An OpenPrescribing signal tells you where to investigate. It does not tell you which individual prescriptions are inappropriate or which clinician should change them. The useful sequence is to understand the measure, examine the underlying numbers, identify a plausible explanation and review a carefully defined group of records.
As checked on 19 September 2026, OpenPrescribing provides England-focused prescribing analysis, including practice and organisational dashboards and a separate hospital offering. Its public information describes routinely updated NHS prescribing data. These are valuable population-level observations, but they are not a substitute for the clinical context held in local records.
Choose a question that the data can answer
Start with a service question rather than a target colour. "Why has our proportion of this medicine group increased?" is answerable. "Why are our clinicians prescribing badly?" assumes the conclusion before the review starts.
Read the measure definition. Identify what is included in the numerator, what is included in the denominator and whether the unit is items, quantity, cost or another specified measure. Check the period, organisational boundaries and any notes about interpretation. Do not treat a count of prescription items as a count of unique patients.
Write the definition into the review plan in plain English. If the team cannot explain what the percentage divides by, it is too early to set a reduction target. A familiar-looking chart can conceal a denominator that is quite different from the one people assume.
A worked example: the percentage rose while the count fell
The following numbers are invented for a measurement exercise, not extracted from a real practice. In period one, a measure has 120 qualifying items out of 1,000 relevant items, or 12%. In period two, it has 112 qualifying items out of 700, or 16%.
The numerator fell, yet the percentage rose. That does not prove either improvement or deterioration. It shows why the team must inspect both parts of the fraction. The changed denominator might reflect altered activity, a service change or a difference in which items are being captured. The record review must establish the explanation rather than choose one because it is convenient.
Add a third column for the practical question: what would have to be true for this change to represent avoidable prescribing? This makes the chart a hypothesis generator. It prevents the meeting from ending with a confident narrative that nobody has checked.
Use comparisons to find questions, not to assign blame
A comparison with other organisations can identify a pattern worth understanding. It does not by itself establish that the populations, services or prescribing responsibilities are equivalent. A local specialist service, shared-care arrangement or change in where prescriptions are issued may matter to interpretation.
Review trends across several periods and inspect the date of any organisational change. Keep a short timeline beside the chart. If prescribing responsibility moved into the practice during the period, annotate it. If nothing relevant is known, record that uncertainty rather than retrospectively inventing an explanation.
The appropriate next step may be a clinical review, a data-quality check or a discussion with another service. Those are different projects. Launching a patient-level intervention to solve a coding or denominator problem wastes time and can create avoidable confusion.
Define a record review that can change something
For the fictional practice, suppose the team wants to understand whether continuing prescriptions have a documented indication and review plan. Define the records eligible for review, the time period and how the sample will be selected. Include a category for records in which appropriateness cannot be judged from the available information.
A practical extraction sheet can be short:
| Field | What the reviewer records |
|---|---|
| Clinical context | Indication and relevant circumstances, without exporting identifiers unnecessarily |
| Decision history | Why the treatment began and which service owns ongoing review |
| Current plan | Evidence of review, intended follow-up and unresolved questions |
| Review conclusion | Appropriate, potentially improvable, or insufficient information |
| Next action | Named team, agreed task and review date |
This is an original review framework, not an OpenPrescribing feature or an official audit standard. Adapt it to the question and local governance. Reviewers should agree what each conclusion means before examining different records, particularly when the distinction between missing documentation and inappropriate treatment is likely to matter.
Close the loop without chasing the graph
A successful review might improve documentation, clarify responsibility or identify a small group needing clinical reassessment. It need not produce an immediate fall in the dashboard measure. Decide in advance which outcomes reflect the work: completed reviews, resolved uncertainties, appropriate follow-up and any unintended consequences.
Use the later dashboard as one part of reassessment, alongside the local findings. If the percentage falls because the denominator grows, do not attribute the change automatically to the intervention. If it remains unchanged after a valuable clarification exercise, do not call the work a failure solely on that basis.
The final report should separate the original signal, the evidence gathered locally, actions completed and questions still open. That is more informative than a before-and-after screenshot with an unsupported claim that prescribing quality improved.
From prescribing analysis to professional learning
iatroX publishes this guide and is included as a complementary clinical-learning resource. It does not replace OpenPrescribing's population analysis or the local clinical record. In the September 2026 product brief, Ask-iatroX offers free source-linked reference, while CPD tools support a learner-reviewed record of an identified development need.
For a GP or pharmacist, the useful connection is specific: the review reveals uncertainty about an indication or monitoring principle; the clinician checks the relevant guidance, practises the reasoning and records what was learned. Do not turn the number of records reviewed into an automatic claim of accredited learning credit.
Use OpenPrescribing when the question concerns prescribing patterns. Use clinical records and appropriate guidance to judge individual care. Use structured learning when the investigation uncovers a knowledge gap. Keeping those roles distinct makes all three more useful.
Frequently asked questions
Does a high prescribing percentile prove poor practice?
No. It identifies a relative pattern that needs interpretation using the measure definition, local context and appropriate record review.
Can a prescribing percentage increase while the number of items falls?
Yes. A larger fall in the denominator can increase the percentage even when the numerator decreases.
What should I measure after the review?
Measure whether the intended actions were completed and uncertainties resolved, then interpret subsequent prescribing data alongside those findings.