Mapping the hospital data landscape beyond the EHR

# Mapping the hospital data landscape beyond the EHR

A 400-bed community hospital generates data in more than 100 distinct software systems. The electronic health record (EHR) is the one everyone knows, but it holds maybe half the story. The other half lives in scheduling engines, cost accounting ledgers, survey vendors, lab middleware, and state-mandated registries that most executives never open. If you want to run analytics in a hospital, you first have to know where the data actually sits and who controls it.

This lesson is a field guide to that terrain.

Why the EHR is not the whole picture

The EHR (the clinical system of record: Epic, Oracle Health formerly Cerner, or MEDITECH in most US hospitals) captures orders, notes, medications, and results. But it was designed for clinical documentation and billing, not for operations, finance, or experience measurement.

Ask a simple question like "what did it cost us to treat this hip replacement, and was the patient satisfied?" and you immediately leave the EHR. Cost lives in a decision-support system. Satisfaction lives with a survey vendor. Joining them is a data integration problem, and that is the whole job.

The four data domains

Think of a hospital's data estate in four buckets. Each has different owners, refresh rates, and quality problems.

1. Operational Data

This is the pulse of the building: who is where, and when.

• ADT feeds (Admission, Discharge, Transfer). Every time a patient is admitted, moved between units, or discharged, the EHR emits an ADT message, usually in the HL7 v2 standard (Health Level Seven, the dominant messaging format for clinical events). ADT is the backbone of census, bed management, and length-of-stay tracking.
• Scheduling and OR systems. Surgical case scheduling often runs in a module of the EHR but may be a separate product. This drives OR utilization.
• Staffing and time-and-attendance. Owned by HR or nursing operations, often in Kronos/UKG or Workday.

Owner: Operations and nursing informatics. Refresh: near real time for ADT, which is why ADT is the most valuable operational stream you have.

A sample ADT message (an A01 admit event) looks like this:

MSH|^~\&|EPIC|HOSP|BEDMGR|HOSP|20260214093000||ADT^A01|MSG00001|P|2.5
PID|1||MRN123456^^^HOSP^MR||DOE^JANE||19550101|F
PV1|1|I|4WEST^412^A|||||||MED||||||||INP

The PV1 segment tells you the patient is inpatient (I) in unit 4WEST, room 412. Downstream dashboards parse thousands of these per day to compute live census.

2. financial data (the data, not the ratios)

We care here about the datasets, not accounting metrics.

• Cost accounting / decision support systems. Products like Strata (Syntellis) or the EHR's own module allocate costs down to the encounter or procedure level. This is where "what did this case cost" is answered, using either ratio-of-cost-to-charges (RCC) or activity-based costing.
• Charge Description Master (CDM or "chargemaster"). The master list of every billable item and its charge. A single hospital chargemaster can have 20,000 to 50,000 line items (widely cited estimate; varies by facility).
• Claims and remittance data. The 837 (claim submitted to payer) and 835 (remittance / payment) files in the X12 EDI standard. These tell you what was billed and what was actually paid.

Owner: Finance and revenue cycle. Quality risk: the chargemaster drifts constantly as codes change; stale entries cause claim denials.

3. Patient-Experience Data

• HCAHPS (Hospital Consumer Assessment of Healthcare Providers and Systems). The standardized US patient-experience survey, mandated by CMS (Centers for Medicare and Medicaid Services). Results are publicly reported and tied to reimbursement. Surveys are typically administered by a third-party vendor (for example Press Ganey or NRC Health), so this data lives outside your walls.
• Real-time feedback and complaint logs. Often a separate system from the mandated survey.

Owner: Patient experience office, often reporting to quality. Quirk: HCAHPS response rates are low (national response rates have often run in the roughly 20 to 30 percent range, an estimate that varies by year and mode), so sample sizes are small and lag by weeks. You can browse the public results yourself at Medicare Care Compare.

4. Clinical Registries

Registries are curated, condition-specific databases, usually maintained by trained abstractors who read charts and code cases to strict specifications. They are higher quality than raw EHR data but slower and more expensive to produce.

• Cancer registry. In the US, hospitals with accredited cancer programs report to a state central registry and often to the National Cancer Database. Data is abstracted by Certified Tumor Registrars, coded to NAACCR standards.
• Trauma registry, STS (Society of Thoracic Surgeons) database for cardiac surgery, GWTG (Get With The Guidelines) for stroke and cardiac care, and disease-specific registries.

Owner: Registry staff inside quality or specific service lines. Refresh: monthly to annual. Value: these are your gold-standard outcomes data, but never mistake them for real time.

Europe: a different shape

In much of Europe the registry culture is stronger and more centralized. The Scandinavian countries run national quality registries that cover most of the population for many conditions. The UK has national audits like the National Joint Registry. National patient-experience surveys (such as the NHS Adult Inpatient Survey in England) play a role similar to HCAHPS. The lesson: outside the US, more of your best data may sit in national systems rather than inside a single hospital.

Who owns what: the governance reality

The single hardest fact in hospital data is that no one person owns it all.

| Domain | Typical system owner | Data stewardData stewardA business-side owner responsible for the quality, consistency and appropriate use of data in their domain.View full definition → |

|---|---|---|

| ADT / census | IT + nursing informatics | Bed management |

| Cost accounting | Finance | Decision supportDecision supportTechnologies and processes that turn raw data into actionable insights via reporting, dashboards and analysis, so teams can decide based on facts rather than intuition.View full definition → analyst |

| Chargemaster | Revenue cycle | CDM coordinator |

| HCAHPS | Patient experience | Survey vendor + PX team |

| Cancer registry | Quality / oncology | Certified Tumor Registrar |

This fragmentation is why hospitals stand up enterprise data warehouses (EDWs) and now cloud lakehouses to bring these sources together. The integration standard shifting the field is FHIR (Fast Healthcare Interoperability Resources), a modern APIAPIApplication Programming Interface: a standardised interface that lets applications communicate and exchange data without knowing each other's internal workings.View full definition →-based standard that increasingly complements older HL7 v2 feeds. US regulation (the ONC information-blocking rules under the 21st Century Cures Act) has pushed adoption hard.

A worked example: joining cost to outcome

Here is the analytics payoff of understanding the landscape. Suppose you want the cost per case for hip replacements and their readmission rate.

You need three sources:

1. ADT + EHR: identify the encounters (by procedure code) and any 30-day readmissions.

2. Cost accounting: pull the fully allocated cost per encounter.

3. Registry (optional): confirm complications.

Say you have 200 hip replacement cases in a quarter. Cost accounting reports a total allocated cost of 2,600,000 US dollars (illustrative figure).

Cost per case = 2,600,000 / 200 = 13,000 dollars.

Now the readmission rate. ADT shows 14 of those patients were readmitted within 30 days.

Readmission rate = 14 / 200 = 7 percent.

Neither number lives in one system. The 13,000 came from finance's cost tool, the 7 percent came from ADT joined to itself across encounters. The join key is the patient identifier, which is exactly where data qualitydata qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.View full definition → bites.