Introducing OpenPrescribing Hospitals measures

We recently launched OpenPrescribing Hospitals, initially making a Submission History feature available. At the Bennett Institute for Applied Data Science, we work openly and develop our tools iteratively based on our users’ feedback.

We’d now like to introduce our next feature, Measures, which we will be releasing shortly. Measures have long been a key feature of OpenPrescribing.net. Now we’re bringing the same functionality to hospital medicines data. In this post we’ll explain what a measure is, how we define them, and why they matter.

What is a measure?

A measure is a way of defining an activity of interest as data. This data can be used to identify opportunities for improvement by considering trends over time, variation between organisations, and deviation from best practice standards. Measures help to highlight variation and spot potential improvements in clinical practice, prescribing safety, and cost-effectiveness.

Measures can simply be a count of the amount of a drug being issued or more commonly a ratio is used, where we need to define a numerator and denominator (more on this below).

For example, the chart below shows the trend over time for an individual Trust, illustrating the percentage of DOACs issued that are not rivaroxaban or apixaban tablets.

To ensure consistency and clarity, all our measures are developed using an editorial and style guide. This means you’ll see a familiar structure and presentation across all measures - making it easier to interpret the data and compare between them.

Why does it matter

Each measure is underpinned by guidance, evidence, or national priorities. To make this clear, we include a “Why does it matter?” section alongside every measure. This explains the context, links to relevant publications or policy documents, and outlines the potential for improvement. We have previously written about “What makes a good OpenPrescribing measure?” and similar principles will apply in OpenPrescribing Hospitals.

Lower is better (usually)

We are often able to make a value judgement - based on guidance, evidence or national priorities underpinning the measure - to determine what represents “better” practice. When we can do this we aim to present measures using a “lower is better” principle to ensure consistency with how data is displayed. To achieve this we may need to design and phrase measures slightly differently to what you may have seen elsewhere. For example, for the “Best value DOAC measure on OpenPrescribing”, instead of showing the percentage of DOACs prescribed as the preferred choice, we display the percentage of DOACs that are NOT prescribed as the preferred choices.

For some measures it is not always possible to determine what is “better,” as this can depend on various local factors, so we cannot apply a value judgement. For such measures, we present the data as it is, allowing users to interpret it in the context of their specific circumstances.

Numerators and denominators

Choosing the right numerator and denominator is essential to creating meaningful measures. Typically, the numerator represents the product or group of products of interest - for example, a particular drug or a class of medicines.

Simple factors like hospital size (e.g. number of beds or patient throughput) can dramatically affect absolute numbers. That’s why we use ratios and percentages wherever possible, to allow fair comparisons between organisations of different sizes. To produce these we use a denominator which will usually define the relevant broader context - such as all items within a therapeutic class - to allow for meaningful comparison.

For the Best value DOAC measure linked above, the denominator is the total number of all DOACs prescribed and the numerator is the number of those which are not best value.

While we could manually specify a list of products to include in either group, this approach is time consuming and the lists can quickly become outdated. Instead, we take advantage of the structure of coding systems like the dm+d to define robust, rule-based inclusion and exclusion criteria – as explained previously in our blog.

SELECT DISTINCT
    vmp.id,
    CASE
        WHEN vtm.vtm IN (
            '775732007', -- VTM code for edoxaban
            '13568411000001103'  -- VTM code for dabigatran
        )
        OR (
            vtm.vtm IN (
            '774624002', -- VTM code for apixaban
            '777455008' -- VTM code for rivaroxaban
            )
            AND
            ofr.name != 'tablet.oral' -- only non-tablet formulations in numerator as these are higher cost
        )
            THEN 'numerator'
        ELSE 'denominator'
    END as vmp_type
FROM viewer_vmp vmp
INNER JOIN viewer_vtm vtm ON vtm.id = vmp.vtm_id
INNER JOIN viewer_vmp_ont_form_routes vofr ON vofr.vmp_id = vmp.id
INNER JOIN viewer_ontformroute ofr ON ofr.id = vofr.ontformroute_id
WHERE vtm.vtm IN (
    '774624002', -- VTM code for apixaban
    '777455008', -- VTM code for rivaroxaban
    '775732007', -- VTM code for edoxaban
    '13568411000001103'  -- VTM code for dabigatran
)
AND vmp.id != '34819111000001102' -- exclude "Rivaroxaban 15mg tablets and Rivaroxaban 20mg tablets" (initiation pack) from the measure as no DDD

Categorising measures

We group measures based on themes such as safety, value and efficiency. This makes it easier for users to browse and filter measures relevant to their interests or needs.

Openness & Transparency

We use Github to host our code, which enables us to make all the code for our measure definitions openly available to view online. This approach invites scrutiny, collaboration, and continuous improvement. Learn more about our commitment to openness and transparency.

Let us know your thoughts

In the following posts we’ll be explaining the measures feature in more detail and looking at some of our initial measures. Your feedback is extremely valuable to us, especially at this early stage.

• Are there OpenPrescribing.net measures you would like to see in OpenPrescribing Hospitals?
• Is there a clinical area or prescribing issue you’d like to explore with hospital data?
• Do you have any other suggestions for measures?

Expect to see new tools and functionality in the coming weeks and months based on the needs of users.

Please share any comments or feedback with us. We would also be happy to join short calls, team meetings etc if you would prefer to give feedback in this manner.