ER Effectiveness Report

Native-data-only ED metrics for a Quality/PI committee: volume, an ED length-of-stay proxy, ED-to-inpatient admit rate, and 72-hour revisit rate, trended monthly - limited to what Synthea's native encounter timestamps can actually support.

Due to the serious gaps and other limitations in the Synthea sample data related to ED reporting, you will notice flat lines and other oddities degrading the report. You will have to envision how this report would be used with more complete data.

About this report

Although the admissions report tracks 30-day inpatient readmissions, this one tracks the emergency department: volume, an ED length-of-stay proxy, ED-to-inpatient admit rate, and 72-hour ED revisit rate - trended monthly for a Quality/PI committee, and deliberately **limited to what native Synthea encounter timestamps can actually support**. Every metric below is computed from ENCOUNTER rows with encounter_class = 'emergency' in the same dataset the Readmission Effectiveness Report loads - see the scope panel below before citing any number on a slide.

Data covers 2025-03 through 2026-08.

What these proxies do and don't tell you — read this before citing any number below

ED length-of-stay is a PROXY, not CMS ED-1/OP-18b. It is computed as encounter start-to-stop for encounter_class = 'emergency'. The official CMS measures are defined across stage-specific timestamps - arrival, triage, provider contact, departure - that native Synthea output collapses into a single interval. This dataset cannot reproduce the official measure at any level of analysis; treat every LOS figure here as "time the encounter record was open," not a certified boarding/throughput measure.

72-hour ED revisit rate is an industry-standard PI metric, not a CMS-branded measure. It is a recognized, widely-used quality indicator, but there is no single official CMS measure ID for this specific window - do not cite it as one.

ED-to-inpatient admit rate is this report's own construction: a same-patient inpatient encounter starting within 6 hours of an ED encounter's departure, not a coded "admit source" field (Synthea's native export has none).

LWBS (left-without-being-seen) rate is unavailable. Left-without-being-seen (LWBS) rate is not shown because it cannot be computed from this dataset: Synthea does not model patients who leave the ED before being seen, so there is no encounter row to derive it from. This is a scope limitation of the source data, not an omission.

459
ED encounters analyzed
1.0 hrs
Avg. ED LOS (proxy)
1.3%
ED-to-inpatient admit rate (6h window)
0.0%
72-hour ED revisit rate

Monthly trend

Most recent 12 months of data. A missing point on the admit- or revisit-rate lines means every ED visit that month was still right-censored for that window (not enough time had passed in the dataset to observe the outcome) - see the scope panel above.

ED volume

What this tells you: how many ED encounter records exist per month in this dataset. What it doesn't: it excludes any patient who left before being seen (LWBS is unavailable - see above), so it is not total ED demand.

ED length-of-stay (proxy)

What this tells you: how long the average ED encounter record stayed open that month. What it doesn't: it is not CMS ED-1/OP-18b - no triage or provider-contact timestamps exist to separate wait time from treatment time.

ED-to-inpatient admit rate

What this tells you: the share of that month's admit-eligible ED visits followed by an inpatient encounter within 6 hours. What it doesn't: it cannot distinguish a true ED-to-inpatient admission from an unrelated later hospitalization that happens to fall in the same window - there is no coded admit-source link in this dataset.

72-hour ED revisit rate

What this tells you: the share of that month's revisit-eligible ED visits followed by another ED visit within 72 hours. What it doesn't: it does not indicate whether the revisit was related to the same complaint, nor does it carry a single official CMS measure ID - cite it as an industry-standard PI metric only.

Month-over-month variance callouts

Rule-based flags only: a swing is listed here when it clears a configured threshold (Edit Target Settings). This reports the size and direction of a change in the numbers - it does not attempt to explain the clinical or operational cause.

MonthMetricChangen
2026-06 → 2026-08 ED volume ED volume fell -67% 1
2026-03 → 2026-06 ED volume ED volume rose +50% 3
2026-01 → 2026-02 ED volume ED volume rose +100% 2
2025-11 → 2026-01 ED volume ED volume fell -50% 1
2025-10 → 2025-11 ED volume ED volume fell -33% 2
2025-09 → 2025-10 ED volume ED volume fell -25% 3
2025-08 → 2025-09 ED volume ED volume fell -20% 4
2025-05 → 2025-08 ED volume ED volume rose +400% 5
2025-04 → 2025-05 ED volume ED volume fell -50% 1
2025-03 → 2025-04 ED volume ED volume fell -50% 2

Case mix by reason

All-time top reason codes by ED visit volume. What this tells you: which complaints drive ED volume. What it doesn't: acuity or severity - reason_description is a free-text admitting reason, not a triage acuity level.

Case mix by day of week

All-time ED visit volume by day of week. What this tells you: which days carry the most ED volume in this dataset. What it doesn't: staffing adequacy - volume alone says nothing about whether a given day was appropriately staffed.

Metric definitions

ED length-of-stay (proxy)
Encounter start-to-stop for encounter_class = 'emergency'. A proxy for CMS ED-1/OP-18b, not that measure - see the scope panel above.
ED-to-inpatient admit rate
Share of admit-eligible (non-censored) ED encounters followed by an inpatient encounter for the same patient starting within the configured admit window (6 hours by default).
72-hour ED revisit rate
Share of revisit-eligible (non-censored) ED encounters followed by another ED encounter for the same patient within the configured revisit window (72 hours by default) - an industry-standard PI metric, not a single CMS-branded measure.
Right-censored
An ED encounter close enough to the end of the available data that its admit or revisit window hasn't fully elapsed yet - excluded from that metric's rate denominator (not just the numerator) for the affected month.
LWBS (left-without-being-seen) rate
Not available - see the scope panel above.

Development notes

This report reuses the Readmission Effectiveness Report's own database and Synthea loader - no separate database, no admin area of its own. It reads encounter_class = 'emergency' rows that were already being loaded alongside the inpatient rows the Readmission Effectiveness Report tracks; only the derivation logic (admit-window, 72-hour revisit, and censoring) is new.

Tech stack

Python 3, Flask, SQLite, Jinja2, Chart.js, python-dotenv, gunicorn, and pytest.