There is a need to build informed, comprehensive, primary care-based medication adherence services which are tailored to patients' needs [1
]. Non-adherence to medication is estimated to affect approximately 30-50% of patients with chronic conditions [2
]. The consequences include a missed opportunity for treatment effect, poor health outcomes and increased healthcare costs. For example, it has been estimated that non-adherence is responsible for 48% of asthma deaths, an 80% increased risk of death in diabetes and a 3.8-fold increased risk of death in the year following a heart attack [3
]. In the United Kingdom National Health Service (NHS), medicines are the biggest expenditure after staff and 71% of the medication budget is spent in primary care. It is estimated that the current cost of unused or unwanted medicines exceeds £300 million annually [4
]. A recent paper that mapped the quality of medicines use in primary care estimated that only between four and 21% of patients achieve effective, error free care, and that an improvement in adherence would be the most significant area to target to improve medicine use [5
If one is to improve some aspect of a service or behaviour it is generally accepted that there needs to be some measure associated with the behaviour, which can be used to assess the success of interventions. All the various managerial continuous improvement processes depend on having key measures which can be taken repeatedly as an indicator of the success of improvement strategies. Improving adherence should be no different. In order to build effective targeted adherence services, a simple, valid and reliable method for detecting the prevalence and type of non-adherence to medicines is needed [6
]. It is important to be able to monitor the extent of adherence in individual patients and populations as part of routine clinical practice. There is a requirement for a method with which non-adherence can be assessed in individuals, appropriate interventions instigated, and the effectiveness of the intervention evaluated. In order to continually improve quality of services provided to a population, there is a need to have a regular working method to monitor the effectiveness of the service and for pharmacists and Primary Care Trusts (local state funded primary care management organisations) to evaluate the effectiveness of changes to the service. In adherence-related services, improvements in adherence of the population using the service is an important marker of quality. In order to monitor adherence, understand the reasons for non-adherence and improve the effectiveness of adherence services there is a need for frequent, regular measurements which are not very time consuming to make.
As it is not possible to measure actual adherence without continually observing patients, the measures that are used may be considered as proxy measures of adherence rather than absolute measures of adherence. Methods that have been used include using a container that has an electronic chip in the lid that records the time of each opening (Medication Event Monitoring System- MEMS), pill counts, pharmacological and biochemical markers, medical and dispensing records and self report. Each of these measures assesses a different stage in the prescribing and use of medicines, and a different time period. Thus medical records measure the amount of medication prescribed, dispensing records the amount of medication dispensed, MEMS measures the opening of the container, pill counts measure the amount of medication removed from the container, pharmacological markers give an indication of when and how much medication the patient has ingested and self report measures the patients' recall of what they have taken. Medication prescribed, dispensed, the opening or emptying of containers or report that medication has been taken do not necessarily reflect what the patient has taken and therefore may be considered as measures of variables indicative of adherence rather than absolute measures of medication use. However, they are described as adherence measures in the literature, and are described as such in this review for consistency.
Therefore rather than there being a 'gold standard', the choice of measure will depend on the specific situation. Many of the measures used in research have limited feasibility for use in clinical practice. For example, accurate pharmacological measures are only available for a limited number of medications and are intrusive and costly. They may therefore be useful when measuring adherence of a specific medication for which there is a pharmacological indicator to enable change during the intervention to be assessed and when there is a limit to the number of times adherence needs to be measured. MEMS is also very expensive relative to the costs of many medicines used in primary care and is not suitable for all formulations and medications. It may be more useful for a single solid dosage medication being delivered and monitored as part of a clinical trial than multiple drug use in routine primary care. Medical records, dispensing records and pill counts may assist with building a picture of patients' adherence over time but will not be able to assess adherence at a more specific point in time and may not be as useful for continuous routine monitoring.
Self report may be considered the most appropriate method to use for monitoring adherence as part of continuous quality improvement in clinical practice. Self report has its disadvantages as patients are known to overestimate their level of adherence [6
]. For this reason in research studies self report is triangulated with other methods of measuring adherence [6
]. However, NICE guidelines have identified that whilst other types of measures are useful for clinical trials of new drugs, self report is an appropriate tool for clinical practice [7
]. Triangulating between methods would not be practical for regular clinical use. In clinical practice we need to measure adherence in a cheap and relatively unobtrusive way which can be used routinely. We also need to distinguish between intentional and unintentional non-adherence, which have different underlying causes and therefore require different interventions. Self report is the only measure which is able to meet all these criteria. Recent reviews have shown that self report has moderate correlation with electronic monitoring [8
], although self reported adherence levels are higher than adherence levels derived from MEMS [9
A self report measure, which can be used for the continuous quality improvement of adherence related services, needs to have optimal pragmatic, psychometric and theoretical properties.
The measure must be low cost, brief and non intrusive so that it can be used to take repeated measures of self reported adherence over time. It would be preferable for the measure to be able to be administered in different ways, for example, by telephone, face to face and by post. It would also be advantageous if the measure was suitable for both patient self administration and for administration by an interviewer. Questionnaires have been found to have better correlation with other measures than interviews [10
] but some patients may require assistance, so flexibility in method of administration is important. In primary care patients are often treated for multiple conditions. Therefore a measure for a primary care adherence service needs to be generic rather than disease specific and should be suitable for patients taking a single medication or multiple medications for different conditions. It is important to include information from carers where appropriate as it is acknowledged they can have significant roles in medication management [11
Validity (a check that the instrument is measuring what it is supposed to measure) should be established. In adherence research, this is often done by checking the measure against other measures of adherence such as MEMS, Pill counts and clinical markers). Reliability (the consistency of the measurement score) and acceptability to respondent are required. The measure must also be sensitive enough to measure change.
The measure also needs to have a theoretical basis that reflects the need for patient tailored adherence services. The accident causation framework can be used to establish the underlying contextual and individual factors contributing to non-adherence [12
]. Thus adherence can be seen as a 'symptom' rather than a 'diagnosis'. In order to target adherence services to support patients' needs, it is necessary to establish the causes of non-adherence. In particular we need to be able to distinguish between intentional and unintentional non-adherence, as they have different underlying causes and therefore require different interventions. In the past, explanatory psychological models (such as the social cognition theories [13
] and the stages of change model [14
]) have been used to identify beliefs and attitudes that are predictive of non-adherence. However, models focussing on beliefs can only explain intentional behaviour, whilst unintentional causes of non-adherence are left unexplained. Therefore, such models are not suitable for the development of self report measures that are able to diagnose the extent and full range of reasons for non-adherence, in individual patients and populations. The accident causation framework is the only model that we have found that has incorporated unintentional non-adherence.
In this paper, we aim to systematically review the literature in order to identify the self report adherence measures currently available which are suitable for generic use and evaluate the extent to which they meet the theoretical, pragmatic and psychometric criteria described above.