Development and piloting of a highly tailored digital intervention to support adherence to antihypertensive medications as an adjunct to primary care consultations

Objectives This paper describes the systematic development and piloting of a highly tailored text and voice message intervention to increase adherence to medication in primary care. Methods Following the Medical Research Council guidance, this paper describes (a) the systematic development of the theoretical framework, based on review of theories and meta-analyses of effectiveness; (b) the systematic development of the delivery mode, intervention content and implementation procedures, based on consultations, face-to-face interviews, think-aloud protocols, focus groups, systematic reviews, patient and public involvement/engagement input, intervention pre-test; and (c) the piloting of the intervention, based on a 1-month intervention; and follow-up assessment including interviews and questionnaires. The mixed-methods analysis combined findings from the parallel studies complementarily. Results intervention development suggested the target behaviour of the intervention should be the tablets taken at a regular time of the day. It recommended that patients could be more receptive to intervention content when they initiate medication taking or they change prescription plan; and more emphasis is needed to patients’ consent process. Intervention piloting suggested high intervention engagement with, and fidelity of, the intervention content; which included a combination of behaviour change techniques, and was highly tailored to patients’ beliefs and prescription plan. Patients reported that the intervention content increased awareness about the necessity to take and maintain adherent to medication, reinforced social support and habit formation, and reminded them to take medication as prescribed. Conclusion Tailored automated text and voice message interventions are feasible ways to improve medication adherence as an adjunct to primary care. Trial registration number ISRCTN10668149.

, and often accompanies other conditions, including type 2 diabetes mellitus (T2DM), coronary heart disease and stroke [2]. These conditions are major risk factors for disability and premature death [2], and medication adherence can significantly lower these risks [1]. For example, in England over ten years, an estimated 7,000 quality of life years could be saved and £120m not spent on related health and social care, if people had better adherence to anti-hypertensive medications [3].
Adherence is defined as taking at least 80% of the prescribed tablets, and/or having medications dispensed for at least 95% days of a prescribed period [4]. However, substantial proportions of patients do not take their medication as prescribed [5]. A recent metaanalysis showed that 40% of people do not adhere adequately to cardiovascular medication and the percentage was similar (41%) for non-adherence to anti-hypertensive medications [6]. Taking into consideration the growing prevalence of HBP, co-morbidities and the ageing population [7], especially in lower and middle-income countries [8], it is likely that there will be an increase in demand for health care resources to support medication adherence. NICE recommends the development of novel and cost-effective interventions to aid patient's decision making about taking medicines, as an adjunct to health care providers' consultations [9,10].
Tailored mHealth interventions, such as text (SMS) and voice (IVR) messaging intervention, is one way to support patients' adherence between their consultations. Review evidence found that such interventions can effectively support adherence to different types of medication, including antihypertensive tablets, when compared to usual care only [11,12], and can potentially be cost-effective if applied to large number of people [13]. Evidence suggest that there is a scope of developing tailored interventions, since these seem to be more effective than non-tailored interventions [14]. However, no such intervention has been developed and piloted in the UK.
The aim of this paper is to describe the development and piloting of a highly tailored text and voice messaging intervention to support adherence to antihypertensive tablets.
Although several guidelines have been proposed to the development of mHealth interventions [15][16][17], this paper is informed by the MRC Framework [18] as it leads not only the process to develop the intervention content and delivery mode, but also the implementation procedures within the primary care National Health Service (NHS), which is an important process in developing and piloting novel interventions.

Review of theory and evidence, interviews, focus groups, consultations and Patients and
Public (PPI) involvement and Engagement (PPE) informed the intervention development process (see Appendix 1); and a one-month pre-post intervention study with follow up assessment informed the piloting of the intervention (see Appendix 2).

Intervention Development
Step intervention was informed by findings from two systematic literature reviews [11,12] and previous empirical evidence [24].
Tailor the theoretical determinants. To develop a highly tailored intervention we produced a tailoring matrix, which consist of a theory-based questionnaire, an algorithm of Integrate theory to intervention. The frequency, combination and sequence of the BCTs were integrated into one intervention. The decision on the above elements was based on the phase participants are in (e.g., uptake or maintain), having as the starting point the collection of the medication from the dispensary.
Step 2. Development of the Intervention Delivery Mode, Content, and Implementation Procedures. Patients and Public Involvement (PPI) and Engagement (PPE) was conducted to assess acceptability of the delivery mode and the theoretical framework, and inform the intervention content. PPI/E members were recruited opportunistically using emails or adverts. In total 100 PPI/E members took part in three events: one meeting with a diabetes

Pre-test feasibility of intervention delivery mode and theoretical determinants.
Four studies informed the acceptability of the delivery mode and the theoretical determinants. These included: think aloud interviews, experiential focus groups, and PPI.
Face-to-face, in depth interviews were conducted with a subsample of patients (n=13) recruited at the study mentioned above. Think aloud protocol was used to assess reliability of the mechanisms of change (i.e., link between the theoretical determinants and the BCTs), and usability of the delivery mode. All interviews were conducted at participants' places. Field notes and think aloud data informed the analysis, and results refined the tailoring of the theoretical determinants [25].
Face-to-face meetings were conducted with two PPI members recruited using emails from the list of the NIHR Cambridge BRC PPI/E Panel and Communications office. Think aloud protocol was used to assess the reliability and sequence of the items to form the measurement of the theoretical determinants (i.e., theory-based questionnaire). Both interviews were conducted at the University. Field notes and think aloud data informed the analysis, and results refined the reliability of the tailoring questionnaire and informed the decision rules of the tailoring algorithm.
Experiential focus groups aimed to obtain participants' views of, and refine, the tailored intervention content (e.g., inclusion of medication name, dose and timings; 'please do not forget to take 1 tablet ramipril 10 mg at 9am') and delivery mode within a social interaction context [26]. Eligibility criteria and recruitment procedure were similar to patients' face-to-face interviews, reported above. Four focus groups took place in primary care practices (n=2) or community centres (n=2). During the focus groups patients (n=12) received examples of SMS and IVR messages at their mobile phones and were asked to provide experiential feedback. Participants were asked their views and recommendations about the combination of the delivery modes and the intervention content, using a thinkaloud protocol.
One experiential focus group with six PPI members aimed to obtain views and recommendations on the intervention content and delivery mode. PPI were recruited using emails from the NIHR Cambridge BRC PPI/E office. PPI members were eligible to take part if The data generated from the experiential focus groups informed the content of the message file and the mapping of the intervention messages to the theoretical determinants.
Develop implementation procedures within primary care. The intervention implementation procedures were informed by systematic review evidence, PPI, and stakeholders' consultations.
Information about the recruitment setting and participants' characteristics at baseline, as well as uptake and retention rates were extracted from the two systematic reviews of trials on medication adherence delivered by SMS and/or IVR messages [11,12].
Due to missing data, we synthesised the findings narratively. Results informed the decisions regarding the recruitment setting and methods, as well as the targeted behaviour and population.
Twelve PPI members provided feedback on the recruitment material (e.g., patients' information sheets, posters, leaflets, flyers, and invitation letters) using emails or face-toface meetings. Their comments informed the content of the recruitment material and the description of the recruitment, informed consent, and intervention procedures in lay language.
Stakeholders' consultations were conducted with health care providers, commissioners, and patients to inform the study design and the implementation procedures within the primary care setting (e.g., recruitment, informed consent, data collection).
Practitioners with experience of advising patients for medications; commissioners for medicine optimisation, prescribing or cardiovascular conditions; and patients with either HBP or both HPB and T2DM were included. Commissioners were recruited informally by email through personal networks. Patients and health care providers (i.e., practice nurses, health care assistants) were recruited through the primary care practices. Those who consented to take part were invited to the consultations. Consultation modes were flexible to maximise participation (n=3 face-to-face, n= 4 email). Stakeholders (N=7) were given a description of the proposed intervention and asked their views and recommendations.
Health care providers (n=4) were consulted about the time to initiate recruitment, and the to track resource use and costs and the evidence needed to inform whether or not to commission such an intervention. Patients (n= 2) were consulted about the delivery mode and the theoretical framework. All stakeholders were asked for their recommendations about the implementation procedures. The data generated from stakeholders' consultations informed the study and intervention implementation procedures.
Step 3. Develop the Prototype Intervention.
A prototype intervention was developed by multiple and iterative syntheses of the data obtained during the intervention development [27]. The intervention is highly tailored to deliver very brief (≥1 minute), theory-based messages at participant's preferred time, telephone to receive the calls (i.e., primary and/or secondary number), and at frequency of calls (e.g., repeat options). The pilot intervention consisted of 29 messages, delivered daily for a duration of one prescription-based month. Two messages were delivered at the first day of the intervention, with feedback tailored to each participant's responses at the baseline questionnaire and prescription plan, and one message per day during the following 27 days of the intervention.
Step 4. Refine the Intervention.
Two PPI members, recruited by the NIHR Cambridge BRC PPI/E office, pre-tested the prototype intervention for one prescription-based month. PPI provided their experiential feedback about the intervention usability and mechanisms of change during four, weekly telephone-based interviews with a researcher, and completed the follow up questionnaire.
The data informed the decision rules of the tailoring algorithm and refined the frequency (i.e., message schedule), combination and sequence of the BCTs (i.e., integration of theory into the intervention), before piloting the prototype intervention within primary care.

Analysis
Data generated from the intervention development and pilot study were integrated into one mixed methods analysis [27], using the technique called "following a thread" [28,29].
The theoretical underpinnings of the intervention were explored across different methods of data collections, data generated from each method informed each intervention

Schedule
The one-month intervention schedule included daily messages and participants could change the frequency or stop them. Patients at follow up interviews, recommended an intervention of longer duration with messages to gradually decrease in frequency and include advice and support for self-monitoring medication taking.

Targeted Behaviour and Participants' Characteristics
Participants reported that it was easier for them to report on the tablets or pills taken per day, rather than the specific type of tablets (e.g., felodipine, statins). However, they reported that this information could be useful for patients who initiate medication taking, and need more education about their prescribed regimens. Patients reported that the timing of medication-taking or the routine associated with the timing (e.g., lunch), rather than the type of medications to take, was easier reminder of behaviour. That was particularly important for those patients taking medication for longer and for more than one health conditions. Health care providers who conducted the recruitment procedures, reported challenges with recruiting non-adherent patients and recommended recruitment strategies delivered at times when patients might be more receptive to uptake such an intervention. In accordance to the systematic review evidence, it was recommended that patients would be more receptive to uptake the intervention when they initiate medication taking or when there is a change to their current prescription plan (e.g., this could be either due to a health event and/or due change to the pharmaceutical product). Otherwise, patients could be reluctant to provide consent and uptake the intervention (table 2). Annual reviews: "annual review is a bit more time … when its annual reviews because we're actually talking about their tablets" practice nurse recommendation about recruitment procedures Practice website: "but you can put a facility on the online forum. A lot of people access our website, so there are quite a lot of other places that people might look maybe more than a poster" practice nurse Medication prescriptions: "but every time you get a repeat prescription list, you can write something on it" practice nurse Note. N=8 health care providers: nurses and health care assistants. Themes have been coded at interviews with health care providers at development (n=5) and piloting of the intervention (n=3). Quotes reported in table are from experiential interviews with health care providers (T2).
Health care providers recommended that recruitment methods could be integrated to medication annual reviews or to patients' prescriptions. They also recommended that nonadherent patients, who do not attend the practice for their medication reviews, either because they do not appreciate the need of these or they are sceptical about the  INTERVENTION DEVELOPMENT AND PILOTING   16   recommended treatment, could be more receptive to the intervention if invited by the   other methods, such as text messages, newsletters, or multimedia (table 2).
However, health care providers also revealed that time constrains prevented them from explaining the implementation procedures to eligible patients during primary care consultations. In line with data from the PPI consultations and patients' interviews, health care providers suggested that there is a lot of information in the recruitment material and suggested that more emphasis is needed to patient's consent process. It was suggested that patients should be made aware about what information will be collected by their practice records, and how these will be utilised from the intervention. To facilitate patients' understanding and standardise the implementation procedures, including the informed consent, we developed videos aligned to written material, and integrated these into a brief face-to-face recruitment appointment.
To reinforce the uptake and integration of the intervention into current NHS, commissioners, patients, and health care providers reported that intervention needs to be of low or no cost. "Would need to be very low cost or free to use" commissioner consultation.

Intervention Uptake and Retention
Twenty participants were recruited. Eighteen provided written informed consent and registered into the study, of whom 17 completed the one-month intervention: one participant was excluded by the research team before registration because he/she was meeting one of the exclusion criteria, one participant withdrew before registration to the IVR, and the other participant dropped out during the intervention. All 17 participants completed the measures at baseline and at the end of the intervention. All primary care practices provided completed baseline (e.g., refill prescription data three months before the start of the intervention) and follow up data, for all participants.

Intervention Fidelity and Engagement
During the one-month intervention on average 29 calls were scheduled, of which 22.52 calls received by participants. On average 37 calls were made, of which 2.8 failed to be made due to technical issues and 13 calls made but failed to go through because they were not picked  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   INTERVENTION DEVELOPMENT AND PILOTING   17 up by the participants. On average three calls were repeated and received by participants. Ten out of the 17 participants had chosen the calls to be repeated if not answered, and five   out of the 17 participants had provided a secondary number for the calls to be repeated, if there was no answered at the primary number. Overall, participants were engaged with the intervention and made inbound calls to change the schedule of the messages or report about intervention content. In summary, there was a good fidelity of the intervention BCTs, with the majority of the BCTs having high fidelity scores (table 3).
Participants suggested that the intervention was acceptable and easy to use, and they provided their views about specific features of the intervention delivery mode and content.
Specifically, the tailored schedule of the messages, the personalisation, and the variation of the content was found to be particularly appealing and were perceived to promote engagement with the intervention (Table 5, Appendix 6). Participants reported being satisfied with the overall experience with the intervention, the availability of the intervention 24/7, and their ability to call in and leave a message (see Appendix 6). They would also recommend the intervention to other people who take prescribed medications (see Appendix 7).

Principal Findings
A highly tailored intervention has been developed to support adherence to medication prescribed for high blood pressure and co-morbidities. The intervention has been systematically developed based on guidance from MRC framework, PPI/E and is based on rigorous theory and evidence. It includes a combination of Behavioural Change Techniques (BCTs) tailored to patients' beliefs (e.g., beliefs about medications, self-efficacy, social norms, perceived regimen complexity) and prescription plan. The the pilot study suggested that the intervention content increased awareness about the necessity to medication adherence, the risk and benefits of maintaining adherent to prescribed medications, reinforced social support and habit formation, and reminded patients to take medications as prescribed. To our knowledge, this is the first medication adherence intervention delivered using automated text and voice messages that has been developed and piloted within the UK primary care setting. The intervention has been proven to be acceptable and feasible, and it is currently evaluated in a randomised controlled trial.

Strengths and Limitations of this study
To date there are studies describing the systematic development of mHealth interventions, however none of them has provided a detailed description based on guidance from the MRC framework, and none has described the methods used to develop and pilot the implementation procedures within the primary care setting. Furthermore, this study has involved PPI/E in the intervention development, design and pilot, adapting a collaborative approach, which minimized the influence of the social desirability effect on the findings, and thus it provided a real-world perspective. As a consequence, the results of this intervention may be better informed and matched to patients' healthcare needs.
A strength of this study is the systematic approach to intervention development and piloting. The study used a rigorous theoretical approach to guide the development of the tailored intervention content. It also designed a flexible and scalable application to deliver messages to support adherence to different type, and complex prescribed medications.
Another strength of this study is the utilisation of a mixed (e.g., qualitative, quantitative) and multi-perspective (patients, health care providers, PPI, stakeholders) methods of data collection. The data were combined complementarily into one comprehensive analysis, using a rigorous analytical approach.
Although this intervention is highly tailored, and participants can provide real time input and further tailor intervention content and delivery, a limitation of the intervention is the self-reported theoretical determinants of medication adherence. Future studies could enable the use of objectively measured theoretical determinants and objectively measured behaviour, using sensing technology. Another constrain of the text and voice messaging is its limited features to facilitate delivery of BCTs visually (e.g., picture of tablets, graphs to show levels of adherence) [30], and future studies could usefully explore whether and how these features could influence intervention engagement and behaviour change.

Future Research
The one-month intervention has been extended to three months intervention and targets all adults (above 18 years old) to increase reach. It also includes another level of tailoring, where participants select between two BCTs to support maintenance of medication taking (e.g., habit formation or self-monitoring). The intervention is currently being evaluated in an RCT, which includes objective behavioural (medication event monitoring system) and clinical Abbreviations: Hospital Open Event -HOP; Science Festival -SF; Focus Groups -FG Scenario: Mark is a 49 year old car mechanic. He was diagnosed with high blood pressure by his GP 6 months ago even though he had no symptoms, was not overweight, did not smoke, ate healthily and had an active job. Initially, he took the medication exactly as his GP prescribed every day. However, after a few weeks he began to experience constipation and read on the leaflet supplied with the tablets that this was a side effect caused by the medication. Now, he only takes his medication every other day and the constipation has cleared up. However, his blood pressure is still too high.

Message Event
Keep on the tablets prescribed to you by your Doctor. High blood pressure is more serious than constipation. HOP His action was foolish -he should have consulted his GP before changing the prescribed medication. HOP This scenario similar to mine. However I do take my meds without fail. I am motivated by my family history which leads me to believe I may be susceptible to heart disease which, of course, is thought to be related to blood pressure. I can understand why people might be concerned about side effects, so they need motivation.
HOP I know that mechanics are frequently asked to work through lunch breaks etc. Please take some fruit and water to work with you to snack on throughout the day. This will help with your constipation, as you should be taking your pills every day. Good luck.

HOP
You need to talk to your GP in order for him to prescribe a laxative to help with the constipation. The need to take your BP medication is crucial in bringing your BP down. HOP BP tablets must be taken every day so consult your GP for medication for constipation. HOP It's stupid not to take your meds! If you're constipated just eat more fibre. It's not worth the risk. HOP Hey Mark. Are you aware of your blood pressure today? Have you taken your meds yet? SF Hi Mark. I sympathise a lot. I used to suffer from blood pressure and constipation, but nowadays I rarely suffer from either. I've had heart surgery but still cycle a lot. Don't give up on the meds. There are different tablets available: ask your GP. But also, use whole brand cereals regularly, even at bedtime! SF Hi Mark. You've sorted the constipation -keep going for the blood pressure! SF Mark, you should keep taking your medication even with the side effects, because it is better for you. SF This is your message for your blood pressure tablets. One easy way to remember your tablets, is to take them with another daily activity, such as your morning cup of tea. If they are always done together, it will be harder to forget.  Beliefs about medications,  self-efficacy to take medications when no health benefits and  emotional state

Habit formation and information about emotional consequences
Please keep taking your tablets as prescribed even if you are well and feeling healthy. High blood pressure is one of those things that unless you actually feel it you're not aware that it is a problem. Taking your medications as prescribed will support you to keep enjoying things or activities that are important to you.
NINA,  beliefs about medications and  emotional state Habit formation and social support (unspecified) Whatever the day may holds, please do not forget to take your tablets. To achieve better control over high blood pressure, you will need to take them every day.
 self-efficacy to cope with  practical social support Habit formation and social support (unspecified) We are calling to see how you are doing with taking your tablets today. Please remember that you can always call the free telephone number if you require more support on how to take your tablets as prescribed. If you require more information about the health condition or the dosage of your medications, please contact your GP practice or pharmacist.  social norms and  selfefficacy to cope with  practical social support Habit formation and social support (unspecified) You may find it helpful to ask people who are important to you to remind you to take your medications as prescribed and maintain your blood pressure under control. Please remember that you can always call the free telephone number if you require more support on how to take your tablets as prescribed.
Note: ↑indicate higher score from the cut-off point, ↓ indicate lower score from cut-off point.

Abstract
Objectives: This paper describes the systematic development and piloting of a highly tailored text and voice message intervention to increase adherence to medication within primary care.
Methods: following the MRC guidance this paper describes: a) the systematic development of the theoretical framework, based on review of theories and meta-analyses of effectiveness; b) the systematic development of the delivery mode, intervention content and implementation procedures, based on consultations, face-to-face interviews, thinkaloud protocols, focus groups, systematic reviews, PPI/E input, intervention pre-test, with c) the piloting of the intervention, based on a one-month intervention; and follow up assessment including interviews and questionnaires. The mixed methods analysis combined findings from the parallel studies complementarily.
Results: intervention development suggested the target behaviour of the intervention should be the tablets taken at a regular time of the day. It recommended that patients could be more receptive to intervention content when they initiate medication taking or they change prescription plan; and more emphasis is needed to patients' consent process.
Intervention piloting suggested high intervention engagement with, and fidelity of, the intervention content; which included a combination of BCTs, and was highly tailored to patients' beliefs, and prescription plan. Patients reported that the intervention content increased awareness about the necessity to take and maintain adherent to medication, reinforced social support and habit formation, and reminded them to take medication as prescribed.
Conclusion: tailored automated text and voice message interventions are feasible ways to increase medication adherence, as an adjunct to primary care.

Strengths and limitations of this study
• This is the first medication adherence intervention for patients with hypertension and co-morbidities that has been developed and piloted within the UK primary care setting.
• The study used rigorous methodology to collect and analyse data, from multiple perspectives.
• This study involved and engaged Patients and Public in all stages of the intervention development and piloting.
• Future studies should include real-time and objective measures of the intervention determinants and outcomes. people had better adherence to anti-hypertensive medications [3].
Adherence is defined as taking at least 80% of the prescribed tablets, and/or having medications dispensed for at least 95% days of a prescribed period [4]. However, substantial proportions of patients do not take their medication as prescribed [5]. A recent metaanalysis showed that 40% of people do not adhere adequately to cardiovascular medication and the percentage was similar (41%) for non-adherence to anti-hypertensive medications [6]. Taking into consideration the growing prevalence of HBP, co-morbidities and the ageing population [7], especially in lower and middle-income countries [8], it is likely that there will be an increase in demand for health care resources to support medication adherence. The National Institute of Health and Care Excellence (NICE) recommends the development of novel and cost-effective interventions to aid patients' decision making about taking medicines, as an adjunct to health care providers' consultations [9,10].
Tailored mHealth interventions, such as text (SMS) and voice (IVR) messaging intervention, is one way to support patients' adherence between their consultations. Review evidence found that such interventions can effectively support adherence to different types of medication, including antihypertensive tablets, when compared to usual care only [11,12], and can potentially be cost-effective if applied to a large number of people [13]. For example, a recent study suggested that such interventions could result in a 2.3% increase in QALYs and overall savings of 21,120 Australian dollars during a patient's lifetime [14]. However, no such intervention has been developed and piloted in the UK primary care setting.
The aim of this paper is to describe the development and piloting of a highly tailored text and voice messaging intervention to support adherence to antihypertensive tablets.
Although several guidelines have been proposed to the development of mHealth interventions [15][16][17], this paper has been informed by the Medical Research Council (MRC) Framework [18] as it leads not only the process to develop the intervention content and delivery mode, but also the implementation procedures within the primary care National Health Service (NHS), which is an important process in developing and piloting novel interventions.

Review of theory and evidence, interviews, focus groups, consultations and Patients and
Public (PPI) involvement and Engagement (PPE) informed the intervention development process (see Appendix 1); and a one-month pre-post intervention study, with follow up assessment informed the piloting of the intervention (see Appendix 2).

Intervention Development
Step . Our decision on the BCTs for this intervention was informed by findings from two systematic literature reviews [11,12] and previous empirical evidence [24].
Tailor the theoretical determinants. To develop a highly tailored intervention we Integrate theory to intervention. The frequency, combination and sequence of the BCTs were integrated into one intervention. The decision on the above elements was based on the phase participants are in (e.g., uptake or maintain), having as the starting point the collection of the medication from the dispensary.
Step 2. Development of the Intervention Delivery Mode, Content, and Implementation Procedures.

Pre-test feasibility of intervention delivery mode and theoretical determinants.
Four studies informed the acceptability of the delivery mode and the theoretical determinants. These included: think aloud interviews, experiential focus groups, and PPI. Face-to-face, in depth interviews were conducted with 13 patients. Think aloud protocol was used to assess reliability of the mechanisms of change (i.e., link between the theoretical determinants and the BCTs), and usability of the delivery mode. All interviews were conducted at participants' places. Field notes and think aloud data informed the analysis, and results refined the tailoring of the theoretical determinants [25].
Face-to-face meetings were conducted with two PPI members recruited using emails One experiential focus group with six PPI members aimed to obtain views and recommendations on the intervention content and delivery mode. PPI were recruited using emails from the NIHR Cambridge BRC PPI/E office. PPI members were eligible to take part if they were prescribed medications for a long-term health condition and were over 40 years old.
The data generated from the experiential focus groups informed the content of the message file and the process to map the intervention messages onto the theoretical determinants. Information about the recruitment setting and patients' characteristics at baseline, as well as uptake and retention rates were extracted from the two systematic reviews of trials on medication adherence delivered by SMS and/or IVR messages [11,12].  about the implementation procedures. The data generated from stakeholders' consultations informed the study and intervention implementation procedures.
Step 3. Develop the Prototype Intervention.
A prototype intervention was developed by multiple and iterative syntheses of the data obtained during the intervention development [27]. The intervention is highly tailored to deliver very brief (≥1 minute), theory-based messages at participant's preferred time, telephone and frequency to receive the calls. The pilot intervention consisted of 29 messages, delivered daily for a duration of one prescription-based month. Two messages were delivered at the first day of the intervention, with feedback tailored to each participant's responses at the baseline questionnaire and prescription plan, and one message per day during the following 27 days of the intervention.
Step 4. Refine the Intervention.
Two PPI members, recruited by the NIHR Cambridge BRC PPI/E office, pre-tested the prototype intervention for one prescription-based month. PPI provided their experiential feedback about the intervention usability and mechanisms of change during four, weekly telephone-based interviews with a researcher, and completed the follow up questionnaire.
The data informed the decision rules of the tailoring algorithm and refined the frequency (i.e., message schedule), combination and sequence of the BCTs (i.e., integration of theory into the intervention), before piloting the prototype intervention within primary care.

Intervention Piloting
A one-month pilot intervention was conducted to assess (a) implementation procedures, (b) uptake, retention, fidelity, and engagement with the intervention, (c) participants' views, understanding, and actions upon the intervention content; and obtain (d) recommendations for improvement.

Patients and Public Involvement
Patients and Public Involvement (PPI) and Engagement (PPE) was conducted to inform intervention development and piloting (see sections above).

Analysis
Data generated from the intervention development and pilot study were integrated into one mixed methods analysis [27], using the technique called "following a thread" [28,29].
The theoretical underpinnings of the intervention were explored across different methods of data collections, data generated from each method informed each intervention component (i.e., intervention content and implementation procedures) iteratively, and questions or contradictory findings generated were followed across the other methods until saturation was achieved.

Results
A summary of the outcomes of the mixed methods analysis are reported in table 1, and explained in more details at this section

Message File
Participants suggested content that informed the intervention BCTs, so that we included more BCTs, such as information about social and environmental consequences.

Targeted Behaviour and Participants' Characteristics
Patients reported that it was easier for them to report on the pills taken per day, rather than the specific type of tablets (e.g., felodipine, statins). However, they reported that this information could be useful for patients who initiate medication taking, and need more education about their prescribed regimens. Patients reported that the timing of medicationtaking or the routine associated with the timing (e.g., lunch), rather than the type of medications to take, was an easier reminder of behaviour. That was particularly important for those patients taking medication for longer and for more than one health conditions.

Implementation Procedures within Primary Care
Health care providers that conducted the recruitment procedures reported challenges with recruiting non-adherent patients and recommended recruitment strategies delivered at times when patients might be more receptive to uptake such an intervention. In accordance to the systematic review evidence, it was recommended that patients would be more receptive to uptake the intervention when they initiate medication taking or when there is a change to their current prescription plan (e.g., this could be either due to a health event and/or due change to the pharmaceutical product). Otherwise, patients could be reluctant to provide consent and uptake the intervention (table 2). Annual reviews: "annual review is a bit more time … when its annual reviews because we're actually talking about their tablets" practice nurse recommendation about recruitment procedures Practice website: "but you can put a facility on the online forum. A lot of people access our website, so there are quite a lot of other places that people might look maybe more than a poster" practice nurse Medication prescriptions: "but every time you get a repeat prescription list, you can write something on it" practice nurse Note. N=8 health care providers: nurses and health care assistants. Themes have been coded at interviews with health care providers at development (n=5) and piloting of the intervention (n=3). Quotes reported in table are from experiential interviews with health care providers (T2). Health care providers recommended that recruitment methods could be integrated to medication annual reviews or to patients' prescriptions. They also recommended that nonadherent patients, who do not attend the practice for their medication reviews, either because they do not appreciate the need of these or they are sceptical about the recommended treatment, could be more receptive to the intervention if invited by other methods, such as text messages, newsletters, or multimedia (table 2).
However, health care providers also revealed that time constrains prevented them from explaining the implementation procedures to eligible patients during primary care consultations. In line with data from the PPI consultations and patients' interviews, health care providers suggested that there is a lot of information in the recruitment material and suggested that more emphasis is needed to patient's consent process. It was suggested that patients should be made aware about what information will be collected by their practice records, and how these will be utilised from the intervention. To facilitate patients' understanding and standardise the implementation procedures, including the informed consent, we developed videos aligned to written material, and integrated these into a brief face-to-face recruitment appointment.
To reinforce the uptake and integration of the intervention into current NHS, commissioners, patients, and health care providers reported that intervention needs to be of low or no cost. "Would need to be very low cost or free to use" commissioner consultation.

Intervention Uptake and Retention
Twenty patients were recruited. Eighteen provided written informed consent and registered into the study, of whom 17 completed the one-month intervention: one patient was excluded by the research team before registration because he/she met one of the exclusion criteria, one patient withdrew before registration to the intervention, and the other participant dropped out during the intervention. All 17 patients completed the measures at baseline and at the end of the intervention. All primary care practices provided completed baseline (e.g., refill prescription data three months before the start of the intervention) and follow up data, for all participants. During the one-month intervention an average of 29 messages were scheduled, of which 22.52 were received by patients. On average, 37 calls were made, of which 2.8 failed to be made due to technical issues and 13 calls were made but failed to go through because they were not picked up by the participants. On average, three calls were repeated and received by participants. Ten out of the 17 participants had chosen the calls to be repeated if not answered, and five out of the 17 participants had provided a secondary number for the calls to be repeated, if there was no answered at the primary number. Overall, patients were engaged with the intervention and made inbound calls to change the schedule of the messages or report about intervention content. In summary, there was a good fidelity of the intervention BCTs, with the majority of the BCTs having high fidelity scores (table 3).

Understanding and Actions upon Intervention Content
Participants reported that the intervention content increased their awareness about medication adherence and the risk and benefits of maintaining adherent to prescribed medications for long term, reinforced social support and habit formation, and reminded them to take medications as prescribed (table 4, Appendix 7).  Participants suggested that the intervention was acceptable and easy to use, and they provided their views about specific features of the intervention delivery mode and content.
Specifically, the tailored schedule of the messages, the personalisation, and the variation of the content was found to be particularly appealing and were perceived to promote engagement with the intervention (Table 5, Appendix 7). Participants reported being satisfied with the overall experience with the intervention, the availability of the intervention 24/7, and their ability to call in and leave a message (see Appendix 8). They would also recommend the intervention to other people who take prescribed medications (see Appendix 7).

Principal Findings
A highly tailored intervention has been developed to support adherence to medication prescribed for high blood pressure and co-morbidities. The intervention has been systematically developed based on guidance from MRC framework, PPI/E and is based on norms, perceived regimen complexity) and prescription plan. The the pilot study suggested that the intervention content increased awareness about the necessity of medication adherence, the risk and benefits of maintaining adherent to prescribed medications, reinforced social support and habit formation, and reminded patients to take medications as prescribed. To our knowledge, this is the first medication adherence intervention delivered using automated text and voice messages that has been developed and piloted within the UK primary care setting. The intervention has been proven to be acceptable and feasible to older adults with high blood pressure or co-morbidities, and it is currently being evaluated in a randomised controlled trial.

Strengths and Limitations of this study
To date there are studies describing the systematic development of mHealth interventions, however none of them has provided a detailed description based on guidance from the MRC framework, and none has described the methods used to develop and pilot the implementation procedures within the primary care setting. Furthermore, this study has involved PPI/E in the intervention development, design and pilot, adapting a collaborative approach, which minimized the influence of the social desirability effect on the findings, and thus it provided a real-world perspective. As a consequence, the results of this intervention may be better informed and matched to patients' healthcare needs.
A strength of this study is the systematic approach to intervention development and piloting. The study used a rigorous theoretical approach to guide the development of the tailored intervention content. It also designed a flexible and scalable application to deliver messages to support adherence to different type, and complex prescribed medications.
Another strength of this study is the utilisation of a mixed (e.g., qualitative, quantitative) and multi-perspective (patients, health care providers, PPI, stakeholders) methods of data collection. The data were combined complementarily into one comprehensive analysis, using a rigorous analytical approach.
Although this intervention is highly tailored, and participants can provide real time input and further tailor intervention content and delivery, a limitation of the intervention is the show levels of adherence) [30], and future studies could usefully explore whether and how these features could influence intervention engagement and behaviour change.

Future Research
The one-month intervention has been extended to three months intervention and targets all adults (above 18 years old) to increase reach. It also includes another level of tailoring, where participants select between two BCTs to support maintenance of medication taking (e.g., habit formation or self-monitoring). The intervention is currently being evaluated in an RCT, which includes objective behavioural (medication event monitoring system) and clinical   Mark is a 49 year old car mechanic. He was diagnosed with high blood pressure by his GP 6 months ago even though he had no symptoms, was not overweight, did not smoke, ate healthily and had an active job. Initially, he took the medication exactly as his GP prescribed every day. However, after a few weeks he began to experience constipation and read on the leaflet supplied with the tablets that this was a side effect caused by the medication. Now, he only takes his medication every other day and the constipation has cleared up. However, his blood pressure is still too high.

Message Event
Keep on the tablets prescribed to you by your Doctor. High blood pressure is more serious than constipation. HOP This is your message for your blood pressure tablets. One easy way to remember your tablets, is to take them with another daily activity, such as your morning cup of tea. If they are always done together, it will be harder to forget.  Beliefs about medications,  self-efficacy to take medications when no health benefits and  emotional state

Habit formation and information about emotional consequences
Please keep taking your tablets as prescribed even if you are well and feeling healthy. High blood pressure is one of those things that unless you actually feel it you're not aware that it is a problem. Taking your medications as prescribed will support you to keep enjoying things or activities that are important to you.
NINA,  beliefs about medications and  emotional state Habit formation and social support (unspecified) Whatever the day may holds, please do not forget to take your tablets. To achieve better control over high blood pressure, you will need to take them every day.
 self-efficacy to cope with  practical social support Habit formation and social support (unspecified) We are calling to see how you are doing with taking your tablets today. Please remember that you can always call the free telephone number if you require more support on how to take your tablets as prescribed. If you require more information about the health condition or the dosage of your medications, please contact your GP practice or pharmacist.  social norms and  selfefficacy to cope with  practical social support Habit formation and social support (unspecified) You may find it helpful to ask people who are important to you to remind you to take your medications as prescribed and maintain your blood pressure under control. Please remember that you can always call the free telephone number if you require more support on how to take your tablets as prescribed.
Conclusion: tailored automated text and voice message interventions are feasible ways to increase medication adherence, as an adjunct to primary care.
Adherence is defined as taking at least 80% of the prescribed tablets, and/or having medications dispensed for at least 95% days of a prescribed period [4]. However, substantial proportions of patients do not take their medication as prescribed [5]. A recent meta-analysis showed that 40% of people do not adhere adequately to cardiovascular medication and the percentage was similar (41%) for non-adherence to anti-hypertensive medications [6]. Taking into consideration the growing prevalence of HBP, co-morbidities and the ageing population [7], especially in lower and middle-income countries [8], it is likely that there will be an increase in demand for health care resources to support medication adherence. The National Institute of Health and Care Excellence (NICE) recommends the development of novel and cost-effective interventions to aid patients' decision making about taking medicines, as an adjunct to health care providers' consultations [9,10].
Tailored mHealth interventions, such as text (SMS) and voice (IVR) messaging intervention, is one way to support patients' adherence between their consultations. Review evidence found that such interventions can effectively support adherence to different types of medication, including antihypertensive tablets, when compared to usual care only [11,12], and can potentially be cost-effective if applied to a large number of people [13]. For example, a recent study suggested that such interventions could result in a 2.3% increase in QALYs and overall savings of 21,120 Australian dollars during a patient's lifetime [14]. However, no such intervention has been developed and piloted in the UK primary care setting.
The aim of this paper is to describe the development and piloting of a highly tailored text and voice messaging intervention to support adherence to antihypertensive tablets. Although several guidelines have been proposed to the development of mHealth interventions [15][16][17]

Methods
Review of theory and evidence, interviews, focus groups, consultations and Patients and Public (PPI) involvement and Engagement (PPE) informed the intervention development process (see Appendix 1); and a one-month pre-post intervention study, with follow up assessment informed the piloting of the intervention (see Appendix 2).

Intervention Development
Step 1. Systematic Development of the Tailored Intervention.  Integrate theory to intervention. The frequency, combination and sequence of the BCTs were integrated into one intervention. The decision on the above elements was based on the phase participants are in (e.g., uptake or maintain), having as the starting point the collection of the medication from the dispensary.

Step 2. Development of the Intervention Delivery Mode, Content, and Implementation
Procedures.  studies. Data generated from PPI/E informed the content of the message file (for example, see Appendix 4). Overall, PPI/E members found more acceptable NINA case studies and seemed to generate more messages for NINA rather than INA case studies.

Pre-test feasibility of intervention delivery mode and theoretical determinants.
Four studies informed the acceptability of the delivery mode and the theoretical determinants.
These included: think aloud interviews, experiential focus groups, and PPI.
Face-to-face, in depth interviews were conducted with 13 patients. Think aloud protocol was used to assess reliability of the mechanisms of change (i.e., link between the theoretical determinants and the BCTs), and usability of the delivery mode. All interviews were conducted at participants' places. Field notes and think aloud data informed the analysis, and results refined the tailoring of the theoretical determinants [25].
Face-to-face meetings were conducted with two PPI members recruited using emails from the list of the National Institute of Health Research (NIHR) Cambridge BRC PPI/E Panel and Communications office. Think aloud protocol was used to assess the reliability and sequence of the items to form the measurement of the theoretical determinants (i.e., theorybased questionnaire). Both interviews were conducted at the University. Field notes and think aloud data informed the analysis, and results refined the reliability of the tailoring questionnaire and informed the decision rules of the tailoring algorithm.
Experiential focus groups aimed to obtain patients' views of, and refine, the tailored intervention content (e.g., inclusion of medication name, dose and timings; 'please do not forget to take Ramipril 1 tablet 10 mg at 9am') and delivery mode within a social interaction context [26]. Eligibility criteria and recruitment procedure were similar to patients' face-toface interviews, reported above. Four focus groups took place in primary care practices (n=2) or community centres (n=2). During the focus groups patients (n=12) received examples of SMS and IVR messages on their mobile phones and were asked to provide experiential feedback. Patients were asked their views and recommendations about the combination of the delivery modes and the intervention content, using a think-aloud protocol [27].
One experiential focus group with six PPI members aimed to obtain views and recommendations on the intervention content and delivery mode. PPI were recruited using emails from the NIHR Cambridge BRC PPI/E office. PPI members were eligible to take part The data generated from the experiential focus groups informed the content of the message file and the process to map the intervention messages onto the theoretical determinants.
Develop implementation procedures within primary care. The intervention implementation procedures were informed by systematic review evidence, PPI/E, and stakeholders' consultations.
Information about the recruitment setting and patients' characteristics at baseline, as well as uptake and retention rates were extracted from the two systematic reviews of trials on medication adherence delivered by SMS and/or IVR messages [11,12]. Due to missing data about the characteristics of the targeted population, the recruitment setting, recruitment methods and material, we synthesised the findings narratively. Results informed the decisions regarding the recruitment setting and methods, as well as the targeted behaviour and population of the intervention.
Twelve PPI members provided feedback on the recruitment material (e.g., patients' information sheets, posters, leaflets, flyers, and invitation letters) using emails or face-to-face meetings. Their comments informed the content of the recruitment material and the description of the recruitment, informed consent, and intervention procedures in lay language.
Stakeholders' consultations were conducted with health care providers, commissioners, and patients to inform the study design and the implementation procedures within the primary care setting (e.g., recruitment, informed consent, data collection). Practitioners with experience of advising patients for medications; commissioners for medicine optimisation, prescribing or cardiovascular conditions; and patients with either HBP or both HPB and T2DM were included. Commissioners were recruited informally by email through personal networks.
Patients and health care providers (i.e., practice nurses, health care assistants) were recruited through the primary care practices. Those who consented to take part were invited to the consultations. Consultation modes were flexible to maximise participation (n=3 face-to-face, n= 4 email). Stakeholders (N=7) were given a description of the proposed intervention and asked their views and recommendations. Health care providers (n=4) were consulted about the time to initiate recruitment, and the recruitment methods and procedures. Commissioners (n=1) were consulted about how best to track resource use and costs and the evidence needed to inform whether or not to commission such an intervention. Patients (n= 2) were consulted about the delivery mode and the theoretical framework. All stakeholders were asked for their consultations informed the study and intervention implementation procedures.

Step 3. Develop the Prototype Intervention.
A prototype intervention was developed by multiple and iterative syntheses of the data obtained during the intervention development. The intervention is highly tailored to deliver very brief (1 minute), theory-based messages at participant's preferred time, telephone and frequency to receive the calls. The pilot intervention consisted of 29 messages, delivered daily for a duration of one prescription-based month. Two messages were delivered at the first day of the intervention, with feedback tailored to each participant's responses at the baseline questionnaire and prescription plan, and one message per day during the following 27 days of the intervention.
Step 4. Refine the Intervention.
Two PPI members, recruited by the NIHR Cambridge BRC PPI/E office, pre-tested the prototype intervention for one prescription-based month. PPI provided their experiential feedback about the intervention usability and mechanisms of change during four, weekly telephone-based interviews with a researcher, and completed the follow up questionnaire. The data informed the decision rules of the tailoring algorithm and refined the frequency (i.e., message schedule), combination and sequence of the BCTs (i.e., integration of theory into the intervention), before piloting the prototype intervention within primary care.

Intervention Piloting
A one-month pilot intervention was conducted to assess (a) implementation procedures, (b) uptake, retention, fidelity, and engagement with the intervention, (c) participants' views, understanding, and actions upon the intervention content; and obtain (d) recommendations for improvement.
Patients with HBP were recruited by primary care practices within the Cambridgeshire and Peterborough CCG, and health care providers who consulted patients for taking medications.

Analysis
Data generated from the intervention development and pilot study were integrated into one mixed methods analysis [28], using the technique called "following a thread" [29,30]. The theoretical underpinnings of the intervention were explored across different methods of data collections, data generated from each method informed each intervention component (i.e., intervention content and implementation procedures) iteratively, and questions or contradictory findings generated were followed across the other methods until saturation was achieved.
Patient and Pubic Involvement. PPI/E was included in all parts of intervention development and piloting.

Results
A summary of the outcomes of the mixed methods analysis are reported in table 1, and explained in more details at this section

Message File
Participants suggested content that informed the intervention BCTs, so that we included more BCTs, such as information about social and environmental consequences.
"Do you, sort of, remind them of the costs and things? Is that in with medication and things, if they're missing it, or every time they lose blood pressure medication or throw it in the bin, what is the cost? " Practitioner consultation

Schedule
The one-month intervention schedule included daily messages and participants could change the frequency or stop them. Patients at follow up interviews, recommended an intervention of longer duration with messages to gradually decrease in frequency and include advice and support for self-monitoring medication taking.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  Patients reported that it was easier for them to report on the pills taken per day, rather than the specific type of tablets (e.g., felodipine, statins). However, they reported that this information could be useful for patients who initiate medication taking, and need more education about their prescribed regimens. Patients reported that the timing of medication-taking or the routine associated with the timing (e.g., lunch), rather than the type of medications to take, was an easier reminder of behaviour. That was particularly important for those patients taking medication for longer and for more than one health conditions. Q: "How many tablets do you take per day?" R: "Oh, about 17". Q: "How many different types of medication do you take per days?" R: "Yeah. That takes more counting and a bit more thinking ... I suppose, as well, if you have so many, you just get used to the times…Morning, lunch, evening, night time" patient Patients who took part in the intervention development studies and the pilot study were recruited from different areas of IMD, were older adults, had either HBP or co-morbidities (see appendix 6, for participants' demographic characteristics), and self-reported being prescribed complex medication regimens (number of tablets per day: mean=6, sd=4.6).

Implementation Procedures within Primary Care
Health care providers that conducted the recruitment procedures reported challenges with recruiting non-adherent patients and recommended recruitment strategies delivered at times when patients might be more receptive to uptake such an intervention. In accordance to the systematic review evidence, it was recommended that patients would be more receptive to uptake the intervention when they initiate medication taking or when there is a change to their current prescription plan (e.g., this could be either due to a health event and/or due change to the pharmaceutical product). Otherwise, patients could be reluctant to provide consent and uptake the intervention (table 2).  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  Annual reviews: "annual review is a bit more time … when its annual reviews because we're actually talking about their tablets" practice nurse recommendation about recruitment procedures Practice website: "but you can put a facility on the online forum. A lot of people access our website, so there are quite a lot of other places that people might look maybe more than a poster" practice nurse Medication prescriptions: "but every time you get a repeat prescription list, you can write something on it" practice nurse Note. N=8 health care providers: nurses and health care assistants. Themes have been coded at interviews with health care providers at development (n=5) and piloting of the intervention (n=3). Quotes reported in table are from experiential interviews with health care providers (T2).
Health care providers recommended that recruitment methods could be integrated to annual reviews or medication reviews. They also recommended that non-adherent patients, who do not attend the practice for their annual or medication reviews, either because they do not appreciate the need of these or they are sceptical about the recommended treatment, could be more receptive to the intervention if invited by other methods; like notes into their prescriptions, text messages, newsletters, or multimedia (table 2).
However, health care providers also revealed that time constrains prevented them from explaining the implementation procedures to eligible patients during primary care consultations. It also prevented them from facilitating patients' motivation to medication adherence and uptake to the digital intervention.
In line with data from the PPI consultations and patients' interviews, health care providers suggested that there is a lot of information in the recruitment material and suggested that more emphasis is needed to patient's consent process. It was suggested that patients should be made aware about what information will be collected by their practice records, and how these will be utilised from the intervention. To facilitate patients' understanding and standardise the implementation procedures, including the informed consent, we developed videos aligned to written material. The content of the videos aimed to also provide normative information about taking medication and medication non-adherence, and increase motivation to initiate behaviour change. The videos were integrated into the study invitation material (e.g., text message invitations to eligible patients) and brief consultations with the health care providers.
To reinforce the uptake and integration of the intervention into current NHS, commissioners, patients, and health care providers reported that intervention needs to be of low or no cost.

Intervention Uptake and Retention
Twenty patients were recruited. Eighteen provided written informed consent and registered into the study, of whom 17 completed the one-month intervention: one patient was excluded by the research team before registration because he/she met one of the exclusion criteria, one patient withdrew before registration to the intervention, and the other participant dropped out during the intervention. All 17 patients completed the measures at baseline and at the end of the intervention. All primary care practices provided completed baseline (e.g., refill prescription data three months before the start of the intervention) and follow up data, for all participants.

Intervention Fidelity and Engagement
During the one-month intervention an average of 29 messages were scheduled, of which 22.52 were received by patients. On average, 37 calls were made, of which 2.8 failed to be made due to technical issues and 13 calls were made but failed to go through because they were not picked up by the participants. On average, three calls were repeated and received by participants. Ten out of the 17 participants had chosen the calls to be repeated if not answered, and five out of the 17 participants had provided a secondary number for the calls to be repeated, if there was no answered at the primary number. Overall, patients were engaged with the intervention and made inbound calls to change the schedule of the messages or report about intervention content. In summary, there was a good fidelity of the intervention BCTs, with the majority of the BCTs having high fidelity scores (table 3).   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59   Note: Data from patients' (n=17) face-to-face interviews at the end of the one-month pilot intervention (T2).

Understanding and Actions upon Intervention Content
Participants also recommended that some messages could be better received if delivered by either IVR or SMS, depending on the timing and the content of the messages. For example, they recommended text messages at the time of medication taking with simple reminders of the behaviour to reinforce habit formation; whereas they recommended voice messages at a time of their preference with information about how to set up a routine with medication taking.
They also suggested messages that target patients' INA beliefs and coping plans, to be delivered less frequently than reminders of the behaviour, as they were more personal and therefore easier to be perceived as invasive.
"So maybe a text message … 'cause you're waiting for the phone and you've got to do something, which, you know, I think is a good thing, in that you've got to do something and you have to respond. You can't just put your phone in your pocket and walk away, but at night time, I think it would probably be a bit intrusive" patient 01062T, follow up interview.

Principal Findings
A highly tailored intervention has been developed to support adherence to medication prescribed for high blood pressure and co-morbidities. The intervention has been systematically developed based on guidance from MRC framework, PPI/E and is based on rigorous theory and evidence. It includes a combination of Behavioural Change Techniques (BCTs) tailored to patients' beliefs (e.g., beliefs about medications, self-efficacy, social norms, perceived regimen complexity) and prescription plan. The pilot study suggested that the intervention content increased awareness about the necessity of medication adherence, the risk and benefits of maintaining adherence to prescribed medications, reinforced social support and habit formation, and reminded patients to take medications as prescribed. To our knowledge, this is the first medication adherence intervention delivered using automated text and voice messages that has been developed and piloted within the UK primary care setting. The intervention has been proven to be acceptable and feasible to adults with high blood pressure or co-morbidities, and it is currently being evaluated in a randomised controlled trial.

Strengths and Limitations of this study
To date there are studies describing the systematic development of mHealth interventions, however none of them has provided a detailed description based on guidance from the MRC framework, and none has described the methods used to develop and pilot the implementation procedures within the primary care setting. Furthermore, this study has involved PPI/E in the intervention development, design and pilot, adapting a collaborative approach, which minimized the influence of the social desirability effect on the findings, and thus it provided a real-world perspective. As a consequence, the results of this intervention may be better informed and matched to patients' healthcare needs.
A strength of this study is the systematic approach to intervention development and piloting.
The study used a rigorous theoretical approach to guide the development of the tailored intervention content. It also designed a flexible and scalable application to deliver messages to support adherence to different type, and complex prescribed medications. Another strength of this study is the utilisation of a mixed (e.g., qualitative, quantitative) and multi-perspective  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n l y (patients, health care providers, PPI, stakeholders) methods of data collection. The data were combined complementarily into one comprehensive analysis, using a rigorous analytical approach.
Although this intervention is highly tailored, and participants can provide real time input and further tailor intervention content and delivery, a limitation of the intervention is the selfreported theoretical determinants of medication adherence. Future studies could enable the use of objectively measured theoretical determinants and objectively measured behaviour, using sensing technology. Another constrain of the text and voice messaging is its limited features to facilitate delivery of BCTs visually (e.g., picture of tablets, graphs to show levels of adherence) [31], and future studies could usefully explore whether and how these features could influence intervention engagement and behaviour change.