SUMMARY
Rationale
Stroke is a common and costly condition where an effective early reatment may be expected to affect patients’ future quality of life, the cost of acute medical treatment, and the cost of rehabilitation and any supportive care needed for their remaining lifetime. To assist in informing discussions on early adoption of potential treatments, economic analyses should accompany investigations that seek to improve outcomes for stroke patients.
Aims
The primary aim is to assess whether IV/IA rt-PA therapy is cost-effective at 3 months compared to IV rt-PA, and provides cost-savings or is cost-neutral by 12 months.
Design
Cost-effectiveness of the two treatment arms will be measured at months 3, 6, 9, and 12. Cost-effectiveness will be calculated using 1) standard cost-effectiveness methodology (Incremental Cost-Effectiveness Ratios), and 2) an econometric model to assess multiple outcome measures while controlling for multiple subject and treatment-related factors that are known to affect both outcomes and costs.
Study Outcomes
Total cost for the initial hospitalization of treating stroke subjects randomized to either IV/IA or IV rt-PA treatment arms will be measured, as will differences in types of resource utilization over 12 months between the two arms of the trial. Quality-of-life data (EuroQol EQ-5D) will be collected over a 12-month period and quality adjusted life years (QALYs) will be used as a morbidity-adjusted measure of effectiveness. Subgroup analyses will include dichotomized NIH Stroke Scale (<20, ≥20), country, time between onset and randomization, and IA devices.
Keywords: Economic Evaluation, Quality of Life, Cost-effectiveness, Resource Utilization
INTRODUCTION
Stroke is a common and costly condition where an effective early treatment may be expected to affect patients’ future quality of life, the cost of acute medical treatment, and the cost of rehabilitation and any supportive care needed for their remaining lifetime. The mean lifetime direct cost of ischemic stroke in the United States was estimated at $140,048 in 1996 (1), with inpatient hospital costs for an acute stroke event accounting for 70 percent of the first year post stroke costs, and the first 30 days post-stroke costing nearly $20,346 for severe ischemic strokes (2). Room charges (50 percent), medical management (21 percent) and diagnostic costs (19 percent) were the major cost drivers (3). The estimated total direct and indirect cost of stroke in the US in 2004 was projected to be over $50 billion. To assist in informing discussions on early adoption of potential treatments, economic analyses should accompany investigations that seek to improve outcomes for stroke patients. Because the cost of stroke care may accrue to a number of different organizations and payers, we must examine more than one economic parameter over multiple time periods to identify the relevant economic benefits and/or burdens.
Fagan and colleagues evaluated the economic implications of rt-PA when given to patients with ischemic stroke (7). Their results revealed that rt-PA treated patients had significantly shorter length of hospital stays than placebo-treated patients (10.9 versus 12.4 days, p=0.02) and more rt-PA patients were discharged to home than to inpatient rehabilitation or nursing home (48% versus 36%; p=0.002). Their Markov model estimated an increase in hospitalization costs of $1.7 million and a decrease in rehabilitation costs of $1.4 million and nursing home costs of $4.8 million per 1,000 eligible treated patients. Thus, treating acute ischemic stroke patients with rt-PA within 3 hours of symptom onset may be expected to promote both clinical benefits to the patient, and potential net cost savings to the health care system. A pooled analysis from the ATLANTIS, ECASS, and NINDS rt-PA Study Group Investigators (2004) further reveals that giving patients rt-PA in a timely manner significantly increases the benefit to the patient, especially if started within 90 minutes (8) (for additional reviews of cost-effectiveness analyses related to stroke, see Holloway et al, 1999 and Evers et al 2000) (4–6.
The general aim of economic evaluations is to provide information on the incremental cost effectiveness of a new treatment as compared to current practice. This single measure, the incremental cost effectiveness ratio (ICER), or “value for money” for a new therapy is generally considered to be “good value” for the US health system if it is below $50,000 per quality-adjusted year of life gained. In addition to an ICER, the economic study proposed here will be able to project what changes in the level and allocation of expenditures to expect should the IV/IA approach to recanalization be adopted. It is important for third party payers to understand if part or all of a potential increase in their initial expenditures will be partially or fully offset by a decrease in future expenditures (i.e., repeat hospitalizations, long-term care services).
Primary Economic Hypothesis
IV/IA rt-PA therapy is cost-effective at 3 months compared to IV rt-PA, and provides cost-savings or is cost-neutral by 12 months.
Economic Outcome Measures
Total cost for the initial hospitalization of treating stroke subjects randomized to either IV/IA or IV rt-PA treatment arms;
Differences in types of resource utilization over 12 months between the two arms of the trial, including hospitalizations, procedures, length of stay, time to first hospitalization/procedure, diagnostic procedures, medications, rehabilitation, and outpatient visits, personal patient costs;
Cost-effectiveness of the two treatment arms over 12 months using 1) standard cost-effectiveness methodology, and 2) an econometric model to assess multiple outcome measures while controlling for multiple subject and treatment-related factors that are known to affect both outcomes and costs; and
Key economic performance measures (cost and cost-effectiveness), broken-out by health care system (United States and Canada), over 12 months of the treatment arms.
METHODS
Economic Analysis
The IMS III economic analysis will determine the cost-effectiveness of the IV/IA approach to recanalization compared to subjects treated with standard IV rt-PA in the setting of optimal medical therapy. Comprehensive information will be gathered on resource utilization and subjects’ valuation of their quality of life over 12 months for all subjects in IMS III. This will allow us to estimate expected differences in hospital costs, cumulative health care costs over the first year, and differences in subjects’ value of their quality of life over 12 months for all subjects. From these measures we will calculate ICERs and cost increases/cost offsets due to the differences in therapy.
Cost will be evaluated over 12 months from a societal viewpoint as best as possible. While it will be attempted to include all health care costs, costs of the initial stroke can be expected to dominate over a 12-month period. The reason for including all costs is that it is sometimes difficult to distinguish stroke-related from non-stroke events. However, given the randomized nature of this study, once past the initial hospitalization, the long term impacts of the types of non-stroke episodes of care should occur evenly between the groups. In addition the effectiveness of each recanalization strategy through clinical endpoints and quality of life measures will be examined. Quality adjusted life years (QALYs) will be used as a morbidity-adjusted measure of effectiveness. A cost-effectiveness analysis will be used to compare the 2 recanalization strategies at 3, 6, 9 and 12 months in dollars per QALY gained. Standard cost-effectiveness methodology will be used as recommended by the task force of experts organized by the U.S. Public Health Service (PHS) (9). For the 3-month cost-effectiveness analysis (primary economic hypothesis), the mean costs for each arm over the initial 3 months of the study will be used in the numerator and the mean QALYs of each treatment arm will be used in the denominator. The median costs and QALYs will also be employed as an alternative estimate of the cost-effectiveness ratio. These analyses will provide the best estimates of the cost-effectiveness of IV/IA approach relative to IV rt-PA over a 3-month period. Cost-effectiveness will also be measured at months 6, 9, and 12. The results of the economic analyses may be used by health care decision makers to compare the cost-effectiveness of IV/IA and standard IV rt-PA therapy to other health care interventions.
As a secondary analysis, an econometric model (10, 11) to assess the cost-effectiveness within IMS III at months 3, 6, 9, and 12 will be developed. This model will permit assessment of the cost-effectiveness using multiple outcome measures and for subgroups in a multivariate manner not possible with standard cost-effectiveness analysis. The econometric model assumes multiple causes affect multiple indicators through an unobservable health effect. Because the true health improvement is not directly observed, multiple outcome indicators are used as proxies for health improvement. Interaction terms will be included in the analysis. They will capture the separate effect of, for example, IV rt-PA alone on subgroups of subjects identified by specific exogenous variables. Subgroup analyses will include dichotomized NIH Stroke Scale (<20, ≥20), country, time between onset and randomization, and IA devices. If it turns out that the average results suggest that IV/IA is more effective, the interaction terms could identify certain types of patients who benefit more than average from IV rt-PA alone. If, for example, the interaction term of age is sufficiently positive, it would imply that older patients receive more health improvement with IV rt-PA alone than with IV/IA, other things being equal. The intent of this approach is to systematically define and analyze how the clinical and subject variables interrelate to obtain a better understanding of which factors are most powerful or the best predictors in explaining treatment costs and outcomes. By systematically analyzing the major factors which contribute to the cost and outcomes of alternative therapies and comparing these results on the dimensions of time, sample subsets, and variable subsets, not only will the overall understanding of the relationship between the cost and outcomes of these therapies be illuminated, but it will also systematically evaluate which combinations of variables are most appropriate and useful. Comparison of the overall effectiveness and the individual outcome group effectiveness will provide a richer understanding on how the different factors contribute to the overall outcome of the therapy.
Thus, the cost, cost/effectiveness, econometric analyses should be seen as complementary forms of analysis, with the econometric model looking internally within IMS III, and the cost-effectiveness and cost analyses providing our most generalizable measures.
IMS III Participant Resource Utilization
Initial Hospitalization
Hospital - United States
One problem common to all empirical studies evaluating the cost effectiveness of any medical procedure that involves hospital services is that the actual cost of the services is not available. The typical solution to this problem in the literature is to approximate the cost of a patient's hospitalization using charge information obtained from hospital billing datasets. Estimates of the total cost of a hospital episode are determined by adjusting charge information using hospital wide cost-to-charge ratios (9, 12, 13). In this approach, the total cost of each hospitalization is calculated as the product between total billed charges during the patient’s hospital episode found in the hospital billing database and the hospital's overall cost-to-charge ratio available from the Medicare cost report.
For this study, the cost of the initial hospitalization from U.S. sites will be estimated by collecting UB-04 hospital billing forms for each initial subject’s hospitalization. This is a uniform billing statement used by all third party carriers. All available codes will be used in this study. Costs will be derived from charges using the appropriate hospital based Medicare cost-to-charge ratio.
Hospital – Canada
For subjects treated in Canada, initial hospitalization costs will be estimated using multivariate regression analysis. The strategy is to use data from US hospitals and subjects to create a cost regression model that captures the financial effect of various patient characteristics and events on initial hospitalization costs. Since no province in Canada has an itemized billing system, for the purpose of uniform costing of resources for patients in the two countries, the estimated cost weights calculated for US subjects will be assigned to Canadian subjects with similar clinical characteristics and events. In other words, we will attempt to estimate the initial hospitalization cost for Canadian subjects as if they were treated in the US. Because teaching hospitals have historically had higher costs, a categorical variable will be included to capture this effect. The effects of this costing method will be assessed in a sensitivity analysis that uses published cost weights from the province of Ontario for all resources used by trial participants.
Physician
Estimating physician utilization during the initial hospitalization for therapeutic and diagnostic procedures will be straightforward. For example, if a subject receives a CT scan during the trial, the standard fee specified for this procedure will be applied. Physician utilization for subject consultations however will require some additional considerations. In addition to diagnostic and interventional procedures, physicians can claim for consultations and assessments given to subjects while they are in hospital. To assess the total number of physician consultations or assessments provided to subjects, a committee of study neurologists will be consulted to produce a protocol of the type and number of consultations / assessments normally claimed for ischemic stroke hospitalizations in the US and Canada. This protocol will be applied to each hospitalization and will vary according to select patient characteristics and events. Length of stay will be considered by country to account for the historically different management of stroke rehabilitation between the US and Canada (US patients are traditionally transferred from acute care to rehabilitation or nursing facility; Canadian patients typically receive rehabilitation in the initial hospital).
Follow-up
There is good support in the literature for patient recall for up to 4 months for health resource use (office visits, diagnostic examinations, treatments, hospital admissions etc) for studies of complex chronic conditions (14–16). However there is general agreement in the field of health economics that 12-month recall data are biased against recall of outpatient visits and treatments. Normally one would collect data only 4 months back at an annual visit and then annualize the results. However, for complex chronic conditions with episodic use of health resources (e.g. stroke) this is unlikely to be a valid approach. Therefore, actual event data will be collected on IMS III participants based on 3-month recall.
Resource utilization data in the form of answers to survey questions will be collected on separate CRFs by the study coordinator or nurse who interviews subjects accrued within this study. The Baseline Visit CRF and the Standard Follow-up CRF (3 month recall) are one and two pages respectively, and capture information on hospitalizations, physician/professional visits, employment information, and personal subject costs at baseline, 3-, 6-, 9-, and 12 months. All subjects/proxies will fill out each form. Select questions were constructed from forms used in the Agency for Healthcare Research and Quality and RAND co-sponsored HIV Cost and Services Utilization Study, and an NIH sponsored study comparing nelfinavir and ritonavir for HIV patients (17, 18). Minor modifications in wording have been made to allow for a condensed questionnaire.
IMS III Participant Valuation of Quality of Life
The EuroQol Group's EQ-5D is a brief, preference-based health status measure designed for use in evaluative studies and policy research, and has been recommended for use by the US Public Health Service’s Panel on Cost-Effectiveness in Health and Medicine (9, 19–21). For the IMS III Economic study, the EuroQol EQ-5D will be collected over a 12-month period, at baseline and each 3-month follow-up visit to assess participant’s valuation of their health state. For the baseline visit, the Resource Utilization Survey and the EuroQol EQ-5D will be collected from both the participant and proxy at Day 5 or discharge. This will allow for the participant to have maximum cognitive capability for filling out the forms. If the patient cannot complete the survey due to aphasia, cognitive change, or death, a proxy will be utilized. At all time periods, the EuroQol EQ-5D will be collected immediately following the administration of the Resource Utilization Surveys.
Telephone Administration of Surveys
Resource Utilization Surveys (including Concomitant Medications) and the EuroQol EQ-5D will be administered face-to-face at baseline and 3 month follow-up, and by telephone to the subject and a proxy (see below) at months 6, 9, and 12. The EuroQol EQ-5D has not been validated for phone administration. Thus, for a sample of IMS III subjects, the EuroQol EQ-5D will be validated for phone administration. Agreement of responses on the “face-to-face” and telephone interviews will be made by Cohen’s simple and weighted Kappa statistic for each of the items on the EuroQol EQ-5D. Logistic regression will be applied to identify factors (e.g., age, race, sex, order of administration of face-to-face versus telephone interview) affecting the concordance of the two responses.
Proxy
A proxy will be asked to complete 1) the Resource Utilization Surveys, 2) Concomitant Medication Form, and 3) the EuroQol EQ-5D assessments, and attempts will be made for the same person to serve as the subject’s proxy at baseline and every follow-up visit as much as possible. The proxy should be someone well acquainted with the subject’s health and healthcare utilization. The proxy will be asked to keep in mind to respond from the participant’s perspective. In the event that a subject is incapable or unable to self-administer the Resource Utilization Surveys, Concomitant Medication Form, or the EuroQol EQ-5D during the study, the proxy’s response will serve as an appropriate measure of the subject’s utilization of resources and valuation of their health state. There will be a proxy identification form at each administration of the EuroQol EQ-5D even if the same proxy has responded previously.
Valuation of Participant Survey Response Events
Primary Method for Resource Utilization - Archival Billing Databases
Patient-oriented resource utilization data will be linked to IMS III clinical data and a set of standard cost weights developed from archival billing data sources from patients with strokes. The resources reported by the study participants (including Medications) will be converted to dollar values (a proxy for opportunity cost) using standard cost weights as follows: 1) medications will be valued according to the Red Book Average Wholesale Price (AWP) value, 2) as a common unit of measure, health care services cost weights will be constructed using charges from the South Carolina (SC) Medicaid program, and hospital admissions records from both the national Hospital Cost and Utilization Project (HCUP) database and the SC Medicare database, and 3) the opportunity cost of missed days from work will be estimated based on subjects’ self-report of income range and missed days. The SC Medicaid database includes billing data for all Medicaid patients. This data set includes all billing data elements for provider visits, outpatient surgeries, hospital admissions, nursing home stays, and prescription drug coverage. The SC Medicare dataset includes claims from inpatient hospitals, physician/suppliers, outpatient-care facilities, skilled nursing facilities (SNFs), home health agencies, and hospice care. HCUP includes the largest collection of longitudinal hospital care data in the United States, with all-payer, discharge-level information beginning in 1988. HCUP databases include inpatient data from a national sample of over 1,000 hospitals, and data from ambulatory care encounters.
For IMS III, events and Episodes-of-Care will be constructed from the SC Medicaid, Medicare and national HCUP databases by ICD-9 or 10 codes and appropriate cost weights applied to the clinical and follow-up data for IMS III participants. These data will be used to provide regional comparisons of costs, as well as provide ranges appropriate for preliminary sensitivity analysis. All cost estimates will be made in constant dollars, with the adjustment factor being the medical component of the Consumer Price Index (CPI). No discounting of costs are necessary as the follow-up period is for 1 year only. Table 1 suggests a hypothetical patient and the sources of data for their economic assessment.
Table 1.
Economic Data Sources for Hypothetical IMS III Patient.
| Timeline | Utilization | Valuation |
|---|---|---|
| Initial Hospitalization | Hospital costs (US) |
|
| Physician costs (i.e., Neurologist; Interventionalist; Neuroradiologist for CT × 2, MR, echo; Rehab consult; Critical Care Physician ×2 days) |
|
|
| Follow-up (Resource Utilization Survey) | Skilled nursing home stay × 4 months |
|
| Re-hospitalization for CHF at 5 months |
|
|
| Outpatient physician visits × 2 at month 4 |
|
|
| Physical therapist visits for 4 months |
|
|
| 150 missed days from work |
|
|
| Equipment purchase-walker |
|
|
| Home modification-ramp |
|
|
| Concomitant medications |
|
Secondary Method for Resource Utilization-Literature for Sensitivity Analysis
To gain further insight into the variation in the resource utilization and costs of IMS III, several sensitivity analyses will be performed. Sensitivity analysis involves systematically altering our assumptions about the cost and effectiveness of the IV/IA approach to recanalization compared to standard IV rt-PA. Two types of sensitivity analyses will be performed. The first set of calculations will be one-way sensitivity analyses. In one-way sensitivity analysis, the assumption about each parameter is varied over a reasonable range of values with all other parameters fixed. Specifically, we will vary one-by-one our estimates of the different cost components of IMS III over the initial hospitalization, and then over the 12-months of follow-up. Robust estimates from the literature will be used to provide ranges (5, 7, 22–25). Specifically, Fagan et al. provide a comprehensive estimate of costs (from the literature) for most of the components of acute ischemic stroke (7). Perhaps most valuable to our analysis is the fact that the estimates are broken down by Rankin Score. These estimates, and others mentioned above, will provide credible values to test the sensitivity of our estimates.
Determining Utility Weights - Valuation of the EuroQol EQ-5D
Valuation of the health states as determine by the EuroQol EQ-5D will be based on the scientifically accepted indexed values determined by Dolan (20, 21). Dolan’s model predicted the mean values for 42 EuroQol EQ-5D health states in terms of level of severity associated with each dimension of the survey. Once IMS III participants complete the EQ-5D survey at baseline and each 3-month follow-up visit, responses will be entered into the database and the mean values will be applied to their responses for an appropriate valuation of their health state.
Economic Power
The number of subjects who will be enrolled in IMS III has been chosen based on the power required to detect a clinically significant difference between the treatment arms at 3 months. Given the importance of rt-PA for the treatment of stroke patients it would not be ethical to enlarge the study sample size in order to enhance the ability to detect economic differences. Furthermore, several economic parameters will be measured that together will be used to describe the expected economic differences between the two therapies. However, it is believed that most decision makers will agree that cost savings, or at least cost neutrality at 12 months is a very desirable economic outcome, given the study population. Thus, the study’s power to identify 10% savings in cost at 12 months for the IA/IV rt-PA treatment group, compared to the IV rt-PA treatment group will be examined. Data from a prior study on a sample of 21 S.C. Medicaid patients hospitalized with a primary diagnosis of ischemic stroke (ICD-9 code of 434) indicate that the average annual cost for patients with stroke was $60,157 in 2000. The study’s power to detect differences in annual costs with 600 subjects in the combined IV/IA arm and 300 in the IV arm (alpha=.05 and a 2 sided test) is estimated in table 2.
Table 2.
Power for Economic Analyses
| Mean Annual Cost | $60,157 |
| Standard Deviation* | $30,078 |
| Reduction in cost (10%) | $6,016 |
| Percent Power | 80% |
assuming that control variables explain 50% of the variance in total cost
Previous cost regressions of clinical trial data from patients with complex chronic conditions have shown that baseline control variables can explain at least 50% of the variance in the annual cost for these types of patients. Given that the clinical study has 600 subjects in the combined IV/IA arm and 300 in the IV arm, at least 80% power to identify a 10% cost difference is expected. Most economists agree that an annual cost reduction of 10% is a meaningful amount of saving for this type of patient.
The bootstrap technique (26, 27) will be used to assess the variation in the cost effectiveness ratios estimated from the study data to obtain a 90% Confidence Interval (CI) around the estimated costs. This technique is appropriate when costs are estimated from the recorded resource use data combined with the standard costs weights.
While there is sufficient power to determine a difference in costs, there are other factors which may affect the power of our econometric model (covariates, unknown regression coefficients, covariance structures). Power associated with the econometric model will be determined based on the statistical associations derived from the model.
In anticipation of missing resource utilization data, it will be assumed all missing data to be missing at random (28). Thus, missing resource utilization data will be imputed using the multiple imputation method. Sensitivity analysis will be performed for additional clarity.
SUMMARY
This prospective economic study is designed to “piggy-back” the IMS III clinical trial of IV/IA rt-PA vs. IV rt-PA while minimizing the burden of additional data collection on clinical investigators and enrolled subjects. In addition, the study will continue quarterly resource utilization and quality of life data collection for 9 months after the end of the 3 month clinical trial. Subjects will be asked by the study nurses to recall their use of major cost drivers for stroke (hospital admission, ER use, procedures, medications, etc.). They will also be asked to value their current health using the EuroQol EQ-5D instrument. These data will be combined with hospital billing data, and cost weights derived from several large medical billing data bases and utility weights from the literature to construct the measures required to assess cost effectiveness and compare expected economic impacts for the therapies over time.
Acknowledgments
Acknowledgements and Funding: The economic sub-study is funded by the National Institute of Neurological Diseases and Stroke (NINDS # U01 NS054630). The primary clinical study is funded by NINDS (NINDS # U01 NS052220). The trial receives rt-PA from Genentech, and devices from Concentric, EKOS, and Johnson & Johnson
REFERENCES
- 1.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. doi: 10.1161/01.str.27.9.1459. [DOI] [PubMed] [Google Scholar]
- 2.Leibson CL, Hu T, Brown RD, Hass SL, O'Fallon WM, Whisnant JP. Utilization of acute care services in the year before and after first stroke: A population-based study. Neurology. 1996;46:861–869. [PubMed] [Google Scholar]
- 3.Diringer MN, Edwards DF, Mattson DT, Akins PT, Sheedy CW, Hsu CY, Dromerick AW. Predictors of acute hospital costs for treatment of ischemic stroke in an academic center. Stroke. 1999;30:724–728. doi: 10.1161/01.str.30.4.724. [DOI] [PubMed] [Google Scholar]
- 4.Holloway RG, Witter DM, Jr, Lawton KB, Lipscomb J, Samsa G. Inpatient costs of specific cerebrovascular events at five academic medical centers. Neurology. 1996;46:854–860. [PubMed] [Google Scholar]
- 5.Evers SM, Ament AJ, Blaauw G. Economic evaluation in stroke research : a systematic review. Stroke. 2000;31:1046–1053. doi: 10.1161/01.str.31.5.1046. [DOI] [PubMed] [Google Scholar]
- 6.Demaerschalk BM, Yip TR. Economic Benefit of Increasing Utilization of Intravenous Tissue Plasminogen Activator for Acute Ischemic Stroke in the United States. Stroke. 2005 doi: 10.1161/01.STR.0000185699.37843.14. [DOI] [PubMed] [Google Scholar]
- 7.Fagan S, Morgenstern L, Petitta A, Ward R, Tilley B, Marler J, Levine S, Broderick J, Kwiatkowski T, Frankel M, Brott T, Walker M. NINDS rt-PA Stroke Study Group. Cost-effectiveness of tissue plasminogen activator for acute ischemic stroke. Neurology. 1998;50:883–890. doi: 10.1212/wnl.50.4.883. [DOI] [PubMed] [Google Scholar]
- 8.Hacke W, Donnan G, Fieschi C, Kaste M, von_Kummer R, Broderick JP, Brott T, Frankel M, Grotta JC, Haley EC, Kwiatkowski T, Levine SR, Lewandowski C, Lu M, Lyden P, Marler JR, Patel S, Tilley BC, Albers G, Bluhmki E, Wilhelm M, Hamilton S. Association of outcome with early stroke treatment: pooled analysis of ATLANTIS, ECASS, and NINDS rt-PA stroke trials. Lancet. 2004;363:768–774. doi: 10.1016/S0140-6736(04)15692-4. [DOI] [PubMed] [Google Scholar]
- 9.Gold MR, Siegel JE, Russell LB, Weinstein MC. Cost-effectiveness in health and medicine. New York: Oxford University Press; 1996. [Google Scholar]
- 10.Lee JS, Bailey MJ, Jeong J, Mauldin PD, Weintraub WS. A study on the cost-effectiveness of coronary revascularization: introducing the simultaneous mimic health status model. Health Econ. 1997;6:613–623. doi: 10.1002/(sici)1099-1050(199711)6:6<613::aid-hec305>3.0.co;2-6. [DOI] [PubMed] [Google Scholar]
- 11.Mauldin PD, Jeong J, Weintraub WS, Becker ER. A multiple outcome model to assess cost-effectiveness: results from the Emory Angioplasty vs. Surgery Trial (EAST) Biometrical Journal. 1999;41:3–23. [Google Scholar]
- 12.Mauldin PD, Weintraub WS, Becker ER. Predicting hospital costs for first-time coronary artery bypass grafting from pre-operative and post-operative variables. Am J Card. 1994;74:772–775. doi: 10.1016/0002-9149(94)90432-4. [DOI] [PubMed] [Google Scholar]
- 13.Ashby J. The accuracy of cost measure devied from Medicare cost report data. Hospital Cost Management and Accounting. 1992;3:1–8. [Google Scholar]
- 14.Epstein AM, et al. Costs of medical care and out-of-pocket expenditures for persons with AIDS in the Boston Health Study. Inquiry. 1995;32:211–221. [PubMed] [Google Scholar]
- 15.Tourangeau R, Rasinski KA. Evaluation of Data Collection Frequency and the Use of a Summary in the National Medical Utilization and Expenditure Survey. Medical Care Utilization and Expenditure Survey, Series A. Washington, DC: U.S. Government Printing Office; 1987. [Google Scholar]
- 16.Killeen TK, Brady KT, Gold PB, Tyson C, Simpson KN. Comparison of self-report versus agency records of service utilization in a community sample of individuals with alcohol use disorders. Drug and Alcohol Dependence. 2004;73:141–147. doi: 10.1016/j.drugalcdep.2003.09.006. [DOI] [PubMed] [Google Scholar]
- 17.Shapiro MF, Morton SC, McCaffrey DF, et al. Variations in the Care of HIV-Infected Adults in the United States Results From the HIV Cost and Services Utilization Study. JAMA. 1999;281(24):2305–2315. doi: 10.1001/jama.281.24.2305. [DOI] [PubMed] [Google Scholar]
- 18.Perez G, MacArthur RD, Walmsley S, et al. A randomized clinical trial comparing nelfinavir and ritonavir in patients with advanced HIV disease (CPCRA 042/CTN102) HIV Clinical Trials. 2004;5(1):7–18. doi: 10.1310/N11F-NK93-MUMR-A1VV. [DOI] [PubMed] [Google Scholar]
- 19.The EuroQol Group. EuroQol-a new facility for the measurement of health-related quality of life. Health Policy. 1990;16:199–208. doi: 10.1016/0168-8510(90)90421-9. [DOI] [PubMed] [Google Scholar]
- 20.Dolan P, et al. Center for Health Economics. York, England: The University of York; 1995. A social tariff for EuroQoL: results from a UK general population survey. Discussion paper 138. [Google Scholar]
- 21.Dolan P. Modeling valuations for EuroQol health states. Med Care. 1997;35:1095–1108. doi: 10.1097/00005650-199711000-00002. [DOI] [PubMed] [Google Scholar]
- 22.Sandercock P, Berge E, Dennis M, Forbes J, Hand P, Kwan J, Lewis S, Lindley R, Neilson A, Wardlaw J. Cost-effectiveness of thrombolysis with recombinant tissue plasminogen activator for acute ischemic stroke assessed by a model based on UK NHS costs. Stroke. 2004;35:1490–1497. doi: 10.1161/01.STR.0000126871.98801.6E. [DOI] [PubMed] [Google Scholar]
- 23.Porsdal V, Boysen G. Costs of health care and social services during the first year after ischemic stroke. Int. J. Tech. Assessment Health Care. 1999;15:573–584. [PubMed] [Google Scholar]
- 24.Lipscomb J, Ancukiewicz M, Parmigiani G, Hasselblad V, Samsa G, Matchar DB. Predicting the Cost of Illness: A Comparison of Alternative Models Applied to Stroke. Medical Decision Making. 1998;18:S39–S56. doi: 10.1177/0272989X98018002S07. [DOI] [PubMed] [Google Scholar]
- 25.Holloway RG, Witter DM, Lawton KB, Lipscomb J, Samsa G. Inpatient costs of specific cerebrovascular events at five academic medical centers. Neurology. 1996;46:854–860. [PubMed] [Google Scholar]
- 26.Efron B, Tibshirani RJ. An Introduction to Bootstrap. New York: Chapman and Hall; 1993. [Google Scholar]
- 27.Chaudhary MA, Stearns SC. Estimating confidence intervals for cost-effectiveness ratios: an example from a randomized trial. Stat Med. 1996;15:1447–1458. doi: 10.1002/(SICI)1097-0258(19960715)15:13<1447::AID-SIM267>3.0.CO;2-V. [DOI] [PubMed] [Google Scholar]
- 28.Little RA, Rubin DB. Statistical Analysis with Missing Data. New York: Wylie and Sons; 1987. [Google Scholar]