Skip to main content

Healthcare — hospital operations & patient flow

The problem. A hospital runs at the edge of capacity. Beds free up and fill minute by minute, the emergency department backs up, and a ward that quietly hits 100% occupancy turns into a corridor of waiting patients. The operations team needs a single live picture of capacity across every ward — and an alert the moment a unit is about to overflow, while there's still time to act.

This scenario is about live operational monitoring: model the hospital, stream occupancy in, and drive a capacity wall-board off a push connection. It uses only operational signals — bed counts, wait times, census — never patient records.

Set up demo data​

New workspace? Run this once (Python) to create a ward's occupancy series, create the subscription before we listen, and push it over the 95% line. The hospital graph is created by step 1. Safe to re-run.

import datahub_sdk, numpy as np, pandas as pd

client = datahub_sdk.DataHubClient.from_env()

client.timeseries.create([datahub_sdk.TimeSeries(external_id="ward_icu_occupancy_pct", name="ICU occupancy %", unit="pct", value_type="float")])
client.subscriptions.create([datahub_sdk.Subscription(
    external_id="bed_capacity", name="Bed capacity", timeseries=["ward_icu_occupancy_pct"])])
idx = pd.date_range(end=pd.Timestamp.now(tz="UTC"), periods=60, freq="1min")
occ = np.full(60, 88.0); occ[-10:] = 96.0          # crosses the 95% capacity threshold
client.timeseries.insert_from_lists(timestamps=idx, values=occ, ts="ward_icu_occupancy_pct")

1. Model the hospital​

A hospital contains wards, a ward contains beds. The same hierarchy you'd model for any physical estate, which lets a ward-level alert roll up to the whole site.

ResourceForm hospital = new ResourceForm();
hospital.setExternalId("hospital_central");
hospital.setName("Central Hospital");
hospital.setLabels(List.of("Hospital"));

ResourceForm ward = new ResourceForm();
ward.setExternalId("ward_icu");
ward.setName("Intensive care");
ward.setLabels(List.of("Ward"));

RelForm contains = new RelForm();
contains.setName("contains");
contains.setFromExternalId("hospital_central");
contains.setToExternalId("ward_icu");

client.resources().create(List.of(hospital, ward), List.of(contains));

Capacity signals become series: ward_icu_occupancy_pct, ward_icu_free_beds, ed_wait_minutes.

2. Drive the capacity wall-board live​

A wall-board polling every minute is already stale. Subscribe to the occupancy series and update the moment a reading lands — and raise a capacity_warning event when a ward crosses its threshold, so charge nurses are paged before it's full. See Consume live data for the listener lifecycle.

import ai.intellistream.datahub.sdk.subscriptions.SubscriptionMessage;

try (var stream = client.subscriptions().listen(List.of("bed_capacity"))
        .stream((SubscriptionMessage msg) -> {   // auto-acks after each message
            updateWallboard(msg.payload());
            if (overThreshold(msg.payload())) {            // e.g. occupancy > 95%
                EventModel warning = new EventModel();
                warning.setExternalId("capacity_warning_icu_" + System.currentTimeMillis());
                warning.setType("capacity_warning");
                warning.setStatus("open");
                warning.setMetadata(Map.of("ward", "ward_icu", "occupancy_pct", "96"));
                warning.setEventTime(ZonedDateTime.now());
                client.events().create(List.of(warning));
            }
        })) {
    awaitShutdown();   // your app lifecycle; closing the stream ends delivery
}

3. Review the day's pressure points​

Hourly-average occupancy per ward shows planners when the squeeze happens, so they can shift staff and elective schedules. See Query & aggregate.

See the result​

The ICU crossing 95% trips the loop:

capacity_warning_icu_…  → open   (occupancy 96% — page the charge nurse)

See also​