Evaluation of Differences in Percentage of International Normalized Ratios in Range Between Pharmacist-Led and Physician-Led Anticoagulation Management Services

Abstract

Background:

The safety and efficacy of warfarin depend on maintaining the international normalized ratio (INR) in an established range.

Objective:

The purpose was to determine whether a coordinated pharmacist-led approach improved percentage of INRs in therapeutic range in comparison to a physician-led anticoagulation management service (AMS).

Methods:

A retrospective chart review was conducted for patients at a multisite primary care organization. INR data for patients receiving warfarin management by a physician were collected from December 1, 2009 to May 31, 2010. These were compared to INR results from December 1, 2010 to May 31, 2011, during which patients received warfarin management from a pharmacist. The primary end points were percentage of INRs within a goal range of 2.0 to 3.0 and an expanded goal range of 1.8 to 3.2 for the physician-led group versus the pharmacist-led group.

Results:

The percentage of INR results within the goal range (2.0-3.0) was greater among patients in the pharmacist-led group (n = .130) than the physician-led group (n = 96; 57.5% vs 50.0%, respectively; P = .0004). The percentage of INR results <1.5 (7.3% vs 5.1%) and >3.5 (11.4% vs 7.1%) was also statistically significant in favor of the pharmacist-led AMS, with P values of .03 and .0004, respectively.

Conclusion:

A pharmacist-led AMS improved the percentage of INRs in range, with significantly less out-of-range results.

Keywords

ambulatory care anticoagulation pharmacist-led warfarin

Background and Significance

Warfarin is the most commonly used anticoagulant in the United States and its effectiveness in preventing and treating thromboembolism is well established.¹ With a narrow therapeutic window, safety and efficacy are dependent on maintaining the international normalized ratio (INR) in an established range.¹ Use of warfarin requires individualized dosing and close laboratory monitoring to achieve optimal outcomes while minimizing the risk of serious adverse bleeding events. Maintaining the INR within the narrow therapeutic range is complicated by various food and drug interactions of warfarin, which may cause large fluctuations in INR.

Numerous studies have compared anticoagulation management service (AMS) provided by a pharmacist with AMS provided by a physician.^2

–7 Studies have suggested that coordinated care with a systematic approach to anticoagulation results in an increase in the number of patients with greater time in the therapeutic range (TTR) and a decrease in the number of adverse events.^2

–7 A meta-analysis of 67 studies involving 50 208 patients followed for a total of 57 155 patient-years showed that the absolute difference in TTR between a pharmacist-led and a physician-led AMS was 8.3% (95% confidence interval, 4.4-12.1), favoring pharmacist-led AMS.⁸ Pharmacist-led AMS may also provide an economic benefit; for example, in one analysis, more than US$1600 were saved in direct health care costs per patient at a pharmacist-run anticoagulation clinic compared to AMS provided by physicians.²

In an effort to increase the TTR and decrease both cost and adverse events for the patient, a multisite primary care organization, located in Rhode Island, established a pharmacist-based virtual AMS. This large, privately owned patient centered medical home (PCMH) group practice offers a wide array of medical services including anticoagulation services, for their patients.

Initial diagnosis and prescribing of anticoagulation therapy, determination of INR goal, and duration of anticoagulation was made by the primary care provider (PCP). The patient was referred to the clinical pharmacist who designed the treatment plan based on patient-specific treatment goals set by the PCP and guided by patient-specific factors and the American College of Chest Physicians (ACCP) guidelines.⁹

During the initial face-to-face visit, the clinical pharmacist assessed the patient’s most recent prothrombin time/INR, understanding of anticoagulation therapy, adherence to prescribed drug therapy, adverse events, clinically significant changes in diet, changes in concomitant drug therapy, and presence of medication-related problems. Patient education regarding the safe and effective use of the medication was provided. After the initial visit, patients were followed up by telephone for dosing instructions and laboratory follow-up after routine PT/INR measurements.

Each afternoon, the clinical pharmacists blocked a few hours of their time to provide AMS to patients. The PT/INR test results for patients who went to the laboratory early in the day were available starting around 12:00 pm, via fax, phone, or electronically. An electronic warfarin flow sheet was assigned to each patient, which included the date of the INR test result, the assigned warfarin dosing schedule, the INR test result, any complications reported by the patient, date of the next scheduled INR lab, and initials of the pharmacist who contacted the patient. Upon receiving the laboratory results, the pharmacist documented the appropriate information in the flow sheet and calculated a new dosing schedule per the ACCP guidelines⁹ as well as based on the treatment algorithm included in the collaborative practice agreement between the pharmacists and the medical director (see Figure 1). The pharmacist then contacted the patient, provided an interpretation of the INR test results, and inquired about any changes to their general health, medications, or dietary changes that may have impacted anticoagulant therapy. The most commonly encountered reason for INR >5.0 was warfarin drug interaction with various antibiotics. Most common reasons for INR <1.8 included interruption of warfarin therapy due to dental procedure, sickness, hospitalization, hematuria, and patient self-adjustment of warfarin dosing. Dose readjustments were often made to the anticipated dose based on the information that the patient provided. The patient received detailed instructions regarding dose adjustments, if any, and was provided scheduling for date of next required INR testing, as consistent with the ACCP guidelines.⁹ Warfarin education was reinforced and the patient was counseled to follow a diet with consistent amounts of vitamin K. The provider was contacted if a patient reported signs or symptoms of thrombosis, serious bleeding episodes, significant adverse drug reactions, or in the case of a significantly subtherapeutic or supratherapeutic PT/INR (usually >5.0). All patient interactions, whether at the office or virtual, were documented in the electronic health record and were communicated with the PCP.

Research on the impact of a virtual AMS in a private group setting is limited as previous research has generally focused on point-of-care testing. The purpose of this study was to evaluate pharmacist-led AMS versus physician-led AMS to determine differences between TTRs at 1 practice location with a telephonic anticoagulation service. The hypothesis was that patients prescribed warfarin at this practice site would achieve superior INR control with a coordinated, pharmacist-led virtual AMS in comparison to AMS by a physician without the involvement of clinical pharmacists.

Methods

Data for this cohort study were obtained using paper records of PT/INR results and a secure electronic health record (EHR). A retrospective chart review was conducted for patients receiving warfarin at 1 office. INR data for all patients prescribed warfarin and followed by a physician were collected for a 6-month period, from December 1, 2009 to May 31, 2010. INR results from this time period were compared to those from December 1, 2010 to May 31, 2011, during which INR for all patients prescribed warfarin was monitored and the warfarin dose was adjusted by a pharmacist.

The cohort included patients at least 18 years old who were prescribed warfarin during the physician (MD)-led AMS and pharmacist-led AMS time periods, as specified in Figure 2. For analyses of INR results, only patients having an INR goal range of 2.0 to 3.0 were included. All patients were included in the analysis, regardless of whether warfarin was initiated prior to or during the study period. A waiver of informed consent was obtained from the institutional review board at the University of Rhode Island in accordance with the principles of the Declaration of Helsinki.

Statistical and Data Analysis

The primary end point was the percentage of INRs within the therapeutic range of 2.0 to 3.0 and the expanded goal range (target INR ± 0.2) for the physician-led group versus the pharmacist-led group. Expanded goal range reflects clinical practice, where providers may permit a ± 0.2 deviation from the target INR before making changes to the warfarin dose. Secondary end points included the percentage of INRs <1.8, >3.5, and ≥5.0 for noninitiators during the studied time period and the number of patients with INRs <1.8, >3.5, and ≥5.0 at least once and at least twice. INRs <1.8 and >5.0 have been used as cutoff in multiple previous studies,^2,3,5,7 as the risk of thromboembolism and the risk of bleeding have consistently shown to be elevated. INR >3.5 was selected because in clinical practice, a warfarin dose is generally not omitted until INR is >3.5, at which point the risk of bleeding is considered to be significant enough to warrant a change.

A subgroup crossover analysis was also conducted, which included patients who had indications for long-term warfarin therapy. These patients were required to be followed by both a physician during the 6-month duration from December 1, 2009 to May 31, 2010, and then followed by a pharmacist during the 6-month duration from December 1, 2010 to May 31, 2011. In addition, the mean percentage of INR results within the therapeutic range was stratified by age group, gender, and other patient characteristics to determine whether these factors affected overall time in range.

Statistical significance of categorical variables was determined using the chi-square test; for continuous variables, the statistical significance of mean differences was determined using the student’s t test for independent samples and analysis of variance with the Tukey test for pairwise comparisons. SAS statistical software, version 9.1, was used to perform these analyses.

Results

A total of 237 subjects were prescribed warfarin between the specified time periods. Of the 102 patients in the physician-led group from December 1, 2009 to May 31, 2010, 6 were excluded based on their target goal range. Thus, 96 patients were included (Figure 1). Of the 135 patients in the pharmacist-led group from December 1, 2010 to May 31, 2011, 5 were excluded based on their goal range. Baseline characteristics were similar between the 2 groups, with no statistically significant differences (Table 1). About 54.0% of patients in each group were males, with a mean age of 75 years. The most common indication for anticoagulation was atrial fibrillation, with a mean CHADS₂ score of 2.4 in both of the groups. No statistically significant differences in risk factors (Table 2), which could increase the risk of either thromboembolism or bleeding, existed between the 2 groups.

Table 1.

Baseline Characteristics of the Study Population.

Characteristics	Physician led, N = 102	Pharmacist led, N = 135	P value
Males, n (%)	55 (53.9)	73 (54.0)	.981
Age, mean ± SD	75.4 ± 12.5	74.5 ± 13.4	.62
Aspirin use, n (%)	27 (26.5)	27 (20.0)	.24
Clopidogrel use, n (%)	2 (2.0)	2 (1.5)	.78
Warfarin indication
AFib, n (%)	72 (70.6)	91 (67.4)	.60
DVT/PE, n (%)	35 (34.3)	48 (35.6)	.84
CVA, n (%)	10 (9.8)	10 (7.4)	.51
Bileaflet valve, n (%)	4 (3.9)	5 (3.0)	.93
Other mechanical valve, n (%)	2 (2.0)	2 (1.5)	.78
Other, n (%)	5 (4.9)	2 (1.5)	.12
Warfarin initiation < 1 month, n (%)	9 (8.8)	19 (14.1)	.31
CHADS₂ score, mean ± SD	2.4 ± 1.1	2.4 ± 1.2	.86
INR target, n (%)
2.0-3.0	96 (94.1)	130 (96.3)	.43
2.5-3.5	3 (2.9)	2 (1.5)	.44
Other^a	3 (2.9)	3 (2.2)	.73

Abbreviations: AFib, atrial fibrillation; CVA, cerebrovascular accident; DVT, deep vein thrombosis; INR, international normalized ratio; PE, pulmonary embolism; SD, standard deviation.

^a Other includes less stringent INR range, such as 1.5 to 2.0 or 1.8 to 2.5, as determined by the primary care physician based on the patient’s bleeding risks.

Table 2.

Risk Factors for Stroke.^a

Risk factors	Physician led, N = 102 (n, %)	Pharmacist led, N = 135 (n, %)	P value
Previous CVA	13 (12.7)	19 (14.1)	.77
Previous DVT/PE	36 (35.3)	48 (35.6)	.97
Protein C/S deficiency	1 (0.1)	1 (0.1)	.84
Factor V Leiden	3 (2.9)	6 (4.4)	.55
Current smokers	1 (0.1)	4 (3.0)	.29
Former smokers	19 (18.6)	25 (18.5)	.98
DM	24 (23.6)	33 (24.4)	.87
HTN	84 (82.4)	106 (78.5)	.46

Abbreviations: CVA, cerebrovascular accident; DM, diabetes mellitus; DVT, deep vein thrombosis; HTN, hypertension; PE, pulmonary embolism.

^a Adapted from Guyatt et al.¹

For the primary end point, a statistically significant difference existed between the 2 groups in the percentage of INRs within 2.0 to 3.0 range, 50.0% versus 57.5% (P = .0004) for the physician-led group versus the pharmacist-led group, respectively (Table 3). The number of INRs <1.5 and >3.5 was also statistically significant favoring the pharmacist-led group, with P values of .03 and .0004, respectively.

Table 3.

INR Results Within and Outside Therapeutic Range According to Physician-Led and Pharmacist-Led AMS.

# INR results	Physician led, N = 96 (n, %)	Pharmacist led, N = 130 (n, %)	P value
Total # of draws	933	1324	.12
In range 2.0-3.0	466 (50.0)	765 (57.5)	.0004
In expanded range 1.8-3.2	600 (64.3)	968 (72.8)	<.001
<1.5	68 (7.3)	68 (5.1)	.03
<1.8	169 (18.1)	202 (15.2)	.06
>3.5	106 (11.4)	94 (7.1)	.0004
>5.0	20 (2.1)	25 (1.9)	.66

Abbreviation: INR, international normalized ratio.

Similar statistically significant differences were seen between the 2 groups when only including patients who were not newly initiated on warfarin during the specified 6-month time period for the physician- and pharmacist-led groups (Table 4). The percentage of INRs within 2.0 to 3.0 range was 50.5% for the physician-led group and 58.1% for the pharmacist-led group (P = .0008). Statistically significant differences existed between the 2 groups for the percentage of INRs in the expanded range of 1.8 to 3.2 (P < .001), INRs <1.5 (P = .009), and >3.5(P = .001). The results from a comparison of crossover patients (Table 5) also displayed a statistically significant difference favoring the pharmacist-led group for the percentage of INRs in the 2.0 to 3.0 range (P = .002), the expanded range (P = .0001), and INRs >3.5 (P = .003). Additionally, we calculated the frequency of occurrence of particular INR test result values (ie, using the INR test result as the unit of analysis). The frequency of observed INR results as a percentage of all INR test results obtained was graphed for crossover patients, denoting results when led by a physician (Figure 3) versus when led by a pharmacist (Figure 4). This visual portrayal displays a greater concentration of INR results within the range of 2.0 to 3.0 when led by a pharmacist as compared to physician-led AMS. INR results for physician-led patients were more frequently scattered outside the 2.0 to 3.0 range.

Table 4.

Analysis Excluding New Warfarin Initiators: INR Results Within and Outside Therapeutic Range According to Physician-Led and Pharmacist-Led AMS.

# of INR results	Physician led, N = 88 (n, %)	Pharmacist led, N = 110 (n, %)	P value
Total # of draws	842	1112
In range 2.0-3.0	425 (50.5)	646 (58.1)	.0008
In expanded range 1.8-3.2	549 (65.2)	820 (73.7)	<.001
<1.5	64 (7.6)	53 (4.8)	.009
<1.8	147 (17.5)	159 (14.3)	.06
>3.5	97 (11.5)	80 (7.2)	.001
>5.0	19 (2.3)	22 (1.1)	.68

Abbreviations: AMS, anticoagulation management service; INR, international normalized ratio.

Table 5.

Analysis of Cross-over Patients: INR Results Within and Outside Therapeutic Range According to Physician-Led and Pharmacist-Led AMS.

# of INR results	Physician-led, N = 86 (n, %)	Pharmacist-led, N = 86 (n, %)	P value
Total # of draws	859	917
In range 2.0-3.0	435 (50.6)	531 (57.9)	.002
In expanded range 1.8-3.2	555 (64.6)	670 (73.1)	.0001
<1.8	156 (18.2)	139 (15.2)	.09
>3.5	93 (10.8)	63 (6.9)	.003
>5.0	17 (2.0)	14 (1.5)	.47

Abbreviations: AMS, anticoagulation management service; INR, international normalized ratio.

In clinical research, the patient is generally the focus of interest when comparing outcomes of various treatments. Data that include multiple observations on the same individual may inflate sample size and thus lead to false statistical significance.¹⁰ Even when unit of analysis was shifted from INR results to patient-level results (Table 6), a statistically significant difference existed in the number of patients with an INR >3.5 at least once (P = .01) and at least twice (P = .03) as well as an INR >5.0 twice (P = .005) between the physician-led and the pharmacist-led groups.

Table 6.

Number of Patients With Out-of-Range INR Results According to Physician-Led and Pharmacist-Led AMS.

# of patients with INR result	Physician-led, N = 96 (n, %)	Pharmacist-led, N = 130 (n, %)	P value
<1.8 once	61 (62.3)	81 (62.8)	.99
< 1.8 twice	47 (48.5)	48 (37.2)	.09
>3.5 once	34 (35.1)	26 (20.2)	.01
>3.5 twice	29 (29.9)	23 (17.7)	.03
>5.0 once	18 (18.6)	18 (13.9)	.34
>5.0 twice	3 (3.1)	4 (3.1)	.005

Abbreviations: AMS, anticoagulation management service; INR, international normalized ratio.

Tukey’s pair-wise comparisons revealed no statistically significant differences between the age categories and CHADS₂ scores (Table 7), while a statistically significant difference was found when comparing the percentage of INRs in range between patients with and without history of prior deep vein thrombosis or pulmonary embolism (53.1% vs 60.0%, respectively, P = .046).

Table 7.

Mean Percentage of INR Results Within Therapeutic Range Stratified by Age Group, Gender, and Other Patient Characteristics.

Category	Mean ± SD, N = 226	P value
Age	57.6	.45
<50, N = 8	63.8 ± 30.0
50-64, N = 35	60.8 ± 25.4
65-74, N = 47	60.0 ± 25.0
≥75, N = 136	55.6 ± 24.1
Males, N = 102	56.3 ± 23.8	.48
Females, N = 124	58.7 ± 25.3
Aspirin use, N = 54	62.8 ± 25.6	.07
No aspirin use, N = 172	56.0 ± 24.2
Clopidogrel use, N = 4	50.9 ± 11.3	.58
No clopidogrel use, N = 222	57.7 ± 24.8
AFib, N = 161	57.9 ± 25.6	.80
No AFib, N = 65	57.0 ± 22.1
Tx DVT/PE, N = 76	53.4 ± 24.2	.07
Not DVT/PE Tx, N = 150	59.7 ±24.6
Tx CVA, N = 20	55.7 ± 26.9	.72
Not CVA Tx, N = 206	57.8 ± 24.5
Tx Bileaflet valve, N = 9	71.1 ± 25.0	.09
Not Bileaflet valve Tx, N = 217	57.0 ±24.5
Initiation <1 month, N = 28	52.0 ± 23.6	.20
Initiation >1 month, N = 198	58.4 ±24.7
CHADS₂ score 2-6, N = 128	56.6 ±25.9	.48
Other CHADS₂ score, N = 98	58.9 ± 23.0
Prior CVA, N = 32	51.3 ± 26.1	.12
No prior CVA, N = 194	58.7 ± 24.3
Prior DVT/PE, N = 79	53.1 ± 24.2	.046
No prior DVT/PE, N = 147	60.0 ±24.6
Past smoker, N = 43	54.6 ± 21.9	.38
Nonsmoker, N = 183	58.3 ± 25.2
DM, N = 57	55.7 ± 23.0	.50
No DM, N = 169	58.3 ± 25.2
HTN, N = 183	58.2 ± 25.7	.42
No HTN, N = 43	54.9 ± 19.5

Abbreviations: AFib, atrial fibrillation; CVA, cerebrovascular accident; DM, diabetes mellitus; DVT, deep vein thrombosis; HTN, hypertension; PE, pulmonary embolism; SD, standard deviation; Tx, treatment.

Discussion

As consistent with the meta-analysis by van Walraven et al,⁸ the results of this study consistently demonstrated a statistically significant improvement in the percentage of INRs in range, 57.5% versus 50.0% for pharmacist-led versus physician-led AMS, respectively. The difference between the 2 groups for the expanded goal range of 1.8 - 3.2 was of even greater magnitude, 72.8% for pharmacist-led versus 64.3% for physician-led AMS. Increasing the frequency of INRs within range and decreasing the occurrence of INR results >3.5 decreases the risks of adverse events either due to subtherapeutic or due to supratherapeutic effects of warfarin.

The time elapsed between the 2 study populations allowed for a period of transition and INR stabilization in patients who were previously monitored by their primary care physicians. The changes in season during the year may affect INR results that need to be maintained in a narrow therapeutic range. Therefore, studying the time period from December through May in both populations helped to reduce bias from lifestyle modifications due to changes in season. Several different methods exist and have been validated to calculate TTR.¹¹ We used the percentage of INRs in range method to calculate TTR. The method by Rosendaal et al has been used most often in previous studies. However, this method has not shown to better reflect the adequacy of anticoagulation than the method used in this study. Both methods have advantages and limitations, but the percentage of INRs was determined to be a more accurate measure, given the small sample size of this study. It allowed all patients to be included in the study, even with 1 INR result, and limited the effects of outliers on the overall results.

Various reasons exist to explain the differences in the primary end point. A clinical pharmacist adjusted the doses by following an established evidence-based protocol. In contrast, the practice's physicians did not have a standardized protocol and, as a result, changes among patients in the physician-led group were influenced by prescriber experiences and biases. Since the pharmacist utilized the same algorithm for all patients with either a subtherapeutic or supratherapeutic INR, all warfarin doses were adjusted accordingly to quickly return INR back to the normal range.

Prior to a pharmacist-led service, primary care physicians adjusted warfarin doses solely based on the INR results and the medical assistant relayed the new dose to the patient, with no further inquiries made regarding the patient’s health status. With a pharmacist-led service, patients received a call from a medical provider who confirmed dose, inquired about changes in health status, missed doses, the addition of any new medications, or any other clinically important change. Pharmacists made warfarin dosing decisions using the INR results and further refined them based on new clinical information provided by the patient. Pharmacists generally blocked 2 to 3 hours each day to provide this service to the patients. It is our belief that the differences in INR results between the 2 groups would likely be smaller if the physicians were directly speaking to the patients and had availability of the same clinical information or standardized protocol as the pharmacists utilized.

Along with a shift to a pharmacist-led clinic, the documentation of INR results was switched from a paper-based chart to a warfarin flow sheet that was integrated into the EHR. The flow sheets, which improved access to medical information, together with the increased inquiry regarding patient-specific information helped to improve the coordination of care based on patient needs.

Several limitations to the study also exist. The study only assessed a 6-month time period with a small sample size. Maintaining the INR within the narrow therapeutic range is crucial to preventing thromboembolic and hemorrhagic events. The study did not measure differences in event rates between the physician-led and the pharmacist-led groups. For patients who had transitioned in and out of either a hospital or a nursing facility during the 6-month period, INR data and dose changes may not have been available. Thus, the impact of the transition on the results of the study could not be adequately determined. It is also important to note that a pharmacist-led model may contribute to the level of patient satisfaction with their care, as pharmacists were able to educate patients regarding their medications, not limited to warfarin. Although this satisfaction was not measured, future research could assess whether the benefits of service extend beyond just increasing TTR. Additional research may also determine event rates and utilize pharmacoeconomic data to evaluate whether pharmacists are able to decrease adverse events in a cost-effective manner.

The study was able to provide evidence regarding the efficacy of the model and the benefits of pharmacist-led anticoagulation services in a private group of physician practices. The results will be used to expand the role of pharmacists to provide anticoagulation services to other offices.

Conclusion

Prior studies have shown a benefit with a systematic and coordinated approach to anticoagulation services with pharmacist involvement. This benefit exists with a virtual model, even in the absence of regular face-to-face evaluation and point-of-care testing. The pharmacist-led AMS emphasizes patient education regarding use of warfarin, systematic INR testing, consistent monitoring, and tracking of laboratory results via electronic flow sheets in the patient’s EMR, use of evidence-based guidelines to determine dosage adjustment algorithms, follow-up, and appropriate patient communication to discuss necessary changes. The pharmacist-led clinic encourages collaboration with the primary care physicians, leading to improved continuity and integration of care.

Footnotes

Authors' Note

Dr Gupta was a PGY-1 Resident practicing at Coastal Medical and University of Rhode Island at the time this research was conducted.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Guyatt

Akl

Crowther

. Executive summary: Antithrombotic therapy and prevention of thrombosis, 9th ed: American college of chest physicians evidence-based clinical practice guidelines. Chest. 2012;141(2 suppl);7S–47S. DOI: 10.1378/chest.1412S3.

Bungard

Gardner

Archer

. Evaluation of a pharmacist-managed anticoagulation clinic: Improving patient care. Open Med. 2009;3(1):e16–e21.

Chiquette

Amato

Bussey

. Comparison of an anticoagulation clinic with usual medical care: anticoagulation control, patient outcomes, and health care costs. Arch Intern Med. 1998;158(15):1641–1647.

Rosendaal

Cannegieter

van der Meer

. A method to determine the optimal intensity of oral anticoagulant therapy. J Thromb Haemost. 1993;69(3):236–239.

Wilson

Wells

Kovacs

. Comparing the quality of oral anticoagulation management by anticoagulation clinics and by family physicians: a randomized controlled trial. CMAJ. 2003;169(4):293–298.

Witt

Sadler

Shanahan

. Effects of a centralized clinical pharmacy anticoagulation service on the outcomes of anticoagulation. Chest. 2005;127(5):1515–1522.

Young

Bishop

Twells

. Comparison of pharmacist managed anticoagulation with usual medical care in a family medicine clinic. BMC Family Practice. 2011;12:88.

van Walraven

Jennings

Oake

. Effect of study setting on anticoagulation control: a systematic review and metaregression. Chest. 2006;129(5):1155–1166.

Ansell

Hirsh

Hylek

. Pharmacology and management of the vitamin K antagonists: American college of chest physicians evidence based clinical practice guidelines (8th edition). Chest. 2008;113(6 suppl):160S–198S.

10.

Altman

Bland

. Statistics notes. Units of analysis. BMJ. 1997;314(7098):1874.

11.

Schmitt

Speckman

Ansell

. Quality assessment of anticoagulation dose management: Comparative evaluation of measures of time-in-therapeutic range. J Thromb Thrombolysis. 2003;15(3):213–216.