Cutting surgical-suite downtime with probabilistic scheduling: methods, tools, and measured gains
Many perioperative teams are wrestling with unpredictable case lengths that create gaps, overtime, and cancellations. Looking for concise, evidence-focused reports of real-world deployments of statistical/stochastic scheduling to reduce unused minutes and raise room occupancy.
Please share first-hand, anonymized experience only (no PHI). Relevant details to include:
- Role and context: (e.g., OR manager / data scientist / vendor; hospital size; case mix; single site vs. system).
- Models & algorithms used: distributions or estimators for case length, bootstrapping, robust buffers, sequencing/assignment rules, optimization formulations, online updating, stochastic programming, simulation heuristics, etc.
- Data & preprocessing: required fields, cleaning, outlier handling, clustering by CPT/surgeon/team, feature engineering, sample sizes and lookback windows.
- Software & tools: languages, libraries, schedulers, simulators (e.g., SimPy, Arena, AnyLogic), dashboards, integration with EHR/ADS systems, compute/hosting notes.
- Change management & rollout: pilot design, stakeholder engagement (surgeons, anesthesia, nursing, scheduling), training, governance, barriers and how they were addressed.
- Concrete before/after metrics (please report time window and sample size): average unused minutes per room/day, percent of scheduled time utilized, overtime hours, cancellations, turnaround times, staff satisfaction or survey results, financial impact where available.
- Artifacts you can share: anonymized benchmark datasets, simulation models, code snippets, dashboard screenshots, and playbooks (please confirm deidentification).
- Vendor comparisons: if you evaluated or used third-party products, list them and outcomes.
Suggested short response format for comparability:
Role | Hospital type | Period & N
Model/Tool
Data/prep
Rollout notes
Key metrics before ➜ after
Main pitfalls / lessons
Artifacts (link or note if available)
Please keep responses practical and specific. Contributors: tag your role and whether you can share code/data. Questions and requests for collaborators (e.g., data scientist to help build a simulator) are welcome.
