Related Experiment Videos
Prospective Validation of Multiweek Inpatient Census Forecasting for Pediatric HSCT and Cellular Therapy
Ezra Porter1, Richard S Hanna1, Timothy S Olson2
1Cell and Gene Therapy Informatics, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania.
Transplantation and Cellular Therapy
|July 17, 2026
Summary
Accurate inpatient census forecasting for blood and marrow transplantation (BMT) programs is now achievable using a component-based model. This method integrates scheduled transplants with predicted admissions and discharges, improving operational planning.
Area of Science:
- Hematology
- Transplant Medicine
- Health Services Research
Background:
- Accurate inpatient census forecasting is crucial for blood and marrow transplantation (BMT) programs to manage bed capacity and staffing.
- Existing forecasting methods often focus on short-term predictions, lacking evidence for the 15- to 30-day horizons vital for BMT operational planning.
- Forecasting BMT census is complex due to the mix of planned (transplants) and unplanned (complications) admissions.
Purpose of the Study:
- To develop and prospectively validate a component-based model for forecasting pediatric hematopoietic cell transplantation and cellular therapy inpatient census 15 to 30 days in advance.
- To create a forecasting approach that reflects the operational structure of BMT care by modeling distinct census-determining processes.
- To provide accurate medium-range census forecasts for BMT operational planning.
Main Methods:
- Developed a component-based model decomposing future census into expected admissions and discharges.
- Integrated known transplant schedules with predicted unplanned admissions using historical data.
- Estimated patient-level discharge probabilities using gradient-boosted decision trees.
- Performed retrospective model development followed by prospective validation of the deployed workflow.
Main Results:
- The component-based model demonstrated strong performance across clinically relevant forecast horizons (15-30 days).
- Achieved mean absolute percentage errors (MAPE) of 13% at 15 days and 16% at 30 days in retrospective evaluation.
- Prospective validation confirmed stable performance in real-world use, demonstrating feasibility for routine operational forecasting.
Conclusions:
- Accurate 15- to 30-day inpatient census forecasting for pediatric hematopoietic stem cell transplantation (HSCT) and cellular therapy programs is achievable.
- A component-based framework integrating scheduled transplants, modeled unplanned admissions, and patient-level discharge prediction is effective.
- This approach supports reliable capacity planning for HSCT/cellular therapy programs and may be adaptable to other complex inpatient services.