Patient volume forecasting across a four-hospital group

Context

A private group of four hospitals covering 18 specialties. Rosters were set weekly on a fixed pattern that did not respond to real fluctuations in patient volume.

The challenge

Some time slots were severely overloaded while others sat nearly empty. Average waiting time during peak hours reached 78 minutes.

What BigAI delivered

A forecasting model per specialty and per hourly slot, using three years of history plus day of week, public holidays, seasonal illness patterns and weather.

Forecasts feed a rostering support tool. The head nurse remains the decision-maker and can override any suggestion — the model proposes, it does not decide.

Results

Average peak-hour waiting time fell from 78 minutes to 33. The share of overloaded slots dropped from 31% to 9%.

Results

Swipe to see the full table

Technology used