D-CAT Group ↗ Contact ENTRDEESAZ
12 Questions Start a 4-week pilot
Start a 4-week pilot

Solutions · by industry · healthcare providers

Hospitals: the bed, the payer and the revenue line, read every morning

CIO / CMIO (data stays on-premise; row-level scope; personal-data masking) · CFO (revenue dynamic, payer mix, discount, target) · Chief Medical Officer (flow and quality) · COO / outpatient director (access and capacity). The buyer is usually the CIO with the CFO; the daily reader is the medical director.

How the three agents on this page read one semantic modelOne semantic model on the left; three scheduled agents in the middle - Chief Medical Officer Agent, Hospital CFO Agent, Outpatient Access Agent; each publishes findings, recommendations and a brief on the right.semantic modelimported from Axoria Data Studiopatient_countencounter_countprocedure_countgross_revenuenet_revenueChief Medical OfficerAgentweekly · Monday 05:30FfindingRrecomm.BbriefHospital CFO Agentmonthly on the 3rd · 06:00FfindingRrecomm.BbriefOutpatient Access Agentdaily · 06:00FfindingRrecomm.Bbriefevery number in a record carries an evidence address

Questions

The three questions a hospital board asks every month

  • Is occupancy up because we admit more or because patients stay longer?
  • Which payer is carrying revenue growth, and what did it cost in discount?
  • Where are we losing patients before we see them — at the ED door or in the appointment book?

In conversation

Sample questions people type into the conversation

deepen path — the finding's evidence is carried frozen; a new number needs a new query, and the agent says so

  • "Break the General Surgery ALOS drift down by admission type."
  • "Is the night-shift walk-out rate the same on all three campuses?"
  • "Which payer grew fastest in August, VAT excluded?"
  • "Show cardiology no-shows by booking channel."
  • "How current is the data?"

Personas

Three example agents

The agents below are examples for this industry, not a fixed set: Chief Medical Officer Agent · Hospital CFO Agent · Outpatient Access Agent. Each one’s scheduled run is shown in the example sets that follow.

Set 1

Chief Medical Officer Agent

Audience: the medical director, ward leads, the ED lead.

Set 2

Hospital CFO Agent

Audience: CFO, board, campus directors — people who already see the number but not what made it.

Set 3

Outpatient Access Agent

Audience: outpatient operations manager, clinic coordinators, call-centre lead.

An agent's name says which measures it reads and for whom. It holds none of the role's authority: it does not decide, approve or act.

Agents for the topics you choose

An agent is a job description, a bound semantic model and a schedule — not code — so the topics it watches are yours to set: a payer, a service line, a campaign, a supplier. We write the first ones with you.

Build your agents with us →

Example sets

Three example sets, one run each

Each set is one agent persona and one scheduled run’s output — the brief, the recommendations and the findings they rest on — in the product’s own record envelope. The first set is open; the other two are collapsed. The model and the measures the sets assume are listed at the foot of the page.

IllustrativeSet 1Chief Medical Officer Agent
Persona
Reads clinical flow and quality for a 380-bed private hospital group with three campuses in one metropolitan area; separates "more patients" from "longer stays", and "more arrivals" from "more walk-outs". · Audience: the medical director, ward leads, the ED lead. · Tone: direct, number-led, no adjectives; where the data cannot separate two explanations it says so. · Output language: en
Signals it watches
(1) Length-of-stay drift without a case-mix shift — bed-days rising while discharges fall. (2) ED walk-outs by shift — the LWBS rate and door-to-provider minutes side by side. (3) Readmission watch line — 30-day readmissions by service line against an 8 % watch line, full months only. (4) Theatre time — minutes used against minutes available, and same-day cancellations. (5) Occupancy composition — occupancy change split into admissions change and stay-length change.
Thresholds it was given
finding 15 % · critical 40 % · ratio metrics (ALOS, rates, shares) 8 % · minimum share 2 % of the network's bed-days · readmission watch line 8.00 % · severity rule: medium when a threshold is crossed, high when a ratio metric moves more than 1.25× its threshold or a count metric more than 1.5× the finding threshold, critical above the critical threshold; an ongoing topic keeps the severity it was given

Run header

  • Scheduled weekly, Monday 05:30, Europe/Istanbul
  • Period: August 2026 (full month, 31 days, 21 weekdays) vs August 2025 (31 days, 21 weekdays); readmissions on July 2026 index discharges (the 30-day window must close)
  • Data complete through 2026-09-19, measured via max(DISCHARGE_DATE), 6 hours ago
  • Scope: this data model has row-level access rules; the agent runs under the clinical operations scope (all campuses, no payer amounts)
  • Model: the workspace's approved model, written onto the run
  • Cost: recorded on the run
Brief

August 2026: General Surgery Bed-Days Up on Fewer Admissions and Longer Stays — Night-Shift ED Walk-Outs at 5.89 %, Internal-Medicine Readmissions Ongoing

Assumptions strip

You did not state a period; last full month (August 2026) was assumed. Readmissions were measured on July 2026 index discharges so that the 30-day window is closed. Payer amounts are outside this agent's scope and do not appear.

The network's August occupancy of 79.90 % (9,412 of 11,780 bed-days) came with fewer, longer stays in General Surgery: the ward admitted 1.96 % fewer patients and discharged 3.16 % fewer than a year earlier but used 7.79 % more bed-days, taking its average stay from 4.86 to 5.41 days (+11.30 %) and its ward occupancy to 84.89 %. At the emergency door the picture is the reverse — more arrivals (+4.60 %) and more people leaving before being seen: the walk-out rate rose from 2.14 % to 3.47 %, and the rise sits in the night shift, where 5.89 % of 3,412 visits ended without a provider contact and mean door-to-provider time was 61 minutes. Internal Medicine's 30-day readmission rate remained above the 8.00 % watch line at 8.85 %, the third month in a row; the network rate is 6.10 %. The theatre axis was not updated this run.

covers: Finding 1, Finding 2, Finding 3

— Run: weekly, the workspace's approved model, data complete through 2026-09-19. Guard Ledger: held 1 — candidate f-05 (theatre utilisation), reason numeric_provenance: token 71 % resolved to no envelope in this run; not_selected 0; duplicate 0.

Recommendation 1

Split General Surgery bed-days by admission type and surgery group for August against their own August 2025 — the drift is in three groups out of five

based_on: Finding 1 · owner: Chief of Surgery, with discharge planning

The stay-length drift is a ward-level fact, not yet a cause. If the ALOS rise is confined to emergency admissions, the front door's acuity changed; if it sits in elective admissions, the discharge process is where the next measurement goes. Both cuts are in this model and cost one query each. A case-mix index is not in this model; if the surgery groups do not settle it, that is the metric to add before the next quarter's review.

The findings this rests on

Finding 1

Stay-Length Drift: General Surgery ALOS 4.86 → 5.41 Days (+11.30 %) While Discharges −3.16 % — Bed-Days +7.79 % on Fewer Admissions

severity highdirection upnovelty newtrust Abreakdown SERVICE_LINE = General Surgery

In August 2026 General Surgery discharged 399 patients against 412 in August 2025 (−3.16 %), yet its bed-days rose from 2,002 to 2,158 (+7.79 %). Average length of stay (bed_days / discharges) moved from 4.86 to 5.41 days (+11.30 %), 1.41× the 8 % ratio threshold — severity high. Admissions in the same window were 401 against 409 (−1.96 %), below any threshold, so the extra bed-days did not come from more patients entering. General Surgery holds 22.93 % of the network's 9,412 August bed-days, so the drift is visible at network level: surgical-ward occupancy (occupied / available bed-days) stands at 84.89 % (2,158 of 2,542) against 78.76 % (2,002 of 2,542) a year earlier; network occupancy is 79.90 % (9,412 of 11,780). The drift is not uniform: three of the five surgery groups show it, two are flat. This data cannot separate a change in case complexity from a delay in the discharge process — the model carries no case-mix index; the surgery-group breakdown is the nearest cut and it does not settle the question.

Recommendation 2

Put the night-shift walk-out rate on a weekly watch by campus, with door-to-provider minutes beside it

based_on: Finding 2 · owner: ED lead and nursing director

The whole-month walk-out rate hides a night-shift rate almost three times the day rate. A weekly watch by campus will show whether one campus carries the rise or all three do; the agent will stamp the topic ongoing rather than alert a third time. The staffing roster is not in this model — the nursing director reads that beside the record.

The findings this rests on

Finding 2

Night-Shift Walk-Outs: ED Left-Without-Being-Seen 2.14 % → 3.47 % of Visits (+62.08 %) While ED Visits +4.60 % — the Rise Sits in the 20:00–08:00 Shift at 5.89 %

severity criticaldirection upnovelty newtrust Abreakdown SHIFT = Night

ED visits reached 9,638 in August 2026 against 9,214 a year earlier (+4.60 %). Patients who left before seeing a provider rose from 197 to 334, so the walk-out rate moved from 2.14 % to 3.47 % — a +62.08 % change in a ratio metric, above the 40 % critical threshold. The rise is not spread across the day: the night shift carried 3,412 visits (35.40 % of the month) and 201 walk-outs (5.89 %), while the day shift's rate is 2.14 % (133 of 6,226) — the same as last year's whole-month figure. Mean door-to-provider time on the night shift is 61 minutes against 34 on the day shift. The model carries arrivals and departures by shift but no staffing roster; whether the night shift was short-staffed cannot be read from this data.

Also published in this run, with no recommendation resting on it

Finding 3

Readmission Watch Line: Internal Medicine 30-Day Readmissions at 8.85 % for the Third Consecutive Month Above 8.00 % — Network 6.10 %

severity mediumdirection upnovelty ongoingtrust Abreakdown SERVICE_LINE = Internal Medicine

Of 486 Internal Medicine patients discharged in July 2026, 43 were readmitted within 30 days (8.85 %). June stood at 8.60 % (41 of 477) and May at 8.42 % (39 of 463); the line has been above the 8.00 % watch line since May and the agent reported it in each of the last two runs. The network rate for July discharges is 6.10 % (134 of 2,198). Internal Medicine's share of the network's July discharges is 22.11 %. This data cannot say whether the readmissions are the same diagnosis groups each month — the diagnosis-group breakdown is in the model but was not opened in this run because the query budget was spent on the ED axis.

What this run could not see.

The theatre-time queries were not run — the budget was spent on the ED axis after the walk-out signal appeared, and the agent said so rather than writing the theatre sentence from memory (the one it wrote was held). A case-mix index and the staffing roster are not in this model. Mortality metrics were excluded from this agent's scope by the workspace administrator.

IllustrativeSet 2Hospital CFO Agent
Persona
Produces the revenue-performance briefing for the CFO and the board; separates the patient count, procedure intensity, payer mix and service mix behind the revenue line; explains deviations with managed metrics and an evidence chain. · Audience: CFO, board, campus directors — people who already see the number but not what made it. · Tone: formal, measured, executive-summary language; short, exact sentences. · Output language: en
Signals it watches
(1) Revenue–volume scissors — revenue rising while patients or procedures fall. (2) Revenue per patient shift — real price move or a mix shift toward costlier services; the breakdown decides, and the agent does not say "prices rose" before it has separated the two. (3) Payer mix — out-of-pocket vs insured shares on the VAT-excluded base; a mix shift changes cash timing without changing revenue. (4) Discount growth — discount_amount growing faster than revenue means the same revenue is bought with more discount. (5) Target deviation — actual vs target, full months only, with the source of the gap (which campus, which payer). (6) Concentration — dependence on a single institutional payer.
Thresholds it was given
finding 15 % · critical 40 % · scissors 10 points · ratio metrics 8 % · minimum share 1 % of gross revenue · severity rule: medium when a threshold is crossed, high when a ratio metric moves more than 1.25× its threshold or a count metric more than 1.5× the finding threshold, critical above the critical threshold; an ongoing topic keeps the severity it was given

Run header

  • Scheduled monthly on the 3rd, 06:00, Europe/Istanbul
  • Period: August 2026 (full month) vs August 2025 — equal calendar days
  • Data complete through 2026-09-20, measured via max(ENCOUNTER_DATE) (encounter date, not posting date; late-posted charges are outside the edge), 9 hours ago
  • Scope: no row-level rule is bound to this agent — the run was produced on every row
  • Model: the workspace's approved model, written onto the run
  • Cost: recorded on the run
Brief

August 2026: Revenue +14.53 % on Price and Mix While Patients −3.32 % — Out-of-Pocket Share Down to 55.44 %, Discount Ratio 6.20 % → 8.25 %, Campus B −12.38 % Against Target

Assumptions strip

You did not state a period; last full month (August 2026) was assumed. Payer shares are computed on the VAT-excluded base and labelled so; the total-revenue sentence uses the VAT-inclusive metric. No row-level rule is bound to this agent; the run was produced on every row.

August revenue of 11.27 M € was 14.53 % above August 2025, but 3.32 % fewer patients and 2.28 % fewer procedures produced it: revenue per patient rose 18.47 %; at these counts the growth is not volume-driven, and the model does not separate tariff from intensity or mix. The mix moved toward insured payers — out-of-pocket fell from 61.78 % to 55.44 % of VAT-excluded revenue while public insurance rose from 24.09 % to 29.74 % — and the discount amount grew 52.46 %, taking the discount ratio from 6.20 % to 8.25 %. Against target the network stands at −5.29 %, and Campus B alone (−12.38 %) carries 80.00 % of the combined shortfall of the two campuses below target. The cost side could not be reported this run.

covers: Finding 1, Finding 2, Finding 3

— Run: monthly, the workspace's approved model, data complete through 2026-09-20. Guard Ledger: held 2 — candidates f-06 and r-03 (cost side), reason source_execution_error on metric protocol_cost (envelope e7 empty); duplicate 1 — candidate f-04 (payer share restated by campus) merged into Finding 2 on a fingerprint match (metric + period + cell); not_selected 0.

Recommendation 1

Break Campus B's August revenue down by payer type against its own target, and put its patient count beside it — the campus carries 80.00 % of the two campuses' combined shortfall

based_on: Finding 3, Finding 1 · owner: CFO with the Campus B director

The network shortfall is Campus B's shortfall. If Campus B's patient count fell in line with its revenue, the gap is volume and the question belongs to patient access; if its patient count held while revenue fell, the gap is payer or service mix. Both cuts are in this model. The target's method is not — the finance team should confirm that the three campus targets were set on the same basis before the board reads this record.

The findings this rests on

Finding 3

Target Deviation Concentrated in One Campus: Network −5.29 % Against August Target While Campus B −12.38 %, Campus C −5.00 %, Campus A +0.39 %

severity highdirection downnovelty newtrust Abreakdown CAMPUS = B

Against an August 2026 target of 11.90 M €, gross revenue reached 11.27 M € (−5.29 %). Campus A delivered 5.12 M € against 5.10 M € (+0.39 %); Campus B 3.68 M € against 4.20 M € (−12.38 %); Campus C 2.47 M € against 2.60 M € (−5.00 %). Campus B's deviation is 1.55× the 8 % ratio threshold for target attainment — severity high; the network figure (−5.29 %) is below the threshold and is reported as the frame, not as a finding of its own. Campus B accounts for 80.00 % of the two campuses' combined shortfall (0.52 M € of 0.65 M €). The comparison is made on a full month only, as the agent's rules require; September is partial and is not compared. Whether Campus B's target was set on a different basis from the others cannot be read from this data — the target metric carries no note of its own method.

Finding 1

Revenue–Volume Scissors: Gross Revenue +14.53 % YoY While Patients −3.32 % and Procedures −2.28 % — 17.85-Point Gap; Revenue per Patient +18.47 %

severity highdirection upnovelty newtrust Abreakdown — (network)

Gross revenue for August 2026 was 11.27 M € against 9.84 M € in August 2025 (+14.53 %). Over the same month the patient count fell from 41,230 to 39,860 (−3.32 %) and the procedure count from 188,400 to 184,100 (−2.28 %). The gap between the revenue change and the patient-count change is 17.85 percentage points, 1.79× the 10-point scissors threshold — severity high. Revenue per patient moved from 238.66 € to 282.74 € (+18.47 %). At the counts shown, the growth is not volume-driven; whether a tariff change, higher service intensity per patient or a shift toward costlier service lines produced it is not separated by this record — the service-line breakdown in this model assigns 80 % of revenue to an "unassigned sub-type" member, so the cut that would settle it is not reliable here.

Recommendation 2

Put the discount-to-revenue ratio on a monthly watch, and ask for the discount metric to be broken down by payer in the next model version

based_on: Finding 2 · owner: pricing and contracts lead, with the data-model owner

Discount grew four times faster than revenue in August. The ratio is the metric to watch monthly; a payer breakdown of discount is what would say whether the growth is contractual (public insurance) or commercial (out-of-pocket campaigns). That breakdown is not in this model today; adding it is a Data Studio change, not an agent change.

The findings this rests on

Finding 2

Payer Mix Shift With Discount Growth: Out-of-Pocket Share 61.78 % → 55.44 % of VAT-Excluded Revenue While Public-Insurance Share 24.09 % → 29.74 % — Discount Amount +52.46 %, Discount Ratio 6.20 % → 8.25 %

severity criticaldirection downnovelty newtrust Abreakdown PAYER_TYPE = Out-of-pocket

On the VAT-excluded base, out-of-pocket revenue was 5.91 M € of 10.66 M € in August 2026 (55.44 %) against 5.77 M € of 9.34 M € a year earlier (61.78 %) — a 6.34-point loss of share while the absolute amount still grew. Public-insurance revenue rose from 2.25 M € (24.09 %) to 3.17 M € (29.74 %); private-insurance revenue from 1.32 M € (14.13 %) to 1.58 M € (14.82 %). In the same month the discount amount grew from 0.61 M € to 0.93 M € (+52.46 %), above the 40 % critical threshold, taking the discount-to-gross-revenue ratio from 6.20 % to 8.25 %. The two facts stand side by side and this record does not link them: the insured segment carried the revenue growth, and the discount amount grew faster than revenue. Which payer received the discount cannot be read here — the discount metric is not broken down by payer in this model.

What this run could not see.

The cost and margin metrics in this model return a source execution error and are marked broken in the model note; every cost sentence the agent drafted was held. Discount is not broken down by payer. The appointment side of the model returns empty, so a booked-versus-walk-in revenue comparison cannot be built here — that is a gap in the model, not a finding.

IllustrativeSet 3Outpatient Access Agent
Persona
Reads the appointment book and the outpatient door: no-shows, late cancellations, lead times, walk-in share and clinic load; tells "patients did not come" from "patients could not get in". · Audience: outpatient operations manager, clinic coordinators, call-centre lead. · Tone: plain and operational. · Output language: en
Signals it watches
(1) No-show rate by clinic and booking channel, month over month and against the same month last year. (2) Lead time — days from booking to appointment — placed beside the no-show rate of the same clinic, without a causal claim. (3) Late cancellations (inside 24 hours) as a separate line from no-shows. (4) Walk-in share of all outpatient visits. (5) Clinic slot utilisation where the slot metric exists.
Thresholds it was given
finding 15 % · critical 40 % · ratio metrics 8 % · minimum share 2 % of booked appointments · severity rule: medium when a threshold is crossed, high when a ratio metric moves more than 1.25× its threshold or a count metric more than 1.5× the finding threshold, critical above the critical threshold; an ongoing topic keeps the severity it was given

Run header

  • Scheduled daily, 06:00, Europe/Istanbul
  • Period: August 2026 vs July 2026 (month over month — August has 21 clinic days, July 23; rates are per booked appointment, so the day count does not bias them), with August 2025 as the second comparison where the metric exists
  • Data complete through 2026-09-21, measured via max(APPOINTMENT_DATE), 5 hours ago
  • Scope: the agent runs under the outpatient scope (three campuses, no revenue metrics)
  • Model: the workspace's approved model
  • Cost: recorded on the run
Brief

August 2026: No-Shows 8.98 % → 10.51 % of Booked Appointments, Cardiology at 14.20 % With an 11.4-Day Lead Time — Walk-Ins Now 31.20 % of Visits

Assumptions strip

Month-over-month comparison (August vs July 2026) was used because the appointment metrics begin in 2026-01 and have no year-earlier value. The third-next-available metric returned no rows for two campuses; this is recorded as a data state, not a finding.

One in ten booked appointments in August was not attended and not cancelled: 6,140 of 58,420 (10.51 %), up from 8.98 % in July, with late cancellations unchanged at 5.41 %. Cardiology stands out at 14.20 % and also carries the longest wait from booking to appointment, 11.4 days against a network mean of 4.8; the two lines move together this month, and the data does not say which drives which. Meanwhile the outpatient door carried more of the load: walk-ins were 31.20 % of 71,400 visits, up from 28.79 %, while attended appointments held flat. The number the agent wrote for "visits lost to no-shows" was held — it had no query behind it.

covers: Finding 1, Finding 2

— Run: daily, the workspace's approved model, data complete through 2026-09-21. Guard Ledger: held 1 — candidate f-03, reason numeric_provenance: token about 900 (visits lost) has no address and no operation; structural drop 1 — candidate r-03 carried no based_on link (gate 0); data state 1 — third_next_available returned 0 rows for two campuses; duplicate 0.

Recommendation 1

Break August no-shows down by booking channel (call centre · app · referral) for Cardiology and Dermatology side by side

based_on: Finding 1 · owner: outpatient operations manager, with the call-centre lead

If the cardiology no-show rate is uniform across channels, the lead time is the place to look; if it is concentrated in one channel, the channel's confirmation step is. The booking-channel dimension is in this model. The reminder-sent flag is not — until it is, "forgot" and "gave up" cannot be told apart from this data.

The findings this rests on

Finding 1

No-Show Rate 8.98 % → 10.51 % of Booked Appointments (+1.53 Points, +17.03 %) — Cardiology at 14.20 % With an 11.4-Day Mean Lead Time Against a 4.8-Day Network Mean

severity highdirection upnovelty newtrust Ainterpretivebreakdown CLINIC = Cardiology

Of 58,420 appointments booked for August 2026, 6,140 were not attended and not cancelled (10.51 %), against 5,110 of 56,900 in July (8.98 %); the ratio moved +17.03 %, 2.13× the 8 % ratio threshold — severity high. Late cancellations are a separate line and stayed at 5.41 % (3,160). Cardiology carries the highest rate — 612 no-shows of 4,310 bookings (14.20 %) — and the longest mean lead time from booking to appointment, 11.4 days against a network mean of 4.8. Dermatology, with a 3.1-day lead time, stands at 6.90 % (297 of 4,304). Across the nine clinics, the four with a mean lead time above 7 days all have no-show rates above 12.00 % and the five below 5 days are all below 9.00 % — the clinic-level result is in the run. That the two lines move together is an interpretive reading and is badged so; this data does not show that one causes the other, and the reminder-sent flag that would separate "forgot" from "gave up waiting" is not in this model.

Recommendation 2

Put cardiology lead time and cardiology no-show rate on one weekly watch, and stop reading either alone

based_on: Finding 1, Finding 2 · owner: clinic coordinator, Cardiology

The two lines moved together in August; a weekly watch will show whether they keep doing so when the lead time changes. The agent will stamp the topic ongoing on subsequent runs rather than raise it as new each week.

The findings this rests on

Finding 2

Walk-In Share 28.79 % → 31.20 % of Outpatient Visits (+8.37 %): Walk-Ins +11.68 % While Attended Appointments −0.47 %

severity mediumdirection upnovelty newtrust Ainterpretivebreakdown VISIT_TYPE = Walk-in

Outpatient visits totalled 71,400 in August 2026: 49,120 attended appointments and 22,280 walk-ins (31.20 %). In July the walk-in share was 28.79 % (19,950 of 69,300). Attended appointments moved from 49,350 to 49,120 (−0.47 %), below any threshold, so the growth in visits is the walk-in line alone (+11.68 %); the share moved +8.37 % in a ratio metric, just above the 8 % threshold — severity medium. Read beside Finding 1, the pattern suggests that patients who cannot get a near appointment are coming to the door instead — that sentence is interpretive and is badged so; the model has no patient-level link between a missed appointment and a later walk-in, and the agent does not claim one.

Data state

Data state (not a finding): the third-next-available appointment metric returned no rows for Campus B and Campus C in August. No severity, no recommendation rests on it; the model owner is notified.

What this run could not see.

The reminder-sent flag and the patient-level link between a missed appointment and a later walk-in are not in this model. The third-next-available metric is empty for two campuses. Revenue is outside this agent's scope by design — the CFO agent reads that.

Every number carries an address

In the product each figure binds to an evidence address — envelope, row, cell, check digit — and a number without one is held, not published. The anatomy of a record →

Derived figures show their operation

A change, a share or a gap is computed by the engine and carries its operands; the model never divides. How every number is proved →

AI governance is the publication layer

Eight deterministic gates, the trust tier, badged interpretation, an assumptions strip that cannot be switched off, and the Guard Ledger. Where the model can and cannot reach →

FAQ

Four questions this page is usually asked

01Does the agent see patient names?

No. Patient name and neighbourhood are marked personal in the model; the agent cannot break down by them, and labels reduce to initials by rule.

02Does the data leave the hospital?

No. The runtime installs next to your warehouse; the gateways are read-only; credentials stay on your servers. The model receives sealed query results, never the tables.

03Can it read our theatre system and our HIS together?

Only what the imported semantic model joins. If the two are separate stars with no bridge, the agent says so and builds the comparison in the sentence, never with a proxy join.

04What if a metric is broken in our warehouse?

The model note marks it broken; the agent never queries it; every sentence that would have used it is held with its reason, visibly.

Free data discovery study

See which of these signals your own warehouse can carry.

Send us the name of your data platform and the domain you argue about most. We read your semantic model with you and answer in writing: which of the signals on this page your model carries today, which need a change in Axoria Data Studio, and what a four-week pilot would measure. No cost, no sequence — a person replies within two working days.

Request a free data discovery study → Start a 4-week pilot

What the examples assume

The data, the model and the measures behind the three sets

Nothing in a set rests on a source outside the model described here. Open what you want to reconcile.

01Assumed data sources — data a hospital already has

Hospital information system / EHR encounters (admissions, discharges, transfers, ED arrivals and departures), billing and revenue cycle (charges, payer, discounts, targets, claims and denials where a revenue-cycle system exists), theatre scheduling, outpatient appointment book, laboratory and imaging orders, HR roster (headcount by shift, never by person in the model). Typical warehouse grain: one row per encounter or per charge line, daily.

02Assumed semantic model — the minimum for the three personas
Measures
patient_count · encounter_count · procedure_count · gross_revenue · net_revenue (VAT-excluded) · discount_amount · target_revenue · payer revenue components (out-of-pocket, public insurance, private insurance, international) · revenue_per_patient (engine-derived ratio) · bed_days · available_bed_days · admissions · discharges · alos (engine-derived: bed_days / discharges) · ed_visits · ed_left_without_being_seen · ed_door_to_provider_minutes (avg) · readmissions_30d · theatre_cases · theatre_minutes_used · theatre_minutes_available · appointments_booked · appointments_attended · no_shows · late_cancellations · walk_in_visits · appointment_lead_time_days (avg).
Dimensions
date (admission / discharge / encounter) · campus · department / clinic · service line / surgery group · payer type · admission type (elective / emergency) · shift (day / night) · nationality (domestic / international) · age band · gender · diagnosis group · booking channel. Marked personal and never broken down by: patient name, neighbourhood, physician name (use physician group if needed).
Data edge
measured by the engine via max(discharge_date) or max(encounter_date); the model declares which.
03Assumed KPIs — what hospital executives track
KPIDefinitionUnitTypical bandUsual breakdown
Bed occupancyoccupied bed-days / available bed-days%75–85 %campus · ward · service line
Average length of stay (ALOS)bed-days / dischargesdays4–6 acute, varies by serviceservice line · admission type
ED left-without-being-seen rateED departures before provider contact / ED visits%< 2 % target; 2–5 % commoncampus · shift
ED door-to-provider timemedian or mean minutes from arrival to first provider contactmin20–40 minshift · campus
30-day readmission ratereadmissions within 30 days / index discharges%6–15 % depending on case mix and definitionservice line · diagnosis group
Theatre utilisationminutes used / minutes available (in-session)%70–80 %theatre · surgery group
Same-day surgery cancellation ratecancelled on the day / scheduled%2–8 %surgery group · reason
Outpatient no-show rateno-shows / booked appointments%5–12 %clinic · booking channel
Appointment lead timedays from booking to appointment (avg or third-next-available)daysclinic-specificclinic
Revenue per patient / per encounternet revenue / patientscurrencymodel-specificpayer · service line
Payer mixshare of net revenue by payer type%country-specificpayer · campus
Discount-to-revenue ratiodiscount_amount / gross_revenue%model-specific; direction matters more than levelpayer · service line
Target attainmentactual revenue / target revenue, full months only%—campus · service line
Night-shift shareencounters in night shift / all encounters%25–40 %campus · department

Related

Where these pages lead

The departments these personas belong to

On this site