The data that drives OpenPrescribing is described briefly in our FAQ. It is supplied by NHSBSA and NHS Digital, and a few other sources.

Over the years we have come to understand the limitations of this data. We’re sharing them here, so researchers can take them into account when carrying out analyses.

When using Practice Level Prescribing Data, bear in mind:

  • The data relates to primary care prescribing only. Secondary care prescribing is not included. In many cases however, ongoing care is largely managed in primary care, so meaningful analyses are still possible
  • The data originates from reimbursement claims from dispensing contractors (such as pharmacies), and therefore does not include prescriptions which are issued but never dispensed
  • As the data is from reimbursement claims, items are recorded for the month in which their costs were claimed by the contractor, which may be several months after they were prescribed (blog)
  • As the data is captured for reimbursement use, there is no way of knowing the indication of the prescription for patients
  • The data is for items which were prescribed by practices in England and dispensed in the UK
  • The data is aggregated to practice level. No data is released for tracking at a patient or GP level
  • The data only describes what was on the prescription form, not what was actually dispensed. For example, items which were prescribed generically may be dispensed as brands.
  • The quantity field is occasionally expressed inconsistently (e.g. sometimes doses, sometimes packs) (notes)
  • Item figures do not provide any indication of the length of treatment or quantity of medicine prescribed. Patients with a long term condition usually get regular prescriptions. While many prescriptions are for one month (28 or 30 days supply), items will be for varying length of treatment and quantity
  • BNF codes change over time. We have attempted to account for this (notes)
  • It can be important to know whether an institution is a standard GP practice, or a different kind of institution (for example, a homeless service, or a drop-in centre). However, in the data provided, there is a small but significant number of obvious errors in coding, such as classification of Care Homes and Violent Patient Services as standard settings, and arbitrary numbers given where the list size is less than 100 or unknown (notes)
  • Some CCGs have “rebates” agreed with pharmaceutical companies, and therefore the net cost will be less. This is not reflected in the prescribing data, as it is a separate income stream to the organisation
  • Closures and mergers are not tracked. Therefore it can be difficult to know where patients have gone when a practice is closed, particularly if there is not a single merger. If analysing at a CCG level it is usually safe to assume the majority of patients remain within the same CCG, though this is not always the case
  • CCGs and practices therein change over time; to project back in time consistently, we show current CCGs and the practices they currently contain (rather than the CCG they were in at the time)

When using list size (e.g. as denominator):

  • List size data has historically not been updated every month (it started being published monthly in April 2017), whereas prescribing data is always updated monthly
  • List size data is not weighted for “need”, e.g. population age, deprivation and disease prevalence. This can have an impact on the nature of prescribing in a practice. STAR-PUs are available for some conditions, but these should also be interpreted with caution
  • Items may be dispensed against practices which are currently closed (and therefore have a list size of zero)(blog)
  • Practices are only marked as “closed” in the data some time after they stop operating; sometimes they remain permanently “dormant.(blog)