The value of education and self-monitoring in the management of warfarin therapy in older patients with unstable control of anticoagulation


Dr Farhad Kamali, Wolfson Unit of Clinical Pharmacology, School of Clinical and Laboratory Sciences, University of Newcastle, Newcastle upon Tyne NE2 4HH, UK. E-mail:


Of 125 patients aged 65 years or over, with atrial fibrillation taking warfarin for at least 12 months, with a standard deviation (SD) of prothrombin time, expressed as the International Normalized Ratio (INR) >0·5 over the previous 6 months, 40 were randomized to continue with usual clinic care and 85 to receive education about warfarin. Of these, 44 were randomized to self-monitor their INR and 41 returned to clinic. Compared with the previous 6 months there was a significant increase in percentage time within the therapeutic range for the 6 months following education [61·1 vs. 70·4; mean difference 8·8; 95% confidence interval (CI): −0·2–17·8; P = 0·054] and following education and self-monitoring (57 vs. 71·1; mean difference 14·1; 95% CI: 6·7–21·5; P < 0·001), compared with those patients following usual clinic care (60·0 vs. 63·2; mean difference 3·2; 95% CI: −7·3–13·7). Using the same comparative periods, the INR SD fell by 0·24 (P < 0·0001) in the group allocated to education and self-monitoring, 0·26 (P < 0·0001) in the group receiving education alone and 0·16 (P = 0·003) in the control group. Inter-group differences were not statistically significant (intervention groups 0·26 ± 0·30 vs. control 0·16 ± 0·3, P = 0·10). Quality-of-life measurements and health beliefs about warfarin were unchanged (apart from emotional role limitation) with education or education and self-monitoring. Patient education regarding anticoagulation therapy could be a cost-effective initiative and is worthy of further study.

Despite the efficacy of dose-adjusted warfarin therapy in reducing stroke risk in atrial fibrillation (Atrial Fibrillation Investigators, 1994), this treatment is under-prescribed (Sudlow et al, 1998; White et al, 1999). One reason for the under use is concern about the drug's safety in older patients because of instability of anticoagulation control and monitoring problems. Risk of life threatening bleeding complications, including subdural haematoma and intracranial haemorrhage, is increased in those with unstable anticoagulation control, and should therefore be reduced by measures that improve stability of control (Fihn et al, 1993).

Much of the intra-individual variability in response to warfarin therapy remains unexplained. Concordance is one factor that directly affects anticoagulation control. Patients’ attitudes have been demonstrated to have a significant influence upon concordance with anticoagulation, with non-concordant patients having perceived less potential benefit and more barriers to treatment than concordant patients (Arnsten et al, 1997). As warfarin has a narrow therapeutic index, and the pharmacodynamic response to it is unpredictable, the recognition of factors that affect patients’ attitudes and concordance towards their therapy will be a prerequisite in reducing the adverse effects of warfarin therapy associated with unstable anticoagulation.

Patients in the UK have traditionally had their response to warfarin monitored at hospital-based anticoagulation clinics, and more recently in general practice. The introduction of small portable devices for near-patient testing, and the provision of monitoring strips by the National Health Service, has enabled patients to either self-monitor (i.e. measure own International Normalized Ratio (INR) but have dosing decisions made by anticoagulant clinic staff), or self-manage (i.e. self-measurement of INR and self-dosing of warfarin) their therapy. Studies have shown that patient self-monitoring or self-management of anticoagulation therapy improves anticoagulation control, probably as a result of more frequent monitoring and dose adjustment and improved patient concordance (Cromheecke et al, 2000). Treatment-related quality-of-life also improves in some patients as a direct consequence of their increased control over their treatment (Sawicki, 1999; Cromheecke et al, 2000). Better INR control ultimately results in fewer thromboembolic or bleeding events (Beyth et al, 2000; Heidinger et al, 2000). Patients participating in studies of self-monitoring of anticoagulation receive a thorough education about the therapy and its indications. The extent to which subsequent improvement in anticoagulation control can be attributed to enhanced patient concordance as a result of education or near-patient testing is not clear.

Studies of near-patient testing for anticoagulant control have, in the main, been non-randomized, and have tended to include younger patients, selected on the basis that they are likely to be highly motivated. The impact of self-monitoring of anticoagulation on the stability of anticoagulation control in the older patient with non-valvular atrial fibrillation has not been evaluated, but is particularly important as this patient population is at a greater risk of adverse effects associated with unstable control of anticoagulation (Fihn et al, 1993).

We therefore set out to assess the effect of an anticoagulation education programme with or without self-monitoring in patients with atrial fibrillation, aged ≥65 years, upon stability of anticoagulant control and treatment-related quality-of-life measures. The study had the approval of the Newcastle and North Tyneside Local Research Ethics Committee.


Sample size calculation

Based on our clinic database, for those patients with atrial fibrillation with a target INR of 2·5 (range 2–3), the median standard deviation (SD) of INR values over 6 months was 0·3. The 20% of patients with the most unstable control of anticoagulation had an INR SD ≥ 0·5. It was calculated that, with this study design, 105 patients (35 in each of the control, education only and education plus self-monitoring groups) were required to detect a difference of 0·32 units in the SD of the INR (significance level 0·05, power 0·8) between the groups.


The study design and patient numbers are shown in Fig 1. A total of 249 patients in the anticoagulation service were selected by computer and identified as eligible on the basis of the following criteria: indication of atrial fibrillation with target INR range of 2–3, taking warfarin for at least 12 months, INR SD ≥ 0·5 over the previous 6 months and aged 65 years or over. Fifty-five patients were excluded because of general frailty, poor hearing or eyesight, impairment of hand function caused by disabling arthritis, stroke or tremor, dementia, or living in institutional care which precluded the use of a CoaguChek System, or when the cause of instability of anticoagulation control was apparent (e.g. drug interactions) and would not be altered by access to the CoaguChek System. 40 (27 males) continued with the usual anticoagulation clinic care, as the control group. They were unaware of their participation in the study as it was deemed that their prior knowledge of this could affect their behaviour. For the intervention arms of the study, 154 patients were approached and the study explained to them individually. Sixty-nine patients declined to take part for reasons of fear of self-testing, needles or preference to continue to attend the anticoagulant clinic.

Figure 1.

Flow chart for patient recruitment.

The 85 participating patients gave written informed consent. They attended, in groups of two or three, a 2-h education session based on Roche educational material led by a doctor. This session perceived health status and quality-of-life using the 36-item United Kingdom Short Form Health Survey (UKSF-36) (Garratt et al, 1993) and European Quality of Life questionnaire (Euroqol) (The Euroqol Group, 1990) and health beliefs specific to warfarin, using an instrument designed to measure quality-of-life in anticoagulated patients (Lancaster et al, 1991), were assessed at interview by a research worker, who was not a member of the anticoagulant clinic staff. Patients were then told about atrial fibrillation and the clinical benefits and risks of anticoagulation in their condition. Information was given about the pharmacology of warfarin, and factors that affect the INR, particularly concordance, drug interactions and diet. The sessions were interactive with opportunities for discussion and questions. Written information covering the issues raised was provided.

The patients were then randomized according to a computer-generated program, using a random numbers table, into two treatment groups: anticoagulant clinic monitoring or weekly self-monitoring of the INR (Fig 1). All patients participating in the study, including the controls, attended the anticoagulation clinic for INR measurements at intervals determined by the stability of their INR and with dosage changes determined through a computerized dosage program (Poller et al, 1998). The median age of patients in the control group was 73 years, range 65–93 years, in the education and clinic care group it was 75 years, range 65–87 years, and in the education and weekly self-monitoring group it was 71 years, range 65–91 years.

The patients (n = 44) who were allocated to self-monitoring underwent training in self-measurement of capillary INR using the CoaguChek System (Coaguchek®; Roche Diagnostics, Mannheim, Germany). They then measured their INR weekly at home. The study co-ordinator spoke with each patient by telephone, and based on the patient's INR value, gave appropriate advice about warfarin daily dosage for the next 7 d.

The patients actively participating in the study also attended the anticoagulation clinic at 6, 12 and 24 weeks to take part in an interview to determine attitudes to anticoagulation and treatment-related quality-of-life (The Euroqol Group, 1990; Lancaster et al, 1991; Garratt et al, 1993), and to have their venous INR values measured as an external quality control measure. The manufacturer's internal quality control measures for Coaguchek were performed each time a new test kit was used or if an INR was unexpectedly high or low, to confirm that the device was functional. The number and type of bleeding episodes, classified according to Landefeld and Goldman (1989), and thromboembolic episodes including stroke and episodes of hospital admission for the two intervention groups were recorded. The reasons for any patient withdrawing from the study before its completion were recorded.


Venous INR.

These tests were performed on citrated samples, using either the MDX or MDA CoaguChek System (Biomerieux, Cambridge, UK) with the company's specific reagents (Simplastin HTF, International Sensitivity Index = 1·15). Three levels of commercial quality control material (normal, slightly abnormal and very abnormal) were applied three times daily to all machines.

Capillary INR.

The Capillary INR was performed on Thrombotrak equipment (Biomerieux) using Manchester reagents (Thrombosis Reference Centre, Manchester, UK) and controlled by the daily application of abnormal quality control material from the same company. Both types of test were controlled externally as part of the National External Quality Assurance Service (Sheffield, UK) and Central Quality Assurance Scheme (Birmingham, UK) profiles. This laboratory has full Clinical Pathology Accreditation [Clinical Pathology Accreditation (UK) Ltd, Sheffield, UK].

Data analysis

The percentage of time at which the INR was within the target range of 2–3 for each patient was calculated by the method of linear interpolation (Azar et al, 1994). The main focus of interest was whether there was a change in mean value of a variable (e.g. time in range, SD of INR) as a result of the interventions. For each variable, the change in score was calculated for each patient and the mean change tested against zero assuming approximate normality, which was found to be an acceptable assumption. When two groups were compared, this was performed using independent group t-tests. The assumptions of normality and equal variance were checked. To generate odds ratios for the seven questions assessing health benefits and barriers relating to anticoagulation, patients’ ratings were dichotomized by combining the response ‘a great deal’ with ‘some’ and the response ‘not at all’ with ‘hardly’ and ‘not sure’. The reason for combining the latter was based on the assumption that patients who do not know whether the intervention is affecting their quality-of-life in a specified way (by responding as ‘not sure’) are not experiencing any significant impact on their health (Arnsten et al, 1997). An exact binomial test was performed on responses that changed from week 0 to week 24. Analyses were carried out on data from those patients who completed the study.


Four patients dropped out of the self-monitoring group: one was unable to perform self-monitoring competently at home, and three discontinued warfarin therapy, one because of a haemarthrosis. Two patients in the education and clinic care group did not complete the study as their warfarin was discontinued. Demographic details of the study population are shown in Table I. Individuals who refused to participate did not differ in age or gender distribution from the studied group. WE did not record the educational level of non-participants.

Table I.  Demographic details of the study population.
 Weekly self-monitoring (n = 40; male:female 26:14)Education and clinic care (n = 39; male:female 19:20)P*
  1. Values within parenthesis are expressed as percentage. NS, not significant.

  2. *Based upon adjusted chi-square test for differences between groups.

Current smoker3 (7)2 (5)NS
Ex-smoker24 (60)22 (56) 
Non-smoker13 (32)15 (38) 
In paid employment3 (7)0 
Education after school-leaving age18 (45)9 (23)NS
Degree or equivalent professional qualification8 (20)4 (10)NS

There was a good correlation between capillary INR, as measured by Coaguchek, and venous INR (r = 0·75; P < 0·001) with any differences between absolute values being small and of no clinical significance.

The quality of anticoagulation control, measured as percentage of time within the target INR range, showed a significant improvement as a result of intervention particularly in the group with education and self-monitoring. For the group participating in education alone, the percentage time within the therapeutic range increased from a mean ± SD 61·1 ± 15·1 during the 6 months prior, to 70·4 ± 24·5 during the 6 months after the study began (mean difference: 8·8, 95% CI: −0·2–17·8, P = 0·054). For patients participating in education and self-monitoring, the percentage time within the therapeutic range increased from 57·0 ± 17·0 to 71·1 ± 14·5 (mean difference: 14·1, 95% CI: 6·7–21·5, P < 0·001). For the control (non-intervention) group there was no significant difference in the percentage time within the therapeutic range between the 6 months prior (60·0 ± 18·8) to the six months after the study began (63·2 ± 25·9) (mean difference: 3·2, 95% CI: −7·3–13·7, P > 0·5). The time within the INR range during the study periods in the three groups is shown in Figs 1 and 2.

Figure 2.

Time in range in the 6 months preceding the intervention (- - - -) and the following 6 months (——–) in the control group (A), those participating in education alone (B) and those participating in education and self-monitoring anticoagulation (C).

Comparison of the results between the 6 months prior to and the 6 months after the study commencement demonstrated a fall in the SD of the INR of 0·24 (P < 0·0001) in the group allocated to education and self-monitoring, 0·26 (P < 0·0001) in the group allocated to education alone and 0·16 (P = 0·003) in the control group (Table II). Although the intervention groups had a larger decline in SD than the control group, the differences between the groups were not statistically significant (intervention groups 0·25 ± 0·30 vs. control 0·16 ± 0·30, P = 0·12). The differences in percentage time within the therapeutic range and in the INR SD between the groups were not statistically significant because of the large amount of variability.

Table II.  Standard deviation (SD) of the INR in the control and the intervention groups.
 SD of the INR, 6 months before interventionSD of the INR, 6 months after interventionMean difference (95% CI for mean difference)P (testing mean = zero)
Education plus self-monitoring0·72 ± 0·200·48 ± 0·150·24 ± 0·22 (0·16, 0·32)<0·0001
Education only0·69 ± 0·170·42 ± 0·300·26 ± 0·38 (0·14, 0·39)<0·0001
Control0·70 ± 0·150·54 ± 0·310·16 ± 0·31 (0·06, 0·26)  0·003

Age was not a relevant factor within this cohort; there was no association between time within range or SD and age either before or after the intervention. Similarly, the changes in time within range and SD did not correlate with age (P > 0·01 for all comparisons).

Those performing self-monitoring made an average of 4 ± 2 dose changes during the study (median 4, range 0–9), compared with 1 ± 1 (median 1, range 0–5) in the education and clinic care group and 2 ± 2 (median 1, range 0–7) in the control group (P < 0·0001) in the 6 months after the study commencement.

Of the parameters recorded by the UKSF-36, there were no significant differences in scores for bodily pain, social function, mental health, emotional function or physical function between the two intervention groups at the start of the study and thereafter at weeks 6, 12 and 24 (Table III). Only one variable changed significantly from baseline to week 24 (emotional role limitation – in the education and self-monitoring group; difference 13·33, 95% CI 0·85, 25·81, P = 0·04). The Euroqol scores were similar between the two groups and did not change during the study (Table III).

Table III.  UK SF-36 and Euroqol scores in the group receiving education alone and the group receiving education and self-monitoring of anticoagulation at enrolment and at week 24.
 EducationEducation and self monitoring
BaselineWeek 24BaselineWeek 24
Physical functioning52 ± 2953 ± 2961 ± 3057 ± 29
Physical role limitation48 ± 3952 ± 4257 ± 4445 ± 44
Bodily pain69 ± 3065 ± 3170 ± 3165 ± 34
General health perceptions55 ± 2456 ± 2152 ± 2053 ± 23
Vitality48 ± 2452 ± 2155 ± 2253 ± 23
Social functioning70 ± 3272 ± 2878 ± 2671 ± 32
Emotional role limitation62 ± 4163 ± 4681 ± 3663 ± 42
Mental health76 ± 1976 ± 2080 ± 1678 ± 18
Euroqol five dimension score0·74 ± 0·270·70 ± 0·290·82 ± 0·020·75 ± 0·27
Euroqol percentage69666967

Perceived benefits and barriers to anticoagulation were not significantly different between the group receiving education and self-monitoring and the one receiving education alone (Table IV).

Table IV.  Perceived benefits and barriers to anticoagulation.
  Education (n = 39)Education and self-monitoring (n = 40)
Week 0 n (%)Week 24 n (%)
Odds ratio

Week 0 n (%)Week 24 n (%)Odds ratio (95% CI)
  1. Odds ratios are for two group comparisons (some/a great deal versus not at all, hardly, not sure).

  2. P calculated by exact binomial test on those that changed from week 0 to week 24. NS, not significant.

  3. *Frequencies too small to calculate odds ratio.

My health is better since I started taking warfarin than it was before17 (44)22 (56)0·6NS17 (43)20 (50)0·7NS
Taking warfarin restricts my lifestyle3 (8)5 (13)0·6NS1 (3)2 (5)0·5NS
I worry a lot about the side-effects of warfarin5 (13)8 (21)0·6NS5 (13)5 (13)1·0NS
Having frequent blood tests bothers me0 (0)4 (10)*NS1 (3)0 (0)*NS
I am less physically active since taking warfarin4 (10)6 (15)0·6NS2 (5)4 (10)0·5NS
Taking warfarin bothered me before in the beginning than it does now13 (33)13 (33)1·0NS8 (20)5 (13)1·8NS
Taking warfarin will prevent clots35 (90)33 (85)1·6NS36 (90)36 (90)1·0NS
Taking warfarin will make my health better in the future24 (62)26 (67)0·8NS26 (65)27 (68)0·9NS

One patient in the self-monitoring group developed a haemarthrosis, his INR being 1·7 at the time and two patients in each of the two intervention groups reported minor nose bleeds. No thromboembolic complications were reported. We did not monitor adverse events in the control group.

Of the 40 patients performing self-monitoring who completed the study, 30 reported no problems with the technique and would have wanted to continue with self-monitoring for its ease and convenience. Ten of the sample, however, reported some difficulty with the technique, finding it difficult to place the fingertip blood drop on the right position on the test strip and, in consequence, having to use multiple strips to achieve a single reading. One patient did not complete the study because of this difficulty.


It has been demonstrated that 50–60% of patients can be expected to remain in their target range if monitoring of INR occurs monthly, 77–85% if monitored weekly and up to 92% if monitored every 3 d (Oral Anticoagulation Monitoring Study Group, 2001). It has been postulated that a threshold may exist beyond which there is no further beneficial effect of increased testing and this has been suggested to be between 2 and 4 d (Samsa, 2000). It would be unrealistic to expect an anticoagulation clinic to monitor patients with such frequency, but home monitoring does allow for this.

Near-patient testing would be a suitable model of oral anticoagulation therapy management if it could be shown to improve therapeutic control, thereby decreasing complication risk, improving patients’ quality-of-life and increasing cost-effectiveness when compared with the current management strategies. Studies examining the quality of anticoagulation control have indicated that patients performing self-monitoring spent significantly more time within their therapeutic range compared with control groups (Sawicki, 1999; Viller et al, 1999; Cromheecke et al, 2000). It is generally concluded that this is probably a result of increased frequency of testing, leading to better dosage advice and improved concordance. The major criticisms of such studies, however, are that patients had not been randomly selected and not all data had been collected prospectively. Extrapolation of the results of such studies is also hampered by the diversity of anticoagulation management of the control groups.

There is little information about the outcomes of self-monitoring or self-management in older adults. Two previous studies in anticoagulated patients aged 65 years or over reported that the percentage time in therapeutic range was higher in the self-monitoring patients, at 56% (Beyth et al, 2000) and 69% (Heidinger et al, 2000), than in those with usual care. These values are rather low in comparison with studies of groups of younger patients (Oral Anticoagulation Monitoring Study Group, 2001).

Because the unstable patients were selected from the top 20% of the SD of the INR in our clinic population, it was anticipated that, for some patients, the SD had been overestimated by chance. Thus we expected an average reduction in the INR SD because of the regression to the mean effect. This may explain the decline in SD in the control group.

A major outcome of our randomized study in patients with unstable control of anticoagulation was that education alone produced a similar improvement in time within the therapeutic range as did education plus self-monitoring of anticoagulation. Evidence suggests that patients having a good contact with health care professionals and more knowledge about their disease and treatment adhere better to their therapy (George et al, 1983; Viller et al, 1999). Not all doctors and pharmacists provide adequate information about anticoagulation therapy, and where it is provided, it may be forgotten or poorly understood by patients (McMahon et al, 1987). Attempts are made in our monitoring service to inform patients about their warfarin therapy, both through written information and with verbal support through answering questions when asked, thereafter. Studies have shown that, in spite of having been given information, patients’ knowledge about warfarin is generally poor, and is less in older patients (Griffin & Griffin, 1996; Lambert & Wynne, 2003). The education programme implemented in our study was a more rigorous way of informing elderly patients about anticoagulation, and led to the improved stability of control, perhaps through better concordance. How the change was effected, how much it depended on education or the fact that the patients were aware that they were participating in a study, the acceptability of the programme in the clinical setting and the relative importance of the key factors of verbal and written information in this education are the subjects of further study.

Some published studies have indicated an improvement in some quality-of-life measures and patient satisfaction as a result of self-management of anticoagulation, but these have focused upon younger patients who have more concerns about warfarin therapy and the limitations it imposes upon their lives (Ansell et al, 1995; Kulinna et al, 1999; Sawicki, 1999; Cromheecke et al, 2000). Some report a patient preference for self-management through empowerment to organize their time more efficiently and increased independence (Kulinna et al, 1999). As with other multi-item tests, the presence of a marginally significant difference in emotional role limitation found by the UKSF-36 in our study groups is probably the result of chance. The difference between the intervention groups and the control group is also partially explained by the high baseline value in the self-monitoring group, with the 24-week values being similar. Differences in patient age does influence attitudes to anticoagulation monitoring; patients over 65 years are less likely to view attendance at a hospital-based clinic as inconvenient than are younger patients, with many describing their visits as enjoyable, providing an opportunity for socializing and receiving support (Siat Chin Leong Sit & Wynne, 2003).

Our failure to demonstrate a statistically significant difference in time within the therapeutic range between the groups is because of the large amount of variability within each group and also possibly due to the limited number of patients that were studied. We aimed at recruiting those patients whose control was the most unstable, being at increased risk of thromboembolic or bleeding complications, and therefore having most to gain in terms of risk reduction by improved control. However, half of those approached declined to participate. Reasons given were physical constraints, which patients judged precluded them from the use of near-patient testing, lack of confidence or being unwilling to experience the uncertainty or perceived risk of the study and not wanting to change their current, satisfactory situation. Concomitant medication or identification of factors contributing to instability rendered ineligible only a minority of patients with unstable control. Forty-three of 44 older patients of this self-selected group did perform self-monitoring competently, the majority wishing to continue, to avoid travelling to clinic or to allow more frequent monitoring. Based upon the free text views, some older people had initial difficulty with the technique of blood sampling but learnt with practice, and all found it convenient. They felt it would prove most suitable for older patients who do not wish, or have difficulty, travelling to an anticoagulant clinic, particularly those with erratic INR control that require frequent monitoring, and for frail people who have a carer who could perform the test at home. Although our experience suggests that older patients will not be lobbying for near-patient testing to take over from clinic-based monitoring in the UK, the case for increased education about anticoagulation being made available to this group should be supported. The significant improvements in anticoagulation control noted after just one education session suggest that this could be a cost-effective initiative worthy of further study.


We acknowledge the financial support of The BUPA Foundation for this study.