This is a typical scenario. It shows how Remolda would approach AI support for patient navigation and scheduling in a regional health network in Ontario, what we would build and what we would measure.
The situation
Picture a network of hospitals and community clinics under steady pressure. Patient volumes rise faster than staffing. The network's patient information line is busy, and nurses who answer it spend part of every shift on questions about clinic hours, locations and visit preparation. Emergency departments see patients who could have been served elsewhere if they had known where to go. Clinics lose appointment slots to no-shows, and reminder calls are done by hand when staff have time.
Clinical expertise is available. The mismatch is between what patients need and the access point they use.
The approach
The scenario covers three parts that support each other: a patient information assistant, volume forecasting for emergency departments and smarter appointment reminders.
Audit (4 weeks). We analyze call logs, wait-time records, appointment data and referral paths, and interview nurses, coordinators, navigators and clerical staff. The baseline is the share of information-line calls that are non-clinical, hold times, no-show rates by appointment type and forecast error of current staffing plans.
Strategy (6 weeks). The design is written with the clinical governance committee, emergency leadership and information-line management. Clinical safety is set as the first constraint: the assistant answers only non-clinical questions and sends anything with clinical content to a nurse at once. The privacy office reviews the data flow under Ontario's Personal Health Information Protection Act before any build, including where each vendor stores and processes data.
Implement, wave 1 (2 months): patient information assistant. The assistant answers questions about hours, addresses, parking, visit preparation and how to reschedule. Any mention of symptoms, medication or a health concern goes to a nurse through an escalation protocol written with information-line leadership. The clinical governance committee reviews escalation logs weekly for the first eight weeks.
Implement, wave 2 (2 months): emergency volume forecasting. Models built on the network's historical visit data, calendar effects and seasonal patterns produce rolling 48-hour forecasts per site. They are validated on held-out history before managers use them for staffing.
Implement, wave 3 (3 months): appointment reminders. A no-show risk score by appointment type, time slot and attendance history decides the reminder: higher-risk appointments get a personal reminder by the patient's preferred channel, two-way rescheduling and a waitlist offer for the freed slot.
Empower (parallel). Separate sessions for information-line nurses, clinical coordinators who use forecasts and clerical staff who handle reminder exceptions.
What we would measure
Each item below is a target we would set with the network and measure against the baseline from the audit. The size of each target is set only after the baseline is measured.
- Hold time on the information line and the share of calls answered without a nurse.
- Escalation accuracy: calls with clinical content that reached a nurse, reviewed weekly by the governance committee.
- Forecast error against actual emergency volumes, per site.
- No-show rate by appointment type, and slots refilled from the waitlist.
- Patient feedback on the assistant, collected in the channel.
Rules that frame this scenario
Ontario's Personal Health Information Protection Act governs personal health information held by health information custodians, and a 2005 federal exemption order declares it substantially similar to Part 1 of PIPEDA for health information custodians. A tool that gives clinical advice could fall under Health Canada's medical device rules, which is one reason the assistant stays non-clinical. HIPAA matters only if the network handles records of US patients. The healthcare sector page lists sources.
Key lessons
1. Design for the whole network. Each part helps on its own. Together they move demand to the right access point, which a single-department project cannot do.
2. Safety and privacy come before the specification. Writing the escalation rules and the data flow first makes the system narrower and easier to approve.
3. Clinicians write the protocols. Escalation rules come from nurses, physicians and the governance committee. The build team implements them and reports on them.
For related work, see AI for hospitals, predictive analytics and customer support assistants.