literature and include population-based studies selected according to criteria designed to result in an estimate reflective of the US population. Where data were deficient, hospital-based studies were used to supplement the data.
A MEDLINE search of all English language articles published since 1980 was conducted to identify all studies of stroke incidence. These articles and any studies referenced in them were obtained and reviewed according to strict inclusion criteria that were defined before the review process. These criteria were based on those suggested by Malmgren et al19 as appropriate for an “ideal” stroke incidence study. In addition, we added the following criteria: (1) CT scanning or necropsy was performed in at least 80% of cases to ensure accurate subtypes; (2) studies were performed in 1980 or later; and (3) populations were chosen to reflect the US population and included North America, Australia, New Zealand, the United Kingdom, and France. If age- and sex-specific rates were not included, this information was requested from the authors. Because of limited data for hemorrhagic stroke, we included one study in which some of the data were collected before 198020 and two studies with CT scanning or necropsy performed in fewer than 80% of cases.20 21 There were six studies that met the inclusion criteria,20 21 22 23 24 25 26 27 but two of these included only hemorrhagic stroke.21 25
Incidence studies were combined using age-specific rates for the same 10-year age groups and weighted according to the number of person-years of follow-up. For ISC, sex-specific rates were similarly combined. For SAH, the relative risk was higher in women, and a constant ratio of 1.6 was applied across all age groups; for ICH, the risk was higher in men, and a ratio of 1.1 was used across all age groups.
For racial differences, data were limited to one incidence study in young adults for ISC.28 These data were supplemented by mortality and hospitalization data29 30 31 to estimate a ratio to correct for black/white differences, which declined with increasing age. For hemorrhagic stroke, information concerning racial differences was limited to one study.25 A ratio to correct for racial differences was uniformly applied across all age groups for SAH, and a variable ratio that declined with age was used for ICH.
The aggregate age-, sex-, and race-specific incidence rates for each subtype of stroke (Table 1⇓) were applied to the 1990 resident US population to estimate the number of first-ever strokes predicted in 1990.
Review of both population-based and hospital-based series revealed no differences in survival for any subtypes of stroke according to sex or race, so only age-specific survival rates were determined. For ISC, a single population-based study was used in which age-specific survival rates for 5 years after ISC were available.32 For hemorrhagic stroke, data concerning survival rates beyond 1 year were very limited, so we assumed no difference in survival after 1 year compared with the general population. For SAH, an unweighted average across population-based studies was determined.20 21 23 26 33 34 35 36 A trend in age-specific survival was obtained from a clinical series37 and was applied to the population-based rate. For ICH, an unweighted average survival rate was obtained by combining both population-based and clinical series,20 28 35 36 38 39 40 41 and age-specific rates from the Oxfordshire study28 were used to apply this trend to the overall rate. Survival rates for the subtypes of stroke are presented in Table 2⇓. For SAH and ICH but not ISC, a 10% rate of sudden death (before hospitalization) was included.42 43
Data were too limited to assess recurrence of hemorrhagic stroke at 3 years or beyond, so we assumed no recurrence for these subtypes. For ISC, both population- and hospital-based studies44 45 46 47 were assessed to determine whether recurrence was influenced by age, sex, or race, and no differences were found. Two population-based studies44 45 reported survival at 3 years or beyond, but only the Oxfordshire study45 gave recurrence rates for each of 3, 4, and 5 years, so these data were used in the model.
Direct Medical Cost Model Overview
Direct costs were disaggregated into acute-care costs incurred in the first 2 years following a stroke, long-term ambulatory care costs incurred 3 or more years following a stroke, nursing home costs, and costs attributable to stroke recurrence.
Two-Year Direct Costs
We used health insurance claims data covering a large, nonelderly population and Medicare claims data covering the elderly population to estimate first- and second-year costs associated with stroke. We used different data sources because it was not apparent that Medicare claims data could be generalized to the nonelderly population.
In estimation of the direct cost of stroke from the perspective of society as a whole, the conceptually appropriate measure of cost is what economists refer to as “opportunity cost,” the value of resources used in their next best use. Observed market prices are often used as measures of cost by economists. In health care, however, it is generally acknowledged that market imperfections make observed market prices poor measures of opportunity costs.48 49 The most commonly available financial accounting measures, such as provider charges or insurer payments, have serious limitations as measures of cost. Published charges are often inflated to counteract third-party discounts or may reflect expenses unrelated to the direct provision of care (eg, medical education). Insurer payments exclude patient out-of-pocket expenses. In the case of Medicare hospital inpatient stays, hospital charges are often adjusted by the Medicare cost-to-charge ratio in an effort to derive a direct measure of cost.49 Our overall strategy in estimating direct medical care costs was to estimate cost directly, using cost-to-charge ratios where possible, and to use Medicare-allowed charges as a proxy for costs for ambulatory-care services. Where allowed charges were not available, we used total payments (insurer reimbursements plus patient out-of-pocket expenditures) as our measure of cost.
Population Under 65 Years of Age
Computerized insurance claims data were supplied by MEDSTAT Systems Inc (Ann Arbor, Mich) for 1782 individuals hospitalized in 1990 with a principal diagnosis of stroke. These patients were followed up for 2 years after the initial 1990 hospitalization, and medical care utilization and expenditures were analyzed. Patients were classified by type of stroke according to the principal diagnosis for the first stroke hospitalization in 1990.
We used total payments as a proxy for costs for the population under age 65 years. Total payments were defined as the sum of all third-party payments plus patient cost-sharing (deductibles and coinsurance) payments. We did not have access to the data for this population that would permit direct estimation of hospital costs, as is possible with Medicare data. A comparison of MEDSTAT total payments for stroke with total charges estimated from the Medicare data suggested that the two were very similar for SAH and ISC. We treated claim payments reported for the under-65 population as equivalent to Medicare charge data and adjusted the MEDSTAT hospital payments by the mean effective Medicare cost-to-charge ratio estimated for all stroke cases.
Population Over 65 Years of Age
Data for the population over 65 years of age were obtained from the 5% sample public-use Medicare claims file maintained by the Health Care Financing Administration. This file contains computerized discharge abstract information for six types of service claims: hospital inpatient and outpatient, physician/supplier Part B, skilled nursing facility, home health agency, and hospice. Because we undertook a separate analysis of nursing home costs, estimates of the cost of skilled nursing facility stays were excluded from the analysis of Medicare costs in the 2 years following a stroke. Data files covering calendar years 1990 through 1992 were used in the analysis. From this 5% sample of Medicare beneficiaries, we identified 18 768 individuals aged 65 years and older with a 1990 hospitalization with a principal diagnosis of stroke.
We randomly selected 31 692 (2%) Medicare beneficiaries in the 5% sample with at least one medical claim in 1990 to serve as a control population. This represents a 1/1000 sample of the Medicare population with claims in 1990. Individuals with a 1990 hospitalization with a principal diagnosis of stroke were excluded from the control group.
The first hospitalization in 1990 with a principal diagnosis of stroke was treated as the index event. Some of these cases represent recurrent cases or rehospitalizations for beneficiaries who had a stroke in earlier years. In a separate analysis, we identified all stroke admissions in 1992 using the same stroke classification scheme. Of the estimated 425 680 Medicare beneficiaries hospitalized for stroke in 1992, 34 100 (8%) had a stroke admission in either 1990 or 1991. Thus, while our estimates of the direct costs of incident strokes per person include costs for some prevalent cases, the impact is likely to be small.
Medicare beneficiaries identified as having a stroke in 1990 were followed up for 2 years. The 5% sample hospital records supplied by the Health Care Financing Administration show only the quarter in which the patient was admitted to the hospital. We followed up patients for seven quarters beyond the quarter of the index hospitalization. The index quarter in 1990 for control subjects was randomly selected to ensure that approximately the same percentage of control and stroke cases would be followed up into 1993. Charges and payments incurred more than 1 year after the index hospitalization were deflated using the Consumer Price Index.
For hospitalizations, we multiplied total charges by the hospital-specific Medicare operating cost-to-charge ratio to estimate costs. Hospital-specific Medicare operating cost-to-charge ratios were obtained from the Provider Specific File maintained by the Health Care Financing Administration and merged with the 5% sample claims data using the hospital provider number. Operating cost-to-charge ratios were not available for all hospitals because some types of hospitalizations (eg, rehabilitation) are excluded from the Prospective Payment System. Therefore, we calculated an average cost-to-charge ratio based on the ratio of estimated costs to charges for all admissions for which cost-to-charge data were available. For outpatient claims, we estimated actual payments by adding back the 20% coinsurance paid for most outpatient services. This highly simplifies the complex reimbursement rules for outpatient services but should serve as a reasonable proxy for outpatient costs. For Part B claims, we used allowed charges as our measure of cost because submitted charges are likely to overstate the true cost of providing services, and the primary difference between Medicare payments and allowed charges in Part B reflects out-of-pocket expenses incurred by Medicare beneficiaries. Because patients pay very little out of pocket for home health care, we used Medicare payments as our estimate of cost for these claims.
On the basis of our review of the literature and evaluation of both the MEDSTAT and Medicare data, we assumed no recurrence beyond the first year for either type of hemorrhagic stroke. We based our estimates of recurrence rates for ISC on data from the Oxfordshire Community Stroke Project.45 Stroke recurrence is implicit in the claims data for the first 2 years after an initial stroke. For subsequent years, the costs associated with ISC recurrence were estimated as the product of annual probability of recurrence after an initial stroke for years 3 (5.0%), 4 (3.3%), and 5 (1.3%) multiplied by the estimated annual cost associated with recurrent ISC. We assumed that the annual cost for someone having another stroke in the 3rd, 4th, or 5th year after a first stroke was the same as the annual cost for the initial stroke.
Prescription Drug Costs
Estimates of prescription drug costs associated with treatment of stroke were obtained from the Household Component of the 1987 NMES.50 The Household Component of the NMES is a national probability sample of households in the United States designed to provide estimates of medical expenditures for noninstitutionalized persons. We used the prescription drug and medical sundries component of the Consumer Price Index to update 1987 drugs costs to 1990 dollars. We identified 618 survey respondents with a self-reported history of stroke. Mean drug expenditures by age category, weighted by the weights used to generate national expenditure projections, were calculated for those reporting a history of stroke and those not reporting a history of stroke. Age-specific differences in spending between the stroke and nonstroke groups were used as estimates of prescription drug spending attributable to stroke.
Long-term Follow-up Costs
To estimate long-term follow-up costs incurred by survivors in year 3 and later, we used an approach similar to that used to estimate prescription drug costs. We excluded 13 (of 618) patients in the NMES who reported being hospitalized in 1987 for stroke. This was done because individuals in the NMES with a stroke hospitalization were likely to have been incident cases, and we were attempting to estimate expenditures in the years following the initial stroke. We used the medical care component of the Consumer Price Index to update long-term follow-up costs to 1990. Long-term ambulatory spending attributable to stroke by age category was estimated as the difference in the age-specific ambulatory spending for stroke versus that in nonstroke cases in the NMES.
Nursing Home Costs
To estimate the nursing home expenditures of individuals admitted to nursing homes after a stroke in 1990, we used a three-step process. First, we estimated the discounted lifetime cost of a nursing home stay per person admitted to a nursing home because of stroke. Second, we estimated the probability of nursing home admission for individuals having a stroke in 1990. Estimates of the lifetime nursing home costs associated with stroke were obtained by multiplying age-, sex-, and stroke-specific incidence rates times the probability of nursing home admission times the discounted cost of a nursing home stay.
We used data from the 1987 NMES-IPC to estimate length of stay and expenses for stroke nursing home residents discharged in 1987.51 The NMES-IPC is a national probability sample of current residents and persons admitted to nursing and personal care facilities in 1987. It is designed for use in producing estimates of US nursing home utilization and expenditures.
NMES-IPC contains information on primary and secondary diagnoses associated with the nursing home stay obtained from the medical record. It was not possible to estimate length of stay and expenditure data by type of stroke because of the limited number of codes reported other than ICD-9-CM code 436. Approximately 90% of nursing home residents with a primary diagnosis of stroke were coded as ICD-9-CM code 436 (acute, but ill defined, cerebrovascular disease).
We estimated length of nursing home stays for stroke residents by estimating the mean length of stay for 197 stroke residents discharged in 1987. We grouped discharges by years of stay and estimated the frequency of nursing home discharge by years of stay. Mean length of stay and mean per diem expenditures for persons discharged in 1987 were estimated by years of stay.
Probabilities of nursing home admissions for persons hospitalized for stroke by age and stroke type were estimated using data from the 1988, 1990, and 1991 National Hospital Discharge Survey,31 a national probability sample of acute hospital discharges. Hospital discharges with a principal diagnosis of stroke were selected, and the proportion of cases with a discharge status indicating discharge to a nursing home was used to estimate the probability of nursing home admission after a stroke hospitalization. We averaged rates across the 3 years to improve the reliability of the estimates.
Estimates of nursing home admissions based solely on discharges directly from hospitals to nursing homes are likely to seriously underestimate total nursing home admissions because they fail to account for admissions from nonhospital settings such as the home. Our initial estimate of the number of nursing home admissions directly from hospitals was 73 100. To account for stroke nursing home admissions from nonhospital sources, we compared this estimate to an estimate of the number of persons admitted to nursing homes with a stroke diagnosis. We used the 1987 NMES-IPC sample of nursing home admissions and identified 308 nursing home admissions with a primary diagnosis of stroke. Using weights provided for the NMES-IPC, these 308 cases projected to 101 900 stroke-related nursing home admissions nationally in 1987. We used the ratio of NMES-IPC stroke-related nursing home admissions to the model-estimated number of nursing home admissions directly from hospitals (101 900/73 100) to adjust our age- and stroke-specific nursing home admissions by 1.4 to ensure estimates of total stroke-related nursing home admissions consistent with the NMES-IPC estimates. This ex post facto adjustment was necessary to ensure model predictions of nursing home admissions consistent with national estimates.
Indirect Cost Model
The present values of indirect costs were measured according to the human capital approach. The human capital approach measures indirect costs in terms of the value of foregone productive activity attributable to premature mortality and morbidity. Productive activity can be either “market” or “nonmarket.” Market activity refers to all productive activities (eg, work) that are remunerated. Nonmarket productive activities include those activities, such as volunteer work or work around the house, that reflect productive activity but for which there is no direct remuneration.
As detailed in the “Appendix,” we define the indirect costs associated with stroke as the difference between the age- and sex-specific present value of lifetime productive activity (market plus nonmarket) for stroke and nonstroke populations. Indirect costs arise for two reasons in this model: premature mortality and reduced productivity for stroke survivors. Indirect cost associated with stroke mortality is measured as the present value of future lifetime productive activity that would have been realized in the absence of a stroke (eg, for the nonstroke population). Indirect cost associated with reduced productive activity among stroke survivors is estimated as the difference in lifetime earnings for stroke survivors compared with an age- and sex-matched nonstroke population.
The expected value of lifetime market production for the nonstroke population (lost earnings associated with stroke mortality) was estimated as the product of the age- and sex-specific labor force participation rate52 multiplied by age- and sex-specific mean annual earnings for individuals with any earnings.53 Mean annual earnings for the general population are defined according to the US Census Bureau and are assumed to grow at a 1% real rate.12 Annual market earnings are greater for men, peak in the 45- to 54-year age range and diminish considerably after age 65.
To estimate the impact of stroke on mean annual earnings for stroke survivors, we have compared earnings estimates of individuals reporting a prior stroke with those not reporting a prior stroke using data from the Household Component of the 1987 NMES. Age- and sex-specific mean annual earnings estimates for stroke survivors declined with age and ranged from 21% to 58% of mean annual earnings for the nonstroke population.
Nonmarket production includes such activities as child rearing, housekeeping, home maintenance, volunteer work, and other productive activities for which no payment is received. Valuing nonmarket activities poses methodological and empirical challenges. Failure to estimate the value of nonmarket production understates the value of production by all groups. This bias is particularly severe for women and the elderly, who have lower labor-force participation rates than nonelderly males.
Estimates of the value of lost nonmarket production have usually been based on studies of the market value of household services. The most recent and most comprehensive estimates of the value of nonmarket activities have been developed by Herzog and Morgan and colleagues.54 55 We use the “market price” estimates of Herzog and Morgan as the basis for estimating the value of nonmarket production for child care, housework, and home maintenance. Because volunteer work encompasses a wide range of activities, estimated earnings of volunteers were used as the measure of the dollar value of volunteer services. These 1986 estimates were updated to 1990 dollars based on the nominal growth in male and female median earnings over the period. Projections of future nonmarket earnings assume the same 1% productivity growth rate assumed for market earnings. Nonmarket earnings were estimated to be considerably less than market earnings during prime earnings years, but they exceed average market earnings at postretirement ages. Nonmarket earnings are nearly twice as great for women compared with men, peak during child-rearing years, and diminish thereafter.
Direct Cost of Stroke
The estimated annual direct costs associated with stroke, by type of stroke, age, and type of cost, are given in Table 3⇓. Inpatient hospital costs comprised the majority of first-year costs, accounting for over 70% of first-year direct cost (exclusive of nursing home costs).
We estimated 101 900 nursing home admissions attributable to strokes occurring in 1990, at a net present value of $29 296 per nursing home admission. The mean length of stay for stroke patients admitted to nursing homes was 432 days.
Lifetime Costs Per Person
We estimated the mean lifetime cost of stroke per person to be $228 030 for SAH, $123 565 for ICH, and $90 981 for ISC (Fig 1⇓). Averaged across all individuals and types of stroke, the lifetime cost per person is $103 576. The higher cost associated with SAH is a consequence of higher direct cost and an earlier mean age of onset, resulting in much greater indirect costs.
Mean per-person lifetime cost of stroke varies considerably by age and type of stroke, with the greatest cost associated with men who have a stroke in their mid twenties (Fig 2⇓). The higher lifetime cost associated with SAH and ICH at early ages reflects higher mortality rates associated with these subtypes. The mean per-person lifetime cost of stroke is greater for men than for women for all three types of stroke, primarily because of greater market earnings assumed for men (Table 4⇓).
Indirect costs account for the majority of lifetime cost for each type of stroke. The high mortality rates associated with SAH and ISC, coupled with an earlier age of onset, result in indirect costs that are two to three times greater than direct costs.
Aggregate Lifetime Cost
We estimate the aggregate economic burden associated with the approximately 390 000 first strokes in 1990 to have been $40.6 billion (Table 5⇓). SAH accounted for 13.7% ($5.6 billion), ICH for 14.8% ($6.0 billion), and ISC for 71.5% ($29.0 billion). Direct costs accounted for 42.0% ($17.1 billion) of the total economic burden in 1990 (Fig 3⇓). Acute-care costs incurred in the first 2 years after a stroke account for 45% of all direct costs, with long-term ambulatory accounting for 35.1% and nursing home accounting for 17.5%. The costs associated with stroke recurrence taking place in the 2 years after a first stroke are included in the first 2-year costs. Recurrent strokes occurring in years 3 through 5 accounted for 2.9% of the total direct costs. Lost earnings attributable to premature mortality account for 56.0% of total indirect costs, with the remainder attributable to lost earnings for stroke survivors.
We varied key parameters in the model to assess the robustness of the underlying assumptions. Using a 10% discount rate resulted in a 22% reduction in the aggregate lifetime cost of stroke, with a 27% reduction in the SAH, a 26% reduction in ICH costs, and a 21% reduction in ISC costs. Given the multiplicative nature of the model, a 10% reduction in incidence rates reduced the aggregated economic burden by a corresponding 10%.
Increasing stroke survival rates by 10% had virtually no effect on the aggregate cost estimates. SAH costs declined only 1%, whereas ICH and ISC costs declined by even less. Even a 50% increase in survival rates reduced total lifetime cost by less than 3%. On the other hand, setting market and nonmarket earnings of stroke survivors equal to the nonstroke population, reflecting prestroke functional status, reduced aggregate lifetime costs for all stroke types by 25.6%.
We provide estimates of the lifetime cost of stroke for three major subtypes of stroke. Our estimates of the aggregate lifetime cost of stroke are approximately 60% to 80% greater than those obtained by Hartunian et al15 even after accounting for the effects of medical sector inflation and population growth. Although our overall estimates are considerably higher, Hartunian et al also estimated indirect costs, which were nearly twice as great as aggregate direct costs. We incorporate new measures of the value of nonmarket activity, which contribute to the higher indirect cost estimates. We believe these measures provide a more accurate measure of the economic contribution of women and the elderly.
Our estimates of the direct costs associated with stroke in the first 2 years after a stroke are consistent with previous research documenting the high costs associated with hospital inpatient costs in the first year. Our estimates of the costs associated with nursing home stays and long-term ambulatory care services suggest that major costs are incurred beyond the first 2 years after a stroke. Caution must be exercised in interpreting these results, however, because we have projected data extracted from 1 year far into the future.
The sensitivity analysis demonstrated that increasing stroke survival had little effect on the aggregate cost estimates. This occurs primarily because the increase in direct costs associated with greater longevity are only barely offset by reductions in indirect costs. The modest decline in indirect cost reflects the lower earnings estimates of stroke survivors assumed in the model. Alternatively, setting survivors' earnings equal to earnings of the nonstroke population, reflecting improved functional status of stroke survivors, had a major impact on lifetime cost by reducing indirect costs associated with stroke. The key to reducing the economic burden of stroke appears to rest in either preventing its occurrence or further reducing stroke mortality and improving functioning and quality of life of survivors.
We used a number of simplifying assumptions in constructing this model of lifetime cost of stroke. The MEDSTAT database covers employees and their dependents of larger employers and underrepresents persons located in the South and West. Patient follow-up was confounded by lack of information on eligibility. It was not possible to identify workers who changed jobs and/or insurers, although efforts were made to minimize this problem by selecting from a population of employers who provided data to MEDSTAT consistently from 1990 through 1992. The Medicare claims data are less likely to suffer from loss to follow-up because few elderly voluntarily withdraw from the Medicare program once they are eligible for benefits. Information on beneficiary race and age, however, was not available for approximately 15% of control cases.
Hypertension, smoking, and diabetes are risk factors for stroke. For a large number of stroke patients, costs associated with treating these underlying risk factors would have been incurred independent of their stroke. We attempted, where possible, to estimate costs attributable specifically to stroke by subtracting age-specific direct costs for patients with stroke from age-specific costs for the rest of the population. Because our control populations were not matched for stroke risk factors (eg, hypertension), our estimates may attribute some of the treatment costs associated with the risk factors to stroke.
Although we believe our reliance on insurance claims data provides the most accurate measure of the direct cost of treating stroke, insurers pay for only part of the direct cost of care. We have also had to make a number of assumptions concerning cost incurred in subsequent years by relying on self-reported stroke in the NMES. In addition, estimating lifetime cost of illness necessarily entails projections far into the future. The estimates presented here reflect patterns of care that are likely to change significantly in the next decade. Finally, we did not attempt to estimate the costs incurred by unpaid caregivers, whose efforts on behalf of stroke survivors represent an important, if as yet unmeasured, component of the lifetime cost of stroke.
Selected Abbreviations and Acronyms
|ICD-9-CM||=||International Classification of Diseases, 9th Clinical Modification|
|NMES (-IPC)||=||National Medical Expenditure Survey (Institutional Population Component)|
The aggregate lifetime cost of first strokes occurring in 1990 is modeled as the product of the incident stroke cases and the lifetime cost for each type of stroke (i) by age at first stroke (y) and sex (s):PVC|<|=|>||<|\Sigma|>| n_|<|i,y,s|>||<|[|>|PVDC_|<|i,y,s|>||<|+|>|PVIDC_|<|i,y,s|>||<|]|>|where ni,y,s is the annual number of new cases of stroke of type i among persons in age group y, and sex s; PVDCi,y,s is the present value of direct costs for new cases of stroke of type i among persons in age group y and sex s; and PVIDCi,y,s is the present value of indirect costs for new cases of stroke of type i among persons in age group y and sex s.
The present value of direct cost is defined as the sum of discounted expected direct costs of medical treatment attributable to stroke from the time of the initial stroke to age 99 years:PVDC_|<|i,y,s|>||<|=|>||<|\sum_|<|t|<|=|>|0|>|^|<|99|<|-|>|y|>||>| \frac|<|P_|<|i,y,s|>|(t)DC_|<|i,y,s|>|(t)|>||<|(1|<|+|>|r)^|<|t|>||>|where Pi,y,s(t) is the probability of surviving t years beyond the initial stroke; DCi,y,s(t) is the direct cost of medical care per person in year t following the initial stroke for persons surviving to year t; and r is the discount rate.
We model the present value of indirect costs as the sum of the discounted differences between the expected value of market and nonmarket production for persons without stroke, EVns, and the expected value of market and nonmarket production for persons after a stroke, EVs:PVIDC_|<|i,y,s|>||<|=|>||<|\sum_|<|t|<|=|>|0|>|^|<|99|<|-|>|y|>||>| \frac|<||<|[|>|P_|<|(|<|\ast|>|)|>|^|<|ns|>|(t)EV_|<|(|<|\ast|>|)|>|^|<|ns|>|(t)|<|]|>||<|-|>||<|[|>|P_|<|(|<|\ast|>|)|>|^|<|s|>|(t)EV_|<|(|<|\ast|>|)|>|^|<|s|>|(t)|<|]|>||>||<|(1|<|+|>|r)^|<|t|>||>|where Pns(*)(t) is the probability of surviving t years beyond age y for a person of sex s who has not had a stroke; EVns(*)(t) is the expected value of market and nonmarket production for persons of age y+t and sex s who have not had a stroke; Ps(*)(t) is the probability of surviving t years beyond age y for a person of sex s who had a stroke of type i at age y; and EVs(*)(t) is the expected value of market and nonmarket production for stroke population t years after first stroke of type i at age y.
We note that there are two sources of indirect costs associated with this model. The first is attributable to losses accruing to excess morbidity, which results in reduced production for survivors compared with the general population [EVs(*)<EVns(*)]. The second is attributable to premature mortality [Ps(*)<Pns(*)(t)].
The expected value of productive activity is defined as the sum of the expected value of market production plus the value of nonmarket production:EV_|<|y,s|>|^|<|ns|>|(t)|<|=|>|LFPR_|<|(y,s)|>|^|<|ns|>|(t) MEARN_|<|(y,s)|>|^|<|ns|>|(t)|<|+|>|NMEARN_|<|(y,s)|>|^|<|ns|>|(t)EV_|<|y,s|>|^|<|s|>|(t)|<|=|>|LFPR_|<|(y,s)|>|^|<|s|>|(t) MEARN_|<|(y,s)|>|^|<|s|>|(t)|<|+|>|NMEARN_|<|(y,s)|>|^|<|s|>|(t)where LFPR(y,s)(t) is the the age- and sex-specific labor force participation rate; MEARN(y,s)(t) is the mean annual market earnings for persons of age y+t and sex s with any earnings; and NMEARN(y,s)(t) is the mean annual value of nonmarket production.
This research was supported by grants from Pharmacia & Upjohn, Inc, Kalamazoo, Mich. We are indebted to Harold Adams, MD, Steven Hass, PhD, and Nancy O'Brien, ART, for their helpful advice and comments concerning various aspects of this project.
Dr Taylor was employed by the Upjohn Company from September 1988 through September 1993. This research project began in March 1994.
- Received November 21, 1995.
- Revision received May 23, 1996.
- Accepted May 23, 1996.
- Copyright © 1996 by American Heart Association
Matchar DB, Duncan PW. Cost of stroke. Stroke Clinical Updates. 1994;5:9-12.
Smurawska LT, Alexandrov AV, Bladin CF, Norris JW. Cost of acute stroke care in Toronto, Canada. Stroke. 1994;25:1628-1631.
Persson U, Silverberg B, Lindgren B, Norrving B, Jadbäck G, Johansson B, Puranen BI. Direct costs of stroke for a Swedish population. Int J Technol Assess Health Care. 1990;4:125-137.
Asplund K, Marke LA, Terent A, Gustafsson C, Wester PO. Costs and gains in stroke prevention: European perspective. Cerebrovasc Dis. 1993;3(suppl 1):34-42.
Drummond M, Ward G. The financial burden of stroke and the economic evaluation of treatment alternatives. In: Rose FC, ed. Stroke: Epidemiological, Therapeutic and Socioeconomic Aspects. London, UK: Royal Society of Medicine Services Ltd; 1986.
Isard PA, Forbes JF. The cost of stroke to the National Health Service in Scotland. Cerebrovasc Dis. 1992;2:47-50.
Adelman MS. Economic impact: National Survey of Stroke. Stroke. 1981;12(suppl 1):I-69-I-78.
Cooper B, Rice DP. The economic cost of illness revisited. Soc Secur Bull. 1976;39:21-36.
Goldstein M. Stroke prevention and health care issues: an American perspective. Cerebrovasc Dis. 1993;3(suppl 1):29-33.
Wiebers DO, Torner JC, Meissner I. Impact of unruptured intracranial aneurysms on public health in the United States. Stroke. 1992;23:1416-1419.
Hodgson TA, Meiners MR. Cost-of-illness methodology: a guide to current practices and procedures. Milbank Q.. 1982;60:429-462.
Bergman L, van der Meulen JHP, Limburg M, Habbema DF. Costs of medical care after first-ever stroke in the Netherlands. Stroke. 1995;26:1830-1836.
Drummond MF, Stoddart GL, Torrance GW. Methods for the economic evaluation of health care programmes. Oxford, UK: Oxford University Press; 1987.
Malmgren R, Warlow C, Bamford J, Sandercock P. Geographical and secular trends in stroke incidence. Lancet. 1987;1:1196-1200.
Broderick JP, Phillips SJ, Whisnant JP, O'Fallon WM, Bergstralh EJ. Incidence rates of stroke in the eighties: the end of the decline of stroke? Stroke. 1989;20:577-582.
Bonita R, Thomson S. Subarachnoid hemorrhage: epidemiology, diagnosis, management, and outcome. Stroke. 1985;16:591-594.
Giroud M, Milan C, Beuriat P, Gras P, Essayagh E, Arveux P, Dumas R. Incidence and survival rates during a two-year period of intracerebral and subarachnoid haemorrhages, cortical infarcts, lacunes and transient ischemic attacks. The stroke registry of Dijon: 1985-1989. Int J Epidemiol. 1991;20:892-899.
Bamford J, Sandercock P, Dennis M, Burn J, Warlow C. A prospective study of acute cerebrovascular disease in the community: the Oxfordshire Community Stroke Project—1981-86, II: incidence, case fatality rates and overall outcome at one year of cerebral infarction, primary intracerebral and subarachnoid haemorrhage. J Neurol Neurosurg Psychiatry. 1990;53:16-22.
Bamford J, Sandercock P, Dennis M, Warlow C, McPherson K, Vessey M, Fowler G, Molyneux A, Hughes T, Burn J, Wade D. A prospective study of acute cerebrovascular disease in the community: the Oxfordshire Community Stroke Project—1981-86, I: methodology, demography and incident cases of first-ever stroke. J Neurol Neurosurg Psychiatry. 1988;51:1373-1380.
Kittner SJ, McCarter RJ, Sherwin RW, Sloan MA, Stern BJ, Johnson CJ, Buchholz D, Seipp MJ, Price TR. Session II: age, race, and sex. Black-white differences in stroke risk among young adults. Stroke. 1993;24(suppl I):I-13-I-15.
National Center for Health Statistics. Health, United States, 1990. Hyattsville, Md: Public Health Service; 1991.
Howard G, Anderson R, Sorlie P, Andrews V, Backlund E, Burke L. Ethnic differences in stroke mortality between non-Hispanic whites, Hispanic whites, and blacks: the National Longitudinal Mortality Study. Stroke. 1994;25:2120-2125.
US Department of Health and Human Services, National Center for Health Statistics. National Hospital Discharge Survey, 1990 [computer file]. Hyattsville, Md: US DHHS, National Center for Health Statistics [producer], 1991. Ann Arbor, Mich: Inter-university Consortium for Political and Social Science Research [distributer], 1993.
Dennis MS, Burn JPS, Sandercock PAG, Bamford JM, Wade DT, Warlow CP. Long-term survival after first-ever stroke: the Oxfordshire Community Stroke Project. Stroke. 1993:24:796-780.
Sacco RL, Wolf PA, Bharucha NE, Meeks SL, Kannel WB, Charette LJ, McNamara PM, Palmer ED, D'Agostino R. Subarachnoid and intracerebral hemorrhage: natural history, prognosis, and precursive factors in the Framingham Study. Neurology. 1984;34:847-854.
Drury I, Whisnant JP, Garraway WM. Primary intracerebral hemorrhage: impact of CT on incidence. Neurology. 1984;34:653-657.
Bamford J, Dennis M, Sandercock P, Burn J, Warlow W. The frequency, causes and timing of death within 30 days of a first stroke: the Oxfordshire Community Stroke Project. J Neurol Neurosurg Psychiatry. 1990;53:824-829.
Schievink WI, Wijdicks EFM, Parisi JE, Piepgras DG, Whisnant JP. Sudden death from aneurysmal subarachnoid hemorrhage. Neurology. 1995;45:871-874.
Meissner I, Whisnant JP, Garraway WM. Hypertension management and stroke recurrence in a community (Rochester, Minnesota, 1950-1979). Stroke. 1988;19:459-463.
Burn J, Dennis M, Bamford J, Sandercock P, Wade D, Warlow C. Long-term risk of recurrent stroke after a first-ever stroke: the Oxfordshire Community Stroke Project. Stroke. 1994;25:333-337.
Hier DB, Foulkes MA, Swiontoniowski M, Sacco RL, Gorelick PB, Mohr JP, Price TR, Wolf PA. Stroke recurrence within 2 years after ischemic infarction. Stroke. 1991;22:155-161.
Lai SM, Alter M, Friday G, Sobel E. A multifactorial analysis of risk factors for recurrence of ischemic stroke. Stroke. 1994;25:958-962.
Finkler SA. The distinction between cost and charges. Ann Int Med. 1982:96:102-109.
Lave JR, Pashos CL, Anderson GF. Costing medical care using Medicare administrative data. Med Care. 1994;32(suppl):JS77-JS89.
US Department of Health and Human Services, Agency for Health Care Policy and Research. National Medical Expenditure Survey, 1987: Household Survey [various computer files]. Rockville, Md: USDHHS, AHCPR [producer], 1993. Ann Arbor, Mich: Inter-university Consortium for Political and Social Research [distributor], 1994.
US Department of Health and Human Services, Agency for Health Care Policy and Research. National Medical Expenditure Survey, 1987: Institutional Population Component [various computer files]. Rockville, Md: USDHHS, AHCPR [producer], 1994. Ann Arbor, Mich: Interuniversity Consortium for Political and Social Research [distributor], 1994.
US Department of Labor, Bureau of Labor Statistics. Employment and Earnings, January 1991.
US Department of Commerce, Bureau of the Census. Money Income of Households, Families, and Persons in the United States, 1990. Series P-60, No. 174, Aug 1991.
Herzog AR, Kahn RL, Morgan JN, Jackson JS, Antonucci TC. Age differences in productive activities. J Gerontol. 1989;44:S129-S138.
Herzog AR, Morgan JN. Age and gender differences in the value of productive activities: four different approaches. Research on Aging. 1992;14:169-198.