Ground ambulance performance & CMS reporting
Translated a large historical ambulance dataset into response, scene, transport, and return-to-service measures for primary and secondary service areas, while surfacing documentation defects that could distort performance reporting.
Several questions. One reporting foundation.
- How long does each part of the ambulance response take—from call / dispatch through scene arrival, transport, and return to service?
- How do response volumes and time intervals differ between the primary and secondary service-area ZIP groupings?
- What mix of emergency / nonemergency responses, transports, and no-transports must be represented for CMS reporting?
- Where do unusually long intervals point to a need for operational review, and where are they caused by incomplete or inconsistent documentation?
- What baseline can the ambulance director use to review performance and compare subsequent reporting periods?
From source records to decision support.
Built the analytical reporting layer from years of historical ambulance records, using complex SQL, CTEs, and joins to assemble event histories and calculate the required time intervals. Defined primary and secondary service-area populations from ZIP groupings and produced a separate set of measures for each. Delivered the results in Excel for the 2022 CMS Ground Ambulance Data Collection work, covering response / transport volumes, time measures, emergency status, no-transports, and service classifications. The work also supplied operational inputs while Finance handled the cost component.
What made the work demanding
The task required more than averaging a time field. It meant aligning call / dispatch, en-route, scene-arrival, destination-arrival, and available-for-service timestamps to the correct response, then checking service-area assignment, response type, and the validity of each interval. Documentation errors can look like poor performance unless the reporting makes those exceptions visible.
What the work enabled
Established a baseline the ambulance director could use to see performance across the response cycle and focus review on long intervals and data-quality problems. Identified 10 records with inconsistent available-for-service documentation, alongside other entry exceptions. The final Excel deliverable supported CMS reporting. Kosi reported broad improvement across most tracked measures in the following year; no percentage improvement or causal attribution to the report is asserted.
What the reporting covered.
| Area | Reporting scope | Decision context |
|---|---|---|
| Response cycle | Call / dispatch → en route → scene arrival | Separate mobilization and response stages. |
| Scene & transport | Scene arrival → departure → destination arrival | Review on-scene and transport duration where valid timestamps support it. |
| Service availability | Destination arrival → available for service | Distinguish return-to-service documentation from the transport itself. |
| Service areas | Primary and secondary ZIP groupings | Calculate parallel measures for the two defined coverage populations. |
| Response mix | Responses · Transports · No-transports · Emergency status | Represent the operational activity behind required totals. |
| Additional scope | Paramedic intercepts · QRU · Service classifications | Support the broader collection requirements and operational context. |
Numbers that can be explained.
These checks make the analytical approach concrete. Where original validation artifacts are unavailable, this section describes the relevant checks rather than implying a published audit trail.
Inspect missing timestamps, negative durations, impossible ordering, and unusually long intervals before aggregating.
Check response identifiers and join cardinality so one response is not multiplied by related records.
Verify ZIP-to-service-area assignment and reconcile the primary / secondary results to the eligible response population.
Keep emergency status, transport / no-transport, and service classifications consistent with collection definitions.
Isolate the 10 inconsistent available-for-service records and separate data-quality findings from potential operational delays.