In September we described how our alert for appointments at risk of a no-show works, in Spanish. Today it is the turn of what we did not cover then: what is really known about no-shows and how far prediction goes.
How many and why
A review of 105 studies across all specialties puts the average no-show rate at around 23%, with wide differences between regions: from 13.2% in Oceania to 43% in Africa. The two factors that recur most are how far in advance the appointment was given and having missed one before. Being a young adult, living far from the clinic and low socioeconomic status also appear (Dantas et al., 2018).
In Spain, a study in the outpatient clinics of a public hospital measured 13.8%, with an estimated cost of more than three million euros. The most frequent reason was forgetting, in 29.6% of cases, followed by communication failures in 16% (Jabalera Mesa et al., 2015).
We have not found a reliable figure for Spanish dental clinics, or an estimate of the cost per no-show in private dentistry. The published dental figures are from other countries and cannot simply be carried over.
Prediction is for ranking, not for getting it right
The dental study with the most data analysed 196,000 appointments at a clinic in Riyadh, where 42.7% of patients did not attend. Its best model told those who miss apart from those who come considerably better than chance, but far from getting it right every time. It confirmed that lead time is the variable that weighs most, followed by the history of no-shows; age and time of day weigh less (Alabdulkarim et al., 2022).
A review of 50 prediction models, by researchers at Universidad Carlos III, found that only two were right more often than saying «everyone will come» (Carreras-García et al., 2020).
The practical conclusion is modest. A model like this does not say who is going to miss their appointment. It serves to rank the schedule from highest to lowest risk and decide whom to call first.
What works is the reminder
- Reminding reduces no-shows. In a review of eight trials with 6,615 participants, attendance went from 67.8% with no reminder to 78.6% with an SMS and 80.3% with a phone call. The authors consider the evidence still insufficient to set policy (Gurol-Urganci et al., 2013).
- Several reminders work better than one. Another meta-analysis measured 15% no-shows with notification compared with 21% without it, and found that voice appeared more effective than text (Robotham et al., 2016).
- In dentistry, the channel matters. In a trial with 318 families in paediatric dentistry, 8.2% of those who received a voice message did not attend, against 17.7% of those who received an SMS (Nelson et al., 2011).
- Calling those at highest risk helps. In a trial with 2,247 high-risk primary care patients, a personal call one week before, added to the automatic reminder, lowered no-shows from 29.2% to 22.8%. In addition, those who cancelled did so with more notice (Shah et al., 2016).
The limits
- No trial of model-targeted reminders has been carried out in dentistry. The evidence comes from primary care and from hospitals.
- It has not been shown that reminding only those at highest risk is better than reminding everyone. A review of eight trials flags it as research still to be done (Oikonomidi et al., 2023).
- Where there is already a reminder, adding another contributes little. In a health system that already called its patients, centralising the call only lowered no-shows from 16.3% to 15.8% (Kheirkhah et al., 2016).
What is best avoided
Using prediction to book two patients at the same time carries a documented risk. In one study, patients from one population group ended up waiting around 30% longer than the rest, because the model assigned them more risk (Samorani et al., 2022).
That risk has not been described for the confirmation call. It is an argument in favour of confirming, rearranging or drawing on the waiting list, and against overbooking.
What we do with this
Dental Brain flags the appointments at highest risk and proposes a reinforced reminder, which reception approves. It does not overbook or decide for the clinic.
With what is known today, what it is reasonable to expect from a tool like this is a better-ordered schedule and fewer unforeseen gaps, not the disappearance of no-shows.
Sources
- Dantas et al. (2018). «No-shows in appointment scheduling: a systematic literature review». Health Policy 122(4). Link
- Jabalera Mesa, Morales Asencio and Rivas Ruiz (2015). «Factores determinantes y coste económico del absentismo de pacientes en consultas externas de la Agencia Sanitaria Costa del Sol». Anales del Sistema Sanitario de Navarra 38(2). Link
- Alabdulkarim et al. (2022). «Predicting no-shows for dental appointments». PeerJ Computer Science 8. Link
- Carreras-García et al. (2020). «Patient No-Show Prediction: A Systematic Literature Review». Entropy 22(6). Link
- Gurol-Urganci et al. (2013). «Mobile phone messaging reminders for attendance at healthcare appointments». Cochrane, CD007458. Link
- Robotham et al. (2016). «Using digital notifications to improve attendance in clinic». BMJ Open 6. Link
- Nelson et al. (2011). «Assessing the effectiveness of text messages as appointment reminders in a pediatric dental setting». JADA 142(4). Link
- Shah et al. (2016). «Targeted Reminder Phone Calls to Patients at High Risk of No-Show for Primary Care Appointment». Journal of General Internal Medicine 31. Link
- Oikonomidi et al. (2023). «Predictive model-based interventions to reduce outpatient no-shows». JAMIA 30(3). Link
- Kheirkhah et al. (2016). «Prevalence, predictors and economic consequences of no-shows». BMC Health Services Research 16. Link
- Samorani et al. (2022). «Overbooked and Overlooked: Machine Learning and Racial Bias in Medical Appointment Scheduling». M&SOM 24(6). Link