Executive workspace ready
Hospital reliability intelligence
The enterprise view of operational reliability.
Review hospital posture, move from pattern to decision, and follow the response system without ranking a unit or measuring an individual.
Prepared review
September 2026
Synthetic demonstration · executive oversight · read-only operations
LWS sees · 100 Healings acts through Pulse
Owner · CNO · executive leadership
Sensus · Executive program review
The master review for the owner and CNO: house supervisor program, charge nurse reporting center, reliability patterns, and the 100 Healings relay — the working document for hospital huddles.
Descriptive measures for the period. This view does not produce a composite enterprise score — distributions, trends, and denominators stand on their own.
01 · Reliability patterns
Shift-level Unit Reliability Index (URI), event load, and resolution patterns across units and shift types. Units are grouped by unit only — never ranked, and never labeled by a band.
Median URI per window with the interquartile spread of qualifying shifts. Each unit reads on its own; charts share a scale so shapes stay comparable. Unit order below is fixed and does not rank performance.
Median across all qualifying shifts per window, with the middle 50% of shifts shaded. A wide band means units and shifts disagree; a falling band means reliability is moving together.
Day, Evening, and Night shift patterns for the period. A contrast is a description, not a verdict on any shift team.
Re-escalation share (first response didn't hold) and carryover share (instability crossing the shift boundary), by window. Line rates are computed on events recorded in each window.
How the instability mix is shifting — which operational problems the hospital is absorbing more of, window by window. Classes use the frozen LWS taxonomy.
How events ended across the period. Resolution statuses use the frozen LWS data dictionary.
02 · House supervisor program
Role-level view of the house supervisor function: response timing across the frozen Identify → Escalate → Acknowledge → Act → Resolve chain, acknowledgment within the frozen 15-minute target, and shift coverage. This program never scores individual supervisors.
Median minutes for each step of the chain, per shift type, for the period. Reference lines are frozen LWS dictionary targets, not new thresholds.
Share of escalated events acknowledged inside the frozen dictionary target, by shift type. Weighted by events per shift.
Shifts with house supervisor coverage, for the period.
Unit-level view of the response chain for the period — the unit, not the person, is the operational object.
03 · Charge nurse reporting center
Program-level view of shift-reliability capture: recording volume, field completeness, and how quickly events land in the log after they start. Reported only by unit — never by person.
Average records captured per shift across the period. Recording volume describes workload and capture, not quality.
Share of records with all frozen required fields, and median minutes from event start to log entry — two dot plots, each with its own scale.
Share of entries classified “Other” against the dictionary rule: if Other exceeds 10% of entries, the taxonomy needs review.
Program metrics per unit for the period.
04 · Where reliability can improve
Observed patterns computed from this period's capture — each with its windows, counts, and a discussion question. These describe what happened; intervention is the executive's decision.
Pre-specified detection rules, descriptive only: change checks compare the first 3 windows against the last 3 (reliability slide ≤ −6 points, lift ≥ +5, carryover ≥ +4pp, re-escalation ≥ +3pp, coverage-gap events ≥ +8pp, shift-level acknowledgment median ≥ 1.25×); a unit is noted as sustained low when its window median sits below 68 in ≥60% of windows; day view shows within-day contrasts. Computed entirely from the synthetic capture above — they prescribe no action and label no unit.
05 · Huddle brief
Generated from this view's period, measures, and detected patterns. Download reads the same as the screen — markdown, for notes or the shared drive.
06 · 100 Healings · Pulse
The 100 Healings relay is threshold-activated: powered by LWS data, dormant until validated thresholds fire. This review reads its performance for the period — how often it stepped in, by which relay action, and how the supervisors judged each step-in. It is judged by its interruptions, not its volume: every activation had to earn the supervisor's attention.
Each bar stacks the window's activations by relay action. Days the relay stood down — no thresholds firing — are design, not missing data.
Activations per frozen relay action, split by the supervisor's disposition on that activation. The relay prepares, never sends.
Supervisor-deemed useful ÷ activations, per window. Windows with no activations carry no rate — silence is not scored.
07 · Pressure test & executive spec
Finished pressure-test research for the Sensus Admin app: every proposed element tested against seven gates, the literature-grounded executive metric set with its values, sources, and hedges, the three-app role contract, and the open items awaiting Jordan — carried in whole as the 2026-10-01 draft.
StatusDraft — for review Dated 2026-10-01 · no build authorized yet
BasisTwo literature surveys: (1) the Sigma concept-gap check, 2026-10-01 (27 constructs); (2) the executive nurse leadership dashboard metrics survey, 2026-10-01.
ScopeLocked: the Admin app stays as a deliverable — the executive layer for directors, CNOs, and presidents. It merges with the Linked Demo's existing CNO view into one executive surface. No duplicate CNO views.
The Admin app earns its existence by answering the three questions only an executive can ask. Every feature must serve one of these three — anything that doesn't is cut.
Panel A answers it: turnover, vacancy, premium labor, span of control.
Panel B answers it: URI trends, event patterns, response timing — aggregate, never real-time.
Panel C answers it: change control, training completion, validation status, evidence base.
Every proposed Admin element was tested against these gates. Any element fails one gate is removed or reframed.
URI / HORI / SRBI terminology intact; house-supervisor-led operating model; non-punitive design.
No causal or outcome claims without live data; synthetic/demo data labeled as such, never presented as findings.
Nothing in the Admin app activates, previews, or implies 100 Healings during the DNP period.
The Admin app is oversight, not command. Executives never acknowledge events, reassign staff, or override supervisor decisions from this view. Real-time operational control stays with the house supervisor.
No unit letter grades, no public unit rankings.
Boundary note carried with the gate: the literature does not peer-review-ban rankings — this is a design choice grounded in just-culture research (blame/shame reduces reporting) plus the frozen non-punitive principle. Stated as our standard, not as a literature mandate.
Patients-per-RN and perceived adequacy are display-only context (“Context, not a score”), fenced off from URI scoring. Already decided 2026-10-01 — the Admin app must preserve the fence, not erode it.
URI is a process-reliability measure, never presented alongside clinical outcomes in a way that implies causation (frozen concept-paper principle).
Draft verdicts from the pressure test — part of the spec awaiting Jordan's review, not build directions.
Dark-space/atom welcome, star-glow SENSUS entry, dissolving letters
Cinematic theater. Executive trust reads as restraint — replace with a quiet, dense executive entry.
Drill-downs (aggregate → operational indicators)
Standard executive dashboard pattern (Karami & Safdari): aggregate first, drill to operational detail on demand.
Plain-language tooltips
AONL 2026: informatics/AI leadership and improvement science are executive competencies — tooltips teach the metric, not just show it.
Day / Week / Month / Year time controls
Executives need trends over weeks/months (umbrella review 2026); this is the correct executive time grain.
R/Y/G status coding
Literature endorses traffic lights; frame red as “needs system support” with drill-down to root causes — never as unit shaming.
Any real-time unit command surface
Fails gate 4. Live boards invite executive override of supervisor decisions. Trends only.
Any unit grading/ranking display
Fails gate 5. Distributions and trends, never grades.
“Is our workforce stable?” — every metric with its definition, values with sources, chart form, data source, and the caveat the research attaches.
% of RNs separated in the period. First-year = 12 months tenure or less.
17.6% national (n = 527 hospitals); ~$60k cost per RN turnover — NSI 2025 data.
NSI National Health Care Retention & RN Staffing Report (annual survey).
Trend line + NSI benchmark line; split voluntary / involuntary / first-year.
Hospital HR.
Benchmarks are vendor survey data, not peer-reviewed — label them as such.
% of budgeted FTEs unfilled; mean days to fill experienced-RN requisitions.
9.6% vacancy; 83-day average fill — NSI 2025 data.
NSI National Health Care Retention & RN Staffing Report (annual survey).
Trend + benchmark; vacancy by unit shown as a distribution.
Hospital HR.
Same labeling condition as A1: vendor survey data, not peer-reviewed.
% of hours from agency/travelers; premium spend trend.
Travel RN ≈ $190k/yr vs. ~$124k staff RN; 80% of hospitals planned decreases — NSI.
NSI National Health Care Retention & RN Staffing Report (annual survey).
Stacked trend: staff vs. premium hours; spend line.
Hospital finance / HR.
Mean direct reports per manager; % of managers above threshold.
Turnover inflections ≈15 and ≈40 reports (Cathcart); +1.6% unit turnover per 10 added reports (Doran); wider span → more med errors/infections (Boned-Galán); 1 in 4 managers with >75 reports (Vizient 2026).
Cathcart et al. 2004 (J Nurs Adm); Doran et al. 2004; Boned-Galán et al. 2023 (integrative review); Vizient 2026.
Distribution + threshold flag; trend.
Hospital HR.
No optimal number exists in the literature — flag thresholds as organizational policy, not evidence-based cutoffs.
Nursing hours per patient day vs. budget; % RN hours.
NDNQI / NQF #0205 (definitions); CDC NHSN staffing protocol.
Variance trend by unit (actual − budget); skill-mix bar.
Hospital finance + staffing.
“Budgeted HPPD” has no published benchmark — internal comparison only.
“Is the operation reliable?” Aggregate and trended only — never the live board. This is the URI the executive is allowed to see.
Weekly/monthly URI distributions — box/distribution plots, never single numbers, never grades.
Trend + distribution.
Honors HRO “reluctance to simplify”: the distribution is the signal, not a score.
Counts by event type over time.
Stacked bars.
Operational pattern recognition.
Ack and action time distributions (median, spread).
Histograms / trend of medians.
Process reliability, not individual performance.
% events carried over; % re-escalated, trended.
Trend lines.
Continuity-burden signal for executives.
Patients-per-RN by unit/shift; end-of-shift perceived-adequacy item.
Small multiples, labeled “Context, not a score”.
Gate 6: fenced from URI scoring — visually fenced in the UI.
Per-shift “care left undone” count (RN4CAST items).
Trend.
Gap #2 closure; routed through change control + DNP chair first — NOT built until approved.
“Are we governed?”
Event-report volume per unit over time.
Edmondson 1990s psychological-safety literature: reporting behavior is a proxy for experienced safety.
Framed as a culture signal, never as a data-quality complaint.
Periodic HSOPS items (especially nonpunitive response, handoffs & transitions).
AHRQ HSOPS v2.0; Lee et al. 2020 (J Nurs Scholarship, Sigma): perceived culture predicts outcomes.
Tests whether non-punitiveness is experienced, not just designed.
Change-control log, training completion by unit, validation status, pre-registration record.
Unique to the Admin app — supervisors don't need it, executives do. Directly evidences the Governance Transfer Control.
The Sigma concept-gap verdicts (covered / partial / gap), with linked sources.
The sensus-concept-gap-check artifact.
Already exists as the Evidence tab — keep the content, demote from primary navigation.
Excluded with reasons, per the spec.
Clinical outcome metrics computed by Sensus (falls, HAPU, CAUTI/CLABSI, FTR)
NDNQI indicators are executive-relevant, but Sensus does not compute them. They may appear ONLY as clearly-labeled external context imported from hospital quality reporting — in a separate panel, with explicit “association is not causation” framing — never adjacent to URI trends in a way that implies Sensus drives outcomes. Gate 7.
Real-time unit status / live event feed
Fails gate 4 (role separation).
Unit grades, rankings, red-list
Fails gate 5.
NEWS2 / physiological scores
Out of scope (gap-check “deliberately not adopt”).
100 Healings activation or threshold preview
Fails gate 3 (DNP firewall).
One shared database. Three role-gated views. The Linked Demo already proved this pattern — extend it, don't build parallel.
Data flow · On-Unit (intake) → shared DB → House Sup (measurement + command) → shared DB → Admin (aggregation + governance)
The Admin app is read-only on operations and write-only on governance. That asymmetry is the entire role contract — it is what keeps the house-supervisor-led model intact while giving executives real oversight.
Draft instructions from the spec — none are build directions until Jordan's review.
Four items are awaiting Jordan's answers. None are assumed, none are resolved — marked plainly Open until she answers them.
OpenQ1
Confirm the metric set (§3) — especially whether hospital HR/finance data feeds are realistic for the pilot site, or whether Panels A1–A5 start as labeled demo content.
OpenQ2
Confirm NDNQI-outcome-as-external-context (currently conditional) vs. excluding outcomes entirely until post-pilot.
OpenQ3
B6 (missed-care capture) and any adequacy instrumentation changes: already routed to change control + DNP chair — not built until approved.
OpenQ4
Merge decision: fold the Linked Demo CNO view in now, or keep both during transition?
Source citations are the literature names as the research names them; the research carried no links, so none are added.
Program orientation
Sensus · Live shift
Accelerated synthetic night · executive observation onlySix units · fixed order · no ranking · one accelerated 12-hour shift
Sensus · CNO simulation