PubMed Health. A service of the National Library of Medicine, National Institutes of Health.

Optimal Warfarin Management for Prevention of Thromboembolic Events in Patients with Atrial Fibrillation [Internet]. Ottawa (ON): Canadian Agency for Drugs and Technologies in Health; 2011 Nov. (CADTH Optimal Use Report, No. 1.2.)


6.1. Quantity of Research Available

The literature search identified 115 articles. One additional article was identified from another source. Of these articles, eight were reviewed in full text and four met the inclusion criteria. Among the included studies were one cost-utility study13 and three cost analyses (Appendix 4).1416

6.2. Review of Included Studies

A. Patient Self-Management

Study description

Regier et al.13,18,19 conducted a cost-utility analysis of patient self-managed versus physician-managed anticoagulation from a Canadian health payer perspective. The authors used a Markov model with five health states (no events, minor or major hemorrhagic event, major thromboembolic event, and death) to simulate the costs and health outcomes of patients receiving chronic warfarin treatment. The probability of moving from one state to another depended on the time that the patient’s INR was in the therapeutic range. Patients who experienced a major hemorrhage or thromboembolic event could be temporarily or permanently disabled. The time horizon was five years and the primary outcome was the incremental cost-effectiveness ratio (ICER).13

The time in the therapeutic range for self- versus physician-managed anticoagulation was taken from a single randomized controlled trial (RCT) conducted at the Vancouver General Hospital that included 140 patients with atrial fibrillation, mechanical heart valve, or venous thromboembolism.20 The data on the probability of a hemorrhagic or thromboembolic event based on time in the therapeutic range were provided by the Italian Study on Complications of Oral Anticoagulant Therapy (ISCOAT) cohort study of 2,745 patients.21 Utility values were taken from several published studies in patients who had experienced a stroke or major hemorrhage. The authors conducted deterministic and probabilistic sensitivity analyses to test the robustness of the model.13


Regier et al. reported that self-management of anticoagulation prevented 3.5 major thrombotic, 0.79 major hemorrhagic events, and 0.12 deaths per 100 patients compared with physician management, over a five-year time horizon (Table 2).13 Self-management was associated with an additional C$989, 0.07 quality-adjusted life-years (QALYs), and an ICER of $14,129 per QALY. Almost all estimates from the probabilistic sensitivity analysis were in the upper right-hand quadrant of the cost-effectiveness plane. There was a 95% probability that self-management was cost-effective if the willingness to pay was $23,800 per QALY. In the deterministic sensitivity analyses, the ICER values ranged from $11,428 to $19,514 when the number of physician visits, probability of disability, discount rate, and utility values were varied. The costs for self-management were high in the first year of therapy due to start-up costs of $1,567 per patient for training and support, and this was reflected in the one-year ICER of $236,667 per QALY.13

Table 2. Summary of Results from Regier.

Table 2

Summary of Results from Regier.


The Regier et al.13 study was limited by the robustness of the clinical outcome data. The authors based the model on a single RCT that reported surrogate outcomes (i.e., time in the therapeutic range). The correlation between time in range and hemorrhagic and thromboembolic events outcomes was extrapolated from a single cohort study. The authors failed to provide detailed descriptions of the patients enrolled or limitations of these studies, and no deterministic sensitivity analyses were conducted to explore the uncertainty in the clinical data. Reporting of data elements was incomplete for several items on the BMJ checklist (Appendix 3). The source for some cost data was not stated and the utilization of several resources was assumed, and not based on actual patient usage. The generalizability of the findings may be limited due to the strict inclusion criteria in the clinical study used for the model. The authors state that self-management is not suitable for all patients, as it relies on the patients’ ability to understand anticoagulation and requires adequate vision and manual dexterity. Thus it may not be possible to extrapolate the findings of this study to the larger population of anticoagulated patients.13

B. Hospital-based Anticoagulation Services

Study 1: Schulman et al.15

Study description

Schulman et al.15 conducted a prospective observational costing study comparing four different models of anticoagulation management (Appendix 5). The authors gathered medical, non-medical, patient, and productivity loss costs from 16 sites across Canada, including hospital-based physician- or pharmacist-managed clinics, community-based family physician-managed care (traditional model), and community-based pharmacist-managed care. Data were collected for three months at each site from consecutive eligible patients who were either new users of warfarin (one month or less therapy) or chronic users (three or more months). Adults with atrial fibrillation or previous venous thromboembolism were eligible to participate. Information was collected from each site on the type of setting, services provided, budget, overhead costs, patient volume, staffing, salaries, procedures for laboratory testing and managing warfarin dosing, and point-of-care testing used. Over the three-month study period, the staff recorded the time and complexity of each patient encounter, including communication with the laboratory and administrative (i.e., charting) duties. Patients were asked to provide background information on their indication for warfarin, drug coverage, complications, and concomitant medications. Using a diary, patients recorded the time, travel, and costs (including lost wages) related to anticoagulation. Caregiver’s costs and lost income were also collected. The unit costs for resources were obtained from government sources, mainly from Ontario (see Appendix 4). The total costs of anticoagulation management from the Ministry of Health perspective included medical consultations, laboratory tests, hospitalization (if applicable), medications, and overhead. The societal costs included the Ministry of Health costs plus patient costs (medication copayments, personal expenses, caregiver costs, home-care costs, cost of patient, and caregiver workdays lost). The average three-month costs for each care model were presented.15


A total of 18 sites were invited to participate in the study and 16 provided data between 2006 and 2008.15 The data from the one community-based pharmacist-managed clinic were incomplete and were therefore excluded from the results, leaving 15 sites reporting results from the three remaining models of care (Appendix 5). A total of 429 patients were included in the three-month study. The patients from hospital clinics were younger than those in community care (hospital, 63 to 66 years; community, 70 years) and used fewer chronic prescription medications. More patients treated in the community had atrial fibrillation (86%) compared with hospital-based physician-managed (59%) and pharmacist-managed (55%) clinics. Patients treated in hospital clinics had more prothrombin time (PT) tests drawn than in community care. No statistical testing was conducted on patient characteristics to test for differences between care models. During the study period, there were five warfarin-related complications, but the costs of these events were minimal and they were excluded from the totals.15

From the Ministry of Health perspective, the total three-month cost of care per patient was $108, $145, and $199 in the hospital physician, hospital pharmacist and the community physician care models, respectively.15 In the hospital-based models, PT tests and other health care professional consultations accounted for the highest proportion of costs (physician: 75%; pharmacist: 87%). Physician consultations accounted for 7% and 5% of costs in the hospital physician and pharmacist models. In the community physician model, physician consultations, other health care professional consultations, and PT tests accounted for 42%, 34%, and 20% of costs, respectively. The proportion of costs for warfarin ranged from 4% to 11% among models.15

When the societal perspective was taken, the total three-month cost of care per patient was $188, $198, and $244 in the hospital physician, hospital pharmacist, and community physician care models, respectively.15 In the hospital physician model, PT tests and lost wages by patients accounted for the highest proportion of costs. In the hospital pharmacist model, other health care professional consultations and PT tests had the highest costs, and in the community model, physician and other health care professional consultations were responsible for the highest portion of total costs.15

Sensitivity analyses were conducted based on data from other sources. The societal three-month costs of care ranged from $203 to $277 when the number of INR tests was increased to 5.2 tests per three months. If the dispensing fee was increased from the Ontario rate (lowest) to the Nova Scotia rate (highest, $10.13), the three-month costs ranged from $229 to $303, and if non-paid caregivers were paid, the total costs ranged from $309 to $503.


The study was limited by the three-month duration, which was insufficient to capture the resources and costs associated with warfarin-related adverse events. There were differences in the patient characteristics between treatment models, which may have had an impact on the total costs of care. The study excluded parking costs, which may be substantial if PT testing was frequent. The authors stated that parking was excluded because these costs were more likely to be related to the size of the municipality than to the anticoagulation service. The authors also state that travel costs may be under-represented, due to the high proportion (~65%) of patients who walked to the laboratory. The authors did not report how sites were selected for inclusion in the study, and whether those that participated were representative of the anticoagulation monitoring services available in Canada. The study excluded patients with more severe comorbidities or non-compliance, who may have higher costs of care. These exclusions may limit the generalizability of the findings.

Study 2: Lalonde et al.14

Study description

Two additional studies14,16 provided some cost data; however, the quality of these estimates may be considered limited. The pragmatic RCT by Lalonde et al.14 compared the quality of anticoagulation, adverse events, use of health care resources, and direct medical costs for a pharmacist-managed anticoagulation service (PMAS) or family physician-managed care (Appendix 6). Patients were eligible if they required six or more months of warfarin treatment. All patients were initially managed by the community hospital PMAS until their INR values were stable. They were then randomized to one of the two care models and followed for six months. Physician management of anticoagulation was not standardized.14

Costs were estimated from the health care payer perspective, using the resource and outcome data collected from the RCT.14 Data on health resources and complications requiring an emergency room (ER) visit or hospitalization were collected from a centralized, networked computer system and administrative databases in Quebec. A blinded adjudication committee reviewed hospital charts to determine the severity of bleeding or thromboembolic events. In the PMAS clinic, the authors reported that each INR test required 6.25 minutes of the pharmacist’s time and five minutes of the secretary’s time. Physicians in Quebec are reimbursed for patient visits but not for telephone follow-up; thus, only the services paid by the government medical insurance plan were included in the estimates. The authors stated that because the number of INR tests and the incidence of treatment complications were similar between groups, these costs were not considered.14


A total of 250 patients were randomized, including 122 women (49%). The mean age was 65 years and 60% of patients had atrial fibrillation.14 The authors reported that both care models provided similar quality of anticoagulation management. There were no statistically significant differences between groups on health-related quality of life measured using two general and one oral anticoagulation-specific questionnaire. Patients in the PMAS group avoided 1.6 family physician visits per year compared with those in the family-physician care group (Appendix 6). The rate of bleeding or thromboembolic events was similar between groups.14

The authors reported that PMAS would require an additional $124 per patient per year compared with family physician anticoagulation management, in patients with previously stabilized warfarin dosing.14 This estimate assumed each patient would require 30 INR tests, 188 minutes of the pharmacist’s time ($109), and 38 minutes of the secretary’s time ($43), and would avoid 1.6 physician visits per year (–$28).14


The study was limited by the six-month follow-up time, which was inadequate to capture differences in bleeding or thromboembolic events. This simple cost analysis used the health care payer perspective and therefore excluded the physician’s or staff’s time spent providing follow-up to patients over the phone. There are opportunity costs associated with these resources that are not captured using the payer perspective. The estimates for the PMAS staff’s time to follow up with patients were not referenced. The estimate of 6.25 minutes per INR did not take into consideration the complexity of the patient’s clinical condition. No sensitivity analyses were conducted to test whether the results were robust. Overhead costs were ignored, as was the small, non-statistically significant difference between groups on the number of INR tests per patient.

Study 3: Bungard et al.16

Study description

The before and after study published by Bungard et al.16 reported the quality of anticoagulation, adverse events, and hospitalization costs for 125 patients receiving chronic warfarin anticoagulation (Appendix 7). Data were analyzed for four months or more before and after referral to a PMAS in a tertiary hospital. Hospitalization and ER visits were collected from the health region database and classified as hemorrhagic-, thromboembolic-, or non-anticoagulation-related events. The resource intensity weight, an indicator of typical resources consumed during the hospitalization for a given admission diagnosis, was multiplied by the unit cost of one resource intensity weight (C$3,500) to determine the cost of each event.16


Of the patients enrolled, the indication for anticoagulation was atrial fibrillation (40%), mechanical heart valve replacement (24%), venous thromboembolism (19%), or another condition (17%).16 The patients had a mean age of 63 years and 42% were female. The mean duration of treatment was 10.7 months and 29.3 months in the before and after periods. The quality of anticoagulation was lower in the before period than in the after period, as measured by the time the INR was in the therapeutic range (49% prior, 67% after, P < 0.0001).16

The rate of ER visits or hospitalization for thromboembolic events was higher in the before period than in the after period (rate ratio 17.6 [95% CI 6.0 to 51.9]).16 The difference between before and after periods was not statistically significant for hemorrhagic events (rate ratio 1.6 [95% CI 0.7 to 3.7]). Total costs of ER visits or hospitalization were $18,050 lower for hemorrhagic events, and $104,100 lower for thromboembolic events during the pharmacist-managed care period compared with prior care.16


This cost analysis16 was limited to costs of hospitalization and ER visits for PMAS and other anticoagulation management. The authors did not describe the providers or the type of anticoagulation management the patients received prior to their referral to the PMAS, nor did they measure any other health care resources or costs for the before and after periods. The higher rate of thromboembolic events in the before period may be related to how patients were selected for inclusion in the study. Patients with a recent thromboembolic or hemorrhagic event may have been more likely to be referred to a specialized service than those who did not have an event. The authors state that the study patients were representative of the larger population of patients requiring anticoagulation, based on a comparison with 502 non-study patients. Their conclusion, however, was based on similar demographics and time in the therapeutic range, and did not assess the rate of hospitalization or ER visits for the two groups. The use of the before and after study design was limited by the omission of a concurrent control group to provide information on temporal trends.

Copyright © 2011 CADTH.
PubReader format: click here to try
Cover of Optimal Warfarin Management for Prevention of Thromboembolic Events in Patients with Atrial Fibrillation
Optimal Warfarin Management for Prevention of Thromboembolic Events in Patients with Atrial Fibrillation [Internet].
CADTH Optimal Use Report, No. 1.2.

PubMed Health Blog...

read all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...