In the rapidly changing medical environment, healthcare efficiency and patient’s satisfaction are usually regarded as important indicators of quality of healthcare services.1 In the ambulatory sector, a well-operated appointment system could save patients’ wait time, enable healthcare providers to allocate resources according to the number of patients, and thus enhance healthcare quality.2 On the contrary, missed appointments, ie, appointments neither attended nor canceled by patients, have a great impact on healthcare efficiency by incurring the inactivity of professionals and equipment and decreasing the availability of resources to other patients.1,3–7 A study in the USA revealed that missed appointments could cause 16% to 54% operating loss for medical insurance.8 Because of different insurance systems and clinic features, the rates of missed appointments (“no-show rates”) varied between 15% and 30%.7,9,10 A 50% no-show rate had even been reported in primary care clinics and 60% in psychiatric clinics.3,4,6,7,10–14
The National Health Insurance (NHI) in Taiwan has been implemented since 1995, with 20 759 contracted healthcare facilities and a 99.6% coverage of the population in 2016. An NHI beneficiary pays an average of 15 ambulatory visits annually, far above the average of 6.9 visits in OECD (Organization for Economic Cooperation and Development) countries.15,16 The high utilization of healthcare services in Taiwan makes a well-operated appointment system indispensable.
To overcome the problem of missed appointments, hospitals in Taiwan often restrict the appointments by patients who have frequently missed appointments but this measure would lead to patients’ dissatisfaction and more work for on-site registrations. The literature indicated that the appointment system should consider the local medical environment and population structure that might affect the efficacy of interventions.11 Although most studies of missed appointments focused on a single specialty with fewer data and shorter periods of observation, no large-scale empirical analysis of the situation in Taiwan was available. The aim of this study was to identify the factors of missed appointments with a large dataset involving multiple specialties of an academic medical center. Our results might serve as the basis for designing a suitable appointment system and facilitate the efficiency of healthcare utilization and the equity of resource allocation.
2.1. Data sources
The data were the registration records of the outpatient department of an academic medical center in Taipei City, Taiwan, from 2014 to 2015. The patients were de-identified for further analysis. Besides regular details, a registration record also contained additional information about the consequences: cancellation or completed attendance. This study has been approved by the Institutional Review Board of Taipei Veterans General Hospital (IRB number: 2017-07-008-AE).
2.2. Study design
This was a cross-sectional study of registrations in 2015. Eligible records did not include the canceled appointments, invalid registrations, and administrative anomalies. The records were categorized into two groups: (1) appointments with completed attendances; and (2) missed appointments. To capture the influential factors on each visit, the patients were assumed to make their decision on attendance to each appointment independently, and all appointments were analyzed at the record level using a logistic regression model. That is, we assumed a single patient makes different decision on each of his appointment with 15 influential variables including patient characteristics, appointment characteristics, and weathers characteristics taken into analysis. Logistic regression was used to calculate the adjusted odds ratio (AOR) of each variable. For nonfirst visits, we further built a logistic regression model with the most influential variables and the additional personal attendance pattern of the previous three appointments.
2.2.1. Patient characteristics
The variables included the patient’s gender and age. The patients’ age were calculated based on the beginning of 2015 (January 1, 2015) and categorized into decades. As birthday is required to make appointments, there are no missing values on age.
2.2.2. Appointment characteristics
The variables included the patients’ past visit history (visits made in 365 days before the current appointment), days of the week, the session of day, the rank on the waiting list and the number of appointments on the same day of the appointed visit, the method of appointment, the interval between dates of appointment making and appointed visit (appointment-to-visit interval, wait time), the department to attend, the first visit or not, and the interval between dates of the appointed visit, and the previous visit to the same department.
According to the list of options on the appointment system, the clinical departments were categorized into medicine, surgery, obstetrics/gynecology/pediatrics, ophthalmology/otorhinolaryngology, affiliated clinics, and others (Supplementary Table 1). The appointments could be made by physicians, on discharge from ward, on the registration counter, on-line and via telephone (voice-guided or with aids of the staff). The first visits in the current study were operationally defined as those without preceding appointment or attendance in the past 365 days.
2.2.3. Weather characteristics
The weather was at first divided into cold (10°C-15°C), cool (16°C-19°C), comfortable (20°C-26°C), warm (27°C-30°C), and hot (above 30°C) according to the 2015 average daily temperatures of Shipai Area in Northern Taiwan that were published by Taiwan Central Weather Bureau. On the basis of the daily precipitation, the weather was additionally divided into no rain (0 mm), rain (0-80 mm), and heavy rain (>80 mm).
2.2.4. Personal appointment patterns for nonfirst visits
For each appointment, the personal appointment pattern was operationally defined as the status of the previous three appointments by the patient. The pattern was displayed with three binary digits where the first digit represented the status of the earliest appointment of the previous three appointments and the third digit represented that of the last appointment. As the notation, 0 represented a missed appointment, 1 represented completed attendance(s), and the underscore ‘_’ represented not available.
2.3. Data processing and statistical analysis
We used Microsoft SQL Server 2016 for data linkage and extraction, Stata release 15 for statistical analysis, and Microsoft Excel 2016 for producing tables and figures. At first, we performed the univariate analysis and calculated no-show rates in each variable. Then, we used the multivariate logistic regression to calculate the AORs of missed appointments in each variable. Within each variable, the item with the minimal value served as the reference. For nonfirst appointments, we further built a logistic regression model with the personal attendance pattern of the previous three appointments and those variables having item(s) with an AOR > 1.5 in the previous regression model. In the current study, we set the heuristic threshold of effect size of AOR at 1.5 to select both statistical significance and clinical meaningful predictors in our model. The 95% CIs were calculated based on normal approximation to represent statistical significances. The adjusted no-show rates were calculated for appointments of each characteristic as they were average on all other characteristics. Those rates were interpreted as the predicted probabilities of missed appointments if they were average in all other respects. A two-tailed level of 0.05 was considered statistically significant.
In 2015, a total of 2 132 577 appointments with 1 771 416 completed attendances were included for analysis. The overall no-show rate was 16.9% (n = 361 161). Among four-patient variables, higher no-show rates were observed in female (17.6%), younger than 40 years (22.0%-29.2%), ≤ 6 visits (19.5%) and a no-show rate higher than 50% in the previous year (39.0%-41.0%). Among nine-appointment variables, higher no-show rates were observed in appointed visits on Fridays and Saturdays (18.4%-19.1%), in the evenings (23.7%), on the last third of the waiting list (17.9%), only one appointment on the same day (17.7%), on-line appointments (20.1%), appointment-to-visit intervals in 7 to 14 days (19.7%), appointments to obstetrics/gynecology or pediatrics (26.2%), first visits (19.0%), and the interval between the appointed visit and the previous visit to the same department ≤7 days (22.0%). Of two weather variables, warm weather (17.9%) and heavy rain (51.4%) had higher no-show rates (Table 1).
Furthermore, we used the multivariate logistic regression model to identify the independent predictors of missed appointments. Of 15 variables of patients, appointments and weathers, only gender had no significant effect on no-show rates (p = 0.877) (Table 2). Five variables had an item with an AOR > 1.5: age, no-show rate in the previous year, appointment-to-visit interval, the interval between the appointed visit and the previous visit to the same department, and rain.
Of all appointments in 2015, nonfirst appointments accounted for 82.9% (n = 1 767 122). For them, we performed the multivariate logistic regression model again by involving the above-mentioned five variables plus the personal appointment pattern. All six variables remained statistically significant after adjustment. The influences in decreasing order on the basis of AOR were heavy rain (AOR: 6.223, adjusted no-show rate: 51.1%), shorter interval (<7 days) between the appointed visit and the previous visit to the same department (AOR: 3.906, adjusted no-show rate: 34.6%), higher no-show rate (75%-100%) in the previous year (AOR: 3.274, adjusted no-show rate: 34.3%), total absence in the personal attendance pattern of the previous three appointments (AOR: 3.051, adjusted no-show rate: 31.9%), longer appointment-to-visit intervals (≥ 57 days) (AOR: 2.537, adjusted no-show rate: 22.1%), and children under 10 years (AOR: 2.005, adjusted no-show rate: 25.0%) (Table 3). The differences of items in each of the six variables were displayed in the Fig. 1.
This study demonstrated that with the exception of patients’ gender all of the other characteristics of patients, appointments, and weathers played roles in no-show rates. We believed that our findings could apply to other academic medical centers in Taiwan and the identified factors could be targeted for interventional measures to improve the appointment system and thus decrease the no-show rates.
The overall no-show rate in our study was 16.9%, comparable with 14% to 20% reported in other academic medical centers in Taiwan.14,17,18 Among the methods of appointments in our study, more than half of appointments were made by physicians, but on-line appointments had the highest no-show rate. This finding was consistent with that reported by Wang et al.17 It was possible that physicians usually judged the patient’s condition and then made appointments for follow-ups. On the contrary, the patients might not have professional recommendations before making on-line appointments. Although medical departments accounted for the majority of appointments in our study, obstetrics/gynecology and pediatrics had the highest no-show rate. One of the reasons might be that medical departments had a larger part of patients with chronic diseases who had frequent visits for refills and thus higher compliance rates.19 In our study, appointments for first visits had a higher no-show rate than those for nonfirst visits. Some studies pointed out that new patients had a longer wait time that lead to a higher no-show rate.3,7 Although most healthcare facilities offered special reservations for new patients, it seemed that such measures did not meet the patient’s needs enough.
In our study, nonfirst appointments accounted for 82.9% of the total appointments and were independently analyzed with the five most influential variables plus the personal appointment pattern of the previous three appointments. We found that younger patients had a higher no-show rate, a finding similar to those in other studies.2,3,13,14 In general, young patients had relatively better health conditions and fewer chronic diseases but less time for attendances because of school, job, and family responsibility.10,13,14,20 Furthermore, missed appointments might become habitual. Some studies revealed that patients with a history of missed appointments were more likely to miss again.17,21 Our study shared similar findings: no-show rate in the previous year had a great influence on the current no-show probability. We also found the personal attendance pattern of the previous three appointments could be another good parameter for identifying habitual no-show patients. The more missed appointments in the previous three appointments, the higher no-show rate in the current appointment. Rachel et al. had also set up a prediction model and found that patients missing three consecutive appointments in 10 appointments had higher no-show rates than those missing three inconsecutive appointments in 10 appointments.5
Among appointment characteristics in our study, a longer wait time for appointments was associated with a higher no-show rate. In a study of referrals to a psychiatric clinic, it was demonstrated that one additional week of wait time for appointments would increase 7% of no-show rate.22 Another study also showed that the no-show rate for appointments with a wait time of more than 2 months could reach 32%.9 Facing a longer wait time, patients might become anxious and turn to other physicians who could offer an earlier appointment,1,3,5,7,13,23 what was also a common phenomenon under Taiwan’s NHI system without strict regulations of referrals. In addition, because the NHI allowed patients with chronic diseases to have refill prescriptions for 3 months, physicians might accordingly make appointments for 3 months later that could be hardly kept by patients. In our study, the shorter interval (<7 days) between the appointed visit and the previous visit to the same department was also associated with a higher no-show rate. The possible scenario could be that after treatment patients felt better and did not consider follow-ups in the short term as necessary.
In our study, the weather, especially rains, exerted a far greater influence on no-show rates than other factors did. After adjustment, the no-show rate could reach 50% in a heavy rain. Past studies had similar findings.2,13,21 However, one other study revealed that typhoons had no significant influence on no-show rates.19 For patients, perhaps typhoons and temperatures were more predictable than rains.
From our study results, we recommended to develop more accurate prediction models and design more appropriate interventional measures by adopting the identified factors of missed appointments. To solve the problem, some literature had proposed the method of overbooking by adding the estimated amount of missed appointments to the upper limit of reservations.2,6,10,13,21 In case of inaccurate prediction, overbooking would prolong wait time, causing dissatisfaction and more missed appointments. It was thus essential to target habitual no-show patients.13 While Giachetti4 suggested a double-booking procedure in which another patient was arranged at the same time of the appointment for a habitual no-show patient, Izard24 suggested arranging high-risk patients behind appointments for regular patients.
Our study had some limitations. First, we only analyzed the related factors of missed appointments and did not ask the patients about the causes of absence. Second, the outpatient registration files did not contain any information about physicians. The patient’s address and sociocultural characteristics were not included, either. Finally, the reasons and diagnoses of visits were, unfortunately, unavailable for analysis. Third, we assumed that each patient makes each decision on attendance independently to his other appointments and analyzed our appointment at record level. Neglecting clustering effect (number of visits made by each patient) would deviate estimation on odds ratio. However, this effect should be minimal to this study because of a small repeated number made by those patients (each patient made only 5.8 visits in average). Moreover, our prediction model did not include any information regarding to contents in an appointment. The inclusion of more information such as medication, procedures, test, and surgery might probably improve prediction power theoretically but lack practical use. Because of operational reason and data protection, the aforementioned clinical information is usually unavailable for a registration system and thus cannot be used as potential predictors.
In conclusion, our study illustrated the diverse factors of missed appointments and especially identified the predictive values of rain, the interval between the appointed visit and the previous visit, and patient’s personal history of missed appointments. The NHI in Taiwan offered better healthcare services in a convenient and affordable way. Yet the highly competitive environment seemed to foster missed appointments because healthcare facilities might refrain from taking restrictive measures. Developing more accurate prediction models could enhance the efficiency and quality of healthcare without sacrificing the patient’s satisfaction.
This study was supported by a grant from the Taipei Veterans General Hospital, Taipei, Taiwan (V107C-019).
APPENDIX A. SUPPLEMENTARY DATA
Supplementary data related to this article can be found at https://links.lww.com/JCMA/A18.
1. Shrestha MP, Hu C, Taleban S. Appointment wait time, primary care provider status, and patient demographics are associated with nonattendance at outpatient gastroenterology clinic.J Clin Gastroenterol201751433–8
2. Yeh JB, Tung CF, Chang WH. The application of the neural network to the forecasting of missed hospital appointments.Taiwan
J Public Health200928361–73
3. Davies ML, Goffman RM, May JH, Monte RJ, Rodriguez KL, Tjader YC, et al. Large-scale no-show patterns and distributions for clinic operational research.Healthcare (Basel)2016415
4. Giachetti RE. A simulation study of interventions to reduce appointment lead-time and patient no-show rate.In: Proceedings of the 2008 Winter Simulation ConferenceDecember 2008IEEE, FL1463–8
5. Goffman RM, Harris SL, May JH, Milicevic AS, Monte RJ, Myaskovsky L, et al. Modeling patient no-show history and predicting future outpatient appointment behavior in the veterans health administration.Mil Med2017182e1708–14
6. Reid MW, May FP, Martinez B, Cohen S, Wang H, Williams DL Jr, et al. Preventing endoscopy clinic no-shows: prospective validation of a predictive overbooking model.Am J Gastroenterol20161111267–73
7. Tseng FY. Non-attendance in endocrinology and metabolism patients.J Formos Med Assoc2010109895–900
8. Huang Z, Ashraf M, Gordish-Dressman H, Mudd P. The financial impact of clinic no-show rates in an academic pediatric otolaryngology practice.Am J Otolaryngol201738127–9
9. Cohen AD, Dreiher J, Vardy DA, Weitzman D. Nonattendance in a dermatology clinic–a large sample analysis.J Eur Acad Dermatol Venereol2008221178–83
10. Cohen AD, Kaplan DM, Kraus M, Rubinshtein E, Vardy DA. Nonattendance of adult otolaryngology patients for scheduled appointments.J Laryngol Otol2007121258–61
11. Daggy J, Lawley M, Willis D, Thayer D, Suelzer C, DeLaurentis PC, et al. Using no-show modeling to improve clinic performance.Health Informatics J201016246–59
12. George A, Rubin G. Non-attendance in general practice: a systematic review and its implications for access to primary health care.Fam Pract200320178–84
13. Sharp DJ, Hamilton W. Non-attendance at general practices and outpatient clinics.BMJ20013231081–2
14. Chou YT, Wu LC, Chen CY. Who fails appointments at psychiatric clinics in a general hospital in Taiwan
15. National Health Insurance Administration, Ministry of Health and Welfare, Executive YuanNational Health Insurance 2016–2017 Annual Report2016TaipeiNational Health Insurance Administration, Ministry of Health and Welfare, Executive Yuan
16. OECDHealth at a glance 2017: OECD indicators2017ParisOECD Publishing
17. Wang WL, Lin WC, Chao CW, Chen CC, Lin HC. Exploring characteristics of ambulatory patients who fail to show for appointments and related problems in a medical center.J Health Sci20035309–20[In Chinese]
18. Wu SS, Chen ML, Hsueh TH, Chang SJ. Decreasing failed appointment rates at a medical center.J MacKay Nurs2011538–48[In Chinese]
19. Wang RH, Yang YC, Lu FH, Chang CJ, Wu CH, Wu JS. Factors associated with non-attendance in patients with follow-up appointment at family practice clinics of a medical center.Taiwan
J Fam Med20122261–9[In Chinese]
20. Miller AJ, Chae E, Peterson E, Ko AB. Predictors of repeated “no-showing” to clinic appointments.Am J Otolaryngol201536411–4
21. Tseng YP. The study of neural network to predict customers’ breaking on appointment of medical industry2001TaipeiMaster thesis, Department of Industrial Engineering and Management, National Taipei University of Technology[In Chinese]
22. Grunebaum M, Luber P, Callahan M, Leon AC, Olfson M, Portera L. Predictors of missed appointments for psychiatric consultations in a primary care clinic.Psychiatr Serv199647848–52
23. Lacy NL, Paulman A, Reuter MD, Lovejoy B. Why we don’t come: patient perceptions on no-shows.Ann Fam Med20042541–5
24. Izard T. Managing the habitual no-show patient.Fam Pract Manag20051265–6