No-Show Predictive Model

The No-Show Predictive Model uses risk profiling to identify Specialist Outpatient Clinic (SOC) patients with a high risk of defaulting on their upcoming appointments, for more focused intervention. This is the first public healthcare initiative in Singapore to use risk profiling to more accurately predict and manage patient no-shows.

Harnessing machine learning to analyse historical patient data points, a predictive model comprising the top 75 factors associated with no-show rates was developed. Each patient is then assigned a risk score, which represents their likelihood of defaulting on appointments. Patients with higher risk scores can be sent reminders as their appointment approaches, and healthcare staff can further contact them to ascertain whether they will turn up, or if there is a need to reschedule the appointment.

The model is being piloted at National Dental Centre Singapore (NDCS) and KK Women’s and Children’s Hospital (KKH). Patient experience has improved, due to quicker access to healthcare services.

On average, one in four patients fail to show up for their appointments without giving prior notice. As a result, healthcare institutions typically over-book appointments to compensate for the no-show rate. However, this could lead to unpredictable and long waiting times for patients, adding further stress to healthcare staff, as there are more patients than expected.

This smarter way of managing outpatient appointments helps institutions optimise the number of appointments for shorter wait times and better patient experience. With predictability of no-show rates, there is better healthcare capacity utilisation as fewer appointment slots go unfulfilled or wasted.

Availability of No-Show Risk Scores means healthcare staff can be more focused in their interventions. This has led to a reduction in manpower previously needed to call and send SMS reminders on appointments.

The No-Show Predictive Model has been implemented at NDCS and KKH, and is being extended to SingHealth Polyclinics (SHP) by “training” the model using relevant data.