Cutting Patient Wait Times by a Third With Demand Forecasting

Vision Medical Centre handles a significant share of medical tourism patients alongside local outpatients, which makes demand far less predictable than a typical clinic.

Appointments were coordinated across three separate systems and a shared spreadsheet. Double-bookings happened most days. Two administrators spent roughly four hours daily reconciling schedules by phone, and patients regularly waited well beyond their appointment time because nobody could see the whole picture.

International patients compounded the problem: their appointments were booked weeks ahead from abroad and often needed several linked procedures across departments on consecutive days, which the existing process could not model at all.

Applying healthcare AI in Amman started not with a model but with eighteen months of appointment history that had never been analysed.

We started with the data rather than the model, because eighteen months of appointment history existed but had never been analysed.

Understanding actual demand

Analysis of historical appointments showed patterns nobody had quantified: consultation duration varied predictably by procedure type and by whether the patient was local or international, and no-show rates differed sharply between the two groups and by day of week.

Scheduling had been treating every appointment as a fixed 20-minute slot, which was the root cause of both the overruns and the idle gaps.

Forecasting and scheduling

We built a duration model that predicts appointment length from procedure type, patient history and department, and a scheduling engine that uses those predictions to allocate slots. Overbooking is applied deliberately where no-show probability is high, rather than accidentally.

For international patients, the system models linked procedures as a single multi-day booking with dependencies, so a delay in one step reschedules the chain rather than silently breaking it.

One source of truth

The three legacy systems were integrated behind a single scheduling service. Departments kept their familiar interfaces during transition, but every booking now writes to one record — which is what eliminated double-booking entirely.

Bilingual throughout: patient-facing confirmations and reminders in Arabic or English by preference, clinical interface in English with Arabic patient names rendered correctly.