Skip to content
Debbie
Back to Project Delivery
Portfolio case study

ClinicFlow

A safer, clearer referral-to-appointment workflow.

Role
Project Manager
Duration
12 weeks
Project context
Healthcare service-improvement portfolio scenario

An end-to-end portfolio project developed to demonstrate project-delivery practice across planning, governance, implementation and handover. No real provider, clinician or patient data is represented.

At a glance

1,200
Monthly referral baseline
41/41
System tests passed
16/16
UAT scenarios passed
79/80
People trained by activation
01

The brief

ClinicFlow is a 12-week healthcare service-improvement portfolio scenario for a fictional community provider. Four approved referral routes fed fragmented re-keying, inconsistent triage hand-offs, status chasing and weak visibility. I structured the work around discovery, a human-clinical-judgement boundary, traceable requirements, controlled testing and staged activation.

The solution did not replace the EPR or automate diagnosis, priority or treatment. It standardised operational status, ownership, completeness checks, exception routing, communication evidence and reporting around authorised human decisions.

02

The starting point

The baseline describes a representative simulated month. Incomplete referrals and administrative rework are separate measures: 276 incomplete referrals sit inside the 372 referrals that needed rework, with 96 further referrals reworked for other reasons.

1,200
monthly referrals
23%
incomplete at receipt

276 referrals

31%
administrative rework

372 referrals · separate measure

4.8d
referral to triage

Working days

18.6d
referral to appointment

Calendar days

430
status enquiries per month
310
aged referrals
210h
duplicate admin per month
03

What I was responsible for

  • I planned discovery, mapped the current state and kept the simulated baseline definitions and overlaps explicit.
  • I facilitated the future-state design while clinical, information-governance and technical authorities retained their decisions.
  • I maintained the 12-week plan, RAID, requirements traceability, change records, budget and status reporting.
  • I coordinated system testing, UAT, operational rehearsal, training readiness, staged activation, hypercare and BAU handover.
  • I kept clinical judgement human and treated patient safety and continuity as constraints on the schedule, not variables to trade away.
  • Process discovery
  • Requirements
  • Clinical boundary
  • Change control
  • Testing
  • Readiness
04

How I structured the project

  1. Weeks 1–2

    Discover

    Map four routes, ownership, failure demand, measures and root causes.

    Stage 1 of 7

  2. Weeks 3–4

    Design and assure

    Baseline the future state, 29 requirements and clinical/IG boundaries.

    Stage 2 of 7

  3. Weeks 5–7

    Configure

    Build 14 components, work instructions and test preparation.

    Stage 3 of 7

  4. Weeks 7–9

    Test

    Close 41 system tests, 16 UAT cases and an operational rehearsal.

    Stage 4 of 7

  5. Week 10

    Prepare

    Assess training, support, data, continuity and sponsor conditions.

    Stage 5 of 7

  6. Week 11

    Activate

    Move a 48-item cohort and two 69-item waves with reconciliation.

    Stage 6 of 7

  7. Week 12

    Stabilise

    Exit hypercare, transfer residual actions and record early indicators.

    Stage 7 of 7
05

The hard part

A stage-age display hid the age that actually mattered.

System testing found that transferring a referral between queues preserved the age of the new stage but obscured its true end-to-end age. That could make a long-waiting referral look new after a hand-off.

I raised the Severity 2 defect and assessed a manual workaround, a correction before UAT and a schedule move. I recommended the correction. The Sponsor released £1,200 contingency; the fix and seven age-control tests passed without compressing clinical or information-governance assurance.

06

Key decisions

Decision 01

Keep clinical judgement outside automation

Context
The workflow needed to surface incomplete information and supplied urgency markers without diagnosing or prioritising patients.
Options
Automate clinical decisions, leave routing fragmented, or automate only operational controls.
Decision
I defined a boundary where the system checks presence, displays supplied markers and routes work; authorised clinicians decide priority and disposition.
Why
Operational consistency should support clinical judgement, not replace it.
Consequence
The future state carries ownership and timestamps while the clinical record and decisions remain human-controlled.

Decision 02

Add age-based exception control

Context
The baseline included 310 aged referrals but inconsistent status labels prevented dependable stage-level trending.
Options
Rely on manual trackers, wait for a larger replacement, or add a limited report and alert control.
Decision
I recommended existing-system configuration plus age-based exception reporting under CR-001.
Why
It addressed visibility without pretending to solve clinical capacity or replace the EPR.
Consequence
£4,800 contingency funded five configuration days and two test days; seven related tests passed.

Decision 03

Correct total-age logic before UAT

Context
DEF-CF-001 could reset the apparent age after transfer.
Options
Use a workaround, accept residual risk, or correct and regress before UAT.
Decision
I recommended correction and retest before business acceptance.
Why
The defect weakened the central safety and visibility objective.
Consequence
The Sponsor approved £1,200; the two-working-day system-test variance did not move later milestones.

Decision 04

Use staged activation

Context
A whole-service switch would have concentrated operational and continuity risk.
Options
Activate all queues at once or use a controlled cohort followed by waves.
Decision
I planned a 48-item cohort and two 69-item waves with reconciliation and abort criteria.
Why
It made support demand, data position and queue behaviour visible before broad activation.
Consequence
All 186 transition items reconciled with zero unexplained variance and no abort trigger.
07

Project control

  • The 12-week baseline contained 12 milestones. A one-week design variance and two-day testing variance remained visible; downstream gates were not silently rebaselined.
  • Nine risks, five issues, six assumptions and five dependencies were closed or transferred with four BAU residual actions.
  • Budget closed at £93,000 against a £96,000 envelope after £6,000 controlled contingency actual, leaving £3,000 unused.
  • The readiness gate reported 12 pass, four pass-with-condition and zero fail domains before Sponsor GO WITH CONDITIONS.
08

Testing and assurance

Requirements and the clinical boundary were traced through 41 system tests, 16 UAT cases, seven age-control tests, training competence, a 24-item synthetic rehearsal, 12 downtime items and three staged activation groups.

  • System testing: 34 initial pass, six fail and one blocked; final 41/41 passed.
  • UAT: 15/16 passed with one S4 condition at Week 8; final 16/16 after focus-order correction.
  • Training: 79/80 competent; one scheduler remained access-restricted with trained cover.
  • Implementation: 48 + 69 + 69 = 186 items, zero unexplained variance and no open incident at exit.
09

A controlled cohort, then two measured waves

Activation began with 48 items before two 69-item waves. The project reconciled queue counts and age-report outputs after each step, kept support cover active and retained abort criteria throughout.

Hypercare exited in Week 12 with 27 support contacts, no open incident and four BAU actions. Early measures cover only the first five full-service working days, so they remain directional rather than mature benefits.

10

What happened

Simulated final result

41/41

System tests

Simulated final result

16/16

UAT

Simulated implementation actual

79/80 · 98.75%

Training

Five-day early indicator · 45/300 · target ≤10%

15.0%

Incomplete referrals

Five-day early indicator · 70/300 · target ≤18%

23.33%

Administrative rework

Early sample · 46/50

92.0%

Workflow compliance

11 · Evidence gallery

These are readable web-native representations of the source artefacts—not fake software screenshots. Open any item for a larger view and its source reference.

12

What I learned

The most important design decision was the operating boundary: the system could make ownership, age and exceptions visible without making clinical decisions.

Process improvement metrics need exact populations and time windows. Incomplete referrals, rework and early indicators can tell a useful story only when their definitions remain attached.

Previous case Perah Studio Delivery SystemBack to DeliveryNext caseAI Essay Marker