Indiana University Health Arnett Hospital, consisting of a full-service acute care hospital and a multispecialty clinic, faced poor statistics because the number of no-show patients (those who don’t show up for their scheduled appointments) rose dramatically to 30%. This was primarily connected to the fact that clinic schedules were driven by individual preferences of the medical staff, which led to increased variations in scheduling rules. To eliminate the problem, the client wanted to develop a scheduling methodology that would benefit the clinic, doctors, and patients. The contractors from Texas A&M University were asked to make a predictive scheduling system to optimize doctors’ schedules and decrease the amount of no-shows. They also aimed for:
- Increasing physician efficiency
- Increasing facility utilization
- Keeping down physician overtime
- Decreasing waiting time for patients
To address the challenge in appointment scheduling, the contractors developed a discrete event simulation model using AnyLogic software.
The model’s input screen was used to insert parameters, including clinic capacity, no-show rates, patient mix, and more. The following data on the clinic was entered on this screen:
- Number of appointment requests per hour the clinic can have at each day stage.
- No-show rate, taking variability into account.
- Patient characteristics, including five types of patients. High priority ones were supposed to have insurance, as opposed to those of low priority.
- Decreasing waiting time for patients:
• Same-day sick patients (people who do not have an appointment and need to be urgently seen)
• New patients of high priority
• Re-check of high priority
• New patients of low priority
• Re-check of low priority
- Sick patients’ field showed the share of patients diverted to a different doctor or nurse.
- Doctors’ working schedule included time of availability and number of patients they were able to help per day. It was also possible to limit the amount of sick and new patients a doctor could see each day.
The interface showed how patients mix, depending on treatment time for patient types (it is assumed that new patients have a longer appointment time than re-check patients) and seasonal factors.
The user could change these capacity parameters to see what changes would help increase medical staff utilization and reduce working time for physicians and waiting time for patients.
The discrete event model showed the following sequence of operations:
- The appearing patients are divided into the four types.
- Once the type is preassigned, the patients waited at home for their scheduled time of visit.
- When they arrived at the clinic on the day of the appointment, the model calculated the no-show rate based on probability specified by the input data.
- Patients were seen by doctors or nurses, and after that they left the hospital.
Output screen showed the model results and performance measures for a simulation run. Data included:
- Number of treated patients for patient visit type.
- Number of no-show patients for patient visit type.
- Appointment lead time for patient visit type.
- Doctors’ overall utilization rates in the facility and the overtime they spent in the clinic each day.
- Proportions of discharged patients per doctor, nurse, or peer/urgent care.
- Maximum daily clinic capacity.
After the users got the results, they could adjust physicians’ schedule in the model and run it again to see how these changes could impact performance measures, including:
- Average patient request-to-appointment time
- Clinic utilization
- Physician overtime for up to two years
The developers chose AnyLogic for several reasons. First, the AnyLogic software has multimethod modeling capabilities that allowed the developers to extend the model in the future if needed. In addition, user-friendly interface, and its engaging options, made it easy for other users to experiment with the model and change the input parameters without additional training.
The AnyLogic simulation model offered various ways of improving the clinic’s operational efficiency and patient satisfaction. The model did not require special skills to use and provided detailed output statistics that included:
- Clinic’s daily capacity
- Staff time utilization and overtime amount
- Patients’ distribution among medical staff
- Patient waiting time, and more
The obtained data allowed users to see how the schedule affects the clinic’s working process and provided insight to choose better staff management policies.
AnyLogic presented a method to test theories before implementing them in the clinic and gave different forecasts. For example, with the model the users could predict when the clinic would reach its maximum capacity. In addition, the existing model could be expanded with agent-based and system dynamic approaches if needed, making the model more adjustable to design a predictive appointment scheduling system in other outpatient clinics with similar settings.