Missed Appointments in an Outpatient Clinic for Adolescents, an Approach to Predict the Risk of Missing
Article Outline
Abstract
Purpose
To predict the risk of an adolescent patient to miss an appointment, based on the previous appointments and on the characteristics of the patient and the appointment.
Methods
Two thousand one hundred ninety-three (1873 females) patients aged 12 to 20 years having scheduled at least four appointments were included. We assessed the rate of missed nonexcused appointments of each patient. Second, a Markovian multilevel model was used to predict the risk of defaulting.
Results
Forty-five percent of the patients have not missed even once, and 14% of females and 17% of males have missed >25% of their appointments. Females show two types of behaviors (an abstract concept that groups individuals based on a combination of their appointment-keeping and their recorded type of healthcare need) depending on the diagnosis. Somatic, gynecology, violence, and counseling diagnoses are mostly grouped together. In this group, having already missed and having an appointment with a paramedical provider increases the risk of missing. In the second group (eating disorders and psychiatric diagnoses) having already missed and a longer delay between appointments influence the risk of missing, although the risk is lower for this latter group. Males only show one type of behavior regarding missed appointments. Having missed a previous appointment, being older, having cancelled the next to last appointment and the type of diagnosis explain the risk of missing.
Conclusions
Patients who have already defaulted have a higher risk of defaulting again. Means of control regarding missed appointments should consequently focus on defaulters, to decrease the associated workload. Reminders could be a solution for the follow-up appointments scheduled with a long delay.
Keywords: Adolescent health services, Patient compliance, Health service management, Continuity of patient care
Missed appointments are an issue that has health and economical consequences [1], [2], [3] and concerns mostly young adults and adolescents [4], [5]. Each year 17% of patients of a general practice are reported to default, 66% defaulting, however, only once [5]. Regarding adolescents, and noting that it concerns consultations in dentistry, a study reported that 16.4% of the patients have missed or cancelled 20% or more of their appointments [6]. A previous study conducted in our outpatient clinic for adolescents showed that 22.7% of the scheduled appointments do not occur (11.8% of all the appointments are missed and 10.9% are cancelled) [7]. There are several risk factors associated to missed appointments: poor psychological condition [8], low socioeconomic status [4], not having an established provider [9], [10], suffering from an acute illness [11], [12], and living far from the healthcare center [13]. The studies focusing on adolescents also describe risk factors for missed appointments such as the adolescent scheduling the first appointment by himself [14], first appointments (compared to follow-up ones) [15], and consulting for well adolescent care or “noncosmetic” problems (e.g., headache, infection; compared to “cosmetic” ones, e.g., dermatologic, obesity) [16].
Current efforts to reduce missed appointments can be resource intensive and appear not to be all that effective. In our university hospital, the implementation of a policy (since January 1, 2004) for which patients have to pay for unexcused missed appointments did not decrease the rate of missed appointments [7]. In the literature, some other means of control (such as telephone reminders) that allow to decrease missed appointments have been reported [15], [16], [17], [18]. By characterizing those youth more at risk for missing appointments, the resources could be targeted to this specific population.
The objective of this study is to predict the risk that an adolescent has of missing an appointment based on the history of the previous appointments and on the characteristics of both the patient and the appointment. We hypothesize that patients who have already defaulted will have a greater probability to miss again, but also that there are factors such as a time elapsed since the last appointment, being older, or certain diagnoses that will increase this risk.
Methods
Our clinic was created in 1998, and since January 1, 1999, data about attended, missed, and cancelled appointments have been registered in a computerized database. For this research we analyzed the data from our database, since January 1, 1999 to December 31, 2006. Our university hospital outpatient clinic for adolescents aged 12 to 20 years old in Lausanne, Switzerland, is multidisciplinary, and consists of 11 staff members: three senior physicians, a chief resident, three residents, and four paramedical providers (psychologist, dietician, family planning counselor, and nurse).
Measures
Over the analyzed period of 8 years (1999–2006), 3577 (2947 females) patients aged 12 to 20 years old and representing a total of 32,816 scheduled appointments have consulted in our clinic. For this analysis, we included all patients with at least four appointments (1873 females and 320 males, respectively 62% and 51% of all patients).
Appointments were coded as missed (if not excused, dependent variable), cancelled (when adolescents called to excuse themselves), and kept. Data registered are ID number, gender, age, diagnosis at each attended appointment (not available for missed/cancelled appointments), reason for consulting, and the consultations' data (date, provider, intended follow-up). When the diagnosis was not available for an appointment, we considered the diagnosis of the previous attended appointment.
We considered the following covariates: diagnosis (violence, gynecology, somatic, psychiatry, eating disorders [ED], counseling and dummy-code missed appointment [only when the diagnosis was not available for any of the last three appointments]), age of the patient (12–15, 16–17, 18–20), year of the appointment (as an indicator of pre [1999–2003] and post [2004–2006] implementation of a charge for missed (nonexcused) appointments), time elapsed since the previous appointment (<15 days, 15–30, 31–90, >90), cancellation (three dichotomous variables indicating whether each of the three previous appointments was cancelled or not), and type of provider (four dichotomous variables indicating whether the provider for the present appointment and of each of the last three appointments is a physician or a member of the paramedical staff). Patients' age and diagnoses are reported in Table 1.
Table 1. Distribution of the sample
| Females N (column %) | Males N (column %) | |
|---|---|---|
| Total | 1873 | 320 |
| Age | ||
 12–15 | 604(32.2%) | 114(35.6%) |
| 16–17 | 829(44.3%) | 132(41.3%) |
| 18–20 | 440(23.5%) | 74(23.1%) |
| Diagnosisa | ||
| Violence | 59(3.2%) | 12(3.8%) |
| Gynecology | 702(37.5%) | 0(0.0%) |
| Somatic | 544(29.0%) | 159(49.7%) |
| Psychiatric | 192(10.3%) | 113(35.3%) |
| Eating disorders | 288(15.4%) | 14(4.4%) |
| Counseling | 88(4.7%) | 22(6.9%) |
aPatients can have different diagnoses during their follow-up. To have an idea of the repartition of each diagnosis by gender, we took the diagnosis of the first attended appointment for each patient. |
Covariates to be used at the hidden and visible levels of the model were chosen through a stepwise procedure. The criterion to discriminate between models was the Bayesian Information Criterion [19].
Analyses
Following preliminary results showing a gender difference in behavior, and also because we have much more female than male patients, all analyses were conducted separately by gender.
To predict the probability of missing, the weight of the history of previous appointments, and their impact, we built a statistical model. The outcome was a dichotomous variable indicating whether each appointment was missed or not. In a preliminary step, we used homogeneous Markov chains to determine the number of past observations of the outcome variable that had to be taken into account to explain the probability of missing the present appointment. If we fixed the number of past events used in the prediction to, for example, k appointments, then all patients with less than k +1 appointments would have been lost for the analysis. Consequently, we chose a maximal dependence order equal to three past consultations, the best arbitrage between the number of available data and the predictive power of the resulting model, and we computed models using from zero to three past appointments (t − 1, t − 2, and t − 3) to predict the present one.
First, we assessed for each patient the rate of nonattendance, being the ratio of the missed appointments through all scheduled appointments of the same patient. We then used a Double-Chain Markov Model [20], [21] to describe as accurately as possible the complex process of missed appointments. The Double-Chain Markov Model is used to predict hidden or unknown events in the future from previous events and descriptive variables. This is a two-level model with a hidden level used to discriminate between several “general types of behaviors” (called States, an abstract concept that groups individuals based on a combination of their appointment keeping and their recorded type of healthcare need) regarding missed appointments, and a visible level allowing a precise description of the reasons leading to missing or not missing an appointment. This model has been preferred to logistic regression for two reasons: first, preliminary results showed that the probability of missing an appointment was more accurately estimated by a Markov chain than by a logistic regression, the latter being only a rough approximation of the former model. Second, the two-level structure of the Double-Chain Markov Model allows us to classify our patients on the basis of both their probability of missing and their personal characteristics.
The state corresponding to an appointment at time t depends on the value of the state at the time of the previous appointment of the same patient (State t − 1), and on covariates (Figure 1). The role of the hidden level is to estimate for each appointment the probability that a patient belongs to each type of behavior, and then to assign the patient to the most likely one. A patient is generally mainly associated to only one of the states, but if his general behavior regarding missed appointments changes during the period of observation, the patient is susceptible to also switch to another state.
Figure 1.
The Double-Chain Markov Model is a two-level probabilistic model with a nonobserved hidden part and a visible level. The hidden level is used to classify data into a number of general types of behaviors regarding missed appointments. Each appointment of a patient is linked to the most probable type of behavior in function of the behavior of the same patient during the previous appointment and of covariates. At the visible level, each type of behavior is then fully described through a Markovian regression-like model using covariates and the fact of having or not missed each of the last three appointments to predict the probability of missing the current appointment.
To each state corresponds a different model at the visible level. The observed outcome variable Y, indicating whether an appointment was missed or not, is explained by the same variable observed for the last three appointments. In addition, factors that might influence the attendance (covariates) are used to improve the prediction of missing an appointment. A Markovian regression model called the Mixture Transition Distribution model [22], [23] is used to link past observations of missed appointments and covariates to the probability of missing the next appointment (Figure 1). This model allows a very precise description of the role and importance of each explanatory factor on the outcome variable.
We used the MARCH v 3.0 package (A&A Berchtold, Switzerland) for the computation of all Markovian models. SPSS 14.0 (SPSS Inc., Chicago, IL) was used for additional computations.
As our study was a quality improvement initiative, the Ethics Committee at the University of Lausanne did not require the submission of a protocol.
Results
Rate of nonattendance
Forty-five percent of female patients have attended all their appointments, 41.2% have missed up to 25%, and 13.9% have missed more than 25% of them. For males, rates are, respectively, 44.5%, 41.4%, and 15.9%.
Homogeneous Markov chains
For females, the independence model predicting the probability of missing an appointment without using information from previous appointments was clearly rejected, indicating a dependence process between successive appointments. Among models using data from the previous appointments, the best model according to the Bayesian Information Criterion [19] measure was the one using the last three appointments. Similar results were achieved for males.
Double-Chain Markov Model
We used the Double-Chain Markov Model to incorporate covariates into the model and to obtain a more precise explanation of the process leading to missing an appointment. For females, the optimal model has two hidden states representing two different types of general behavior. The state corresponding to an individual rarely changes, and is mostly similar to the state at the previous appointment. Out of the 1873 females, 1022 (55%) have always been in State 1, 833 (44%) in State 2, and only 18 (1%) have changed state once during the observed period, 13 times from State 2 to State 1, and 5 times from State 1 to State 2. These 18 females are characterized by a very large average number of appointments (39.8) compared to both other groups (9.3 for females associated to state 1 only and 11.6 for females associated to group 2 only), all differences being highly significant (p < .001). The diagnosis is the only covariate that has a significant influence on the state. Somatic, gynecology, violence, and counseling diagnoses are mostly associated to State 1, whereas psychiatric and ED diagnoses are mostly associated to State 2 (Table 2).
Table 2. Hidden level of the Double-Chain Markov Models for females
| Factor | Modality | Appointment at time t | ||
|---|---|---|---|---|
| Weight (95% CI)a | Nonweighted probability of being in State 1 (95% CI)b | Nonweighted probability of being in State 2 (95% CI)b | ||
| State at time t (−1) | 1 | 95%(±1%) | 100%(±1%) | 0%(±1%) |
| 2 | 0%(±1%) | 100%(±1%) | ||
| Diagnosis | Violence | 5%(±1%) | 100%(±37%) | 0%(±37%) |
| Gynecology | 92%(±8%) | 8%(±8%) | ||
| Somatic problems | 71%(±8%) | 29%(±8%) | ||
| Psychiatry | 24%(±12%) | 76%(±12%) | ||
| Eating disorders | 0%(±8%) | 100%(±8%) | ||
| Counseling | 85%(±22%) | 15%(±22%) | ||
| Missed/cancelled | 100%(±8%) | 0%(±8%) | ||
aPercentage of the total explanation attributable to each explanatory factor. |
bNonweighted probability of belonging to each state related to each modality of explanatory factors. |
At the visible level, for the model corresponding to state 1 (somatic diagnoses), the risk of missing an appointment is increased by the fact of having already missed one of the three previous appointments, the amount of time elapsed since the last appointment, not having cancelled the last appointment, having an appointment scheduled with a paramedical provider, and having consulted a paramedical provider at the last appointment. The most prominent of these factors is the intended provider (47% of the total explanation).
For State 2, which is mostly associated to ED and psychiatric diagnoses, having missed one of the three previous appointments, a longer delay and having cancelled the next to last appointment are factors explaining the risk of missing an appointment. Having missed any of the three previous appointments is the most prominent factor with, respectively, 30%, 26%, and 31% of the total explanation. Among girls, the probability of missing an appointment is lower for those with a psychiatric diagnosis. Detailed results are reported in Table 3.
Table 3. Visible level of the Double-Chain Markov Models for females
| Factor | Modality | Model 1 | Model 2 | ||||
|---|---|---|---|---|---|---|---|
| Weight (95% CI)a | Nonweighted probability to miss (95%CI)b | Weighted increase in probability to miss (95%CI)c | Weight (95% CI)a | Nonweighted probability to miss (95%CI)b | Weighted increase in probability to miss (95%CI)c | ||
| Missed at t (−1) | No | 10%(±1%) | 11%(±3%) | Ref | 30%(±1%) | 0%(±2%) | Ref |
| Yes | 75%(±6%) | 6.4%(±0.9%) | 4%(±14%) | 1.2%(±4.8%) | |||
| Missed at t (−2) | No | 2%(±1%) | 11%(±3%) | Ref | 26%(±1%) | 0%(±2%) | Ref |
| Yes | 75%(±6%) | 1.3%(±0.2%) | 4%(±14%) | 1.0%(±4.2%) | |||
| Missed at t (−3) | No | 8%(±1%) | 11%(±3%) | Ref | 31%(±1%) | 0%(±2%) | Ref |
| Yes | 75%(±6%) | 5.1%(±0.7%) | 4%(±14%) | 1.2%(±5.0%) | |||
| Delay | <15 days | 12%(±1%) | 8%(±5%) | Ref | 10%(±1%) | 1%(±9%) | Ref |
| 15–30 days | 44%(±8%) | 4.3%(±1.6%) | 20%(±13%) | 1.9%(±2.2%) | |||
| 31–90 days | 65%(±7%) | 6.8%(±1.4%) | 36%(±14%) | 3.5%(±2.3%) | |||
| >90 days | 0%(±10%) | −1.0%(±1.8%) | 79%(±20%) | 7.8%(±2.9%) | |||
| Cancelled at t (−1) | No | 15%(±1%) | 60%(±3%) | Ref | NS | — | — |
| Yes | 55%(±9%) | −0.8%(±1.8%) | — | — | |||
| Cancelled at t (−2) | No | NS | — | — | 3%(±1%) | 46%(±11%) | Ref |
| Yes | — | — | 83%(±33%) | 1.1%(±1.3%) | |||
| Designed Provider | Physician | 47%(±1%) | 6%(±2%) | Ref | NS | — | — |
| Paramedical | 17%(±3%) | 5.2%(±2.4%) | — | — | |||
| Provider at t (−1) | Physician | 6%(±1%) | 37%(±6%) | Ref | NS | — | — |
| Paramedical | 45%(±9%) | 0.5%(±0.9%) | — | — | |||
aPercentage of the total explanation attributable to each explanatory factor. |
bNonweighted probability of missing the appointment related to each modality of explanatory factors. |
cCorresponding weighted probability with respect to the reference modality. |
For males, preliminary analyses have shown no differences at the hidden level for the general type of behavior; thus, the model is reduced to only one visible regression-like model. The probability of missing an appointment is explained by having missed or cancelled the next to last appointment, having missed the antepenultimate appointment, being older, diagnosis, and time elapsed since last appointment. The most important of these factors is having missed the next to last appointment with 39% of the total explanation. Interestingly, having or not missed or cancelled the immediate previous appointment has no effect and has been excluded from the model. Males have globally a higher probability to miss than both female groups. Results for males are reported in Table 4.
Table 4. Visible level of the Double-Chain Markov Model for males
| Factor | Modality | Weight (95%CI)a | Nonweighted probability to miss (95%CI)b | Weighted increase in probability to miss (95%CI)c |
|---|---|---|---|---|
| Missed at t (−2) | No | 39%(±2%) | 0%(±3%) | Ref |
| Yes | 39%(±10%) | 15.2%(±5.1%) | ||
| Missed at t (−3) | No | 17%(±2%) | 0%(±3%) | Ref |
| Yes | 39%(±10%) | 6.6%(±2.2%) | ||
| Cancelled at t (−2) | No | 8%(±2%) | 0%(±9%) | Ref |
| Yes | 50%(±30%) | 4.0%(±3.1%) | ||
| Age | 12–15 | 13%(±2%) | 0%(±12%) | Ref |
| 16–17 | 30%(±11%) | 3.9%(±3.0%) | ||
| 18–20 | 53%(±12%) | 6.9%(±3.1%) | ||
| Diagnosis | Violence | 18%(±2%) | 48%(±41%) | −1.1%(±9.5%) |
| Somatic problems | 39%(±9%) | −2.7%(±3.8%) | ||
| Psychiatry | 34%(±11%) | −3.6%(±4.1%) | ||
| Eating disorders | 0%(±26%) | −9.7%(±6.8%) | ||
| Counseling | 42%(±29%) | −2.2%(±7.4%) | ||
| Missed/cancelled | 54%(±12%) | Ref | ||
| Delay | <15 days | 5%(±2%) | 36%(±17%) | Ref |
| 15–30 days | 53%(±20%) | 0.9%(±1.9%) | ||
| 31–90 days | 21%(±22%) | −0.8%(±2.0%) | ||
| >90 days | 0%(±31%) | −1.8%(±2.4%) |
aPercentage of the total explanation attributable to each explanatory factor. |
bProbability of missing the present appointment corresponding to each modality of explanatory factors. |
cCorresponding weighted probability with respect to the reference modality. |
Discussion
In our study, the rates of patients having missed more than 25% of their appointments represented 14% of females and 16% of males. Nevertheless, our rate of patients having never missed is lower than those reported in studies with adults [4], [5], confirming that adolescents are, with young adults, at greater risk of missing.
However, the fact that our sample only takes into account patients with four or more appointments could partly explain this lower rate. As a referral center, our clinic receives many patients referred by their primary care provider for a specialized evaluation, and the probability of missing is probably lower, because of the specific status of these appointments. In most cases, these patients, who are followed by their provider after assessment, have less than four appointments and were not included in the model.
For females, ED and psychiatric diagnoses are associated to state 2, which has the lowest risk of missing, even if the patients have previously done it. Although we found no study addressing the attendance of patients with ED, owing to this diagnosis is the self-oriented perfectionism [24], which could explain the particular attendance of these adolescents. Contrary to Cashman et al [8], who have shown that the presence of at least one psychological diagnosis is associated with missing an appointment, our patients with a psychiatric diagnosis have a behavior similar to patients with ED and tend to miss less than females with a somatic diagnosis. The fact that we are not a psychiatric clinic is a possible explanation: patients with complex or more severe comorbidities, who could be more subject to miss, are referred to psychiatry.
All males have a similar behavior concerning missed appointments. This could be explained by the fact that ED diagnoses mostly concern females [25], [26], and that males less often report a need concerning psychological problems and a willingness to use mental health services [27], [28]. Moreover, psychiatric diagnoses are not necessarily similar for both genders. For males, these diagnoses are more frequently linked to substance abuse or dependence, which are probably associated with an increased risk of missing, as demonstrated for drugs and tobacco use [8]. Consequently, few males have the diagnoses that are associated by females to the state 2 (mainly ED and depression).
In all cases, one of the most important factors determining the risk of missing is having missed one or several previous appointments. A possible explanation is the lack of direct adverse consequences for the defaulters. Even if in our clinic adolescents should pay for missed appointments since January 1, 2004, a previous study showed that it does not decrease the rate of missed appointments [7]. The fact that the dichotomous variable “year” distinguishing the periods before and after the introduction of a charge for missed appointments is not significant in any of the models in our study seems to corroborate that having to pay for nonexcused appointments has no effect.
On the other hand, having cancelled one previous appointment influences all models, although the increase in the probability to miss is only significant for males. There are probably different types of behaviors concerning cancellations: some patients cancel a long time in advance because they will not be able to attend, whereas others do it at the last minute because they have forgotten the appointment. Moreover, since January 1, 2004, some patients cancel also at the last moment just to avoid having to pay. We are unfortunately not able to distinguish these different behaviors, what limits our possibilities to explain cancellations.
For females with a somatic diagnosis, delays between 15 and 90 days are associated with a higher probability to miss than shorter delays. When scheduling an appointment with a long delay, adolescents have probably a greater risk of forgetting the appointment, which is a common cause for missing [29], [30]. They also have more time to plan other activities, which will possibly be preferred to the scheduled appointment. Reminders, which have been shown to be efficient concerning missed appointment for adolescents [15], [18], could possibly be used more specifically for appointments occurring after a determined delay. Moreover, focusing on these “high-risk appointments” could decrease the required workload involved in strategies to diminish missed appointments. Delays greater than 90 days, however, are not associated with an increase in risk of missing in this group. We hypothesize that appointments scheduled after such a delay are most probably taken for a new reason for consultation that, similarly to what has been described in adults [31], adolescents would be less likely to miss. That is not always the case for follow-up appointments, which are probably mostly purposed by the provider, as patients may feel the follow-up appointments more frequently unnecessary than the providers, similarly to what has been reported by Wick et al [32] in a study conducted in general practices.
For females belonging to the state of psychiatric diagnoses, a delay longer than 90 days is associated with an increased risk of missing. These patients have a closer follow-up. Consequently, a longer delay possibly means health improvement and a spacing of the appointments more than a new reason for consultation. This improvement could explain that they perceive the appointment as more unnecessary. For all patients, stating very clearly the reason for scheduling a follow-up and avoiding unnecessary appointments could probably decrease missed appointments.
The fact that the intended provider is a member of the paramedical staff significantly increases the risk of missing for females with a somatic diagnosis but not for those in the psychiatric model. Our hypothesis is that, as paramedical providers mainly work part time in our clinic, it is probably more difficult to find a convenient time slot for the appointment. Moreover, the fact that health insurances do not always fully cover all our paramedical providers (and consequently patients have to pay) could also be an explanation. It is also possible that the care relationship established between the psychiatric patients and the paramedical providers (who are mainly the psychologist or the dietician), are different than the relationships with the family planning counselor or the nurse who are probably more frequently consulted by females in the somatic state. Furthermore, concerning the family planning counselor, an increase in the risk of missing might partly be because of the fact that the consultation is more an accessory service offered to the patient than a consultation to treat their health issues. Thus, the patient could feel the appointment as unnecessary.
Paramedical providers meet males in a smaller proportion than females. Although a substantial part of all males' diagnoses are psychiatric diagnoses, substance misuse represents a greater proportion than for females. Thus, they probably consult physicians more frequently than females for psychiatric diagnoses. It is also possible that we did not have enough power to show an influence of the provider for males because of the small number of appointments they scheduled with paramedical providers, and also to the smaller number of males included in the model.
The risk of missing an appointment is greater for older males, in concordance with published research indicating that young adults are at greater risk of missing an appointment than adolescents [4], [5]. However, the same is not true for females. Although a gender difference has been inconstantly reported regarding missed appointments [3], [4], [8], [33], we found no study comparing the age influence separately by gender and we have no explanation for this finding.
Strengths and limitations
The present study is innovative in the sense that, as far as we know, the prediction of the risk of missing an appointment, taking the patients' background into account, have never been studied. Moreover, taking into account cancellations, which are a form of nonattendance, allows us also to have a more precise model. However, some limitations need to be stressed. First, the diagnosis is not registered for the appointments that have not been attended. Thus, in these cases, we had to rely on the diagnoses of the previous consultations. Second, fewer males have consulted in our clinic, what limits our statistical power for males and could limit the establishment of different states at the hidden level. Third, because of the structure of our prediction models, patients with less than four appointments were not taken into account. However, we still consider a large proportion of patients over a period of 8 years. Moreover, our approach was able to put into evidence the influence of past appointments, which would not have been possible otherwise. Finally, Markovian models are mostly used when observations are equally spaced in time, what is not the case here, although similar approaches were already successfully used in the past [34].
Conclusions
As reported in the literature, a majority of our adolescent patients of both genders default at least once. Females have two distinguishable behaviors separating somatic from psychiatric diagnoses. Males show a unique behavior, which is similar to the somatic behavior of females. This should be taken into account for the management of missed appointments. Moreover, providers should be aware of the fact that adolescents could feel an appointment unnecessary; thus, they should always explain the reason for scheduling a follow-up appointment and avoid appointments that are not really necessary. As they have the highest risk of missing the next appointment, our efforts should be focused on those who have already defaulted. Therefore, strictly implemented means of control focusing on those at risk for missing appointments could be a solution that would have the advantage of decreasing the workload. Finally, for long delays between consultations, reminders could be an appropriate solution.
References
- . Understanding non-attendance in outpatient paediatric clinics. Arch Dis Child. 1990;65(2):192–195
- . Prospective controlled study of psychiatric out-patient non-attendance (Characteristics and outcome). Br J Psychiatry. 2000;176:160–165
- . Time and money: effects of no-shows at a family practice residency clinic. Fam Med. 2001;33(7):522–527
- Missed appointments in general practice: retrospective data analysis from four practices. Br J Gen Pract. 2001;51(471):830–832
- . Defaulters in general practice: who are they and what can be done about them?. Fam Pract. 2000;17(3):252–253
- . Dental anxiety and dental avoidance among 12 to 18 year olds in Norway. Eur J Oral Sci. 1999;107(6):422–428
- Missed appointments in an adolescent outpatient clinic: Descriptive analyses of consultations over 8 years. Swiss Med Wkly. 2007;137(47–48):677–681
- . Patient health status and appointment keeping in an urban community health center. J Health Care Poor Underserved. 2004;15(3):474–488
- . Broken appointments. Pediatrics. 1964;34:127–132
- . A field experiment to evaluate various outcomes of continuity of physician care. Am J Public Health. 1974;64(11):1062–1070
- . Failure to keep appointments at a community health centre (Analysis of causes). Scand J Prim Health Care. 1990;8(3):151–155
- . Medical consequences of missed appointments. Arch Intern Med. 1984;144(6):1163–1166
- . Analysis of the patients who missed their appointments at the neurology clinic of the Marina Alta. Rev Neurol. 2002;34(8):701–705
- . Appointment-keeping behavior in adolescents. J Pediatr. 1981;99(5):799–802
- . Telephone reminders improve adolescent clinic attendance: a randomized controlled trial. J Paediatr Child Health. 2002;38(1):79–83
- . Lowering broken appointment rates at a teenage health center. J Adolesc Health Care. 1982;3(2):110–113
- . Effectiveness of telephone reminders in improving rate of appointments kept at an outpatient clinic: a randomized controlled trial. J Am Board Fam Pract. 2001;14(3):193–196
- . Telephone call reminders and attendance in an adolescent clinic. Pediatrics. 1998;101(6):E6
- . Bayes factors. J Am Stat Assoc. 1995;90(430):773–795
- . The double chain Markov model. Commun Stat Theory Methods. 1999;28(11):2569–2589
- . High-order extensions of the double chain Markov model. Stochastic Models. 2002;18(2):193–227
- . A model for high-order Markov-chains. J R Stat Soc B. 1985;47(3):528–539
- . The mixture transition distribution model for high-order Markov chains and non-Gaussian time series. Stat Sci. 2002;17(3):328–356
- . Psychosocial profiles of young adolescent females seeking treatment for an eating disorder. J Adolesc Health. 2007;40(4):351–356
- . Partial eating disorders among adolescents: a review. J Adolesc Health. 2005;37(5):417–427
- . Eating disorders. Lancet. 2003;361(9355):407–416
- Santé et Styles de Vie des Adolescents Âgés de 16 à 20 Ans en Suisse (2002). Lausanne: Institut Universitaire de Médecine Sociale et Préventive; 2004;
- . Stigma starts early: gender differences in teen willingness to use mental health services. J Adolesc Health. 2006;38(6):754–758
- . Appointment-keeping behavior in adolescents: factors associated with follow-up appointment-keeping. Pediatrics. 1993;92(1):20–23
- Reasons for and consequences of missed appointments in general practice in the UK: questionnaire survey and prospective review of medical records. BMC Fam Pract. 2005;6:47
- Missed appointments at a Swiss university outpatient clinic. Public Health. 2007;Jun [Epub ahead of print]
- . Views of patients and physicians on follow-up visits: results from a cross-sectional study in Swiss primary care. Swiss Med Wkly. 2005;135(9–10):139–144
- . Missed and cancelled appointments among 12-18-year-olds in the Norwegian Public Dental Service. Eur J Oral Sci. 1998;106(6):1006–1012
- . Markovian models for the developmental study of social behavior. Am J Primatol. 2002;58(3):149–167
PII: S1054-139X(08)00088-8
doi:10.1016/j.jadohealth.2007.12.017
© 2008 Society for Adolescent Medicine. Published by Elsevier Inc. All rights reserved.
