Regression forecasting of patient admission data

Area of Science:

• Health Services Research
• Operations Research
• Biostatistics

Background:

• Accurate forecasting of patient admissions is critical for effective hospital management.
• Key areas impacted include elective surgery scheduling, bed management, and staff resourcing.
• Existing forecasting methods may not fully capture complex admission patterns.

Purpose of the Study:

• To analyze historical patient admission data from Emergency Departments.
• To forecast future patient admissions using various regression techniques.
• To evaluate the accuracy of different forecasting models for hospital resource planning.

Main Methods:

• Utilized five years of Emergency Department admissions data from two distinct hospitals.
• Employed regression techniques, including linear regression with dummy variables for monthly variations.
• Compared forecast data against observed admissions over a 6-month prediction horizon.

Main Results:

• Linear regression incorporating 11 dummy variables for monthly variation achieved the lowest Mean Absolute Percentage Error (MAPE) of 1.79%.
• A 2-year averaging method demonstrated comparable forecasting performance.
• The findings support the efficacy of regression-based forecasting for patient admissions.

Conclusions:

• Regression analysis, particularly linear regression with monthly variation modeling, is a highly effective tool for patient admission forecasting.
• The accuracy achieved supports its application in optimizing hospital management functions.
• Further research at finer time scales is warranted to refine forecasting precision.