Angelien Sieben

2.2 Methodology
First, patients may not have been exposed enough to the intervention (see paragraph 3.1). Second, in this study we used pharmacy refill data (PRD) as adherence measure. Refill adherence rates have extensively been used for the evaluation of medication adherence. Compared to electronic monitoring, refill data provide researchers with a relatively simple method for investigating adherence to medication in large populations. However, due to the increasing availability of automatic refills in the Netherlands, this measure may represent high adherence levels while patients do not necessarily take their medication. Since there is no ideal medication adherence measure, it is appropriate to use more than one measure. This recommended multi-measure approach was applied in this study and both showed medium to high adherence rates. The results of the MMS also showed no differences in medication adherence between the groups. One can consider to perform adherence research with more direct measurements as measurement of the medication in body fluids, such as blood or urine or the presence of a biological marker given with the drug or a direct observation of patient’s medication-taking behavior. Even though these measures are considered to be the very accurate and can be used as physical evidence to prove that the patient has taken the medication, there are also concerns regarding their use. They simply generate a yes or no result without revealing any pattern of nonadherence. Also drug metabolism should be taken into account while considering using these methods. Furthermore, direct measures are very expensive and difficult to perform. It will also make the patient aware of the observation of his medication adherence which can be seen as an intervention in itself. However, much more research in adequate and objectively measuring medication adherence is needed because even after decades of research, it is difficult for healthcare professionals and researchers to choose the most suitable adherence measures. Third, all patients in our study received the same structural cardiovascular risk assessment, lifestyle intervention and best medical treatment and supportive care according to the European Guidelines of prevention of cardiovascular diseases. All patients in our study received the same usual cardiovascular care from the nurses. Although medication adherence was not a structural approach in our usual care setting, the attention and screening on CVD-risk factors may have influenced adherence to medication positively. A previous evaluation of our nurse-led cardiovascular vascular risk program showed that a structural multidisciplinary evaluation and initiation of the best medical treatment in combination with addressing unhealthy lifestyle reduces cardiovascular risk as indicated by a reduction in smoking, alcohol consumption, unhealthy eating, blood pressure, and LDL-cholesterol level. A reduction in mortality and morbidity by these nurse-led programs is shown more often. Moreover, nurses are achieving results, equal or even better than GPs, for the management of cardiovascular risk factors. Nurses do have a key role in understanding and addressing patients’ beliefs during consultations about their medication. Nurses are close to the patient and often his family, so there can be a therapeutic partnership that is respectful of the beliefs and choices of the patient in determining when and how treatment is followed. The CVRM at our hospital is performed by specifically trained nurses. They are trained in knowledge about cardiovascular diseases and their risk factors. They are also trained in motivational interviewing to support patients to a healthy life style. And they are successful in it. Nonetheless, the present study did not evaluate the actual performance of the nurses in this matter, and quality and individual differences in nurses’ communication are likely to exist.

2.3 High adherence in all groups
Because of the high adherence rates in all groups, it was difficult to demonstrate a difference between the groups. A power calculation was executed to establish the number of patients needed to demonstrate a difference in adherence. But contrary to the level of poor adherence we found in our literature search, the overall adherence rate in our total population was much higher (50-60% in the literature to 80% in our population). So, patients who participate in an adherence study may be patients who already are more adherent. This will be discussed further in paragraph 3.5.

2.4 Pharmacy Refill Dates (PRD)
Trying to get the data needed to calculate the PRD turned out to be a major challenge. There was no central database we could use. There was no information about refill dates of our patients available for the prescribing clinicians. We had to ask every single pharmacist (more than a hundred) if they would be willing to provide us with the refill dates of the patients participating in this study. Some pharmacy databases were easy to use for the calculation of the PRD. Other databases needed to be cleaned up first because all kinds of supplements were documented like non-prescription medication and medical aids. Eventually we were able to calculate the PRD for over 80% of the patients. It made us aware that in the circle around the patient nobody knows exactly if and how often a patient’s medication is supplemented. A prescribing clinician or a pharmacist can only rely on the information a patient reports and has no other tools at his disposal. Personalized feedback using data obtained from electronic devices who monitor the patient’s adherence also tends to have a positive effect on medication adherence.

2.5 Patient’s perceptions and beliefs
In order to enhance adherent behaviour, the perceptions and beliefs of the patient need to be taken into account. Using the answers of the BMQ enabled the nurses to individualize their consultations. A shift in the categories showed from ambivalent towards accepting in the intervention group was shown. This was not observed in the usual care group. was observed in this study. In the ambivalent group, necessity beliefs were high but concern beliefs too. In the acceptance group necessity beliefs were high but the concern beliefs were low. In order to know if the change in the necessity category for the intervention group would have a more positive effect on the adherence rate over time, we need to measure the adherence rate more than one year after the intervention, because one year after the intervention, adherence rates were still high in both groups.

2.6 Effect on cholesterol and blood pressure levels
Although overall adherence was relatively high at 12 months follow-up, only 20% of all patients had a systolic blood pressure within target. The mean blood pressure (BP) was even higher than it was at baseline. We did not expect this result at all. Several major studies did, however, demonstrate that nonadherence is a major determinant for not reaching target levels, for lipid lowering medication as well as for antihypertensive medication. There are several ways to elaborate on this finding. The PRD and the MMS, even if in combination, may not have been sensitive enough to detect nonadherence in our population. We discussed this in paragraph 3.1. Another explanation could be that failure to achieve the target systolic blood pressure level is not due to nonadherence. It sometimes seems to be difficult to establish the relationship between adherence and BP. Also, all BP-measurements were done at the outpatient clinic. A 24-hour ambulatory home BP monitoring would have been more accurate.

2.7 Adherence in (non) participants of a randomized controlled trial
Because of the unexpected and surprisingly high adherence rates at baseline of our study population we wanted to explore the suggestion that patient recruitment methods in randomized controlled trials could have resulted in inviting a population with high adherence to medication. Our study of the differences in medication adherence among patients who did or did not consent to participate in an RCT showed that patients who were willing to participate in the RCT to evaluate the effect of an intervention to improve medication adherence had a comparable adherence levels to patients who declined to participate. However, the participants were younger and were more educated. Because these characteristics are known as prognostic characteristics for patients who are willing to participate in a clinical trial and for a high adherence level as well, it was expected that these characteristics could explain the assumed higher adherence rates in the participant groups. We could not confirm this hypothesis. Also, next to the high adherence rate, a high mean NCD score was present in all groups. This is congruent with earlier studies showing that medication beliefs can be a more powerful predictor of medication adherence than clinical and socio-demographic factors. However, we did not observe a relationship between NCD and trial participation. This suggests that a population representative in adherence levels was included in our RCT evaluating the effect of an intervention to improve medication adherence. An alternative explanation for the high adherence rates in both groups could be that we started inclusion shortly after the initial cardiovascular event, a period in which the importance of being adherent is emerging in most patients. Yet, as the impact of the event fades and symptoms subside, adherence levels may decline. Studies with a long follow up are needed to establish a difference in adherence between participants and non-participants over time.

2.8 Identification of patient groups at risk for nonadherent behaviour
In this study, cardiovascular patients who were relatively young, used a limited number of medicines and had an unhealthy lifestyle were identified as patients who are at risk for non-adherent behaviour. It has been speculated that medication adherence is a marker for other health choices, the so-called “healthy adherer effect”. Patients with low medication adherence may have an unhealthy lifestyle. It is suggested that factors associated with an unhealthy lifestyle such as poor health knowledge and low self-efficacy to change behaviour, also lead to poor adherence. But one can also easily turn this around. ‘By living as a very healthy person, I don’t need my medication anymore’. This would mean that if someone is living really healthy, one would be worse medication adherence (trade-off). Both groups are described in the literature. Although older age and complexity of drug regimens have previously been identified as a major determinant for non-adherence, we unexpectedly found that a younger age at the time of a TIA/stroke or acute coronary syndrome was associated with a reduced medication adherence. Although this was an inconclusive finding, it suggests that younger patients may be more likely to be nonadherent to preventive medications due to lower perceived risk of another CVD, misconceptions about the duration of the treatment, or concerns about potential harm from statins. The other remarkable finding was that although a complex drug treatment plan is often associated with lower medication adherence in the cluster with the poorest adherence only a small number of medicines were used. This could be explained by the clinical outcomes which already were at target at baseline. The indication for prescribing medication was simply less present. Although cluster three (i.e., patients who are relatively young) showed the poorest adherence, our cluster analysis did not show an association between nonadherence and clinical outcomes at target level. An explanation could be the relatively young age of this group. With aging, the prevalence of metabolic syndrome (including hypertension and dyslipidemia) increases. Younger patients may already have a (sub)optimal level of LDL and BP before the cardiovascular event. By analyzing the single variable age in relation to adherence in this population, no significant difference in adherence was observed between the three age groups. Only when clustering the variables a significant difference between the groups on adherence could be reported. We expected there would also be a difference between the clusters in the outcome of the BMQ. The clusters, however, showed no significant differences in the outcome categories of the BMQ. In this study, we only used the categorical outcomes of the BMQ (the four different belief groups) because categorical outcomes are the preferred measure for a cluster analysis. The difference between the continuous and categorical outcome may explain the absence of an association between the BMQ and the MMS. Identifying non-adherent behaviour in cardiovascular patients by clustering determinants based on the structural cardiovascular screening outcomes can lead to a more effective approach to improve medication adherence. The study showed that isolated established predictors of adherence are often insufficient to identify individual patients who are likely to be non-adherent. Identifying patients who are at risk for nonadherent behaviour should not be simply identified with single measures for assessing medication adherence. Different risk factors should be taken into account to identify a patient at risk.

3. Conclusion
Complex problems and partly unexpected interactions between patient characteristics, beliefs and behaviour underly poor adherence in cardiovascular patients. Because these patients should take their medication for the rest of their lives, the impact of poor adherence is huge. Measuring medication adherence is challenging and labour intensive. Although the prevalence of non-adherence over a longer period after the initial CVD event seems far less than previously thought, there is still a need for interventions to tackle this problem. In developing such an intervention, we were convinced that we took the right design steps, taking existing knowledge into account. By evaluating this study, new insights emerged:

Pharmacy Refill Dates
Improve the way in which data from refill prescriptions are available for the prescribers and pharmacists. In order to improve communication about medication adherence between prescribers, pharmacists and patients, PRD should be easy to access for clinicians. There is a world to win concerning the use of information about medication, even in daily practice.

Development of interventions
To improve behavioural change interventions and tailor them to patient’s needs, involve patients in the development of these interventions. However, recent studies which involved patients in developing interventions did not show a better uptake of the interventions. They considered that patient’s needs can differ from each other so even if you include patients in intervention development, patients who will be offered this intervention can have another, specific need. A recent review of adherence to eHealth technology confirms there still is much to learn about how to adhere patients to eHealth. More research should be done on what is effective in getting patients to use behaviour change programs in general and eHealth in particular.

Cardiovascular risk management (CVRM) by nurses
The key role nurses have in cardiovascular risk programs and improving outcomes should be recognized and stimulated. Recognition of the role of nurses in general in improving health is increasing. However, nurses have a shared concern about staffing problems and inadequate education, training and support. This can result in poor quality care. Moreover, nurses report that they are frequently not permitted to practice their competence to the full and are unable to share their learning, have too few opportunities to develop leadership and fulfill leadership roles. The world needs 9 million more nurses and midwives to achieve universal health coverage by 2030. That is why the World Health Organization (WHO) has designated 2020 as the universal year of the nurse and the midwife. By investing in and developing nursing worldwide, a triple aim should be achieved; greater gender equality, stronger economies and better health.

Risk profile of patients
The group of patients that showed the poorest medication adherence was relatively young and used a limited number of medicines. This is in contrast to the more traditionally known determinants of poor adherence (elderly age and polypharmaceutical use of medications). Further studies could lead to a different approach to improve medication adherence in CVD patients, targeting a different patient group.

4. Recommendations
4.1 Recommendations for further research
In developing an intervention to influence patients’ behaviour, especially e-Health programs, we need more insight in patients’ needs and in how to meet these needs. In defining the use of a technology and selecting valid adherence measures, the goal or the assumed working mechanisms should be leading. Adherence measures can then be standardized, which will improve the comparison of adherence rates to different technologies with the same goal and will provide insight into how adherence to different elements contribute to the outcomes. In order to confirm that the time to initiate an intervention is important, and perhaps should occur much later after a CV event, adherence rates must be measured further in time, up to two years after an event. If the adherent rates are lower, a retrospective study should be done with the same determinants we used in our study to identify even better who are at risk for non adherent behaviour. The cluster analysis which can help identify CVD patients who are at risk for non adherence should be done in a much larger cohort to confirm the findings of this study. To confirm that our regular nurse led CVRM program is also helpful in improving non adherence, we should compare the medication adherence of patients who followed such a program with patients who did not.

4.2 Recommendations for clinical practice
Improve the feedback for prescribing clinicians on the refill data on medications of their patients. This could be very helpful in the communication between clinician and patient about adherent behaviour. Continue with the CVRM program led by nurses. This program is well implemented in the Netherlands and could also be very helpful in reducing medication adherence. Be critical in the kind of patient you think is at risk for nonadherence on the basis of their higher age or large number of prescribed medicines. Although these patients are often described as at-risk patients, in CVD the relatively young patient with an unhealthy lifestyle could be more at risk. Nurses who lead the CVRM programs can help their patients and use their skills to help these group of patients especially about their adherent behaviour.