By Fiona Villamor on November, 19 2019

An Analance™ business case

On its own, a patient’s appointment cancellation might not amount to much. But on a bigger scale, it can result in operational inefficiency, decreased customer satisfaction, and even losses in revenue. As such, it has become a pressing issue that the healthcare industry needs to address.

No-shows are inevitable though. After all, there are many reasons why patients end up cancelling their doctor appointments: financial constraints, long wait times, poor medical literacy, or even logistics issues. While it’s not feasible to totally avoid cancellations, it’s quite possible to anticipate them.

This is where predictive analytics through machine learning comes in. Using historical patient data, build a predictive model that will identify which patients aren’t likely to show up plus the various influencing factors. With these predictive insights, healthcare institutions can implement proactive measures to mitigate the loss of resources that comes with no-shows.

Related: 4 ways the healthcare industry can benefit from Predictive Analytics


970x90-Score and deploy ML Models into production within the same platform


No-shows: A Predictive Analytics use case

Now the question is: how exactly do you incorporate this predictive solution in hospitals, clinics, and healthcare facilities? To predict appointment cancellations for a pain care facility, we used Analance for data analysis and modeling, reporting, dashboarding, and alerts.

We used the pain care facility’s patient appointment data to identify variables/predictors that are significant in predicting patient no-shows. Next, we trained the classification models in Analance and validated the results of all the algorithms in an ensemble mode to identify the best performing model for this case. The Random Forest Classification algorithm was the optimum model and was used to identify the patients who are most probable to cancel their appointment.

The solution was taken a step further with real-time alerts. With built-in automation, data-driven alerts were scheduled to notify administrators when patients are at-risk of missing their appointments—enabling timely corrective action.

“While it’s not feasible to totally avoid [patient] cancellations, it’s quite possible to anticipate them.” Tweet this >


Preparing the data

The pain care facility’s patient appointment dataset had a total of 655,141 observations. We sourced this data through Analance’s SQL Server connector, which allows for real time / live data to be made available inside the platform.

  • Data source – Patient appointment dataset
  • Connector – Analance SQL Server

The dataset had 57 different predictor variables that were considered: appointment category, appointment type, marital status, gender, and employment status, to name a few. All variables available were studied to understand distributions.

Must-Read: Data Collection, Preparation, and Cleaning: A guide

Data was then cleansed by means of handling outlying values, missing values, and looking for interrelationships between predictors before looking to see if any data had a significant relationship with the outcome. After which, a Bivariate Analysis (Chi-Squared) was performed for all predictor-outcome combinations. This helped in restricting the analysis to only those predictors that majorly influence patient no-shows.


The variables chosen for modelling plus their significance in the outcomes.

Running the forecast

Once the data was prepared, the next step was to forecast the key metrics chosen for the prediction. Analance has 41 prebuilt machine learning algorithms, with 8 for classification. We ran the classification model with all the available algorithms in an ensemble mode and analysed the results to identify the best performing model for this scenario.

The Multiclass Random Forest algorithm had the highest accuracy (0.8466) and optimum Recall (0.8087) and F1 score (0.76), compared to the other classification models.

The Receiver Operating Characteristics plot was used to visualize the performance of the model. As seen in the graph below, the area under the curve is 0.9053, which means that there is a 90% chance the model is able to distinguish between cancellations and non-cancellations.


The ROC plot was used to visualize the performance of the model.


The density plot was used to show the distribution of 0 and 1 with respect to their predicted probabilities.

Setting up custom alerts

The algorithm would classify scheduled patients into risk thresholds based on their likelihood of cancelling. When scheduled patients are at risk of missing their appointment, administrators are alerted in real time. This allows the pain care facility to proactively strategize on how to discourage missed appointments.


Analance used the Pain Care Facility’s Patient Appointment data to identify patients most likely to cancel an appointment—with summaries and findings easy to explore through dashboards and reports.

Leveraging AI to manage patient no-shows

With visibility into a patient’s likelihood of cancellation, providers can proactively reschedule appointments, schedule new patients, or put measures in place to mitigate cancellations and incentivize patients to show up for their appointment.

For example, the facility can send timely reminders to busy patients, provide helpful resources to first-time patients, or even provide more strategic clinic hours. Additionally, by detecting no-shows ahead of time, other patients in the queue can be booked or other clinical activities can be planned to optimize operations.


Analance is an end-to-end enterprise platform that delivers AI solutions for various verticals. See how our platform can help you solve your toughest business challenges by requesting a demo today.


Fiona Villamor

Fiona Villamor is the lead writer for Ducen IT, a trusted technology solutions provider. In the past 8 years, she has written about big data, advanced analytics, and other transformative technologies and is constantly on the lookout for great stories to tell about the space.