ebmdatalab / price-per-dose

2 stars 1 forks source link

Some products are not comparable within their generic class #9

Closed sebbacon closed 7 years ago

sebbacon commented 7 years ago

This issue is the inverse of #1.

Our price-per-dose calculations assume that all presentations coded as equivalents (via the "generic equivalent" field in the BNF code) can be substituted for each other.

However, this is not always the case.

For example, the generic presentation named Gluten Free Bread includes white sliced bread, wholemeal rolls, etc. Gluten Free mixes includes cake mix, pastry mix, and bread mix. The result is a very mixed range of possible prices within a single category.

image

In the case of Gluten Free mixes, there are actually generic presentations for each of these non-comparable items (e.g. G/F Cake Mix, G/F Bread Mix, etc) but these are not used for coding the actual products, which end up in a single "G/F Mix" bin (090401030AAAKAK).

We have previously considered that creams and ointments suffer from this problem, but I no longer think this is the case; it's more likely they appear frequently in our top savings lists either because they are prescribed in bulk and come in a wide spread of pack sizes; or that they are NP8 and show genuine variation.

Once identified, we should exclude such codes from our analyses.

How to identify other such classes

These are generic classes which contain a mixture of non-comparable products. The way we identify a generic class is with AA in positions 10 and 11 of the BNF code. The way we can tell if the products in that class are comparable is by looking at their names. When it is just a brand name, we can't really tell.

We could write a query that finds all actively prescribed generics sorted by the number of brands, and then scan these by eye. See https://github.com/ebmdatalab/prescribing-queries/issues/32 for an example.

Proposed classes to drop

On the whole, it seems like the following tend to have high incidence of inconsistent quantity definitions within their classes:

Where things are the same product but multiple pack sizes across one generic presentation, this isn't such a problem, as we already include multiple pack sizes within single codes elsewhere.

The Larvae are the only dressing outside the Dressings pseudo-chapter; we skip that entire chapter as it doesn't include information about generic equivalents. The Larvae should therefore also be skipped.

richiecroker commented 1 year ago

Have also included thickening agents, as these show simply as "generic"