Chapter 6Discussion

Which types of reminder systems are most effective in improving the uptake of health service appointments?

Are there any systems that effectively support the cancellation of appointments?

Currently, there is strong evidence that a personal telephone reminder will increase patient cancellation/rescheduling rates [see Chapter 4, Evidence statement (E): there is strong consistent evidence that reminder systems will promote cancellation of appointments (evidence category Ia)]. Reti⁸⁰ suggests that one of the dominant benefits of telephone reminders may be the ‘promotion of suitable cancellation behaviours’. Telephone reminders carry the inherent advantage that patients who are unable to attend can cancel their appointment at the time of contact from staff. This advantage is further heightened if the same telephone interchange can be used either to rebook the appointment or to establish that the appointment is no longer required. There is less evidence to support the use of SMS messages to promote cancellation and rebooking [see Chapter 4, Evidence statement (E): there is strong consistent evidence that reminder systems will promote cancellation of appointments (evidence category Ia)]; however, this may be because SMS reminders are frequently sent very close to the appointment time, which may leave insufficient time for patients to act on the prompt.⁷⁸ A second reason that SMS may not optimise cancellations could be that when patients receive a reminder they are frequently asked to phone a telephone number rather than simply replying to the text [see Chapter 4, Evidence statement (E.2 and E.5): there is strong evidence that whether or not a patient cancels and rebooks will be influenced by reminder system factors that facilitate/hamper patient intentions to rebook (e.g. nobody answers the telephone) (evidence category Ia)]. There is strong evidence that patients frequently encounter heath-care system factors such as difficulties accessing cancellation lines, which can hamper patient attempts and intention to cancel and rebook [see Chapter 4, Evidence statement (E.2 and E.5): strong consistent evidence indicates that system factors related to appointment systems will hamper patient intentions to cancel/rebook (evidence level Ib)]. Better administrative processes in support of SMS may support cancellation more effectively.

Reminder systems are often employed with increased attendance rates in mind and relatively little attention is given to cancellation and/or rescheduling of appointments. Cancellation of appointments may be considered a desirable outcome, especially when the cancellation occurs in time for the appointment to be reallocated to another patient. Even under time-constrained conditions, cancellation without reallocation of the appointment may still be viewed as a favourable outcome as it reduces uncertainty about whether or not the patient will attend, allows clinical staff to assign the equivalent freed time to other patients or to alternative activities and, therefore, reduce the pressure on crowded appointment schedules. Rescheduling of appointments may also be a desirable outcome, especially when the health need of the patient is great. As there is strong evidence that the timing of a reminder, between 1 and 7 days prior to the scheduled appointment, has no effect on patient attendance behaviour [see Chapter 4,Evidence statement (B.5): there is strong consistent evidence that the timing of a reminder, between 1 and 7 days prior to the scheduled appointment, has no effect on patient attendance behaviour (evidence category Ia)], it would seem appropriate to send all reminders with more time to allow patients to cancel their appointments and for health services to reallocate appointments to other patients. Although there are no studies investigating the framing of reminders [see Chapter 4, Evidence statement (A.2–A.6): there is weak consistent evidence that ‘reminder plus’ are more effective than simple reminders at helping a patient to attend their appointment (evidence category IIIa)], it is recommended that the reminder should be framed with cancellation and/or rescheduling in mind. How the request is framed may increase a patient’s sense of moral obligation to cancel;¹⁰² however, more robust structures to support cancellation are also required. Cancellation procedures that are easy and require minimal effort from the patients are important, e.g. automated responses to SMS messages rather than the patient having to phone a telephone number, or cancellation and rescheduling telephone lines with adequate administrative support so that patients can quickly leave a message. E-mail may also be a convenient cancellation medium for many patients, as there is no need to wait to get through to a receptionist.⁴² Automated methods of cancellation may be perceived as easier than methods that require direct contact, as they offer flexibility to cancel at a time convenient to the patient and also reduce the need to provide explanations for cancellation.^42,206 Following cancellation of appointments, rescheduling of the appointment, if it has not occurred synchronously, also needs to be easy for the patient. For example, it may be sensible to have central booking lines which are open 24 hours a day.

Kitcheman et al.⁶⁹ observed that an effective reminder system will increase the ‘already heavy workload of psychiatric out-patient clinics’. The authors note that ‘even when appointments are missed clinicians do not, as a rule, waste their time’. Alternative activities include the fact that ‘patients are double-booked, dictation is completed, telephone calls made and colleagues consulted’.⁶⁹ This highlights that increased attendance at appointments may be welcomed more by managers, primary carers and patients than by those working in outpatient clinics. As a consequence, each service faces the challenging dilemma of whether to build in assumptions of non-attendance without opportunity for appointment cancellation or reallocation or to build in processes that optimise cancellation and rescheduling. If building in processes to optimise cancellation and rescheduling, then services will also need to consider the impact on staff that frequently utilise non-attendance at appointments as an opportunity to catch up on other health-care-related activities.

Although reminder systems will increase attendance, full appointment attendance is unlikely to be an achievable outcome; therefore, cancellation and rescheduling of appointments should be seen as a desirable outcome. Reminders would appear to have considerably more potential for promoting cancellation of unwanted appointments to allow reallocation of cancelled appointments than the current evidence indicates. Researchers and health services seeking to optimise attendance should not only consider reminder solutions, but should also consider the supporting resources that would be required to facilitate patient cancellation and rescheduling, reallocation of appointments to other patients while simultaneously supporting hard-pressed clinicians to manage administrative tasks that are routinely carried out when a patient misses an appointment.

Do different reminder systems have differential effectiveness for particular subgroups of the population (e.g. by age group, ethnic group, socioeconomic status, gender, etc.)?

There is very little good-quality evidence investigating whether or not reminder systems have differential effectiveness for population subgroups (e.g. gender, socioeconomic status, ethnicity, employment status, mental health, other comorbidities, presence of symptoms, diagnosis, severity of condition) (see Chapter 4, Proposition F: distal/proxy individual attributes). The majority of included studies have not been designed to perform secondary analysis to identify subgroups who are more or less likely to respond to reminders.

What factors influence the effectiveness of different reminder systems for particular population subgroups?

We have added no further comments to this section as we have discussed this under Previous patterns of non-attendance in the context that understanding patient non-attendance can provide insights into how reminder systems could be designed and deployed.

How do the perceptions and beliefs of patients, their carers and health professionals regarding specific types of reminder systems, and patient/carer resources and circumstances, influence their effectiveness?

In general, our review did not generate much information on the perceptions and beliefs of patients regarding specific reminder systems, as opposed to their preference for particular technologies, and we cannot say how perceptions and beliefs of patient or health professionals about reminder systems influence their effectiveness. This reflects the fact that most reminder-related research takes a rather narrow focus. There is evidence to suggest that a personalised interaction with a member of practice staff is perceived as more useful than an automated reminder system. However, there is also conflicting evidence as to whether this would make a patient more or less likely to communicate a cancellation than a less personalised method of delivery. Our review has also identified that an individual’s perception about health-care systems and professionals will influence patient attendance (see Chapter 5, Patient–provider ‘alliance’, communication and ‘engagement’). This could be related to an individual’s view of the quality of care provided, trustworthiness, empathy and cultural sensitivity of staff and issues of mutual respect. However, individuals’ views may also be shaped by the surrounding community in which they live, through a process of ‘storytelling’ whereby negative storylines will contribute to the individual’s negative perceptions of the health system or health professionals.¹⁵⁵ Patients may experience stigmatisation within individual relationships, such as those with family, GPs or other health professionals, which may directly influence an individual’s intention to attend or further colour the individual’s perception of the health service.¹⁴³ This is a difficult area to manage, and wider solutions must also be considered, but there is greater scope to consider how these factors might influence the effectiveness or design of reminder systems interventions. The role of reminder systems in this context is potentially limited given the negative publicity that British health services and health professionals have received through the media. However, the small role that reminder systems might play here in positively influencing individual-level perceptions of health-care services is via their contribution to patient engagement and therapeutic alliances (see Characteristics of the appointment system). Positively framed reminder messages, delivering appropriate and helpful information, could help to counter an individual’s negative perceptions of a health service.

How do external factors (e.g. content, delivery, setting, frequency, notice period) influence the effectiveness of reminder systems?

There are few studies that directly investigate the extent to which characteristics of the reminder system can influence the effectiveness of the reminder system; however, two characteristics stand out as being important. There is strong evidence that factors such as accessibility to the reminder will influence the effectiveness of the reminder and that the timing of the reminder, between 1 and 7 days prior to the appointment, will not influence the effectiveness of the reminder and we suggest that there are good reasons for sending reminders earlier than usual, i.e. 3–7 days prior to the appointment (see Timing of the reminder).

How do organisational factors influence the effectiveness of reminder systems?

No studies investigating whether or not factors about health-care settings influence the effectiveness of reminder systems were identified from this review [see Chapter 4, Evidence statement (C.1–C.6): there are no studies investigating whether or not factors about health-care settings influence the effectiveness of the reminder (evidence category VIIa)]. However, there is a wealth of research that shows that attendance behaviours are linked to health-care settings; however, often the findings are conflicting, indicating that contextual issues may contribute to complexity. Exploration of this evidence offers possible insights into how reminder systems and supporting processes should be designed in order to optimise attendance at outpatient appointments. This section will focus on specific organisational factors that impact on attendance and for which a reminder-based solution might contribute to reducing non-attendance. For clarity, this material is organised into four inter-related themes: (1) administrative errors, (2) characteristics of appointment systems, (3) patient–provider alliance and engagement and (4) clinic accessibility.

What disadvantages (or challenges) should be considered when introducing reminder systems for specific populations for health care and health services?

The principal technologies should not only be considered in terms of their principal characteristics. Technologies may possess additional functionalities that, if activated, may have a differential effect on the effectiveness of the reminder. To take just two examples, SMS technology can be used to deliver a message (one way), to reply with a confirmation of attendance or, in some cases, to effect a cancellation and/or a rebooking. When evaluating text messaging reminders, Leong et al.⁷² did not require participants to confirm receipt of reminders. This carried the implication of uncertainty whether or not all participants had received the text messages. In this circumstance, they assumed that those who did not receive messages were more likely to be non-attenders and, therefore, concluded that their study would probably underestimate the effectiveness of text messaging reminder. Similarly, the effectiveness of a telephone-based reminder system, whether personal or automated, will depend on factors such as whether or not a patient possesses an answering machine and whether or not it is acceptable policy to leave a message without speaking to the patient themselves or to a nominated next of kin. Again, Leong et al.⁷² made no attempt to leave messages on the answerphone/voicemail, stating that the ‘investigators were of the opinion that most people do not check their voice-mail and there is no way to check if the participants had retrieved their messages’.

What are the likely economic impacts of reminder systems?

A full economic analysis of reminder systems is beyond the scope of this review. In the majority of included studies, there is either no or minimal data capture on health economics. However, 11 studies^{25,43,47,53,57,72,100,102,108,208,215} discussed the cost of reminder systems, using different methods of calculating cost savings and providing sufficient data to create a picture of possible economic impacts. All 11 studies support the view that SMS messaging is a cheaper option than other type of reminders. Hasvold and Wootton⁴⁷ conducted a systematic review of 14 studies and found that the mean cost of SMS reminders, although it varied from country to country, was €0.14 per patient, compared with €0.90 for telephone reminders. The cost-effectiveness analysis showed that the cost per attendance for the SMS group (¥0.31) was significantly less than that for telephone group (¥0.48).⁵⁷ Leong et al.⁷² found that text messaging was the cheaper option, as suggested by the total costs, i.e. 87.66 Malaysian ringgit for text messaging reminders and 160.33 Malaysian ringgit for mobile phone reminders’ for clarity. When the authors compared the two reminder systems, the ratio of the cost of text messaging versus mobile phone was 0.55.⁷² A US trial in a physiotherapy setting calculated that 19 SMS reminders needed to be sent to prevent one missed appointment.⁸⁵ The authors suggest that this system of sending reminders to patients may lead to economic savings when the relatively low cost of each SMS reminder is balanced against the cost of employing a physical therapist for the time of a non-attended appointment. However, they do point out that a formal economic evaluation is needed. Milne et al.¹⁰⁸ calculated that the cost of SMS to reduce non-attendance was £7.50 (US$13.13, €10.88) per ‘no-show’ avoided. Based on number needed to text (NNT) analysis, approximately 14 people would need to be sent a SMS reminder to prevent one non-attendance.¹⁰⁰ Perry et al.²⁰⁸ stated that the cost of the SMS reminder is minimal, considering the loss of revenue generated by failed appointments. Text reminders on a large scale would be a cost-effective strategy that could be improved. Battistotti et al.¹⁰² calculated costs on an annual basis and estimated that for 500 SMS/day at a unitary cost of €0.0065, the first-year cost will be €11,500, while next-year cost will be about €8200, accounting for SMS and maintenance. They calculated that this would average out at around €35 per day once the SMS system was established. As the average monetary loss of a dropout is €20 per appointment missed, recovering two dropouts per day would be to sufficient to pay for the system. The annual cost of missed appointments in England is estimated to be close to £575M and the use of SMS reminders could save £55.6–83.5M a year.¹⁶⁵

Reminders over the telephone may be expensive when compared with other approaches.⁴⁹ In a recent systematic review,⁴⁷ the average estimated costs from 14 studies reporting phone reminders was €0.90 compared with €0.14 for SMS messages; however, costs of telephone calls vary from country to country.⁴⁷ Chiu⁵⁸ similarly found that with an absolute reduction in non-attendance rate of 9.4–14.82% as a result of telephone reminders, 7–11 telephone reminders were required to prevent one non-attendance. Computing the cost for each telephone reminder by clerical staff at HK$7.44–15.33, Chiu⁵⁸ estimated that it cost HK$109.37–296.79 for clerical staff to prevent one non-attendance using telephone reminders. In his specific context of radiological procedures, Chiu⁵⁸ estimated that the cost of wasted computerised tomographic scan appointments offset the cost for the delivery of telephone reminders. This cost–benefit ratio would naturally be less favourable for less expensive interventions.⁵⁸ Robert et al.⁸² estimated that the cost of telephoning 200 patients could be offset by preventing one non-attendance.

Perron et al.⁷⁸ reported results from a stepped approach to reminders, comprising telephone call, text and then postal reminders, and demonstrated that this generated 55 additional consultations. The intervention proved to be cost-effective in providing a total net benefit of €1846 per 3 months once the cost of the intervention had been deducted. Bech²¹⁵ concludes that very few studies include more than one intervention. This, combined with the fact that very few report information on the cost of the intervention, enables only vague conclusions about the cost-effectiveness of the interventions to be made.

Charging a fine for missing appointments is unlikely to be cost-effective as the cost of charging and collecting a fine includes personnel time, setting up information technology (IT) facilities and accounting systems, postal and reminders cost.²¹⁵ These costs have recently been estimated in Denmark to be around 40 Danish krone per fine at 2003 prices. At first glance this appears to be more costly per unit than letter and telephone reminders; however, these administrative costs are generated only for non-attendees whereas letter and telephone reminders by their very nature generate a cost for every scheduled patient; therefore, fines may very well be cost-effective because of their low average cost per patient. However, these estimates would need to be tested in various UK settings in order to determine whether such a system is feasible or, indeed, acceptable. Whether or not charging a fine is cost-effective would depend partly on the cost of the appointment and partly on whether or not non-attendances decrease. As non-attendances decrease, so the unit cost of administering the fine increases.

Strengths and limitations of the systematic review

Generally speaking, the systematic review method seeks to provide a precise answer to a tightly focused question. Such reviews tend to be most useful when there is a high degree of homogeneity around the five PICOS elements: the population, intervention, comparison, outcomes and study types. As can be seen from our description of findings in Chapter 5, such homogeneity is not present in this particular review. A wide range of population types is included within the RCTs we identified. This variation by population is seen in the clinical settings, countries and age groups studied in the included studies. Some populations are recruited completely randomly, increasing our confidence in the applicability of results in a real-world setting. However, others either established technology preferences prior to the study or excluded those who did not have a particular technology (e.g. mobile phones). The reader should, therefore, carefully examine the number and nature of participants exiting the study prior to randomisation before drawing conclusions on real-world effectiveness. In addition, two studies targeted non-attenders, but not in comparison with a general population.^64,83 As a consequence, we are unable to advise on specific strategies for the non-attending population – an otherwise viable decision alternative to blanket coverage.

Further variation is observable in the interventions. Although it is not unusual for a systematic review to examine, for example, different drugs within the same therapeutic class, this is quite different from appointment reminder systems for which different technologies employ vastly different mechanisms. Such variation is not simply evidenced in the technologies themselves [e.g. synchronous (SMS text) vs. asynchronous (letter); interactive (personalised telephone call) vs. non-interactive (e.g. automated calls); or mobile (SMS/mobile) vs. ‘static’ (letter/landline)]. Different studies use different features of the technologies, for example reply slips (post),⁵⁶ acknowledgements/confirmations (SMS)⁵⁶ or answerphones (for landlines).^66,84 Included studies also follow different protocols with regard to, for example, the number of attempts to be made and whether or not leaving a message on an answerphone constitutes successful contact. Such heterogeneity makes it extremely challenging to attribute success to a particular technology and/or to a specific component feature.

Notwithstanding a fairly plentiful body of studies, the number of different interventions being compared and, in addition, the number of supplementary questions they simultaneously seek to answer (e.g. comparing nurse calls with doctor calls, 3-day with 7-day intervals, etc.) adds a vast array of comparators with only single studies occupying many of the intervention–comparator matrix cells. Some interventions compare adjunct interventions, e.g. postal reminder plus SMS text reminder, while one intervention in particular has a stepped approach proceeding through telephone, SMS then post.⁷⁸

The domain probably experiencing least heterogeneity is that of outcomes. Attendance rate (expressed alternatively as its converse, the no-show rate) is ubiquitous among the studies. A limited number of studies report cancellations, and an even lower number examine rebookings. Clearly, all three are required for an understanding of interactions and ‘knock on’ effects. This commonality of outcome did help our review team in performing some preliminary comparisons that showed, for example, that attendance rates for non-diagnosis specific situations, such as blood donation, are generally lower and that paediatric studies tend to report higher attendance rates than other populations, particularly compared with adolescents. However, such apparent conformity belies variation in how attendance or cancellations are actually defined. In some cases, actual attendance is collected while, in other cases, an operational definition depends on the timeframe within which a particular service can respond.

With regard to study type, this particular area of investigation is well travelled by systematic reviews but is complicated by differential inclusion of study types. Our systematic review included only RCTs and, as a consequence, it became challenging to compare our included studies with those in several existing systematic reviews.

It is important to note that the scope of the review required a specific focus on appointment reminder systems. While this opened up a fairly substantial body of evidence, it did restrict the team by excluding studies about appointment behaviour. The focus on RCTs further restricted the number and types of question that these studies could address.

Our systematic review has numerous strengths, including a structured search protocol requiring thorough searches of electronic databases, reference lists and citations. As a consequence, we believe that we have assembled the widest possible body of relevant knowledge. Comparison with other systematic reviews demonstrates the improved currency and coverage of our review combined with a sole reliance on properly conducted RCTs, which led to increased rigour and robustness of studies. To extend our coverage in a field that is rapidly changing, we conducted various internet searches to identify new developments. While we acknowledge that innovation must be accompanied by evaluation, this at least ensured that the team was aware of the main directions for future travel. The rigour of the process was ensured by various mechanisms such as the use of multiple reviewers, regular consensus processes and checking of newly identified studies against original sift criteria.

There are also limitations to our review. Our review was limited to English-language literature, but we consider this acceptable given the intended audience for this review. However, it should be noted that published, non-English reports are not identified in our report. Our reporting and assessment of each study was limited to data in each published article with no attempt made to contact the authors for additional information or missing data. Assessments of quality are based on the reporting of study details, which is regarded as an adequate surrogate but may not equate to the quality of the conduct of each study. An additional limitation to this review stems from the potential for publication bias whereby studies that demonstrate positive results in favour of a particular technology are more likely to be published.

We can conclude that the systematic review component followed accepted guidelines for conducting such reviews. It utilised a clearly focused question that was easily operationalisable in terms of inclusion and exclusion criteria. However, as we freely acknowledge, the topic area itself possessed several complexities. Decisions we made at the systematic review stage therefore carry associated ‘knock on’ implications for the review informed by realist principles as discussed below under Strengths and limitations of the review informed by realist principles.

Strengths and limitations of the review informed by realist principles

Methods for realist review (also known as realist synthesis) are still in their relative infancy and standards for conduct and reporting are still subject to ongoing development. Our review team recognised the value of exploring the complexity of appointment behaviour as it determined interaction with appointment reminder systems. However, we had concerns that our overarching aim was to inform practical recommendations for appointment reminder systems, not only to generate explanatory theory. We followed realist principles by starting from outcomes and then seeking to identify upstream determinants that may have influenced appointment behaviour. We consider context to be informed by four principal considerations (personal, organisational, trial and problem context).²¹⁶ Personal context includes personal characteristics such as age, gender, sociodemographic and employment status. These attributes can variously act as ‘baffles’ to damp down the signals from the appointment reminder or as ‘amplifiers’ to give a reminder additional urgency and/or importance. Health-care organisations can create a culture within which attendance, or indeed non-attendance, becomes the norm. Paradoxically, flexibility of appointment systems may be misinterpreted as either a ‘couldn’t care less’ or an ‘it’s not that important’ mentality and could aggravate non-attendance trends. All staff, clinical and administrative, have an influence in creating and/or communicating this culture. The trial context relates to some of the decisions that may be taken in the context of a trial protocol that might not be carried through in a real world situation. This might include the number of attempts that is reasonable or feasible to make to contact a patient via telephone or counting patients as non-attendances when they have actually used the opportunity to rebook. Finally, the problem context refers to the way in which the nature, scale and cause of the problem may shift (with varying and sometimes unknown impact on its size and frequency). Other national and/or local policy factors may determine patterns of attendance, which may result in oversimplification of the mechanisms of effect. Even using the tools suggested by realist synthesis methods such contextual challenges continued to pose a significant challenge.

Realist synthesis itself poses a challenge to systematicity in that its processes are iterative, intuitive and often defy audit and transparency. We used the systematic review principles described above to ensure that we constructed a transparent and well-defined sampling frame. Characteristics of identified studies were plotted on our ‘PopInS’ matrix to map the landscape for our question; however, once we had extracted and analysed all systematic reviews and RCTs to address the questions relating to effectiveness, we faced a decision as to how to prioritise the remaining studies. We decided to focus on UK studies and, in particular, to pay closest attention to those studies that supplied data against the various aspects of the conceptual framework.

Several decisions made in connection with the earlier systematic review had implications for the subsequent review informed by the realist principles. These included:

1. Appointment systems for screening programmes for which a patient is required to respond by booking an appointment. These were characterised as ‘recall’ systems not ‘reminder’ systems and were therefore, excluded from the review. We argue that the action of booking indicates the presence of some extrinsic motivation for fulfilling the appointment thus making attendance atypical from appointments in general. However, such an exclusion may seem arbitrary alongside such systems as Choose and Book, which are included in our review, and indeed the extra interactions required by, for example, systems requiring postcard and/or reply-paid confirmations. It is possible, however, that an exploration of recall systems might contribute to our further understanding of motivations for appointment attendance and how these might be harnessed within a reminder system.
2. Interventions that include not simply a ‘reminder’ but also some therapeutic interaction, typically via the telephone. Clearly there is a substantive difference between a member of administrative staff telephoning with factual details of a time and location of appointment and a clinically qualified staff member offering reassurance, advice and treatment options, etc., along with an exhortation to attend a forthcoming appointment. Such therapeutic interactions may be particularly important in the contexts of counselling, health promotion and psychological therapies, etc. However, it is methodologically challenging to isolate the reminder component from the therapeutic component and, therefore, we chose to exclude such studies from our systematic review. Nevertheless, our review informed by realist principles identifies ‘engagement’ with treatment, service or health provider as a likely contributor to attendance behaviour. It is again possible that an exploration of therapeutic communication may contribute further insights and offer value added services that can be easily integrated with the basic reminder function. Indeed, such a distinction becomes even more blurred where health promotion information is included with a SMS reminder or when an orientation statement is included with a postal reminder (both included).
3. The content of SMS text messages, telephone calls and letters is poorly described in many included studies. Little attention is given to the ‘framing’ of any spoken or written communication. Framing is a significant area of research in disciplines such as psychology and information science and it is likely that assumptions on what is the default position (assume to attend or assume not to attend), to use one example, will have an effect on the effectiveness of reminder systems. We highlight the impact of such framing in our discussion on ‘nudge theory’ (see Nudge theory and custard creams), but only in connection with the collective behaviour of a practice population. However, there is clearly significant potential for further exploration in terms of effects on individual behaviour, both in terms of theory-building and subsequent empirical inquiry.

The strengths of our review informed by realist principles include the strong embedding of our propositions in the extracted data. This stems from the practical orientation of our review and facilitates the production of implications for practice.

A weakness of our review, and yet a corresponding strength, is the fact that, unlike realist synthesis, we did not settle on a single theory as an overarching explanation for what can be observed in the very complex situation of appointment attendance and non-attendance. We engaged with a full range of theory operating at different levels and originating from different disciplines, in a sense problematising the literature as a collective evidence base, analogous to critical interpretive synthesis methods. It is, therefore, not entirely coincidental that we discovered resonance with the work of Dixon-Woods et al.¹⁶⁸ nor, indeed, that those authors chose appointment attendance as one case by which they explained their own findings. Our steering group encouraged us to identify a leading candidate theory in the quest to simplify the subsequent review process. However, the review team strongly believed that no single theory served to explain all aspects of the phenomenon of attendance and the effect of appointment reminder systems. To the contrary, we sincerely considered multiple theories, in total or in part, to stimulate our interpretation of the problem. This is but one reason why we prefer to describe our review as being ‘informed by realist principles’.

The appointment as an event

Ambiguity: importance of appointment as an event will increase the likelihood that patients attend, up to a point. However, beyond this point it may simply increase anxiety. Similarly on the one hand doctors want to appear supportive, reassuring and alleviate anxiety; however, they may want to stress the importance of attending the appointment.

Lest we forget – or forget about forgetting

Forgetting consistently features as the most commonly cited reason for missing an appointment. However, results on reminders, e.g. 24-hour or 48-hour reminders, show no significant differences in effectiveness for these intervals. This suggests that the reminder function is not the most important consideration for the communication between service and patient. If timing of the message is not significant from a patient viewpoint then we can reverse the direction of the imperative and determine the reminder interval by the time period within which a service can best accommodate cancellations and rebooking. Such an approach does seem to fly contrary to the literature, which almost unanimously confirms the orthodoxy that reminder systems are for the express benefit of the forgetful patient. In fact, it is the service that stands to benefit most, not from the reminder itself but from the response that it triggers either in terms of notifying the service of cancellations or initiating rebooking at a more convenient time. However, such a response becomes useful only if notification is received in sufficient time for the service to reallocate the vacant slots to other areas of clinical need.

Further support for this counter-narrative that appointment reminders systems are primarily determined by considerations of service scheduling, not of assisting the forgetful patient, is seen by attitudes to late arrival, which were seen as more disruptive than non-attendance. In the study by Martin et al.,⁵ professionals felt that it was, in fact, even more disruptive if patients turned up late for an appointment than it was if they did not attend at all.

However, from a patient perspective it seemed that patients were unaware of the implications of turning up late for the practice. The impression, from previous experience, of waiting past their appointment time for their actual appointment had misled them into thinking that the practice would be quite relaxed to their late arrival. This reinforces the idea that norms around attendance may be either consciously or subconsciously reaffirmed by health service staff as much as by fellow patients. At this point it should be emphasised that health services staff working in the same service or at the same location will not necessarily communicate unambiguous messages regarding the values associated with attendance. For example, in the study by Martin et al.,⁵ receptionists were more concerned than GPs about missed appointments, possibly because of the greater disruptive effect that non-attendance had on reception staff. Hussain-Gambles et al.¹⁰⁴ also found that receptionists felt that they were most affected by non-attendance and wanted GPs to address this in consultations with patients. GPs were more guarded about this, being more fearful of damaging the doctor–patient relationship.¹⁰⁴

What are desirable outcomes?

Our review informed by realist principles yielded some interesting insights on what might be considered a good or bad outcome. First cancellation, although an irritation to administrative staff, could be seen as a desirable outcome provided that the system has the flexibility to compensate for the non-attendance of the particular patient and the welfare of the patient is not impaired, e.g. there is no disease window of opportunity that is missed by deferral.

A key variable that is not addressed by any of the studies is the underlying assumption around clinic capacity, i.e. is the clinic booked at 100% capacity with cancellations providing much needed respite or is a clinic booked at overcapacity to allow for 100% attendance following the inevitable withdrawals. All recommendations and planned interventions should be moderated by an understanding of this underpinning service rationale.

Such ambivalence could also be observed from the viewpoint of the individual clinician.⁵ Missing appointments may be seen to offer some respite from an often relentless stream of appointments and offered clinicians the chance to catch up with note writing or other administrative tasks.⁵

However, reception staff themselves found missed appointments more frustrating and irritating than GPs. They talked about how hard they had worked to find patients an appointment slot. If this appointment was then wasted, there were fewer appointments available to other patients.⁵

Reaction to a reminder in terms of cancellation can be considered a valuable trigger of a service response by reducing the uncertainty of whether or not the patient is going to attend. A system of telephone reminders, particularly because it offers interactivity, not only reduces the non-attendance rate, but also decreases the uncertainty associated with non-attendance. Therefore, health-care managers become more able to control the workload, plan duty rosters and allocate health-care resources.

There is some evidence to suggest that ‘forgetting’ is readily adopted as a ‘blame-free’ explanation for non-attendance and may therefore be used as an avoidance strategy against articulating other priorities or reasons that may not be so readily accepted.^5,12,56 Qualitative evidence suggests that ‘forgetfulness’ is used as an umbrella explanation for non-attendance when subsequent explanation reveals other more tangible factors.^5,12 For example, Martin et al.⁵ observed that, following an initial explanation of forgetfulness, numerous patients identified a variety of employment and family-related factors which might become a higher priority than attending the appointment.

Nevertheless, figures from our effectiveness review demonstrate a consistent effect of any reminder suggesting that genuine forgetfulness does indeed play a significant, if less prominent, part. Such a ‘smoke screen’, perhaps in response to perceived or actual judgemental behaviour from clinicians or administrative staff, ironically performs a disservice when it comes to understanding appointment behaviours and developing effective reminder strategies. Only if candour, in a non-judgemental blame-free context, can be elicited by researchers and replicated in interactions with health service staff will service managers be able to devise appropriate patient-focused solutions to problems of non-attendance. In the course of our review informed by realist principles we have found that the observations of Dixon-Woods et al.¹⁶⁸ resonate with our own work, namely that attendance behaviour research displays a consistent tendency to problematise the patient, not the service. The phenomenon of ‘convenient forgetting’ should not be allowed to mask the fact that certain sections of the community are particularly prone to genuine forgetting, for example the elderly or those experiencing symptoms of dementia. Clearly it would be retrograde to imprint suspicion of the explanation of forgetting too vigorously in the minds of those managing appointment rescheduling.

Voting with their feet

Non-attendance rates can be viewed as a way of patients expressing their reaction to the perceived inaccessibility or inflexibility of the service. If cancellation is easy (and does not subject a patient to disapprobation), then a service will gain valuable opportunities to reschedule other patients or for clinicians to plan to utilise downtime effectively. If services are flexible in terms of good availability of clinical services then cancellations and rebookings will be low. Very little attention is paid in the literature to attendance and cancellation rates as a diagnostic indicator for a service manager in assessing how their service performs with regard to access and availability. Instead, the research literature focuses on ‘fixing’ the problem rather than on understanding it and re-engineering innovative service solutions.

Another way in which poor attendance rates can be seen as a diagnostic indicator relates to inappropriate referral. Patients may fail to attend an appointment because they no longer consider it necessary, a condition may be self-resolving, a crisis point may have passed and been successfully negotiated or a patient may have arrived at a view that their referral is no longer desirable or necessary. In health-care systems that maximise patient choice, through availability of other health-care providers, a patient may simply have gone elsewhere for resolution of their condition.

Inappropriate referral and self-resolution of symptoms are two possible explanations for the commonly reported phenomenon of high non-attendance for appointments located in the distant future from onset of symptoms. Another factor, which itself may lead to patients concluding that their visit to a clinician is not entirely necessary, relates to signals that conflict with a sense of urgency. As a result of our review informed by realist principles, we hypothesise that any single factor or combination of factors that attributes importance or significance to attendance at a particular appointment serves to create a sense of that appointment being an ‘event’ rather than being routine or mundane. These signals may emanate from the individual patient (e.g. the severity of the symptoms), from the disease trajectory (e.g. the sense of a limited ‘window of opportunity’), from the service (e.g. a battery of tests or procedures in close succession) or from the health staff (e.g. how urgently they treat scheduling of an appointment or how they react to cancellation of an appointment). This ‘event’ hypothesis possesses explanatory power in terms of the frequent use of forgetfulness as a reason for non-attendance – in general, we are less likely to forget those occurrences that have acquired the status of an ‘event’.

A further hypothesis relates to the importance of creating a sense of ‘engagement’, whether this be engagement with the service, the treatment or the therapeutic relationship. Patients who are more engaged are more likely to prioritise their attendance at appointments and to take such appointments seriously. In terms of appointment reminder systems, various degrees of engagement are possible. These include speaking to a member of health service staff rather than an automated message, speaking to the clinician handling your case and interactively acknowledging that you will attend an appointment – the ‘symbolic contract model’ mentioned in Chapter 4, Evidence statement (A.2–A.6): there is weak consistent evidence that ‘reminders plus’ are more effective than simple reminders at helping a patient to attend their appointment (evidence category IIIa).

Within the specific context of help-seeking for mental health problems, commentators have described the interplay of approach–avoidance conflict whereby an appointment holds both the potential of resolution of symptoms and exposure to a stressful situation.²¹⁷ This tension neatly captures some of the ambiguities encountered in the course of the review informed by realist principles whereby an appointment can be seen as both a positive and a negative event leading to markedly alternative ‘futures’.

Carrots, sticks and sermons

It has been suggested that a three-part classification of carrots, sticks and sermons is one way of characterising three different approaches to changing behaviour. This classification could be applied to the content of the reminder messages and how they are framed. Carrots can refer either to financial or other tangible incentives or to positive ‘strokes’ aimed at appealing to personal reward (you will be treated in a timely fashion, you will reduce your uncertainty promptly, etc.) or social good. Sticks can again be tangible (as in financial penalties) or affective (e.g. if you don’t attend then . . .), perhaps in provoking a guilt response. In a health education context, sermons refer to information that tells patients why reminders are good for them. Appointment reminder messages would tell you the health consequences of missing the appointment, particular if there is a relatively narrow window of opportunity with regard to the particular condition. It is our perception that there is a considerable body of literature on these phenomena in a health education context and yet, except when health promotion messages are included with appointment reminders, this evidence is largely divorced from the reminder context.

Do I really need to go?

Except in a mental health context for which patients may have an impaired sense of determining their need for treatment, it may be reasonable to expect patients, particularly those with a chronic self-managed condition, to be able to judge when they need an appointment. However, certain conditions may have a brief and critical window of opportunity which, if missed, can lead to severe consequences in terms of patient health and/or health-care resource utilisation. In addition, if members of health service staff endorse inappropriate over-utilisation they may perpetuate such a situation. Alternatively, if they appear too judgemental, they may discourage utilisation all together. Regular, unmediated appointment intervals may be viewed as an alternative to more sophisticated communication and interaction between patient and clinician and to facilitate informed patient self-determination in decision-making. It also may be viewed as a defensive procedure to minimise risk of negligence. As mentioned under Carrots, sticks and sermons, patients may receive ambiguous messages about the importance of the appointment particularly if there is a juxtaposition of a ‘task-focused’ view from the administrative staff with a ‘person-focused’ perspective of the clinical task. Of course, these roles are equally problematic if the perspectives are reversed.

Nudge theory and custard creams

Current thinking on mechanisms for effecting behavioural change is heavily influenced by the so-called ‘nudge theory’. Nudge theory has its origins in behavioural psychology and economics and is currently being promoted by the ‘behavioural insights team’ at the UK Cabinet Office. Nudge theory suggests that making it easier for people to take a healthier or more socially acceptable course can prove more effective than introducing legislation or penalties. Of particular relevance to the problems associated with appointment attendance are experiments being trialled with regard to payment of taxes or fines. Initiatives, such as sending personalised text messages urging people to pay fines and including images of untaxed vehicles in demands for payment of duties, have demonstrated initial success. Other suggestions include highlighting key information in bold or ‘strong’ colours, using lotteries or prize draws as an incentive for prompt payment of taxes and sending ‘thank you’ letters to people. Clearly some, although not all, of these approaches have direct implications for appointment behaviour and appointment reminder systems. For example, Hayes et al.²¹⁸ describe how they initiated a behavioural contract to increase appointment keeping among low-income families in a child and family therapy programme. Such families received US$30 in coupons at attendance at four consecutive appointments with costs offset against the potential cost of broken appointments.

This potential for social influence or ‘nudge’ theory has recently been explored in the specific context of those who did not attend appointments. Martin et al.¹⁴² describe three interventions tested in GP surgeries in NHS Bedfordshire that could be implemented quickly and appear to be cost neutral.

• Patients calling for an appointment should be asked to repeat back the time and date of their appointment before the call ends.
• When booking follow-up appointments, patients should write down the time and date on an appointment reminder card rather than health-care or reception staff doing it for them.
• Replacing common signs that communicate the number of patients who did not attend appointments in previous months with signs that conveyed the much larger number of patients who do turn up.

These first two interventions correspond to our ‘engagement’ hypothesis in that they seek to increase patient participation, thereby either reducing did-not-attends or at least increasing a sense of responsibility such that those subsequently unable to attend would call and cancel. Surgery staff booking appointments over the telephone asked patients to repeat back the time and date of their appointment before ending the call. This led to a 3.5% reduction in did-not-attends, which is comparable to the minimum gains made by moving from no reminder to using any form of appointment reminder system. The second intervention, asking patients to write the time and date of their follow-up appointment with a nurse on the appointment card themselves, rather than the nurse doing so, led to a reduction in did-not-attends of 18%. This figure equates to some of the highest gains made when introducing reminder systems.

Finally, Martin et al.¹⁴² turned their attention to the idea that publicising the number of patients who did not attend via posters on waiting room walls or TV monitors might increase rather than decrease did-not-attends. By drawing attention to the regrettable frequency of unwanted behaviours they understood that this would be likely to normalise such behaviours thereby resulting in increased occurrence. They replaced these signs with posters that conveyed the much larger number of patients who did attend. When this intervention was combined with the verbal and written commitment-led interventions described above, this resulted in a 31.7% reduction in did-not-attends compared with the average of the previous 12 months. While caution should be expressed at the use of relative, rather than absolute, figures for improvement, particularly when there is no way of knowing what the baseline rates were and whether they are high or low when compared with typical practices, it is clear that low-cost behavioural interventions should at least merit as prominent place in the future research agenda as high-technological ‘fixes’.

Service inertia

One of the attractions of the approaches by Martin et al.¹⁴² is involvement of practice staff in the diagnosis and resolution of problems associated with non-attendance over a coffee break, hence the ‘custard creams’ of the article title. This is particularly to be welcomed given that there is some evidence to suggest that health service staff may act as a barrier to the development of innovative patient-friendly solutions that overcome obstacles to their attendance. In the study by Martin et al.,⁵ patients described difficulties in getting through to the surgery by telephone, as lines were often busy. Interestingly patients did express the perception that getting through to the surgery might have enabled others to take advantage of a cancellation slot, although there was no evidence that the practice offered such flexibility. This suggests that if services were able to actively demonstrate that they were able to reutilise slots freed by cancellations, an ‘altruistic’ subset of patients might choose to phone in to cancel, providing that the structural barriers, such as engaged telephone lines, were removed. However, there was evidence from this particular study site that the surgery problematised the patients, not the service. This attitude may explain an apparent inertia from this surgery in response to the suggestion concerning a dedicated cancellation line or facility to use SMS text messaging as a means of cancelling an unwanted appointment. In fact, staff expressly rejected these suggestions from patients as being impractical to implement or difficult for existing staff to manage.⁵

General implications

In this section we provide an overview of findings that are potentially relevant for all health services. All reminders are effective at improving attendance at appointment. Simple reminders which provide details of timing and location of appointments are effective at increasing attendance at appointments and would appear to be useful for all patients, across all health-care settings, who are at risk of forgetting their appointment [see Chapter 4, Evidence statement (A.1): there is strong consistent evidence that simple reminders which provide details of timing and location of appointments are effective at helping a (forgetful) patient to attend their appointment (evidence category Ia)]. There is also weak evidence that ‘reminder plus’, which provides additional information over and above date, time and location of the appointment, is more effective than simple reminders at helping a patient to attend their appointment [see Chapter 4, Evidence statement (A.2–A.6): there is weak consistent evidence that ‘reminder plus’ are more effective than simple reminders at helping a patient to attend their appointment (evidence category IIIa)]. Orientation information and information about health services and procedures may be helpful at reducing perceived obstacles to attendance, although more high-quality research is required to confirm this. We speculate that ‘reminder plus’ may be particularly useful to facilitate first appointment attendance and attendance at screening appointments, and that simple reminders may be appropriate thereafter for most patients and most health services, most of the time (see Reminder plus), although this hypothesis requires further testing.

Appointments can be scheduled at inconvenient times, become inconvenient owing to a change of circumstances or no longer be required for very many reasons that have been discussed in this review. This indicates that all patients are likely to be unable to attend an appointment (whether it is a new appointment or a follow-up) at some point (see Previous patterns of non-attendance). Unless patients indicate that they do not want to receive a reminder, we recommend that all patients should receive a reminder or ‘reminder plus’, which recognises this as true and actively encourages patients who are unable to attend to cancel their appointment and to reschedule if further appointments are required (see Are there any systems which effectively support the cancellation of appointments?). We further recommend sending the reminder around 3 days in advance because although timing of a reminder between 1 and 7 days prior to the scheduled appointment has no effect on patient attendance behaviour, this will allow sufficient time for patients to cancel and health services to reallocate the cancelled appointment to another patient [see Chapter 4, Evidence statement (B.5): there is strong consistent evidence that the timing of a reminder, between 1 and 7 days prior to the scheduled appointment, has no effect on patient attendance behaviour (evidence category Ia)] or allow the clinician to undertake care-related administrative tasks, e.g. telephone calls, discharge letters, etc. (see Are there any systems which effectively support the cancellation of appointments?).

We found many examples in the literature of inefficient reminder activity (e.g. poor patient accessibility to reminder technology, incorrect patient contact details and human error) leading to suboptimal reminder effectiveness (see Accessibility) and poor resourcing of reminder systems. To optimise attendance, cancellation and rescheduling, there needs to be robust procedures to ensure that patient contact details are up to date. There needs to be easy-to-use, probably multiple, systems for cancelling appointments that suit the needs of the patients and not just the needs of the service, e.g. automated SMS cancellation, answerphone, e-mail, etc. Robust rescheduling procedures need to be in place to allow easy rescheduling of appointments for patients, both within and outside normal working hours to provide patients with flexible opportunities. Finally, an effective reminder system will increase the workload on clinical staff and services will need to build in alternative administrative time for staff that frequently utilise non-attendance at appointments as an opportunity to catch up on other health-care-related activities (see Are there any systems which effectively support the cancellation of appointments?).

We found few studies investigating the differential effectiveness of reminder systems for population subgroups. Consequently, we have provided a narrative overview of literature generated by our review, to identify groups who are at higher risk of non-attendance. In general, contrary to the popular view, our review provides no evidence that age, gender and previous attendance patterns predicts non-attendance but it does show that key groups at high risk of non-attendance are related to deprivation status, ethnicity, substance abuse and comorbidity/illness (see Chapter 5, Patient characteristics). Our findings suggest that reminder system choices need to be carefully considered in order to maximise accessibility for these key patient groups. Simple reminders and automated reminders to attend may be ignored or overlooked and may put these patient groups at a disadvantage compared with general outpatient populations. Reminders with direct personal contact might be appropriate in these groups (see Other reminder characteristics, Format of the reminder). To facilitate attendance in the most at risk, vulnerable groups we have suggested that reminder systems of increasing intensity and interactivity could be introduced to ensure that disparities in health-care opportunities are not compounded. We have introduced the concept of a sequential reminder intervention in order to reach the most number of patients and maximise attendance⁷⁸ (see Substance abuse/mental health/comorbidity and physical illness). The re-engagement of patients with treatment after they have missed their appointment may be important if they have particular health problems that need ongoing treatment. Intensive approaches, such as ‘stepped reminders’ and patient navigators, have been effective at increasing attendance at screening and immunisation programmes in disadvantaged and vulnerable populations and might also be effective at re-engaging similar groups of patients who have dropped out of treatment [see Do different reminder systems have differential effectiveness for particular subgroups of the population (e.g. by age group, ethnic group, socioeconomic status, gender, etc.)?].

Specific implications

The general implications above should be considered by all health-care services; however, this review also identifies various reminder or appointment-related issues that may be relevant to specific health services. Therefore, along with the ideas outlined above, health service managers should consider the following issues that will add a layer of decision making to their choices about reminders. In essence, reminder systems are a complex intervention,²¹⁹ because of the potential number of interacting components within the interventions, and so owing to the requirement for tailoring of the intervention to the health service and the number of difficulties and behavioural changes from those receiving and delivering the reminder to facilitate decision-making, we developed:

• a conceptual framework (see Figure 2)
• a set of clinical scenarios (see Appendix 7)
• reasons for non-attendance and possible reminder solutions and wider solutions (see Appendix 8)
• advantages/disadvantages for various reminder systems (see Appendix 9).

Modify current ways of delivering appointment reminders

A study carried out last year at two Bedfordshire practices¹⁴² revealed how three simple and cost-effective measures could significantly reduce the number of appointments missed by patients. Asking patients to write down their appointments cut did-not-attends by 18% compared with the average of the previous 6 months. Asking patients who booked over the phone to repeat back the time and date of their appointments cut did-not-attends by 3.5% compared with the previous month. Displaying posters stating how many patients had attended their appointments in the previous month, in combination with the other two measures, cut did-not-attends by a total of 31% in comparison with the average for the previous 12 months.

Further recommendations about how services can cut did-not-attend rates are outlined in the ‘did not attends – Reducing Did Not Attends’ Quality and Service Improvement Tool from the NHS Institute for Innovation and Improvement.²²⁰

Another alternative is to integrate appointment reminders within a more wide-reaching approach to caring for the client by utilising other features of the chosen reminder technology. For example, NHS Bolton supplemented a text-based appointment reminder system within the context of an alcohol relapse prevention programme.²²¹ In addition to allowing clients to confirm whether or not they will be attending their aftercare appointments, clients were contacted a few times a week via text message with simple questions about their feelings and emotional state. The response from the clients triggered, in turn, an appropriate service response depending on the answers given. This innovative use of mobile technology as an engagement strategy is concordant with our earlier observations and seeks to enhance self-sufficiency for clients through reminder and feedback on treatment success.

Utilise new ways of delivering appointment reminders

Utilise new ways of appointment booking

Utilise new ways of managing appointment systems

Utilise alternative ways of delivering health care

Changes to appointment opening times

Experience from the Royal Bolton Hospital NHS Trust,²⁴⁰ which responded to perceived capacity issues by holding extra outpatient clinics out of normal hours, at high cost, indicates the importance of a sophisticated understanding of appointment behaviours in using poor performance as an indicator of systemic issues. The trust encountered a significant number of cancellations and rescheduling of appointments, with the consequence that some clinics were under-running. Furthermore, it identified waste within the appointments themselves. As a consequence, the trust has been working with the Health Foundation to apply innovative ‘lean’ methods to the outpatient setting.