Interventions to Increase Initial Appointment Attendance in Mental Health Services: A Systematic Review

Nonattendance at initial appointments in outpatient mental health settings has been reported to be almost twice as high as in other specialties (1–4). Initial appointments in mental health services are also less likely than subsequent appointments to be rescheduled, suggesting that many patients do not successfully access help for their problems (5).

Initial appointment nonattendance has been posited to be associated with such factors as having less family support, being younger, and belonging to a racial-ethnic minority group (6). It has also been reported to result in particularly high loss of resources due to wasted clinicians’ time and high administrative costs (7,8).

Currently, there appears to be no clear understanding of how to best address this problem. Moreover, the processes that underlie it are not well understood, which arguably adds a layer of difficulty to designing and evaluating interventions. A systematic review of interventions to increase initial appointment attendance in mental health services is therefore an important step toward understanding and addressing this problem. This review aimed to answer three important questions. What interventions are likely to be most effective in promoting initial appointment attendance? What is the strength of the available evidence for the superiority of various interventions (for example, prompts, reminders, or opt-in systems) over standard procedures? What causal factors are potentially important to address in order to reduce low initial attendance?

A number of reviews have been conducted in this area (9–12). However, they have evaluated only specific types of interventions (for example, opt-in systems) or have focused more generally on early dropout from treatment. We are aware of only one systematic review in this area (12). However, it focused only on the effectiveness of prompts for participants with severe mental illness. A broader systematic review is needed to answer the three questions.

The primary objective of this review was to assess—by identifying randomized controlled trials in this area and synthesizing the findings—the effectiveness of interventions that aim to increase initial appointment attendance in adult outpatient mental health services compared with standard procedure. Second, the review aimed to assess effects of these interventions on other outcomes (for example, cancelled appointments and treatment completion). Third, the review sought to determine whether studies reported an a priori theoretical model justifying the tested intervention and to record which models were most commonly cited.

Methods

Data sources

Before this review was undertaken, a protocol was registered in an online database (13). The online databases MEDLINE, EMBASE, PsycINFO, CENTRAL, British Nursing Index, and CINAHL were systematically searched on June 18, 2012, from the earliest date of each database. The electronic search strategy combined participant and setting keywords (for example, “mental health” or “psychiatry”) and subject headings (for example, “mental disorders”) with intervention or outcome keywords (for example, “opt-in” or “initial attendance”) and subject headings (for example, “reminder systems”). These keyword and subject heading combinations were used with the Cochrane Highly Sensitive Search Strategy for identifying randomized controlled trials (14). No restrictions were made in terms of language, publication status (peer reviewed or non–peer reviewed), and publication date. Five review papers were also identified (9–12,15), and their reference lists were screened for potential studies.

Selection of trials

All randomized controlled trials (RCTs) that tested an intervention against standard care with the explicit aim of increasing initial appointment attendance (or decreasing nonattendance) in adult mental health services were eligible for inclusion. It was also decided to include “quasi-randomized” controlled trials (in which participants are allocated in a way that is similar to randomization, such as alternation) and to investigate with sensitivity analyses the effect of including them in the review, if possible. Studies were eligible if the comparison group received the standard procedure, which was defined as no contact or standard appointment letter.

Studies in which participants were children were excluded because the factors that affect initial appointment attendance are likely to be different when a caregiver is substantially involved in the process. Studies focusing on initial attendance at group appointments were also excluded because it was thought that the barriers to attending a group session and a one-to-one session might be different (for example, not being able to cancel or reschedule a group session).

Two authors (OS and NA) independently screened the 2,418 titles and abstracts that resulted from the searches. A total of 144 reports were obtained for all studies designated by the reviewers as potentially relevant. Four of the full-text articles could not be obtained and could therefore not be assessed for eligibility (16–19).The reference lists of the 144 full-text articles were also screened for additional studies. Disagreements were settled through discussion, and if a consensus could not be reached, a third author (LA, SB, or SC) acted as an arbiter. As a result, 21 studies were included in the review. [A flow diagram illustrating the search procedures is available as an online data supplement to this article.]

The methodological quality of the studies was assessed with the Cochrane Risk of Bias Tool (20). Each study that was rated “unclear” or “high” on each of the risk domains (selection, information, attrition, and publication bias) was removed when possible in the sensitivity analyses.

Data extraction and analysis

Data extracted from the studies onto a data extraction sheet included sample characteristics and size, intervention and comparison group details, study design, and outcome. Outcome was recorded as the number of participants in each group who attended, did not attend, rescheduled, or cancelled. Data extraction and methodological quality assessment was undertaken independently by three authors (OS, LA, and SB), and any discrepancies in data extraction were discussed by the review team.

Five studies had multiple intervention arms (8,21–24). For these studies, the data from the intervention arm that was deemed to be most resource intensive or most different from the standard procedure were used.

The primary outcome for this review was the number of participants attending the initial appointment session. If a report indicated only the number not attending the initial session, and the number of people who attended could not be obtained from the authors, the study was retained in the review but was excluded from further analyses. The assumption that people who did not fall into the “did not attend” category actually attended their scheduled appointment (as opposed to cancelling or rescheduling) was not deemed to be valid. It is possible that some participants cancelled or rescheduled appointments as a result of the intervention or did not receive an appointment in the first place (that is, they had to opt in). The outcome of attendance was thus chosen specifically to investigate whether these interventions result in increased access to mental health services because studies reporting only nonattendance rates might be filtering out people who are ambivalent about attending. Treatment completion rates, subsequent hospitalizations, and the number of cancelled and rescheduled appointments were also extracted if they were reported.

The primary outcome was analyzed on an intention-to-treat basis in which the probability of attending (all participants who attended in the intervention group divided by the number randomly assigned to that group) was divided by the probability of attending in the control group (all participants who attended in the control group divided by the number randomly assigned to that group). A risk ratio (RR) greater than 1 thus indicated that the intervention increased the probability of attending, whereas a RR of less than 1 indicated that the proportion of participants attending was higher in the control group.

Studies included in the analyses were divided into groups by type of intervention (for example, telephone reminder, written reminder, or opt-in system). The results were analyzed separately within each intervention group, because analyzing all interventions together was not deemed clinically meaningful.

To determine whether meta-analyses (pooling outcome statistics) were appropriate in a given intervention group, statistical heterogeneity was assessed with the I² and Cochrane’s Q statistic. If the I² statistic was <50% and the Q statistic was nonsignificant at an alpha level of .1, a random-effects model was applied. This method of synthesis was chosen because it is reasonable to assume that studies that evaluate complex interventions are estimating related effects but not exactly the same effects (25). If at this point, however, meta-analyses were deemed inappropriate (for example, because of a high degree of heterogeneity), a narrative synthesis was undertaken. This entailed summarizing (but not pooling) effect sizes of each study in a table and in a forest plot. Similarities and differences between studies were then discussed.

Results

The 21 studies included in this review are summarized in Table 1. The studies were conducted in a range of settings (community mental health centers, hospital outpatient settings, substance abuse clinics, and psychotherapy services) and included data from a total of 5,043 participants. Nine studies were conducted in the United States, ten in the United Kingdom, one in Spain, and one in New Zealand. The studies involved several categories of interventions: four tested opt-in systems (in order to receive an appointment, the participant had to first take some action), six tested telephone call prompts or reminders (telephone contact with the participant before the appointment), five tested orientation or reminder letters (written contact with the participant before the appointment), two tested accelerated intake (reducing time between referral and the first session), two tested sending a psychodynamic questionnaire to the client before the first appointment, and two tested other types of intervention. These other types of interventions were implementation intentions (asking clients to formulate an “if-then” plan to deal with ambivalence toward attending) and giving clients the opportunity to choose the therapist’s style. [A table presenting details of the intervention and control conditions in each study is available in the online data supplement.]

Risk of bias assessment

All the included studies were assessed for risk of bias (Table 2). Many of the studies were brief, with little information to assess the risk. For one study, the authors provided additional details (26). A substantial proportion of the studies were rated as having a high risk of bias in terms of random sequence generation and allocation concealment (selection bias), as well as incomplete outcome data (attrition bias). Most of the studies did not address blinding procedures (detection and performance bias), and little information was available (for example, from study protocols) to enable the assessment of selective outcome reporting (reporting bias).

Effects of the intervention

Sixteen of the 21 studies (3,673 participants) were included in the quantitative analyses. Five studies were excluded from the analyses because they reported only the proportion of participants who did not attend the initial appointment (8,21,24,27,28). The results are presented in Table 3.

Meta-analyses were not implemented in the analysis because, even though the accelerated-intake studies and the studies that used orientation and reminder letters fulfilled criteria for sufficiently low statistical heterogeneity, it was decided that the high clinical and methodological heterogeneity (for example, variation across settings and time periods) made pooling the effect sizes unjustified (Table 3). [A figure presenting forest plots of the studies is available in the online data supplement.] Narrative synthesis was thus implemented for all intervention groups. Because this method was used, separate subgroup or sensitivity analyses were not possible, but the similarities and differences between the studies were discussed in the context of the main analysis.

Opt-in systems did not appear to have a significant beneficial effect, although one study from the United Kingdom reported a significant increase in initial appointment attendance (RR=1.64, 95% confidence interval [CI]=1.06–2.55) (29), and one study from the same psychotherapy setting showed a nonsignificant result in the same direction (29). Both studies had a relatively low risk of bias (Table 2). One older study from a U.S. clinic for people with marital and sexual difficulties showed a statistically nonsignificant finding in the direction of decreasing initial appointment attendance (30). This study’s risk of bias was mainly unclear.

Orientation and reminder letters appear to show some promise in increasing initial appointment attendance. Three of four studies indicated a beneficial effect, which ranged from a 9% to 44% increase in attendance (31–33). The three studies were conducted in the United Kingdom. However, only one, with mainly an unclear risk of bias, reported findings that were statistically significant (32); the findings of another study, with mainly a low risk of bias, showed a trend toward significance (33). Notably, one older unpublished U.S. study did not indicate any benefit of an orientation and reminder letters (34). This study also had mostly an unclear risk of bias. It should be noted that these studies were conducted over a span of more than 30 years, which may account for the variation in findings between studies.

Telephone prompts do not appear to significantly increase initial appointment attendance. The results were highly variable across studies, with one Spanish study reporting that the intervention significantly increased attendance by 26% (35). For the findings of two studies, intention-to-treat analyses indicated that telephone prompts and telephone reminders may decrease attendance. Findings in one of these studies, from New Zealand, were significant and reported a decrease of 37% in attendance after an intention-to-treat analysis was conducted (36). This study also had a high risk of bias. The third study, from the United States, which was also the only one in this group that was conducted in a substance abuse service, reported a 5% decrease in attendance, but the difference was not statistically significant and the risk of bias was mainly unclear (37).

Accelerated intake may also be a beneficial method of increasing initial attendance. Two small studies both indicated a beneficial effect, ranging from an increase of 18% to 60% in initial appointment attendance (22,23). However, these increases were not statistically significant. Both studies were conducted in U.S. substance abuse settings and had mainly an unclear risk of bias.

In two U.K. studies participants were sent a psychodynamically formulated questionnaire before their first psychotherapy appointment. One found a nonsignificant effect in the direction of increasing attendance (38), and the other showed an effect in the opposite direction that was nearly significant (39). Both studies had an unclear or high risk of bias on most of the domains.

Studies of interventions in the “other” category reported statistically significant or nearly significant results (Table 3). Findings of a U.S. study that tested the effect of clients being able to choose the therapist’s style showed a trend toward significance (RR=1.57, CI=.99−2.50) (40). This study had mainly an unclear risk of bias. A fairly recent U.K. study that tested sending participants an implementation intention exercise found that initial appointment attendance at a psychotherapy service increased by about a third, and the result was statistically significant (26). This study had a low risk of bias across domains.

Five studies reported the number of cancelled initial appointments (29,30,33,38). No differences were found in the number of cancelled appointments between the control and intervention groups in any of these studies. Furthermore, two studies reported the number of rescheduled initial appointments (31,38). Neither study showed a significant difference between the groups.

Only one study, which used an orientation letter, reported the number of people who completed treatment, as well as rate of subsequent hospitalizations and use of the Mental Health Act (involuntary commitment) (33). The intervention had no significant effect on any of these outcomes. It should be noted that during the follow-up period, very few people in this study were reported to be hospitalized, and no one in the sample was admitted under the Mental Health Act.

Theory

Only three studies explicitly specified using a model to guide the design of the intervention (22,26,34). Sheeran and colleagues (26) used the theory of planned behavior and Witkower (34) alluded to the crisis theory. Stark and colleagues (22) also mentioned a model by Miller (41), which emphasized that clients’ current motivation—more than their stable dispositional traits—was as an important determinant of treatment engagement. The other studies did not explicitly state a model in the introduction or discuss the mechanisms behind attending initial appointments as a basis for their chosen intervention.

Discussion

Summary of findings

The analyses indicate that orientation and reminder letters may increase initial appointment attendance in mental health services. In contrast, the use of telephone prompts does not appear to be significantly more effective than standard procedure. However, the results were mixed, with significant results in both directions (35–37). Similarly, opt-in systems did not increase initial appointment attendance, with one exception reported in a study with a relatively low risk of bias (29). Accelerated intake was also not found to be more effective than standard intake procedures in the two small studies reviewed (22,23). In addition, sending a psychodynamically formulated questionnaire to the participants before the first therapy session appears to have had no significant effect (38), and one study of such an intervention indicated decreased attendance, with a trend toward significance (RR=.87, CI=.76−1.00) (39).

Although the effect was not statistically significant, a choice of therapist style showed some promise as a means of increasing attendance (RR=1.57, CI=.99−2.50) (40). Larger and more rigorously controlled studies are needed to clarify this finding. Moreover, it should be noted that this intervention is directly applicable only in a psychotherapy setting.

One study that used an implementation intention exercise administered by mail found that the intervention increased the likelihood of initial appointment attendance by approximately a third (26). Although this is a promising finding from a study with a low risk of bias (Table 2), further research is needed to investigate whether this type of intervention can be effective across settings.

Only a few studies reported a secondary outcome. The analyses suggest that interventions to increase initial appointment attendance do not have a substantial effect on the number of cancelled or rescheduled appointments. For the other outcomes, data were available only for isolated studies or not at all.

Overall applicability of evidence

The findings of this review suggest that robust evidence is lacking for interventions to increase initial appointment attendance. It should be noted that the most common types of interventions identified in this review (opt-in systems, telephone prompts, orientation and reminder letters, and accelerated intake) showed either small significant or nonsignificant results. However, many of these interventions are low cost and logistically simple. Even small improvements in attendance can prove to be cost-effective. More research is needed. Individual studies that used more novel methods for increasing attendance (for example, an implementation intention induction) show some promise but require replication.

The findings of this review suggest that health care providers should be cautious when investing in these methods because they may not yield substantial benefits. However, some interventions may prove beneficial in decreasing nonattendance at initial appointments and thus free up administrative and clinical resources. For example, appointments would not need to be scheduled for people who would be likely to not attend (for example, opt-in systems). However, such interventions would not address the problems outlined at the beginning of this article, which are related to increasing access to services for those who need them. For this reason, it is crucial that evaluations of these interventions use appropriate outcome measures.

Most of the studies appear to have been designed with no explicit reference to a theory or a model. Therefore, they are of limited utility to research investigating causal factors underlying initial appointment attendance. A model is arguably also crucial when complex interventions are developed and evaluated (42,43).

The broader help-seeking literature would suggest that factors such as stigma (44) and fears regarding treatment (45) are important determinants of propensity to seek help for mental health problems. In addition, it has been argued that increased control over the process of accessing help may be crucial in order to help people reach the goals that are important to them (46). Furthermore, these obstacles and facilitators should be placed in the context of societal and political influences (47). Future interventions could focus on all these aspects.

Strengths and limitations

A strength of this review was that considerable effort was made to counter publication bias by identifying unpublished studies and non–English-language reports. Moreover, compared with previous literature reviews, this review showed that beneficial effects often do not hold when an intention-to-treat analysis is performed and when initial appointment attendance, rather than nonattendance, is used as an outcome (9,11,15). This review also thoroughly evaluated the risk of bias in the studies included, which indicated that the quality of current evidence for each intervention type is generally poor to modest and that better-quality research is necessary. However, a few trials of higher quality were identified, and it is important to acknowledge them and contrast them with the lower-quality studies so that their findings can be used to guide policy or future research.

The number of studies of each intervention type was relatively small. Thus robust evidence in favor or against these interventions could not be obtained. Furthermore, many studies were relatively small, and some included several intervention arms, which led to low power (especially when intention-to-treat analyses were performed). A further limitation with regard to pooling different studies is that the comparison groups differed somewhat across the studies. It also important to note that the studies reviewed span nearly four decades and include participants with different types of mental health and substance abuse problems. Although successful randomization should control for differences within the study, the reasons for low attendance may vary over time and across clients. Generalizing across findings should thus be done with caution. To acknowledge this variation, we implemented a narrative synthesis rather than a meta-analysis. However, this method of analysis, along with the small number of studies, meant that the variance could not be investigated further in subgroup analyses.

Many of the reports were also brief and provided little detail about methods. Attempts to contact the authors for further information were often not successful. Four studies could not be assessed for eligibility because they were unavailable (16–19). Outcomes were not extracted for five studies (8,21,24,27,28). The results of this review may have been different if appropriate outcomes had been obtained for these studies.

Conclusions

The review found no conclusive experimental evidence of the effectiveness of telephone prompts, accelerated intake, psychodynamic questionnaires, and opt-in systems in increasing attendance at initial appointments. Orientation and reminder letters appeared to increase attendance, and this approach may be an attractive alternative for many services because it is simple and relatively low cost. However, because the effect appears to be small, a cost-effectiveness study is recommended. In addition, research to identify mechanisms that influence first appointment nonattendance could help inform development of more effective orientation and reminder letters.

Giving patients more control over the process through procedures such as opt-in systems or therapist choice is in line with the current policy of promoting patient preferences and meeting their needs (48) and may therefore be beneficial from a health policy perspective, even though such interventions may not have a significant effect on attendance. Furthermore, more attention to the societal and political context in which these interventions are implemented is crucial to their effectiveness. Current political and financial policies should thus be taken into account, as well as the extent of stigma in society (47).

More methodologically robust studies are needed to conclusively determine the effectiveness of any of the identified interventions. An important methodological point highlighted in this review is that it is crucial to define—and sufficiently describe—outcomes used in evaluations of such interventions. In addition, more attention should be paid to identifying important causal factors and psychological processes behind the phenomenon of attendance; interventions should be designed and evaluated with explicit reference to these factors and processes. Because many of the conventional methods do not appear to be very effective, the findings of this review suggest that it may be the time for more novel approaches.