Who is responsible for the care of patients treated with warfarin therapy?

Warfarin is the predominant anticoagulant for protection against stroke in patients with atrial fibrillation and/or mechanical valve replacement. It is also used for management of venous thromboembolism. In Australia, warfarin use is rising at a rate of 9% per annum,1 with about 30% of patients commencing therapy being aged over 70 years.2 Despite its efficacy, warfarin remains a major contributor to potentially preventable medication-related adverse events.3,4 Annual rates of major bleeding range from 1% to 7%, with minor bleeding in 16% of patients with atrial fibrillation.5,6 A major determinant of bleeding is over-anticoagulation, as reflected by a high international normalised ratio (INR).

Previous research has focused on drug, genetic and patient factors known to increase risk of either sub- or supra-therapeutic response. However, it is increasingly recognised that system of care and qualitative factors such as medication compliance, cognition, warfarin knowledge, and social support7-10 have a major role in determining the safety of warfarin use.

Anticoagulation therapy with warfarin encompasses a variety of care models, including specialist anticoagulation clinics, pharmacist-assisted care, and point-of-care monitoring. In some Australian states, it typically involves sharing of clinical responsibility between specialists, general practitioners and a pathology service. Within this model, the GP is commonly the care coordinator, while also providing day-to-day management of episodic and chronic conditions.

We proposed to investigate potential weaknesses in management of warfarin therapy, by exploring the possible contributing factors to elevated INRs in community-dwelling patients. The data are qualitative and hypothesis-generating in nature and provide the foundation for a larger project investigating the relative contribution of patient and warfarin care system variables that contribute to over-anticoagulation.

Forty of 56 eligible patients (response rate, 71%) and 36 of their 40 treating doctors (35 GPs and one specialist physician; response rate, 90%) were recruited (Box 1).

Three-quarters of this group of patients demonstrated measurable psychosocial factors presumed to contribute to risk of medication management-related adverse events, such as cognitive dysfunction, depression, social isolation and self-reported medication non-adherence. Although this study could not determine whether the association with elevated INR was causal, it is noteworthy that most of these patients had one or more psychosocial deficits, and this was not recognised by a third of managing GPs.

Specialists recommending warfarin may have limited knowledge of a patient’s psychosocial circumstances. After referring the patient back to the GP for initiation of treatment and ongoing care, management may then be devolved to a pathology provider without transfer of this crucial information. This division of responsibility may be acceptable for many patients, but it allows room for error in high-risk patients who may not have the capacity for managing medication with small safety margins such as warfarin. Our findings suggest a need for clarification of the relative responsibilities in the coordination of warfarin management. All involved parties should ensure they are aware of a patient’s psychosocial circumstances, to appropriately inform the risk–benefit ratio of therapy for the individual. Appropriate patient assessment and collaborative discussion between the GP and specialist before the commencement of therapy could empower both in the clinical decision and enhance coordination of care. In the absence of such discussion, the GP may feel medicolegally obliged to fulfil a treatment plan initiated by a specialist, despite having reservations about the risk.

There is also a need for a more systematic approach to facilitate comprehensive assessment of individual risk, such as establishing a warfarin risk score. Validated bleeding-risk assessments, including CHADS217 and HEMORR2HAGES,18 consider numerous clinical characteristics but neglect psychosocial factors. We are currently developing a Warfarin Suitability Score that takes into account variables including contraindications, age, comorbid illnesses, intellectual function, depression, social supports, complexity of medication regimens, and medication management capability. This score may improve recognition of patients for whom treatment is more likely to be problematic and who may require assistance with warfarin management.

When high-risk patients are identified, the appropriate model of shared care should be considered. Responsibility may best be confined to a single provider who can explore suitable measures, including simplifying medication regimens and enlisting family help and the assistance of pharmacists or other agencies. A practical approach to service provision would be for the GP to take on this role. Interviews highlighted doctors’ time constraints as a factor that made the efficiency and reliability of the pathology service an attractive option. Time limitations are recognised as a barrier to the holistic care of elderly complex patients.19,20 One solution may be to delegate responsibility for education and follow-up of high-risk patients to a practice nurse, rather than referral to an external pathology provider. Patient education currently provided by the participating pathology service includes written information regarding warfarin at the time of initiation and an understanding of the service provided. Appropriate scheduled remuneration of nurses and/or pathology services may improve existing, and facilitate future, quality care systems.

Division of responsibility for providing education must be clear. Although it is plausible that education was provided, almost half the patients in our study perceived they had never received any warfarin education — the majority reported no regular discussion with their doctor, and most showed measurable ignorance about anticoagulation. The possibility of recall bias in the patients as a result of cognitive deficits cannot be excluded. The contribution of insufficient education and knowledge to poor anticoagulation control has been demonstrated,21-23 and patient report of absence of medication information has been identified as an independent predictor of future adverse events.24

Although the service policy of the participating pathology provider is comparable to those of other providers in Melbourne, no formalised contract of care is obtained outside the initial referral. Issues with non-compliance, instability and warfarin reversal are considered the responsibility of the referring doctor and are dealt with by the pathologist in consultation with both the patient and referrer. An appreciation of individual roles and responsibilities may be unclear in the absence of previous consultation or need for patient intervention. Similarly, agreement to patient care plans (infrequently requested of the pathology service) rarely fleshes out issues of responsibility for more complicated patients or those needing non-routine care. Clearer definitions, both at patient enrolment into anticoagulation services and by general practice care plans, could be of value.

Our study is limited by absence of a control arm to assess psychosocial deficits in patients with stable anticoagulation levels. Patients commencing warfarin for atrial fibrillation are generally elderly2 and may be uniformly affected by cognitive and social variables. The relative contribution of these variables will be assessed in a subsequent case–control study by our group. Further, we can make no comparisons or assumptions about patients managed within public pathology services or independently by their primary care doctors, or whether the factors examined contribute to instability of anticoagulation in hospital inpatients or patients in residential care. The system of care discussed and our recommendations for change may not be generalisable to areas where private pathology providers do not assume a similar role (eg, in other Australian states where the service provides INR testing but not dosage) or where there are stated policies for warfarin management. A better study design would have been to randomly select patients from randomly selected GPs. However, although doctors in this sample were identified by their association with patients experiencing an episode of significant over-anticoagulation, they also managed patients whose anticoagulation levels were stable. This suggests their recorded experiences are likely to be representative of other practitioners.

Despite these limitations, to our knowledge this study is the first to investigate community-based warfarin care from both the treating doctor’s and the patient’s perspective, interviewing doctors at their practices and patients in their homes, with most other studies gathering information from databases or medical records.

The emergence of newer anticoagulant alternatives may result in reduced reliance on warfarin, however there are currently insufficient data supporting their use, cementing warfarin as the mainstay of anticoagulation therapy for now. Future directions may need to include pharmacokinetic assessment alongside systematic use of a risk assessment tool such as a validated Warfarin Suitability Score. There is also a need for role clarification when pathology services become more involved, perhaps by default, in warfarin dosing issues in patients they have presumably never met face to face in consultation. Additionally, collaborative assistance from practice nurses for time-poor GPs should be considered in the care of high-risk and elderly patients. This role would incorporate not only psychosocial dynamics, but continual education, monitoring of medication use including over-the-counter and herbal medications, review of regular INR testing, and liaison with the treating GP before provision of repeat prescriptions.

1 Recruitment process of eligible patients with an INR ≥ 6 and their treating doctors

INR = international normalised ratio.