An Economic Model of Adverse Events and Costs for Oral Anticoagulants Used for Atrial Fibrillation

By Leigh, J Paul White, Richard H

Key words: Anticoagulants – Cost-effectiveness – Economics – Warfarin ABSTRACT

Objective: To construct a semi-Markov model to compare health outcomes and medical costs associated with warfarin and a second anticoagulant over 1- and 5-year periods.

Design: We posited a hypothetical cohort of 10 000 identical 70- year-old patients with atrial fibrillation. We posited 20 scenarios for events that included four possibilities for ischemic strokes (mild, moderate, severe, death) and 16 possibilities for hemorrhages. The model allowed for four levels of International Normalized Ratio. Event rates were based on outcomes in clinical trials and observational studies. Costs were estimated from the perspective of the third-party payer.

Results: The greatest cost-generating events were virtually the same for the two drugs and included severe stroke ($1 758 548 for 1 year for both drugs), moderate stroke ($380 355 for 1 year for both drugs), and severe lower gastrointestinal (GI) hemorrhage ($193 804 for 1 year for warfarin and $193 474 for second drug). The least costly events for both drugs were mild intracranial or intracerebral hemorrhage ($7584 for warfarin and $4314 for second drug) and fatal upper GI hemorrhage ($16 781 and $16 752). Total costs for adverse events over 5 years were similar: $18 330 662 for warfarin and $17 102 847 for the second drug. Fatalities for 5 years were 123 for warfarin and 101 for the non-warfarin drug. Varying assumptions for nursing home care and numbers of ischemic strokes and hemorrhages generated the widest variation in costs.

Limitations: We did not account for out-of-pocket expenses, ‘pain and suffering’ costs, or variation across practice settings.

Conclusions: There was substantial variation in numbers and costs of adverse events across 20 scenarios, and for fatalities between the two drugs, but variation in costs between the two drugs was modest.

Introduction

More than two million Americans over the age of 60 have atrial fibrillation1. Roughly 2-15% develop acute ischemic stroke. Use of oral anticoagulants such as warfarin is recommended by current practice guidelines in patients with atrial fibrillation who have risk factors for ischemic stroke1. These factors (age over 65, hypertension, diabetes) are so common that the majority of patients with atrial fibrillation meet American College of Cardiology criteria for taking warfarin1.

Although warfarin is effective, it must be monitored over time to optimize its benefit while simultaneously minimizing the risk of hemorrhages. Medical care providers must vigilantly monitor patients, asking them to provide repeated blood sampling to determine the anticoagulant effect measured using the international normalized ratios (INR) of the prothrombin time. INR values above 3.0 are associated with an exponential increase in risk of bleeding, whereas INR values below 2.0 are associated with increased risk of ischemic stroke2.

Studies have addressed the economics of warfarin and reviews are available3,4. Accurate determination of the costs associated with warfarin is important, particularly when performing cost- effectiveness analyses and making comparisons with other therapies, such as aspirin or one of the new oral anticoagulants now under development5,6. Essentially all of these newer drugs do not require chronic monitoring and, as a result, would reduce the costs to health systems and make life easier for patients. Recent studies have addressed varied issues: comparing costs of warfarin to another drug7, estimating warfarin (only) costs in Finland8, and developing an economic model of stroke in atrial fibrillation9. Our study contributes to the literature. First, compared to existing studies, we created a more comprehensive model that included: 20 adverse events, varying times in the therapeutic INR range, and detailed per- unit costs for two anticoagulants. Second, our model, formulas, and calculations are available for use in future studies by other researchers.

Methods

Our semi-Markov model assumed we followed a hypothetical cohort of 10000 patients with chronic atrial fibrillation for 1 and 5 years (2003-2007). The cohort was comprised of 70-year-old patients who were identical at baseline except for their INR values. We assumed our cohort would be similar to the cohorts in Go et al.10, Hylek et al.n and Douketis et al.12, since figures from these three studies are heavily used in our model. Go et al.10 and Hylek et al.11 rely on the same observational study (not a clinical trial) in which the mean age was 72, percentage female was 43%, and risk factor percentages were: previous ischemic stroke 9.3%, hypertension 51.0%, and congestive heart failure 30.8%. Douketis et al.14 was a clinical trial for which: ages were < 65 years (21-22%, depending on drug), 65-75 (40%), and > 75 (38-39%), percent female was 3031%, and risk factor percentages were 21 % for previous ischemic stroke, and 77% for hypertension. Our age 70 assumption was similar to highly regarded cost studies: ages 65 and 75 for Gage et al.14 and exactly 70 years for O’Brien and Gage7. In addition, we followed Gage et al.14 and O’Brien and Gage7 by not distinguishing between men and women in our hypothetical cohort.

Our model also assumed varying amounts of time in the therapeutic INR range (INR 2.0-3.0), 20 adverse health events, nine categories for hospitalization (nine Diagnostic Related Groups – DRGs), seven physician procedures and encounters within the hospital (seven Current Procedural Terminology – CPTs), varying numbers of days spent in the hospital, varying types of post-hospital care including outpatient physical therapy visits and placement in a skilled nursing facility. Varying probabilities were assigned to the time within INR ranges and likelihood of adverse health events. Per-unit costs were assigned to the DRGs, CPTs, hospital visits, and post- hospital care.

An overview of the model appears in Table 1. The first category accounts for variation in steady-state INR values. We assumed varying amounts of time spent in each of these INRs based on values in the literature10,11. Within our base-case, the following percentages of time applied: 4.2% (INR= 1.0-1.4), 22.6% (INR = 1.5- 1.9), 62.5% (INR = 2.0-3.0), and 10.70% (INR > 3.0). These percentages sum to 100%.

Our model contained 20 possible adverse health events, listed in column 2. The categories were broadly divided into the two groups of ischemic stroke and hemorrhage. Four severity categories of ischemic stroke were posited: mild, moderate, severe, and death. For each category and INR range we assumed a corresponding likelihood of occurance as estimated in Go et al.12 and Hylek et al.10, Several diagnostic categories of hemorrhage were posited including intracranial subdural, intracranial intracerebral, upper gastrointestinal, lower gastrointestinal, retroperiotoneal, urine, epistaxis, and ‘other.’ Three to four severity categories were allowed for hemorrhage: mild, moderate, severe, and death. For each category of hemorrhage, severity, and INR range, we assumed a corresponding likelihood of occurance as estimated by Go et al.10, Hylek et al.11, and Douketis et al.12 We assumed no deaths or severe cases resulted from less serious bleeding involving retroperiotoneal, epistaxis, or urinary tract. Complete descriptions of 20 adverse health events are available in an unpublished appendix from the authors.

We assumed a third-party cost perspective that relied heavily on Medicare and Medicaid data. Column 3 lists nine possible Diagnostic Related Groups (DRGs) from Medicare that applied to the adverse events. These ranged from a craniotomy (DRG 1) to gastrointestinal (GI) hemorrhage with complications (DRG 174) to rehabilitation (DRG 426). For example: a craniotomy (DRG 1) applied to severe intracranial subdural hemorrhage; and DRG 14 applied to severe upper GI hemorrhage.

Seven Current Procedural Terminology (CPT) codes from Medicare were assumed to apply to the scenarios. The CPTs are listed in column 4. Most typical were ‘history and physical’ (CPT 99223) and ‘follow-up visit in hospital’ (CPT 99232). But some unique CPTs were applied such as partial colectomy (CPT 44140) for two lower GI hemorrhage scenarios.

Table 1. Model variables

Days in hospital appear in column 5. Days in hospital ranged from 1.0 for a mild ischemic stroke and for intracranial, intracerebral mild hemorrhage, to 11.3 for a severe lower GI hemorrhage ending in death.

Clinical judgment informed our decisions when selecting the DRGs, CPTs, and some of the expected length of stays in the hospital. Most estimates for days in hospital were drawn from estimates of mean number of days for specific DRGs15. Cost data on DRGs and CPTs were derived from national estimates (2003 US dollars)15,16.

Post-hospital care was posited based on clinical judgment and on Medicare allowances and requirements. We assumed, for example, eight outpatient visits by a physical therapist followed a mild stroke. We assumed Medicare paid $98.54 per visit in 2003. As another example, following a severe stroke, we assumed Medicare paid for 100 days of care at a skilled nursing facility for $276.60 per day in 2003. Estimates for cost of physical therapy were drawn from Medicare and Bureau of Labor Statistics (BLS) studies (Centers for Medicare and Medicaid Services and BLS websites: www.cms.hhs.gov/home/rsds.asp, accessed 12/1/2006 andwww.bls.gov/cpi/, accessed 10/9/2006). Data on skilled nursing facilities were drawn from Medicare studies (same website) and Bodenheimer and Grumbach17. We assumed family members or friends cared for the patient after these 100 days so that thirdparty payer costs were zero. Our method is outlined in Figure 1. Chances of adverse events were multiplied by percent of time in INR ranges to yield patient years for each event which, in turn, were then multiplied by unit costs for medical care to yield total costs for each adverse event. Our method can be most easily understood following a sample scenario, 1 a: mild ischemic stroke in Table 2, which applies to only 1 year (years 2, 3, 4 and 5 were adjusted to reflect attrition due to death and severe hemorrhage). The assumption regarding time in INR range 1.0-1.4 (4.2%) was drawn from Go et al.10 (page 2688). The assumption regarding chance of any ischemic stroke in INR range 1.0-1.4 (7.7%) was drawn from Hylek et al.11 (their Table 5). We assumed that if there was an ischemic stroke in patients whose INR was in the range 1.0-1.4, the chance that the ischemic stroke was mild was 3%11. The value 0.97 in line 5 in the top panel is the number of patients in this category who sustained an ischemic stroke. This 0.97 value is obtained by multiplying the number of cases, times the risk of having a low INR, times the risk of having an ischemic stroke in this INR range, times the chance that the ischemic stroke was mild: 10000 x 0.042 x 0.077 x 0.03.

Continuing further down Table 2, the next two panels apply to the number of ischemic strokes associated with an INR in the range 1.5- 1.9, and INR between 2.0 and 3.0. ‘Sum of sub-totals’ adds the number of patients with stroke: 0.97 + 1.29 + 0.69 or 2.95 patients. This number for patients, 2.95, is then multiplied by costs-per- patient with the DRG categories, CPT categories, and other categories and these products are added to post-hospital care. Costs for hemorrhages were similarly calculated.

Numerical assumptions behind our base-case appear in Table 3 which lists the eight most expensive scenarios and the unique variables within each of these eight scenarios. (A long table listing all 20 scenarios is available from the authors.). In Table 3, to conserve space, we did not list variables in subsequent scenarios if those variables were listed in previous scenarios or in Table 2. For example ‘time in INR range 1.0-1.4 was included in Table 2 so we did not list it in Table 3.

We also calculated costs associated with a second anticoagulant based upon ischemic stroke and bleed risks reported for ximelagatran10’12. Because the literature suggests strong similarities for ischemic stroke risks between warfarin and ximelagatran, we assumed ximelagatran would result in the same number and kinds of strokes associated with warfarin. We also assumed the second anticoagulant generated the same per-unit costs for each type of ischemic stroke event as warfarin. Full descriptions for both drugs, each scenario, with sources, are in an unpublished appendix and spreadsheets for all calculations and are available from the authors.

We constructed a 5-year semi-Markov model. Attrition was allowed for fatalities and severe bleeds. We reasoned that patients with severe bleeds would be taken off any anticoagulant. We retained patients with severe ischemic strokes within the model. We assumed that persons with severe ischemic strokes and severe bleeds would generate greater costs than the average for other patients retained over 5 years. Specifically, we assumed that subsequent annual costs for patients after their severe ischemic strokes and bleeds would equal 20% of the medical costs (excluding nursing facilities) generated by their severe events. These 20% cost amounts were roughly three times the average costs for all hypothetical patients.

We conducted sensitivity analyses based upon reasonable alternative values for the most consequential variables.

Results

Table 3 presents the costs of the eight most expensive adverse health events associated with chronic anti-coagulation therapy for 10000 hypothetical patients followed for 1 year. The greatest cost- generating events were the same for warfarin and a second anticoagulant. These highest-cost events for 1 year were: severe stroke ($1 758540), moderate stroke ($380355), and severe lower gastrointestinal (GI) hemorrhage ($ 193 804 for warfarin, and $193474 for second drug). The least costly events for both drugs were mild intracranial or intracerebral hemorrhage ($7584 for warfarin and $4314 for second drug), and fatal upper GI hemorrhage ($16 781 for warfarin and $16 752 for second drug).

Figure 1. Arithmetic scheme for model

Table 2. Example of a scenario

Table 3. Eight most expensive scenarios, variables, base-case values, first year only (abbreviated list, no variables repeated across scenarios. Complete list in Appendix available from authors)

Table 3. Eight most expensive scenarios, variables, base-case values, first year only (abbreviated list, no variables repeated across scenarios. Complete list in Appendix available from authors)

Table 4 presents similar data, including number of events and costs over 5 years. Data on ischemic strokes for patients appear in the top of the table. The greatest number and cost of events within our separate 20 scenarios occurred for moderate and severe strokes for warfarin and summed to 489.94 events and $11 154508 costs. For hemorrhage, the greatest number and costs occurred for the severe upper and lower gastrointestinal scenarios combined: 222.97 events and $2 546167 costs for warfarin, 223.00 events and $2 545 807 costs for the second coagulant.

For both drugs, the number of deaths and the costs of these fatalities were greater for hemorrhages than for the ischemic strokes. For warfarin, for example, the number and costs were 87.71 and $929905 for all hemorrhages and 35.02 and $213 868 for all ischemic strokes. Again, for both oral anticoagulants, although the number of severe ischemic strokes was similar to the number of severe hemorrhages, the costs were considerably higher for ischemic strokes. For warfarin, for example, there were 256.20 severe ischemic strokes that cost, in total, $9283306, whereas there were 275.9 combined severe bleeds that cost, in total, $3 645 711.

Differences between the costs associated with the two drugs appear in the final column (warfarin minus second drug). Differences range from -$15 755 to +$372144. For ischemic strokes, differences were small. Apart from the ischemic stroke results, the greatest similarities between the two drugs occurred for both upper and lower gastrointestinal events. The greatest differences between the two drugs occurred for both subdural and intracerebral hemorrhage as well as retroperiotoneal and especially ‘epistaxis and other’ bleeds. Using the data reported by Douketis et al.12, we assumed there were a little more than twice the number of warfarin cases that experienced more minor or Other bleeding’ when compared to the second drug. This ‘other’ category included intraocular, subcutaneous hematoma, muscle hematoma, hemoptysis, tongue hematoma, rectal and vaginal bleeding, among others.

Over 5 years, the model calculated 123 fatalities for warfarin and 101 fatalities for the second anticoagulant.

We conducted sensitivity analyses (Figure 2) for significant model assumptions including: time in nursing home, time (or probability) outside therapeutic range, chance of stroke, and chance of hemorrhage. These sensitivity analyses applied only to patients with warfarin. We found the sensitivity analyses applied to patients with the second anticoagulant were strikingly similar.

Our base-case assumed patients with a severe ischemic stroke would spend 100 days in a skilled nursing facility paid by Medicare. For sensitivity analyses, in our first new assumption, we allowed there to be 2 additional years and 9 months spent in a nursing home paid by Medicaid for an increase in cost. The second assumption allowed for the elimination of 100 days in a skilled nursing facility paid by Medicare and led to a decrease in costs. The Medicaid spending added an additional $78 675 per patient or $201 56535 in total spending over 5 years. The ratio of this additional cost to our base-case is roughly 1.10/1.0 indicating a 110% increase over the base-case. This first new assumption by far exceeded the costs of any of the other new assumptions entertained in our sensitivity analyses. The next greatest variation derived from increasing the numbers of ischemic strokes or hemorrhages by 50%. The dollar and percentage changes for the 50% increase and decrease for strokes and hemorrhages were equal because the model was linear, i.e. the model did not include, for example, polynomials or log transformations.

Table 4. Number, costs and pharmaceuticals across adverse events for 5 years

Discussion

Our study estimated the cost of oral anticoagulants in a hypothetical cohort of 10000 senior patients with atrial fibrillation. We posited 20 scenarios corresponding to adverse events involving ischemic strokes and hemorrhages. The model allowed for varying time within four INR ranges: 1.0-1.4, 1.5-1.9, 2.0-3.0, and >3.0. We estimated costs over 1 year ($3386072 for warfarin and $3 158 079 for the second drug) and 5 years ($18330662 and $17049391). In the first year, per-person estimated costs of adverse events for warfarin were $338.61 and for the non-warfarin drug were $315.80. For both the 1-year and 5-year results and for both drugs, the numbers and costs of fatalities were greater in association with hemorrhagic complications compared to strokes. However, severe ischemic strokes were considerably more costly than severe hemorrhages for both drugs. The difference was attributed largely to our assumption that severe strokes resulted in rehabilitation and skilled nursing home stays but that severe bleeds did not. Differences in 5-year costs for strokes were negative for two reasons: we assumed the same annual risk of strokes for the two drugs, but the second drug was returning more patients than warfarin to the model in each year. That is, over the 5 years, warfarin was responsible for more deaths and severe bleeds than the second anticoagulant. The cost of adverse gastrointestinal events for warfarin were quite similar to those for the non-warfarin drug. A sensitivity analysis demonstrated that variation in assumptions regarding nursing home stays generated great variation in costs, but that plausible variation in other assumptions generated only modest variation in costs. Figure 2. One-way sensitivity analysis on influential variables for warfarin, 5 years

We compared our model to two representative studies7,14. Gage et al.14 and O’Brien and Gage7 develop cost-effectiveness models to compare warfarin with aspirin and with ximelagatran, respectively. Like us, both Gage et al.14 and O’Brien and Gage7 posit hypothetical populations (>65 years old for Gage et al.14 and exactly 70 years old for O’Brien and Gage7) involved in therapy for many years. Also, like us, they rely on Medicare costs.

There are three broad advantages of our model over Gage et al. and O’Brien and Gage7, as well as over other published studies20- 25: the comprehensive modeling of the entire spectrum of adverse events that are included in our model, the corresponding variety of unit prices for DRGs and CPTs, and the inclusion of varying levels of INR. Gage et al.14 consider only four scenarios: major GI hemorrhages, hemorrhagic stroke, minor hemorrhages, and ischemic stroke. Whereas O’Brien and Gage7 do allow for the four categories for ischemic stroke we included, they allow for only four hemorrhages.

Our study has some limitations. First, whereas we considered only values and ranges for the number of adverse events that were consistent with the literature, the depth of this literature is not great. Some values and ranges for the incidence rates we used might be higher or lower than the values that will be reported in future research. second, we took the perspective of the thirdparty payers – Medicare and Medicaid. We ignored outof-pocket expenses, any additional provider costs not covered by Medicare and Medicaid payments, and any costs associated with ‘pain and suffering’ that might be accounted for in a societal perspective of costs26. Third, we omitted costs for 101 to 123 fatalities over 5 years. From the strict point of view of Medicare and Medicaid, a death reduces future costs. But from the societal point of view, a death is a great loss. Fourth, we restricted attention to only those patients with atrial fibrillation. Fifth, we used only DRGs and CPTs which we judged most appropriate for each adverse event. Additional DRGs and CPTs that others might view as appropriate are in Appendix C, available from authors. Sixth, we relied on the clinical judgment of only one physician to estimate number of outpatient visits.

Variables for age, gender, and risk factors such as previous ischemic stroke or hypertension were not explicitly included in the model. These were implicitly captured, however, in the values on variables reflecting time in various INR ranges, chances of different severities of ischemic stroke and hemorrhages since these values correspond to cohorts from Go et al.13 and Douketis et al.14. Our age 70 assumption appears to be closer to clinical trial studies (age 69) than patients seen in clinical practice (age 75)1.

A final limitation is that we did not account for any variation across practice settings. Lafata et al.20,21 estimate a monitoring cost differential of $157 (1997 dollars) for ‘usual’ care with a primary care physician and $233 (1997 dollars) for care in an anticoagulation clinic. But when Lafata et al.20,21 combine the costs of monitoring with adverse events, the overall costs are very similar $420 for usual care and $406 clinic care – suggesting that any account of usual and clinic care would not have appreciatively altered our findings. On the other hand, the Lafata et al.20,21 studies were conducted prior to the 2006 Guidelines’ so that this similarity of costs between usual and clinic care may not hold today.

Conclusions

We found considerable variation in costs associated with 20 adverse events. For example, the numbers and costs of fatalities were greater for hemorrhages than ischemic strokes; but the number and costs for severe disabling non-fatal events were greater for ischemic strokes than for hemorrhages. Variation in assumptions for nursing home stays generated the greatest variation in costs. We also found that warfarin generated about 7% more costs and 22% more fatalities than the non-warfarin drug. Despite limitations, we believe the detail associated with our 20-scenario model will prove useful to future researchers attempting to compare the costs of warfarin with other potential pharmaceutical agents.

Acknowledgments

Declaration of interest: JPL and RHW received funds from AztraZeneca Pharmaceutical Inc. in support of this research. JPL received partial funding from the National Institute for Occupational Safety and Health (R01 OH008248-01). Barbara Wickson provided word processing and computer programming assistance. Judith Curtis and Juli Neal provided information on DRG and CPT codes.

References

1. European Heart Rhythm Association and the Heart Rhythm Society, Fusten V, Ryden LE, Cannon DS, et al. ACC/AHA/ ESC 2006 Guidelines for the management of patients with atrial fibrillation – Executive Summary: A Report of the American College of Cardiology/ American Heart Association Task Force on Practice Guidelines and the European Society of Cardiology Committee for Practice Guidelines. Circulation 2006; 114:700-52

2. Hylek EM, Skates SJ, Sheehan MA, et al. An analysis of the lowest effective intensity of prophylactic anticoagulation for patients with nonrheumatic atrial fibrillation. N Engl J Med 1996;335:540-6

3. Szucs TD, Bramkamp M. Pharmacoeconomics of anticoagulation therapy for stroke prevention in atrial fibrillation: a review. J Thromb Haemost 2006;4:1180-5

4. Bushnell CD, Matchar DB. Pharmacoeconomics of atrial fibrillation and stroke prevention. Am J Managed Care 2004; 10:866- 71

5. Weitz JI. Emerging anticoagulants for the treatment of venous thromboembolism. Thromb Haemost 2006;96:274-84

6. Bates SM, Weitz JI. The status of new anticoagulants. Br J Haematol 2006;134:3-19

7. O’Brien CL, Gage BF. Costs and effectiveness of ximiligatran for stroke prophylaxis in chronic atrial fibrillation. J Am Med Assoc 2005;293:699-706

8. Hillinen T, Martikainen JA, Soini EJO, et al. Direct costs of warfarin treatment among patients with atrial fibrillation in a Finnish health care setting. Curr Med Res Opin 2006;22:683-92

9. Caro JJ. An economic model of stroke in atrial fibrillation: the cost of suboptimal anticoagulation. Am J Managed Care 2004;10:S451-8

10. Go AS, Hylek EM, Chang Y, et al. Effect of intensity of oral anticoagulation on stroke severity and mortality in atrial fibrillation. N Engl J Med 2003;349:1019-26

11. Hylek EM, Go AS, Chang Y, et al. Anticoagulation therapy for stroke precaution in atrial fibrillation: how well do randomized trails translate into clinical practice? J Am Med Assoc 2003;290:2685-92

12. Douketis JD, Arneklev K, Goldhaber SZ, et al. Comparison of bleeding in patients with nonvalvular atrial fibrillation treated with ximelagatran or warfarin. Arch Intern Med 2006; 166: 853-9

13. Go AS, Hylek EM, Borowsky LH, et al. Warfarin use among ambulatory patients with nonvalvular atrial fibrillation: the anticoagulation and risk factors in atrial fibrillation study. Ann Intern Med 1999;131:927-34

14. Gage BF, Cardinalli AB, Albers GW, et al. Cost-effectiveness of warfarin and aspirin for prophylaxis of stroke in patients with nonvalvular atrial fibrillation. J Am Med Assoc 1995;274: 1839-45

15. Hart AC, Schmidt K. 2004 DRG Expert 20th Edition. Ingenix, St. Anthony Publishing, Salt Lake City, UT 2004

16. Wasserman Y. Physicians’ Fee Reference 2004. 21st Edition, Yale Wasserman Medical Publishers, Milwaukee, Wisconsin, 2004

17. Bodenheimer TS, Grumback K. Understanding Health Policy: A Clinical Approach, 3rd edn. New York, NY: Lange Medical Books/ McGraw-Hill, 2002

18. 2003 Drug Topics Red Book 10th edn. Montvale, NJ: Thompson PDR, 2003

19. Samsa GP, Matchar DB, Phillips DL, et al. Which approach to anticoagulation management is best?: Illustration of an interactive mathematical model to support informed decision making. J Thromb Thrombolysis 2002;14:103-11

20. Lafata JE, Martin SA, Kaatz S, et al. Anticoagulation clinics and patient self-testing for patients on chronic warfarin therapy: a cost-effectiveness analysis. J Thromb Thrombolysis 2000;Suppl 1:513- 19

21. Lafata JE, Martin SA, Kaatz S, et al. The cost-effectiveness of different management strategies for patients on chronic warfarin therapy. J Gen Intern Med 2000;15:31-7

22. Seto TB, Taira DA, Tsevat J, et al. Cost-effectiveness of transesophageal echocardiographic-guided cardioversion: a decision analytic model for patients admitted to the hospital with atrial fibrillation. J Am Coll Cardiol 1997;29:122-30

23. Gianetti J, Gensini G, De Caterina R. A cost-effectiveness analysis of aspirin versus oral anticoagulants after acute myocardial infarction in Italy – equivalence of costs as a possible case for oral anticoagulants. Thromosis Haemostasis 1998;80:887-93

24. Lightowlers S, McGuire A. Cost-effectiveness of anticoagulation in nonrheumatic atrial fibrillation in the primary prevention of ischemic stroke. Stroke 1998;29:1827-32

25. Marchetti M, Pistorio A, Barone M, et al. Low molecular weight heparin versus warfarin for secondary prophylaxis of venousthromboembolism: a cost-effectiveness analysis. Am J Med 2001;111:130-9 26. Russell LB, Gold MR, Siegel JE, et al. The role of cost-effectiveness analysis in health and medicine. J Am Med Assoc 1996;276:1172-7

CrossRef links are available in the online published version of this paper: http://www.cmrojournal.com

Paper CMRO-3936_7, Accepted for publication: 13 June 2007

Published Online: 24 July 2007

doi: 10.1185/030079907X210822

J. Paul Leigh(a) and Richard H. White(b)

a Center for Healthcare Policy and Research and Department of Public Health Sciences, University of California, Davis, CA, USA

b Division of General Internal Medicine, University of California, Davis Medical Center, CA, USA

Address for correspondence: J. Paul Leigh, PhD, Department of Public Health Sciences, UC Davis Medical School, TB 168, Davis, CA 95616-5270, USA. Tel.: +1 530 754 8605; Fax: + 1 530 752 3239; pleigh@ucdavis.edu

Copyright Librapharm Sep 2007

(c) 2007 Current Medical Research and Opinion. Provided by ProQuest Information and Learning. All rights Reserved.