The analysis of data about adherence to antibiotics has led us to question the concepts of adherence and methods used to quantify the phenomenon.1 In their systematic review of "misuse" of antibiotics, Kardas et al report overall mean 'compliance' of 62%.2 However, compliance is defined as "percent of patients who comply with therapy as defined by the original article." The concept of adherence is usually formulated as "the extent to which patients follow the instructions they are given for prescribed treatments"3 or (often in the case of antibiotics) the "percentage of patients who complete the course."2 Both of these fail to account for the many ways in which instructions can be deviated from. More specific definitions include:
- The percentage of people who self-administer a medicine for a specified period
- The percentage of people who have no dose units left at the end of a specified period
- The number of doses units taken in a period as a percentage of those available to take
- The percentage of dose units taken at the time specified (within certain tolerances)
The first of these may rely on simple self-report, whereas the rest require a pill count and the last requires some sort of log. Independent observation and/or electronic monitoring are normally considered necessary to reduce reporting bias. A popular short questionnaire (MARS) gives an overall score (from 5 to 25) which indicates the extent of 5 types of non-adherence including taking too much.4 Reporting bias is widely discussed in the literature, whereas 'concept bias' is apparent but not usually considered as critically. In a recent study of our own we could use the same data (based on self-reported pill count) to claim adherence ranging from 75% to 95% depending on the definition adopted. Comparison between studies is likely to be even more problematic than widely recognised.
To turn to the causes of non-adherence, regression models on our data indicated that the available (and plausible) independent variables could explain no more than about 10% of variation in adherence. This may be a function of the generally high levels of adherence we observed, but may also indicate unquantified factors in operation. For example, we were able to easily account for dose frequency, course length and side-effects experienced; but not perceived disease severity, the fear of side-effects or level of trust in prescribers.
There is a division in the literature between those (in the bio-medical tradition) still concerned with objective adherence and those (advocates of social medicine) investigating more holistic concepts like concordance. Clearly, in some circumstances objective adherence is important but methodological rigour should perhaps be combined with greater understanding of patients' viewpoints. Practice researchers should discuss:
- The need for an agreed specific definition of adherence
- The importance of objective measurement, which may vary depending on circumstances
- The requirement for further research to aid adherence optimisation
- The impact of patient-led care and expert patients on the concept of adherence

References

1. Jackson C, Lawton RJ, Raynor DK et al Promoting adherence to antibiotics: a test of implementation intentions. Patient Education and Counseling. doi:10.1016/j.pec.2005.03.010
2. Kardas P, Devine S, Golembesky A, Roberts C. A systematic review and meta-analysis of misuse of antibiotic therapies in the community. International Journal of Antimicrobial Agents 2005, 26, 106-113.
3. Haynes RB, Yao X, Degani A, Kripalani S, Garg A, McDonald HP. Interventions to enhance medication adherence. Cochrane Library 2005, Issue 4. DOI: 10.1002/14651858.CD000011.pub2
4. Horne R. The medication adherence report scale. University of Brighton, 1996. Available online at http://www.concordance.org/projects/evaluation-toolkit. Accessed on 19/1/2006.