This intervention design, sequentially using phone calls, SMS and postal reminder, proved to be efficient, cost-effective, and well accepted by patients, in an urban primary care clinic serving a majority of vulnerable population.
The intervention significantly reduced the rate of missed appointments. However, subgroup analysis shows that the intervention's effectiveness was only statistically significant in the primary care and the tobacco cessation consultations. While reasons of the lack of efficacy of the intervention in the HIV and dietician consultations are not clear, one explanation could be attributable to low numbers of patients enrolled in the HIV (n = 303) and in the dietician consultation (n = 127). Other characteristics such as patient profiles or patterns in profiles may also have impacted on such differences. Analysis of sociodemographic and medical characteristics of patients attending either the general/nicotine cessation consultation versus the HIV consultation showed that there was a significantly higher percentage of male, insured patients, follow-up consultations of < 1 year, consultations made by senior doctors in the HIV consultation.
The overall 11.4% rate of missed appointments in the control group was lower than estimated based on prior assessments. This fact could be explained in several ways: the rate of missed appointments may have been previously overestimated on the basis of administrative indicators which did not distinguish between missed and cancelled appointments. Seasonal variations in attendance rate have also been described  and an unpublished report about missed appointments in our clinic showed that, in 2001, rates of missed appointments varied between 23.5%, 35% and 16% at three different periods of the year (June, August and March) . However, most probably, the implementation of the intervention itself may have had a positive influence on both intervention and control patients, through information panels and may have increased alertness and awareness of receptionists and patients. A year after the end of the study, the average rate of missed appointments was back to 14%, which calls for the inclusion of such strategies in the administrative routine of outpatient clinics.
Patients enrolled in the HIV outpatient care and in the tobacco cession program had higher baseline rates of missed appointments. Previous studies showed that overall attendance rates appear to be rather low in such populations: up to 30% of HIV infected patients scheduled for a clinic appointment never turned up for an initial evaluation [20, 21] and 35% of scheduled medical appointments were not honoured in an other setting . Optimal attendance can reach 66 to 83% of patients enrolled in nicotine cessation program [23–25].
The success of such an intervention depends on the mobile phone and phone penetration and recording in a given population [26, 27]. In our population, only 51% of the patients had their mobile phone recorded at the clinic and 6% both a mobile and a fixed phone. It is not possible to estimate on such basis penetration rates of mobile phones among our patient population since patients may prefer to give only one phone number. However, while penetration rates of mobile phone range from 20% to 99% over the world (calculated as the% of total telephone subscribers), in Switzerland, mobile phone subscribers represented 64.6% of total telephone subscribers in 2008 . Recent studies showed that, whereas phone calls and SMS are equally effective in reducing the rate of missed appointments in various settings, SMS are more cost-effective [15, 16]. SMS requires significantly less staff resources: messages can be standardised and sent to a large number of patients at low cost . In our setting, the fact that about half of the patients did not provide us with their mobile phone number may have made an intervention exclusively designed on the basis of a SMS reminder less effective than the sequential intervention used. Thus, careful analysis of mobile and fixed phone distribution among patients is essential before choosing and implementing a reminder system.
Missed appointments were significantly associated with younger age, being an asylum seeker, having substance abuse problems, having a follow-up appointment after more than a year These findings are in accordance with prior studies: factors influencing missed appointments usually include patient related factors such young age [7, 28, 29], poor socioeconomic status, and health insurance coverage [5, 10, 30, 31]. Structural aspects also play a role: a long interval between the booking time and the time of the consultation, frequent changes inside the medical team with lack of interpersonal continuity, and difficulty to contact the clinic to cancel or report the appointment, have been identified as factors predicting missed appointments .
Conclusions are conflicting concerning gender, some studies showing a higher rate of missed appointments in men  while others in women . Although most authors observe that patients from lower socioeconomic level or from deprived areas tend to have an increased rate of missed appointments [4, 7, 26, 28], the fact that being an undocumented immigrant was not associated with missed appointments in our study indicates that low socioeconomic status may hide other difficulties such as linguistic, communication or psychosocial issues, and may not be in itself a predictor of missed appointments. Interestingly and contrary to previous studies, being a new patient or having a post-ED visit was not associated with missed appointments [7, 32]. Junior doctors' patients were more likely to miss their appointments than senior doctors' patients. These findings support the assumption that interpersonal continuity may increase attendance rate and build up loyalty, trust and respect [7, 33]. Junior doctors tend to stay for one year of training at our primary care clinic, while senior doctors represent a more stable staff.
Our study has several limitations. First, we did not compare different reminder systems and, therefore, are not able to identify which one works better than the other and for which population of patients. Instead, we chose to test a pragmatic sequential intervention, taking into account the difficulties in reaching our patient population. Second, we are missing information about 57 (8.7%) patients scheduled for an appointment in the intervention group and 93 (5.4%) patients scheduled for an appointment in the control group. These patients were included in the analysis based on intention to treat. Reallocation of free slots was not recorded for appointments cancelled more than 48 hours before the appointments in both intervention and control groups. Finally, the acceptability survey was made by phone and reached only patients who had a phone and responded at the first attempt. It did not distinguish between patients reached by phone, SMS or mail, but one could expect that people at work would be more disturbed by a phone call than by a SMS. Finally, the economic evaluation, which, proved to be economically efficient, similarly to other randomised studies testing text messaging and/or mobile phone reminders only, [2, 11, 16, 34] was limited. It did not take into account the precise number of missed appointments which led to subsequent rebooking. Economic implications may differ according to the health care system: wasted physician time may be more relevant in countries based on a capitation payment system while financial losses may raise more concerns in health care systems based on a fee-for-service model.