Healthcare

A 40-bed medical ICU monitors patients with continuous vitals (heart rate, SpO2, blood pressure, respiratory rate). The Modified Early Warning Score (MEWS) predicts deterioration 2-4 hours before a code event — but only if computed continuously from real-time vitals, not from nurse spot-checks every 4 hours.

Patient in bed ICU-12B coded at 3:17 AM. Retrospective chart review showed SpO2 trending down from 96% to 89% over 90 minutes, but the 1 AM spot-check recorded 94% and the next check wasn't until 5 AM.
LIVEvitals_stream
bed_idpatient_idheart_ratespo2systolic_bpresp_ratetemp_cts
ICU-12BPT-88421082961281637.12024-03-15T01:00:00.000Z
ICU-12BPT-88421089941321837.22024-03-15T01:30:00.000Z
ICU-12BPT-88421098931382037.42024-03-15T02:00:00.000Z
ICU-12BPT-884210108911452437.62024-03-15T02:15:00.000Z
ICU-12BPT-884210118891522837.82024-03-15T02:30:00.000Z
ICU-07APT-77103872981181436.82024-03-15T02:00:00.000Z
Streaming SQLRunning
Ingest continuous vitals from Kafka
CREATE SOURCE vitals_stream WITH (
  connector = 'kafka',
  topic = 'icu.vitals.monitor',
  properties.bootstrap.server = 'broker:9092'
) FORMAT PLAIN ENCODE JSON;
Compute MEWS and detect score acceleration
deterioration_alertsauto-updating
bed_idpatient_idmews_scoretrendspo2_trendpredicted_eventalert_level
ICU-12BPT-8842107RAPID_RISE-2%CODE_IMMINENTCRITICAL
ICU-12BPT-8842105RISING-1%DETERIORATINGHIGH
ICU-12BPT-8842103RISING-2%NULLMODERATE
RisingWave computes MEWS every 30 seconds from continuous monitor feeds. Bed ICU-12B's score rises from 2 to 7 as SpO2 drops below 91% and heart rate climbs above 115 — rapid response team called 47 minutes before the code would have occurred.
Best-in-Class Event Streaming
for Agents, Apps, and Analytics
GitHubXLinkedInSlackYouTube
Sign up for our to stay updated.