Randomized controlled trial of supervised patient self-testing of warfarin therapy using an internet-based expert system


Stephen Byrne, Pharmaceutical Care Research Group, School of Pharmacy, University College Cork, Ireland. Tel.: + 353 21490 1658; fax: +353 21490 1656.
E-mail: stephen.byrne@ucc.ie


Summary. Background: Increased frequency of prothrombin time testing, facilitated by patient self-testing (PST) of the International Normalized Ratio (INR) can improve the clinical outcomes of oral anticoagulation therapy (OAT). However, oversight of this type of management is often difficult and time-consuming for healthcare professionals. This study reports the first randomized controlled trial of an automated direct-to-patient expert system, enabling remote and effective management of patients on OAT. Methods: A prospective, randomized controlled cross-over study was performed to test the hypothesis that supervised PST using an internet-based, direct-to-patient expert system could provide improved anticoagulation control as compared with that provided by an anticoagulation management service (AMS). During the 6 months of supervised PST, patients measured their INR at home using a portable meter and entered this result, along with other information, onto the internet web page. Patients received instant feedback from the system as to what dose to take and when the next test was due. During the routine care arm, patients attended the AMS at least every 4–6 weeks and were dosed by the anticoagulation pharmacist or physician. The primary outcome variable was the difference in the time in therapeutic range (TTR) between both arms. Results: One hundred and sixty-two patients were enrolled (male 61.6%, mean age 58.7 years), and 132 patients (81.5%) completed both arms. TTR was significantly higher during PST management than during AMS management (median TTR 74% vs 58.6%; z=5.67, P < 0.001). Conclusions: The use of an internet-based, direct-to-patient expert system for the management of PST improves the control of OAT as compared with AMS management.


Chronic disease management has become an enormous challenge for most western countries [1]. Recently, it has been suggested that the cost-effective management of chronic disease will have to rely on the broad application of information technology and expert systems. Ideally, such systems would involve some type of home telehealth system, through which patients could enter data and receive feedback on how to manage their therapy [2,3]. The purpose of this randomized controlled trial (RCT) was to test this concept of home telehealth. Specifically, we assessed the efficacy of an internet-based, direct-to-patient expert system in the management of patients who require long-term oral anticoagulation therapy (OAT) with warfarin.

Warfarin is widely prescribed for the management of a variety of disorders that predispose patients to thromboembolism, including atrial fibrillation (AF), deep vein thrombosis (DVT), pulmonary embolism, and prosthetic heart valves, as well as other miscellaneous disorders. Warfarin has a narrow therapeutic window, and close monitoring of the intensity of anticoagulation [as measured by the International Normalized Ratio (INR)] is essential to minimize adverse thrombotic and hemorrhagic events. Dedicated hospital-based anticoagulation clinics have evolved to manage patients on OAT. This requires frequent, time-consuming visits to the hospital for laboratory measurement of the INR. With the current demand for OAT expected to increase up to six-fold by 2050 [4], enormous pressure will be put on already overstretched OAT clinics. Thus, the management of OAT is an excellent model in which to test the concept of automated telehealth.

Patient self-testing (PST) of the INR has been proposed as one way of improving the success of OAT. With this model of care, patients measure their INR themselves using a portable point-of-care (POC) device and report the result to a healthcare professional who is responsible for dosing advice. This is in contrast to patient self-management, whereby the patient not only measures their INR themselves but makes the necessary warfarin dose adjustments based on an algorithm. Increased frequency of testing facilitated by PST at home has been shown to have a positive impact on anticoagulation control. Consistent follow-up is essential for the safe and effective management of this model of care. Patients measure their INR at home, but they must still communicate test results to the healthcare provider, who must then manage the therapy and communicate dose adjustments and repeat testing instructions back to the patient. Less than 1% of US patients on OAT practice PST, and of those who do, more than 95% phone the anticoagulation clinic with their INR results for dosing advice [5]. Such ‘conventional’ management of PST may be logistically challenging and time-consuming, and verbal communication with patients may increase the risk of medical errors.

Technological innovations have lead to the development of ‘expert systems’; software systems that use expert knowledge to solve problems that normally require human intelligence. However, to date, expert systems have only been utilized to assist healthcare professionals in making clinical decisions [6,7]. A novel concept involves the use of a direct-to-patient expert system that employs rules-based algorithms to automatically inform patients how to adjust therapy and INR testing frequency on the basis of data inputted by the patient. This concept has recently been applied to the home management of OAT (CoagCare, ZyCare Inc., Chapel Hill, NC, USA). CoagCare combines a direct-to-patient expert system accessed via the internet with PST to provide a novel model of caregiver-supervised telehealth. A pilot investigation of the clinical effectiveness of this model of care as compared with routine care by an anticoagulation management service (AMS) was recently conducted at Duke University Medical Center. Sixty patients were managed with the internet-based direct-to-patient system for 6 months, and anticoagulation control was compared with that achieved by an AMS during the previous 6 months. The results showed a significant improvement in time in therapeutic range (TTR) with the use of this system (63% vs. 74.4%, P = 0.004), while requiring less caregiver time [8]. The aim of this study was to conduct an RCT of this direct-to-patient expert system in the management of OAT. These results also have implications for the automated management of other chronic diseases.

Materials and methods

Study design

A prospective RCT was carried out at the AMS of Cork University Hospital (CUH) to test the hypothesis that supervised PST using an internet-based expert system could provide superior anticoagulation control to that provided by routine anticoagulation management. This was a cross-over study; each patient was randomly assigned (computer-generated randomization sequencing implemented via sealed envelopes) to either 6 months of conventional AMS management or 6 months of supervised PST, at the end of which they were switched to the other management regimen. Approval was received from the Clinical Research Ethics Committee of the Cork Teaching Hospitals in December 2005, and the study was conducted according to the CONSORT guidelines for RCTs, 2001 [9].

Outcome measures

The primary outcome variable was the difference in TTR during the 6-month period of AMS management and the 6-month period of supervised PST. Secondary outcome measurements were the number of INR readings indicative of excessive undercoagulation or overanticoagulation (INR < 1.5 or INR > 5.0) and the number of serious adverse clinical events (hemorrhagic or thromboembolic) during both 6-month periods.

Sample size and statistical analyses

To detect a TTR difference of 11.4% between the two groups (in a pilot study that was performed, a mean difference of 11.4% was observed) with a power of 90% and a significance of P = 0.05, a sample of 160 patients was needed.

The control of anticoagulation was assessed by evaluating the TTR. The overall percentage TTR (Rosendaal method [10]) in each arm was first calculated. Then, for each individual patient, the difference in TTR between the two study periods was calculated using paired-sample t-tests or the non-parametric equivalent where appropriate. The numbers of extreme INRs were compared using independent-sample t-tests (or the non-parametric equivalents), with statistical significance defined as P < 0.05. All statistical analysis was performed using SPSS software (Version 15.0 for Windows).

Study population

We aimed to recruit 160 patients from the AMS of CUH. Patients who were on warfarin therapy for at least 2 months, who were expected to be on treatment for the duration of the 12-month study and who had internet access were eligible for enrolment. Patients were excluded if they were unable to use a home INR meter, if they missed more than two clinic appointments in the preceding 6 months, if they were taking anticoagulants other than warfarin, if they had experienced a hemorrhagic complication with a therapeutic or subtherapeutic INR in the preceding 6 months, or if they were unable to attend the hospital clinic at short notice.


All patients attending the AMS were invited (by letter) to participate in the study. Two information evenings were held in the hospital to explain the design and rationale of the study. Some patients were identified by the staff of the AMS as being potentially suitable candidates for the study, and these were subsequently contacted by the research pharmacist. The first 160 patients who provided informed consent and met the inclusion criteria were enrolled by the research pharmacist.

Education and training

Patients attended an education/training session. These were carried out in groups of one to three people by the research pharmacist, and lasted for approximately 90 minutes.

The education program was designed to cover the most important aspects of OAT, and included the theoretical aspects of anticoagulation, the mechanism of action of warfarin, the colors and strengths of the different warfarin tablets, blood tests, the INR and target ranges, potential signs and symptoms of overcoagulation and underanticoagulation, diet and lifestyle, missed doses, drug interactions, etc.

Patients were randomly allocated to the supervised PST or AMS group by the research pharmacist. Prior to commencing the supervised PST arm of the study, patients were shown how to use the CoaguChek XS point-of-care (POC) meter (Roche Diagnostics, Burgess Hill, UK) and how to access the internet-based website by the research pharmacist. They received written information on every aspect of measuring and reporting their INR and contact details of the study personnel. Patients performed at least two reproducible INR measurements under the pharmacist’s supervision, using the POC meter. These results were compared with a laboratory INR measurement from a venous blood sample. If there were wide differences (> 0.5 INR units) between the dual measurements, patients were excluded from the study [11].

Supervised PST management

Initially, patients tested their INR twice weekly using the POC meter. If the INR was therapeutic for two or three consecutive readings, the interval between tests was increased to a maximum of every 2 weeks. Using personal login details, patients accessed the patient interface of the system. They were prompted to enter relevant data (e.g. signs or symptoms of hemorrhagic or thrombotic adverse events, missed doses, dietary changes or changes to regular medication, and concurrent illnesses). They then entered their INR. If a patient reported a therapeutic or only slightly out-of-range INR and no other issues, they received instant feedback from the system regarding their dose and testing instructions. Patients with more extreme INR deviations were asked to take a bolus dose of warfarin (< 1.5) or hold their warfarin (> 5.0) and/or to log in later the same day for additional instructions. If a patient reported a symptom suggestive of a bleed or an embolus, for example chest pain, they were told to seek immediate medical advice.

The research pharmacist accessed the caregiver interface of the program at least once daily. Patient problems were prioritized for review using a predefined severity scale. The system also identified any patient who failed to test their INR, or log in to the program as scheduled. These patients were contacted by telephone the same day. All new dosage recommendations were reviewed and adjusted if necessary. All extreme INRs (< 1.5 or > 5.0) were discussed by the research pharmacist with a consultant hematologist.

Patients were reviewed in the anticoagulation clinic every two months to assess progress. During these visits, a venous blood sample was taken for laboratory analysis of the INR, which was compared against the POC-measured value [11].

AMS management

During the conventional management arm of the study, patients attended the AMS at least every 4–6 weeks. Venous sampling was performed by venepuncture and INR analysis was performed with a Sysmex CA-7000 or CA-1500 coagulometer (Siemens Healthcare Diagnostics, Deerfield, IL, USA). Dosage adjustments were performed by the clinic pharmacist or doctor, who was experienced in OAT management. The APEX (iSOFT, Banbury, UK) computerized decision support system was used to assist dosing and to predict the time to the next AMS visit.

Adverse event reporting

At the end of the AMS period of the study, patients were interviewed by telephone to determine whether they had experienced any serious adverse events during the preceding 6 months. Computerized clinic records were reviewed to identify any inpatient INR recordings that might indicate a hospital admission. During the supervised PST arm, patients could report any signs or symptoms of thrombosis or hemorrhage at each login. Medical notes were accessed to gather information about any serious adverse events reported, and the details were recorded on case report forms. Complications were evaluated and diagnosed by an independent consultant hematologist who was not involved in the study and who was unaware which arm of the study the patient was in at the time of diagnosis.

Patient satisfaction

A patient satisfaction survey was administered to all patients at the end of the 12-month study to evaluate their perceptions of supervised PST.


Study population

One hundred and sixty-two patients were enrolled over a 9-month period (July 2006 to April 2007). This represented 20.6% of the anticoagulation clinic population as a whole (Fig. 1).

Figure 1.

 Identification of eligible patients and patient enrolment. Percentages refer to the percentages of total patients enrolled in the anticoagulation management service (AMS). OAT, oral anticoagulation therapy.

Thirty patients (18.5%) were subsequently withdrawn for the following reasons: warfarin therapy discontinued (n = 8), PST ‘too stressful’ (n = 5), difficulty in accessing the internet (n = 4), decided to stay with usual care (n = 2), left the AMS (n = 5), withdrawn from the study because of poor correlation of the POC meter results with the laboratory-measured INR (n = 3), died (not anticoagulant-related) (n = 2), and difficulty in obtaining a fingerstick sample of blood (n = 1).

One hundred and thirty-two (81.5%) patients completed both arms of the study. Demographic data are summarized in Table 1.

Table 1.   Patient demographics
CharacteristicStudy population (n = 132)
  1. AF, atrial fibrillation; DVT, deep vein thrombosis; INR, International Normalized Ratio; OAT, oral anticoagulation therapy; PE, pulmonary embolism; SD, standard deviation.

 Mean ± SD (years)58.7 ± 14.3
 Male, no. (%)80 (61.6)
Indication for OAT, no. (%)
 AF43 (32.6)
 Prosthetic heart valve49 (37.1)
 DVT/PE29 (22.0)
 Other11 (8.3)
Target INR range, no. (%)
 2.0–3.067 (50.8)
 2.5–3.551 (38.6)
 3.0–4.014 (10.6)

Anticoagulation control

The overall median TTR during the AMS arm was 58.6% (interquartile range (IQR) 45.6–73.1%), which increased to 74% (IQR 64.6–81%) during the PST period (z = 5.67, P < 0.001) (Fig. 2). Paired-sample t-tests were performed for each individual patient to calculate the difference in TTR during both periods; (t131 = − 7.122, P < 0.001).

Figure 2.

 Percentage time in therapeutic range (TTR) in anticoagulation management service (AMS) vs. internet-supervised patient self-testing (PST).

Overall, 87.4% of readings were within 0.75 INR units of the target INR during PST management, as compared with 78.2% of INRs during AMS management.

Sixty patients were randomized to conventional management followed by PST. Median TTR increased by 16.6% (AMS 57.6% vs. PST 74.2%). Seventy-two patients were initially randomized to PST followed by AMS. The increase in TTR was slightly less (12.3%) (AMS 59.7% vs. PST 72%). However, the effect of the order of management on anticoagulation control was not significant (P = 0.412). Daily time required to manage 80 patients ranged from 10  to 45 minutes [23.2 ± 9.5 min (mean ± standard deviation) or approximately 11.6 hours per month] with the expert system.

The PST group measured their INR almost four times more frequently than the AMS group. The mean frequency of INR testing in the PST period was once every 4.6 days, as compared with once every 19.6 days during standard management (Table 2). Table 2 also shows the number of extreme INRs reported during both arms. There were significantly fewer subtherapeutic and supratherapeutic INRs during PST management. The dosage recommendations generated by the expert system were accepted by the caregiver in 97.2% of cases. No technical difficulties relating to the internet website were reported during the study. However, several patients had problems accessing the CoagCare website, because it was not compatible with the Apple Mackintosh web browser.

Table 2.   Number of International Normalized Ratio (INR) measurements, observation period, and extreme INR values
 AMSSupervised PSTP-value
  1. AMS, anticoagulation management service; PST, patient self-testing; SD, standard deviation.

Mean INR tests per patient ± SD (range)10.7 ± 5.2 (5–35)41.7 ± 6.6 (24–60)< 0.001
Mean observation period ± SD (days) (range)185.5 ± 12.2 (151–233)187.9 ± 9.1 (154–239)0.65
Mean frequency of testing (days) ± SD19.6 ± 6.64.6 ± 0.8< 0.001
% of extreme INRs61.7< 0.001
INR < 1.5 (% of total)45.933.3< 0.001
INR > 5.0 (% of total)

Adverse effects

There were two serious thrombotic complications during the supervised PST management period and two serious adverse events during the AMS arm of the study (Table 3).

Table 3.   Adverse events
EventINR on admission to hospitalTTR preceding event (%)Time point in arm
  1. AMS, anticoagulation management service; DVT, deep vein thrombosis; INR, International Normalized Ratio; PST, patient self-testing; TTR, time in therapeutic range.

 Gastrointestinal bleed2.568.5Day 51
 Transient ischemic attack2.278.5Day 122
Supervised PST
 DVT1.664.9Day 164
 DVT1.353.3Day 118

Patient satisfaction with care

One hundred and seventeen patients (88.63%) completed the satisfaction survey. Of the respondents, 99.1% found the POC monitor easy to use, and most patients felt confident with the results that they obtained. All patients (100%) agreed that the CoagCare questions and dosing instructions were clear and easy to understand and follow. Most patients felt that they were monitored more closely by the pharmacist during the supervised PST arm of the study (87.6%), and 94.7% felt that their INR was better controlled. The majority (98.3%) expressed a preference for supervised PST over attendance in the AMS.


This study is the first RCT of an internet-based, direct-to-patient expert system in the management of chronic disease. Management of thromboembolic diseases with OAT was chosen as a model for study, because the technology exists for monitoring therapeutic outcome (INR) at home, and algorithms for managing dosage based on clinical signs, symptoms and INR reading already exist. The results of this study clearly show that an automated, direct-to-patient expert system utilizing this technology can improve anticoagulation control. This result was achieved while reducing clinic workload and staff resources. Thus, direct-to-patient expert systems may contribute to the management of other chronic diseases in our aging population.

In our study, patients spent considerably more time in range while using the direct-to-patient expert system (74% vs. 58.6%; P < 0.001). Although this study was not powered to detect a significant difference in complication rates between the two groups, the improvement in TTR seen during the PST period (15.4%) may considerably reduce the incidence of adverse events. TTR is a recognized surrogate marker for hemorrhagic and thrombotic complications [12]. Data from the SPORTIF III and SPORTIF V studies demonstrated that the incidence of death, major bleeding and stroke may be halved by a 15% improvement in TTR from 60% to 75% [13]. Van Walraven et al. reported an increased risk of mortality [odds ratio (OR) 1.29; P < 0.001), ischemic stroke (OR 1.10; P = 0.006) and other thromboembolic events (OR 1.12; P < 0.001) associated with a 10% increase in time out of range [14]. Four adverse events occurred during the study: TTR was < 70% in three instances. The BCSH (2005) guidelines recommend that 60% of INRs should be within 0.5 INR units of the target and 80% should be within 0.75 INR units [15]. These goals were more than exceeded during the supervised PST period but were not achieved during usual care.

Serious under-anticoagulation (< 1.5) or over-anticoagulation (> 5.0) occurred more frequently during AMS management (6% vs. 1.7%). The strong association between supra-anticoagulation or sub-anticoagulation and complication rates has been reported in many studies [12,14]. The incidence of thromboembolic complications may be as high as 17.5% when the INR is < 1.5, as compared with 2.3% when the INR is between 2.0 and 3.0 [16]. Almost half of hemorrhagic complications occur when INRs are above the therapeutic range [12]. Van Walhaven et al. found that by maintenance of the INR between 2 and 3, one in every four hemorrhagic events and one in every 10 thrombotic events could be avoided [14]. Two patients suffered DVTs during the PST arm, and both had subtherapeutic INRs on admission to hospital (1.6 and 1.3).

Conventional management and oversight of PST can be time-consuming, but expert systems can optimize the efficiency of such programs. The direct-to-patient expert system not only improves anticoagulation control, but also makes OAT management more practical by incorporating a systematic and coordinated approach to data collection, record keeping, tracking, follow-up, reporting, and audit. Furthermore, this type of telehealth reduces the burden on the AMS clinic and staff resources. Only 12 hours per month of professional time was required to manage 80 patients remotely. Phlebotomists, laboratory staff and administrative staff who are usually involved with routine management of patients were not required. Finally, remote management of patients with a direct-to-patient expert system allowed patients greater freedom and convenience. Because the system can be accessed from any location that has internet access, patients and caregivers can travel without any interruption to OAT management. The patient satisfaction survey indicated clearly that patients very much preferred this approach to OAT management to standard clinic management.

Our study had a number of strengths. Each patient was randomly assigned to either 6 months of conventional clinic management or 6 months of supervised PST, at the end of which they were switched to the other management regimen. This cross-over design has the advantage of eliminating all possibility of covariate imbalance, because each patient received both treatments and served as their own control.

A disadvantage of a cross-over study is that the order of the intervention can affect the clinical outcomes. When patients undergo a period of home-testing but then return to routine care in the AMS, the benefits of PST may be ‘carried over’. This is because the benefits of PST may not result purely from increased frequency of testing, and improvements may be attributable to the effects of increased education, awareness, patient empowerment, and enhanced compliance [17,18]. This could influence subsequent anticoagulation control in conventional management [19]. Although we reported a slight trend to improved anticoagulation control in patients attending the AMS after PST, these results were not significant, and the order of randomization did not affect the improvements seen in TTR.

As with all studies of PST, our study may not be representative of the wider population on OAT, or patients with other chronic diseases. The study population was a self-selected group, the majority of whom were male and younger than the clinic population as a whole (58.7 years vs. 65.2 years). Many would agree that patients who volunteer to participate in self-care programs are more motivated and possibly more compliant than the average patient on OAT [17]. Our study was limited to patients on long-term OAT, and excluded those who were on warfarin for < 2 months. The risk of OAT-related adverse events are highest at the beginning of therapy [20], so newly anticoagulated patients have much to benefit from increased supervision. Further work is required to determine the generalizability and cost-effectiveness of this approach, and larger studies with longer follow-up periods are required to investigate the effect of supervised PST on thrombotic and hemorrhagic complications. Traditionally, patients who perform PST contact the caregiver via telephone, e-mail, or fax. In order to assess the direct effect of internet-based PST management, future studies should compare patients using this method of supervised PST with patients who self-test but who do not have access to the expert system.

Not all chronic conditions are as applicable to management with a direct-to-patient expert system as are thromboembolic conditions, where there are clear signs and symptoms as well as a discrete convenient blood test to generate input for a care algorithm. However, certain other conditions may lend themselves to this type of management. Diabetes is an obvious candidate, as is congestive heart failure, where patients’ symptoms (e.g. dyspnea) and biological markers (weight, blood pressure, heart rate, etc.) could be used to drive medication dosage algorithms that can be incorporated into an expert system that is accessible by the patient.


This novel approach to anticoagulation management combines the advantages of a specialized AMS, an expert system to generate changes in therapeutic regimen, and increased frequency of testing of the INR. Our results show that supervised PST using an internet-based direct-to-patient expert system can improve anticoagulation control as compared with traditional AMS management. TTR increased by 15%, and there was a greater than three-fold reduction in the incidence of extreme INRs. Our results suggest that such technology may have the potential to improve our ability to manage patients and help alleviate the increasing burden of chronic illness that confronts our healthcare system.


S. O’Shea and S. Byrne organized the study concept and design. F. Ryan was responsible for performing the study. She acquired, analyzed and interpreted the data. The manuscript was drafted by F. Ryan and revised by S. O’Shea and S. Byrne. F. Ryan had full access to all the data in the study and takes full responsibility for the accuracy and integrity of the data.


The authors would like to thank all the patients who participated in this study, and the staff of the anticoagulation clinic and phlebotomy department and laboratory personnel at Cork University Hospital. We thank Roche Diagnostics for providing the POC testing equipment, and Zycare Inc. for the use of their software program.

Disclosure of Conflict of Interests

Funding for this study was provided under the ‘Building Partnership’ scheme, a peer-reviewed competitive grant application jointly funded by the Health Research Board, Ireland (Grant no. PA 05 16), Roche Diagnostics (Burgess Hill, UK) and Zycare Inc. (Chapel Hill, NC, USA). The study sponsors did not have any involvement in the study design, the collection, analysis or interpretation of the data, the writing of the report, or the decision to submit the manuscript for publication.