GP prescribing / repeat record
Medication list, repeat state, issue history, medication review and reconciliation outcomes where available.
Pilot readiness pack · minimum data specification
Use the minimum data needed to answer the pilot questions safely.
This is a logical data specification, not a demand for one particular NHS system or API. Local discovery should map these fields onto the provider's existing EPR, GP, pharmacy, eMAR, prescribing, dispensing, finance and pathway systems.
PrincipleMinimum
Data minimisation
Collect only fields required for detection, ownership, evaluation and safety.IdentityPseudo
Pseudonymise where possible
Identifiable access only where the live clinical workflow genuinely requires it.ProvenanceAlways
Source stays visible
Conflicts are reconciled, never silently overwritten.RefreshFit
Frequency follows use case
Near-live only where needed; batch is acceptable for many evaluation measures.Logical field set
What the pilot needs and why.
Exact names, formats and interfaces are resolved during local technical discovery.
| Data domain | Minimum fields | Product | Purpose | Class | Priority |
|---|---|---|---|---|---|
| Patient linkage | Stable pseudonymous person identifier | All | Link events across settings without exposing unnecessary demographics. | Pseudonymised preferred | Required |
| Medication identity | dm+d concept / local code, name, strength, form | All | Match the same medicine across prescribing, dispensing and administration systems. | Clinical data | Required |
| Medication state | Active / stopped / changed / held plus effective timestamp | Medicine Loop | Establish the latest authorised state and detect conflicts. | Clinical data | Required |
| Change provenance | Source organisation, author role, timestamp, reason where available | Medicine Loop | Show who changed what, when and why; support reconciliation rather than silent overwrite. | Clinical data | Required |
| Downstream owner | GP / pharmacy / care home / community / other responsible service | Medicine Loop | Assign acknowledgement and reconciliation to the right workflow. | Operational | Required |
| Acknowledgement / closure | Received, reviewed, reconciled, rejected, reason, timestamps | Medicine Loop | Measure whether the medication-change loop actually closed. | Operational | Required |
| Prescription / repeat state | Issue date, quantity, repeat status, prescribed dose where available | WasteGuard | Establish what was authorised and whether an old repeat remains active. | Clinical/operational | Required |
| Dispensing history | Dispensed date, quantity, status, pharmacy identifier | WasteGuard | Calculate cumulative supply and determine whether another item is plausibly required. | Operational | Required |
| Patient-held stock/use | Structured confirmation or locally available proxy | WasteGuard | Improve precision where stock accumulation or non-use is suspected. | Patient-reported / operational | Optional initially |
| Pathway episode | Medicine, indication/pathway, route, formulation, location, administration date | Resource Intelligence | Define the treatment pathway being compared. | Clinical/operational | Required |
| Staff resource | Preparation / administration minutes by relevant staff group | Resource Intelligence | Measure actual resource change rather than assuming published effects apply locally. | Operational | Required for selected pathway |
| Facility resource | Chair time, bed time or treatment-space duration | Resource Intelligence | Measure reusable capacity where relevant. | Operational | Required if claimed |
| Medicine/resource cost | Acquisition cost, consumables and agreed local costing basis | Resource Intelligence / WasteGuard | Classify financial effect transparently. | Finance | Required for £ claims |
| Unused preparation | Prepared, not administered, reason, disposition and value | Resource Intelligence | Measure avoidable preparation loss and root causes. | Operational | Optional workstream |
| Safety / balancing | Incident, urgent supply, escalation, relevant attendance/readmission proxy | All | Ensure apparent productivity does not conceal harm or access problems. | Clinical/operational | Required |
Likely NHS source categories
Map locally rather than assuming one national architecture.
EPR, ePMA and discharge
Authorised medication changes, discharge medicines, administration and pathway activity.
Dispensing / DMS / local systems
Actual supply, quantity, timing, interventions and acknowledgement workflow.
eMAR / care-home workflow
Administration state, stock/use signals and reconciliation actions where included.
SACT / aseptic / treatment systems
Preparation, route, administration, chair/bed use, cancellations and unused preparations.
Finance / costing / workforce
Local cost basis and staff/capacity measures required to validate benefit claims.
Refresh model
Not everything needs real-time integration.
- Medication-change events: near-live or sufficiently prompt to support safe downstream reconciliation.
- Dispensing / repeat signals: daily or event-driven where practical.
- Resource pathway measures: batch extracts may be adequate for baseline and evaluation.
- Finance validation: periodic agreed extracts are usually sufficient.
- Safety outcomes: cadence agreed with the clinical safety/evaluation team.
Fallback principle
Missing data should reduce confidence, not trigger invention.
If a required field is unavailable, the pilot should either use an agreed proxy with its limitation recorded, narrow the use case, or classify the event as not evaluable.
Access and governance
Separate analytics access from clinical-action access.
Prefer pseudonymised datasets
Baseline, opportunity sizing and most evaluation work should not require direct identifiers.
Minimum identifiable access
Where a professional must act on an individual case, access should be role-based, auditable and limited to that workflow.
Provenance is part of the data model
Source, timestamp, author/role, action and resolution history should be retained according to the agreed governance model.
Technical discovery acceptance
Before shadow mode begins, answer these questions.
- Can the same patient and medicine be linked safely across participating systems?
- Can an authorised medication change be distinguished from a local stale state?
- Can the responsible downstream organisation or workflow be identified?
- Can actual dispensing be distinguished from prescribing intent?
- Can the selected resource pathway be measured consistently before and after intervention?
- Can staff workload and safety balancing measures be captured?
- Can finance validate the costing method used for any £ claim?
- Can every data item and decision be traced back to its source?
Pilot readiness gate
If the minimum fields cannot be supplied reliably, narrow the pilot before adding complexity.
A smaller evaluable cohort is more valuable than a large pilot built on uncertain medication state or untraceable benefit assumptions.