A mixed‐methods exploration of non‐attendance at diabetes appointments using peer researchers

Abstract Background Non‐attendance at diabetes appointments is costly to the health service and linked with poorer patient outcomes. Objective Peer researchers aimed to conduct interviews and survey people who miss appointments about their beliefs and perceptions regarding their diabetes and diabetes appointments. Design A mixed‐methods cross‐sectional design with interviews conducted by peer researchers with diabetes and a questionnaire was used. Setting and participants Peer researchers conducted semi‐structured telephone interviews in one health board in Scotland with ten people who had missed diabetes appointments. A further 34 people who had missed appointments completed a questionnaire. The study was informed by two psychological theories (the Theory of Planned Behaviour and the Self‐Regulation Model), and interviews were analysed using thematic analysis. Results Interviewees planned to attend appointments but practical barriers, low perceived value of appointments and the feeling that diabetes had little impact upon their lives’ emerged as key reasons for missing appointments. Questionnaire data supported these findings and showed that respondents perceived diabetes to have only mildly serious consequence and cause limited concern and emotional impact. Participants’ understanding of their condition and perceptions of personal control and treatment control were low. Gender, perceived behavioural control and emotional representations were significantly associated with the number of appointments missed in the previous year. Conclusions These findings highlight the importance of psychological variables in predicting non‐attendance at diabetes appointments and provide avenues for how non‐attendance might be tackled.


| BACKG ROU N D
Many people who have diabetes do not regularly attend their diabetes-related appointments. Estimates of the prevalence of missed diabetes vary but two large studies in the United States reported that 16.2% of people with diabetes missed their last primary care appointment 1 and 12% of people missed more than 30% of schedules primary care appointments in 1 year. 2 Non-attendance at scheduled appointments has consequences both for the health service and for the person with diabetes. Non-attendance increases the cost of delivering care, reduces available appointments and increases waiting times for other patients. 3 People with diabetes who do not attend clinic appointments tend to have poorer glycaemic control, more complications, more frequent hospital admissions and increased allcause mortality. [1][2][3][4][5] Previous research on the topic of non-attendance at diabetes appointments has generally focused on demographic and clinical factors such as age and gender. If non-attendance is to be tackled, it is important to seek the opinions of people with diabetes themselves and to investigate psychosocial factors such as beliefs and attitudes that are potentially amenable to change. 6 Patients have historically played a passive role in health research, but the benefits of a more active role are increasingly being recognized and patient involvement in health-care research is recommended in health-care policy in the UK. 7 Patients can be involved in a range of activities from the design of research to conducting research and presenting the findings, and there is evidence to suggest that patient involvement can help to improve the credibility and relevance of research. 8,9 A systematic review of the impact of patient involvement in research noted that there were benefits at all stages of the research process including increased enrolment rates and development of research questions, questionnaires and interview schedules that were more user focused. 9 To our knowledge, there has been no previous research investigating non-attendance at diabetes appointments that has involved patients who have diabetes in the research process. Training people with diabetes to survey peers who have failed to attend appointments could generate novel and valuable findings because of the shared experience of those carrying out the work and those being surveyed.
Two psychological models of illness/health behaviour have been widely used in research on long-term conditions generally, and specifically to understand non-attendance in diabetes appointments 6 : the Theory of Planned Behaviour (TPB 10 ) and the Self-Regulation Model of Illness Behaviour (SRM 11 ). The TPB states that voluntary behaviours are largely predicted by our intentions regarding the behaviour. Intentions in turn are determined by our attitude towards the behaviour (our judgement of whether the behaviour is a good thing to do), subjective norms (our judgement of what important others think of the behaviour) and perceived behavioural control (PBC; our expectation of how successful we will be in carrying out the behaviour 10 ).
The SRM proposes that people interpret information about a potential illness to create a 'lay' view or representation of the illness. The coping responses employed by an individual, for example, adhering to treatment regimens and attending appointments, are said to be related to the illness representations they hold and to their appraisal of how successful they perceive the chosen coping responses to be. Illness representations are proposed to be formed around six different themes: identity (label or diagnosis of illness), cause (factors believed to have caused the illness), timeline (expected duration of illness), consequences (expected effects of illness on physical, social and psychological wellbeing), control/cure (extent to which illness can be controlled/cured) and illness coherence (how well the person understands their illness 11 ).
Both of these models have been previously been applied to understand non-attendance at diabetes appointments and could therefore provide a potentially valuable framework for a peer-led investigation of this issue. 6 The SRM and TPB offer different approaches to understanding attendance at diabetes clinic appointments with the SRM focusing on patients' beliefs about diabetes, whereas the TPB is concerned with beliefs about the actual act of attending appointments. As there is no clear evidence to suggest which approach might be most appropriate, both models were used to inform the present study. The aim of the study was to conduct peer-led interviews and surveys with people who miss diabetes appointments to assess their beliefs and perceptions about their diabetes and attendance at diabetes appointments.

| Study design
This study used a mixed-methods cross-sectional design to explore non-attendance at diabetes appointments in one health board in Scotland. People with diabetes (peer researchers) were involved in the conception and design of the study and were trained to carry out semi-structured telephone interviews in the first phase of the study. The findings of these interviews informed the content of a questionnaire in the second phase of the study. It was originally intended that the first author would carry out telephone interviews to provide a comparator to the peer researcher interviews but a very low response rate meant this was not possible. Ethical approval was obtained for this study from the West of Scotland NHS Research Ethics Committee 4 (reference number: 13/WS/0177). Participants in the telephone interviews were sent information sheets and provided informed consent verbally. Information sheets were sent out with the questionnaires, and completion of the questionnaire was taken to imply consent. The authors elect to not share data.

| Peer researcher training
Four people with diabetes who were interested in being peer researchers were identified from an existing diabetes patient group and trained to conduct semi-structured telephone interviews but only three of peer researchers were available at the time of data collection to conduct interviews. Two of the peer researchers were female, one was aged between 50 and 60, two were aged over 60, two peer researchers had type 1 diabetes, and one had type 2 diabetes.
Training for peer researchers was developed and delivered by the first author, who at the time of the study was a Registered Health Psychologist and PhD candidate. The training was based on training delivered by colleagues of the first author for a previous research project using peer researchers and on resources from the INVOLVE website (INVOLVE is a national advisory group that supports greater public involvement in health and social care research). Training included content on confidentiality, ethics, interviewing techniques, reflexivity and dealing with distressed participants and provided opportunities for role play to give peer researchers practical experience of these topics (See Table 1 for a full outline of topics covered in training). In total, two evenings and one full day of training were provided, and peer researchers were informed they could request additional training at any point as required, either individually or as a group. Both authors independently assessed the peer researchers' interview skills during role play, and all four met the required standard for the agreed criteria. The first author telephoned peer interviewers before and after they carried out their first interview to check they were comfortable with the interview process and to reflect on their experience of conducting it. Contact after the first interview was made according to the preferences of the peer researcher.

| Participants and recruitment
People with diabetes who had previously failed to attend diabetes clinic appointments were recruited from a single health board in Scotland, UK. English-speaking patients, aged over 18, with type 1 or type 2 diabetes, were eligible for inclusion if they had missed three or more diabetes clinic appointments in the previous 24 months (either in hospital or at their general practice). Twenty General Practices and two secondary care diabetes clinics agreed to identify patients for this study, although not all practices claimed the small remuneration offered to cover the administrative cost of this, so fewer than 20 may have managed to distribute the information. An initial contact letter was drafted for the study and sent out by a member of the existing clinical care team, along with an information sheet about the study. Patients selected to take part in the telephone interviews were asked to return a slip to indicate that they were interested and agreeing to their details being passed on to the research team. Patients selected to take part in the questionnaire part of the study were sent a paper copy of the questionnaire along with an initial contact letter. Different participants took part in the telephone interviews and questionnaire.  Table 2 gives example questions for each topic and shows which constructs from the theoretical framework that these questions relate to. Interviews lasted between 8 and 27 minutes and were audio-recorded with the participant's permission and transcribed verbatim by a professional transcription service. Peer researchers were offered, but chose not to accept any remuneration for their involvement in the study.

| Telephone interviews
Only a few of the practices involved reported how many people they contacted with information about taking part in a telephone interview, but extrapolating from those who did, we estimate that 200 received this request. A total of 14 people completed a return slip to indicate they were willing to be contacted. The peer researchers were able to make contact and interview 10 of the 14 potential participants but problems with recording meant that verbatim transcripts were only used in the analysis for 7 interviews and field notes for the remaining 3. Table 3 lists the participants in the telephone interviews and their characteristics.

| Questionnaire
Questionnaires were sent to patients from June to September 2016 and again from June to November 2017 to boost the response rate. This guide recommends conducting an 'elicitation study' to inform development of the TPB measure. An elicitation study is a qualitative investigation that aims to establish the most salient beliefs about a particular behaviour in the population in question. In this study, the telephone interviews were used to form an elicitation study to identify attitudes and barriers to attendance at diabetes clinic appointments. The most salient attitudes and barriers identified in the interviews were developed into questions to be included in the questionnaire. The format of the TPB questions meant that reliability analysis was not appropriate. 13 Full details of how TPB components were measured in the questionnaire are outlined in Table 4.

| Data analysis
A thematic analysis of the interview data was conducted independently by the first author. 14 Thematic analysis involves the identification of themes in interview transcripts. This is achieved through reading and re-reading the data to become familiar with the content.
Regular recurring experiences and feelings described by participants are manually identified and then formed into themes which give an overall view of the way that participants feel about the service. This method was chosen as it organizes and minimizes data whilst retaining detail. Although no independent analysis of the data was carried out, a draft of the themes was shared and discussed with the peer researchers to check the interpretations made by the first author.
Data collected from the questionnaire were analysed using SPSS for windows version 25. The characteristics of the sample, and TPB and SRM components were analysed using descriptive statistics.
Comparisons of SRM and TPB components across demographic groups were carried out using Mann-Whitney tests. This non-parametric alternative to the independent t test was appropriate with the small sample size and violations of normality found in the data.
Mann-Whitney tests were carried out for each demographic variable (gender, age and type of diabetes) with the demographic variable entered as the independent variable and the SRM and TPB components entered as the dependent variables. Poisson regression analysis was conducted to assess whether TPB and SRM components and demographic variables predicted self-reported attendance at diabetes clinic appointments.

| Telephone interviews
Three reasons for non-attendance at diabetes appointments emerged from the interviews conducted by peer researchers: two of which were explicitly stated by participants and a third was implicit in the interview data. The two explicitly stated reasons for non-attendance were practical barriers to attending and low perceived value of attending. A perception of diabetes as having a limited impact on participants' lives emerged as a third implicit reason for not attending diabetes appointments. The findings regarding these reasons for non-attendance at diabetes appointments are discussed in detail below with supporting verbatim quotes from participants identified by participant number. Table 5 summarizes the main themes and subthemes from the telephone interviews.

| Practical barriers
Many participants mentioned that competing demands for their time, such as work or family commitments, made it difficult for them to attend appointments.
We've got quite a small team, and, um, and over the

1-7
Perceived behavioural control Control beliefs 1. It will be difficult to get transport to my diabetes clinic appointments (strongly disagree/strongly agree). 2. Diabetes clinic appointments are likely to be at a time of day that doesn't suit me (strongly disagree/strongly agree) 3. It will be difficult for me to remember to attend my appointment (strongly disagree/strongly agree).

1-7
Item scores were multiplied as follows: 1*4, 2*5, 3*6. Overall perceived behavioural control score was the sum of the resulting scores. Overall perceived behavioural control scores had a possible range of −63 to +63 with a negative score representing low perceived behavioural control and a positive score high perceived behavioural control for attending appointments.
Perceived power 4. When it is difficult to get transport to my diabetes clinic appointment I am less likely/more likely to attend. 5. When clinic appointments are at a time of day that doesn't suit me I am less likely/more likely to attend. 6. When clinic appointments are at a time of day that doesn't suit me I am less likely/more likely to attend. −3 to +3 Intention 1. I intend to attend all of my diabetes clinic appointments in the next year (strongly disagree/ strongly agree).

1-7
A score of one indicated low intention and a score of seven indicated high intention to attend diabetes appointments There's maybe so many appointments…within the space of…within the space of like maybe two or three months. I don't…sometimes I think there's too many.
There's too many appointments at the one time. (P1) For most participants, travelling to appointments was not a barrier to attending but some participants did experience difficulties particularly with those appointments that were further away from home.
Participant five talked about problems with her mobility and the difficulties she had in getting public transport to hospital appointments.

| Perceived impact of diabetes
Underlying participants' discussion about the reasons for non-attendance at appointments was a sense that their condition was not something that concerned them, and many participants had little to say about their condition which was reflected in the short length of some interviews. Although some participants stated that they did worry about their diabetes and mentioned complications that they had experienced, the majority of participants felt that it had very little effect on their day to day lives.
No. I don't really take it as an illness, to be quite honest with you…just carry on, you know what I mean.
It's just like you never had it, know what I mean, kind of thing.
There was a perception among participants with type 2 diabetes that their condition was controlled by the medication they took, rather than by self-management behaviours. This belief seemed to further reduce concern about their condition.

| Questionnaire
Of the 405 questionnaires distributed, 35 completed ones were returned, although one was excluded from the analysis as the partici- Mean scores for the TPB components across the whole sample outlined in Table 6 show that participants held weakly positive attitudes and perceived there to be moderately positive social pressure to attend diabetes appointments. Although intentions to attend appointments in the next year were high, perceived behavioural control was low. Mean overall illness perception scores indicated a perception of diabetes as neutral/weakly threatening. The mean scores for the individual dimensions of illness perceptions indicate that participants perceived diabetes to have only mildly serious consequences on their lives and cause mild concern and emotional impact. Participants' perceptions of how much personal control they had over their condition were low, and perceptions of how helpful treatment was for controlling their diabetes were lower still. Mean coherence scores show that participants felt they had a poor understanding of their condition, and mean timeline scores indicate that they viewed their condition as being long-lasting. The most common causes of their diabetes reported by participants with type 1 diabetes were genes, poor diet and immune system. The most commonly stated causes of type 2 diabetes were being overweight, poor diet, lack of exercise and stress/depression.
The results of the Mann-Whitney tests comparing TPB and SRM theory components by gender age and type of diabetes showed significant differences in some theory components (see Table 6 for full results). There were significant differences in coherence between men and women with men having lower median scores for both of these constructs meaning they felt they had less control over, and a poorer understanding of their condition. Significant differences were also found in coherence by type of diabetes with people with type 1 diabetes having lower median scores. Significant differences were found in subjective norm, perceived behavioural control and identity between younger and older participants. Younger participants had higher median scores for subjective norm suggesting they felt stronger social pressure to attend appointments but lower median perceived control suggesting that they felt less able to attend appointments. Median identity scores were higher for younger participants meaning that they experienced more symptoms than older participants.
Poisson regression was carried out to predict the number of missed appointments based on gender, type of diabetes, age, TPB components and SRM components. The coefficients, confidence intervals and significance levels for the variables entered in the Poisson regression can be found in Table 7. The overall regression model was significant (P < .01). Individual variables that were significantly associated with an increased risk of missing appointments included female gender, higher emotional representation score, higher attitude score and lower perceived behavioural control.

| D ISCUSS I ON
This study is the first, to our knowledge, to have people with diabetes conduct research to explore reasons for non-attendance at diabetes appointments and builds upon the small body of work exploring the influence of psychological variables on non-attendance.
Understanding factors that influence attendance at diabetes appointments that are potentially amenable to change is vital if non-attendance is to be tackled.
The interviews conducted by peer researchers revealed that although participants generally planned to attend their appointments, practical barriers to attending, low perceived value of appointments and the feeling that diabetes had little impact upon their lives emerged as key reasons for missing appointments. The data collected by the peer researchers accorded well with the data collected in the questionnaire phase of the study where respondents also reported that they intended to attend appointments but perceived themselves to have limited control and did not hold particularly positive attitudes towards attending their appointments. The questionnaire data provided further understanding of the perception highlighted in the interviews that diabetes had little impact on participants' lives.
Questionnaire respondents viewed diabetes as a long-lasting condition but perceived it to have only mildly serious consequences and cause limited concern and emotional impact. Participants' understanding of their condition and perceptions of personal control and treatment control were all low in the present study. Together, these findings highlight a number of practical barriers, beliefs and perceptions underlying non-attendance at clinic appointments that need to be addressed. Significant differences were found by gender, age and type of diabetes in some beliefs and perceptions suggesting that any intervention addressing these beliefs and perceptions may need to be tailored to the individual.
There is only limited previous research using psychological the- This unique finding of the present study suggests that psychological theory may be valuable in helping to understand why people with diabetes struggle to overcome some of the practical barriers to attending appointments.
In the present study gender, emotional representations, perceived behavioural control and attitude were significant predictors of the number of missed appointments in the regression. A recent systematic review of factors associated with attendance at diabetes appointments included no studies that assessed emotional representations, perceived behavioural control or attitudes. 18 The review reported men were found to be more likely to miss appointments in some studies but that gender was not associated with attendance in most of the included studies. 18 Gender was found to be a significant predictor of the number of missed diabetes appointments in the present study, but in the opposite direction to previous research with women being more likely to miss appointments than men.
However, a study of non-attendance at GP appointments found that women were more likely to miss appointments until the number of appointments made was controlled for, and then the opposite pattern was observed. 19   but none has been identified using those of the TPB. The findings of this study therefore need to be interpreted with caution. The low response rate also meant that comparator interviews could not be conducted by the first author as planned but impressions formed by the first author during analysis were that interviews conducted by peer researcher provided useful data on the topic with the shared experience between peer researcher and interviewee seeming to assist in rapport building. It was noted though that peer researchers sometimes did not probe or follow up interesting statements made by participants which resulted in shorter, less in-depth interviews than might have been achieved by a more experienced researcher. However, without comparator interviews these impressions cannot be confirmed and we cannot assess the effect of peer researchers on response rates.
People who do not engage with health care are known to be difficult to engage in research and often considered inaccessible. 15 Much of the research investigating non-attendance at healthcare appointments uses routinely collected health-care data 21,22 or includes both attenders and non-attenders 23 thus overcoming difficulties with recruitment. However, research using routinely collected health-care data is limited as it can only provide clinical and demographic information. Although the response rate was low and sample size small in the present study, the data collected provide an in-depth exploration of psychological factors that influence attendance that was led by people who have diabetes themselves, and focused only on those who miss appointments. As such, it is hugely valuable in helping to understand people who do not attend appointments and the steps that might need to be taken to tackle non-attendance.
Another limitation in the present study was that we asked participants to report the number of appointments they had missed as an absolute figure rather than a proportion. As non-attendance was self-reported, it was felt that it would be too difficult for people to calculate or recall the proportion of appointments that were missed.
Although there will have been differences between participants in the number of appointments they were due to attend, it is unlikely that this difference would have been large as participants were all located in a single health board and would be receiving care based on the same clinical guidelines.

| CON CLUS IONS
The findings of the present study highlight the importance of psychological variables in predicting non-attendance at diabetes appointments and suggest that psychological variables, such as those from the TPB and SRM, could be of value in applied settings for identifying people with diabetes who are at risk of not engaging in health care relating to their diabetes. Whilst interventions to improve diabetes appointment attendance should address practical barriers to attending appointments, such as forgetting and the time and day of appointments, there may also be a need to take account of the underlying perceptions about diabetes, the emotional impact and perceived lack of value in attending appointments.

ACK N OWLED G EM ENTS
Thanks to peer researchers for their input throughout the study and thanks to all the people who took the time to participate in the research.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The authors elect to not share data.