Abstract
By implementing a big data simulation model, we have achieved a cost-neutral reduction of the mean total waiting time (TWT) by 21%. Big data simulation enables users to evaluate variations to an existing system before implementation into clinical practice. Various models for improving patient flow or reducing capacity loads can be evaluated cost-effectively.
Introduction
In healthcare environments, patient satisfaction is intrinsically linked to the Total Waiting Time (TWT). However, optimizing clinic flow is challenging due to the complexity of medical workflows, which often involve multiple bottlenecks such as registration, diagnostic testing (e.g., OCT scans), resident examinations, and consultant reviews. Traditionally, clinics attempt to maximize "doctor occupancy," which frequently results in overbooked slots and significant patient backlogs. This study explores the use of big data and Discrete Event Simulation (DES) to optimize these workflows without the risks associated with live trial-and-error.
Methods
The researchers employed Discrete Event Simulation (DES) to create a "digital twin" of the University Eye Hospital (LMU Munich). The methodology involved several key steps:
- Data Extraction: Timestamps for 4,401 patient visits were pulled from a clinical data warehouse linked to Electronic Health Records (EHR).
- Process Mapping: Manual time studies (n=15 to 49) were conducted for specific tasks not captured by EHR, such as fundoscopy or anterior segment examinations, to define statistical distributions for process times.
- Model Building: Using FlexSim Healthcare 3D, the team constructed a simulation mirroring the clinic’s physical layout and patient journey.
- Validation: The model was validated by comparing simulated wait times (225 min) against historical real-world data (229 min).
- Experimentation: Nine different scheduling modalities were tested, varying slot lengths (10, 15, or 20 minutes) and the distribution of double or triple bookings.
Key Results and Insights
The "Longer is Shorter" Paradox
The most significant finding was that increasing individual appointment slots from 10 to 20 minutes (Model 2) actually reduced overall wait times. While counter-intuitive, this change smoothed the arrival of patients at diagnostic and consultant bottlenecks, preventing the compounding delays seen in denser scheduling models. The optimal model utilized these 20-minute intervals with "triple" appointments only at the start of the day (7:30 am and 10:00 am) to keep the clinical pipeline full without overflowing it.
Real-World Impact
Following the implementation of the optimized model in clinical practice, the mean TWT dropped from 229 minutes to 183 minutes—a 21% reduction. This improvement was achieved cost-neutrally, requiring no additional staff or equipment.
Discussion and Limitations
This study demonstrates that big data simulation is a powerful, risk-free "sandbox" for operational efficiency. However, several limitations remain. The simulation began at the registration desk, failing to capture the "hidden" wait time at initial ticket machines. Furthermore, the model's accuracy depends on clinicians' timely data entry into EHRs. While the DES approach is universally applicable, the specific success of the 20-minute slot model is dependent on the unique resource ratios of the LMU Munich clinic. Future work may involve integrating real-time location systems (RTLS) to track patient movement with even greater precision.
Summary updated