Is your question not listed below? Send us a message here and we'll get back to you
We use NHS Business Services Authority's (BSA) Detailed Prescribing Information report from their Information Service Portal. The BSA have a release calendar for this dataset, although the data is sometimes released early. We aim to update OpenPrescribing by the Wednesday following the release date, but do try to update earlier if the BSA publish sooner. The data is two months behind, so for example January's prescribing data is published in March.
We have written a blog post on how to download data from the Trends and Analyse sections of the site
It is now possible to download data extracts from our dashboard pages. However, if you are unable to download the data you need get in touch and we may be able to send you an extract.
You are welcome to use data or graphs from this site in your academic output with attribution. Our methods paper will be published shortly, until then please cite "OpenPrescribing.net, EBM DataLab, University of Oxford, 2017" as the source for academic attribution.
If you use data or images from this site online or in a report, please link back to us. Your readers will then be able to see live updates to the data you are interested in, and explore other queries for themselves.
We have data from the past 5 years on OpenPrescribing.net. We do have the prescribing dataset from August 2010 onwards in our databases, and national data going back to 1999. If you require this information get in touch, as we may be able to help.
See our About page.
No, the prescribing dataset we use does not contain any patient information.
NHS Digital define a prescription item in their FAQ (PDF) as:
‘A prescription item is a single supply of a medicine, dressing or appliance written on a prescription form. If a prescription form includes three medicines it is counted as three prescription items.
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.’
NHS Digital define quantity in their FAQ (PDF) as:
'The quantity of a drug dispensed is measured in units depending on the formulation of the product, which is given in the drug name. Quantities should not be added together across preparations because of different strengths and formulations.
• Where the formulation is tablet, capsule, ampoule, vial etc the quantity will be the number of tablets, capsules, ampoules, vials etc
• Where the formulation is a liquid the quantity will be the number of MLS
• Where the formulation is a solid form (eg. Cream, gel, ointment) the quantity will be the number of grammes'
NHS Digital define actual cost in their FAQ (PDF) as:
‘The Net Ingredient Cost (NIC) is the basic price of a drug, i.e. the price listed in the Drug Tariff or price lists. Actual Cost is the estimated cost to the NHS, which is usually lower than Net Ingredient Cost. Actual Cost is calculated by subtracting the average percentage discount per item received by pharmacists (based on the previous month) from the Net Ingredient Cost, but adding in the value of a container allowance for each prescription item.’
Chemicals can occasionally be re-classified from one BNF code to another. An example of this is Linaclotide, which changed from 0102000AH to 0106070B0 in 2013.
We have aimed to normalise the data so that past prescribing is always listed under the current BNF code. This means if you look at Linaclotide’s trends page, you will see the prescribing pre-2013 even though the code changed.
We have done this by mapping the old codes to the new codes, by using a map the NHS Business Services Authority provided us in a personal communication (after correcting for obvious typos etc.). You can read how we did this on github.
You can read more about BNF codes and how they are structured here.
We use the current CCG membership of a practice in our analyses. If a GP Practice moved CCG, the data from before the move will be included in the data for their current CCG. Their former CCG will not include that practice's data. This means that when you look at our prescribing measures, you will be looking at the past prescribing of the current GP practices in that CCG.
When CCGs merge, we create a new dashboard for the newly formed CCG and remove the old CCG dashboards. We use the prescribing data for the current practices in the new CCG in our measures. Any data for practices in the old CCGs that closed prior to the merger will be included in the new CCG's data
Data from our analyse and trends pages also use current practice membership of a CCG when aggregating the data by CCG
Instructions are found here.
We use this site a lot for our own work and our own research. When we use the analysis page, we find it useful to choose denominators cautiously.
You can use "list size" which tells you how many patients a practice covers, but this can be problematic, because different practices will have different kinds of patients, some with lots of older people, and so on.
To account for this the NHS uses imperfect but useful "adjusted" denominators called STAR-PUs, which try to account for the age and sex structure of the practice's population. These STAR-PUs are specific to specific disease areas, because they try to account for different rates of usage -- in different age bands of the population - for specific treatment. So for the STAR-PU for cardiovascular disease prevention prescribing, for example, gives you extra points for every man aged 40-50, even more for men aged 50-60, and so on; but less for women in the same bands, and very little for younger people.
Generating these STAR-PUs for each practice, each disease area, and each month, takes coder time, so we currently only have the STAR-PU for antibiotics.
When using the data ourselves we tend to use more thoughtful approaches to try to "bake in" population prevalence or need for a particular condition, or to explore different prescribing patterns. For example, we often use whole classes of drug as the denominator in our analyses, as in the video walkthroughs; or we compare the use of one drug against the use of another. When looking at whether a practice is using a lot of Nexium (an expensive "proton pump inhibitor" pill for treating ulcers) we might look at "Nexium prescribing" versus "all proton pump inhibitor prescribing" (example).
Play around and let us know if you find anything interesting, or develop any interesting methods.
Most appliances are currently not available to search for on the OpenPrescribing site. We are working to add this, but in the meantime you can contact us to ask for a data extract.
Just because a practice or CCG is an outlier for high or low use of a particular treatment, that doesn't necessarily mean they are good or bad prescribers. These are measures rather than indicators, and they need to be interpreted judiciously.
For example, a practice that prescribes a lot of benzodiazepines may have a lower threshold for prescribing them, or may run a specialist service for people with substance misuse, or have one doctor with an interest in - and long list of - such patients.
There can be huge variation in the price a practice or CCG pay for a treatment, even for the same drug at the same dose. It is well known that branded and generic versions of the same treatment will have different prices; but different specific “brands” of “branded generic” may also have different prices; and there are many other similar sources of variation in price. Every month we estimate what could be saved if every organisation were prescribing as well as the best 10%, for each item prescribed. Read our detailed FAQ on the subject here.
Ghost-branded generics are generic items which have unintentionally been prescribed with a manufacturer name. For example, Naratriptan 2.5mg Tablets is the correct generic name; a ghost-branded version is Naratriptan 2.5mg Tablets (Teva UK Limited). When an item is prescribed generically, the dispenser is reimbursed at the price in the Drug Tariff; but when a manufacturer is stated, the reimbursement price is usually more expensive.
In some cases, generics will be deliberately prescribed with the manufacturer name, for example where a patient needs to have the same colour pill consistently, or in items with a narrow therapeutic index. However, we assume these are a tiny majority of the total ghost-branded prescribing that happens.
The prescribing data released by NHS BSA aggregates all generic and ghost-generic prescribing together, so it is not possible for us to break down the data precisely.
In addition, the published Drug Tariff prices are not necessarily the ones used by BSA for reimbursement (see our technical notes here for an explanation of this). Therefore, to estimate the size of the problem for individual practices or CCGs, we first have to estimate the Drug Tariff price for a generic, which we do by taking the median price paid for that generic item across the entire country. This gives us a price that should be paid for all prescriptions of that generic. We then compare this with the price that was used for reimbursement. This difference gives a reasonable estimate of the total possible savings if a generic were prescribed correctly. We only include savings of more than £5. Note that we use Net Ingredient Cost for our calculations; the true reimbursement cost will typically be 7% less due to bulk discounts.
You can read more about the background of this issue on our blog.
The spreadsheet download on ghost-branded generics page for each CCG or practice is sorted by total possible savings. The practices and items appearing near the top of this list are likely to be prescribing the highest level of ghost-branded generics. ePACT2 now carries data at an AMPP level for ghost-generics (termed "premium price generics"), meaning you can identify individual scripts which are at fault; PrescQIPP has also added ghost-generic reports and searches.
In general, the fix will be dependent on your prescribing software. It should be possible to configure it so only true generics can be picked, at least for the most egregious items.
We are currently gathering feedback from practices who have implemented such fixes, and will publish a summary on our blog in due course.
OpenPrescribing uses the prescribing dataset published monthly by NHS Digital, which includes the past prescribing for practices that have merged or closed. Data will still show for a closed practice if a prescription prescribed by that surgery was dispensed in the community after the closure date. The historical data for closed practices will continue to be shown on OpenPrescribing, and any practice with a current status of Dormant or Closed will be labelled as such. They will continue to be shown under the CCG they were last recorded under.
This tool uses Prescription Cost Analysis (PCA) data back to 1998, which shows all products dispensed in the community (or personally administered by doctors) in England, whereas the data used elsewhere on the site shows how items were prescribed. This means that, in this dataset:
This data cannot be split down to individual months or practices.
All figures are shown per 1000 population (adjusted for England's mid-year population size each year) and costs are corrected for inflation (using consumer price index), both obtained from the ONS. We also normalised the data by drug name and classification, where possible, such that drugs moving around Chapters, or changing name, do not disrupt the trends. We did this by mapping all drug names to their current position in the BNF. Drug names not matched exactly to a currently available item were assigned appropriate classifications via approximate matching, ensuring that drugs were not inappropriately moved across categories, e.g. those with multiple uses. Some chemicals which are no longer available will not have been assigned an up-to-date chemical code, but will have a chemical name (as supplied in the original dataset) and a product name (derived from the drug name). Efforts were concentrated on Chapters 1-6 and 10 so other Chapters may have more inconsistencies, particularly Chapters 9 (Nutrition) and 13 (Skin). The details are explained further in our paper, currently under review with BMJ Open, and in the code can be found here.
Scroll to the bottom right, then:
You can download the entire dataset, either the compiled original data or our normalised data, from FigShare.
The annual data is released in March each year; however, ONS figures for population size are not available until June.
We accept no liability for any errors in the data or its publication here: use this data at your own risk. You should not use this data to make individual prescribing decisions.