Closed Miriam-S-git closed 2 years ago
As discussed, we will use just the PRODCODES as these are well-validated and less subject to bias
Code list kept as it is.
After discussion at the meeting we were worried that: 1) People who had received 4 or more prescriptions for an acute pain event (e.g. following an accident) would be classified as having a chronic painful condition.
We therefore decided to adjust the code as follows: 1) 4 or more prescriptions of prodcodes in a 12 month period. 2) The length of time between the first and last prescription in the 12 month period should be >6 months. (this aligns with the definition of a chronic condition being one that lasts > 6 months).
Algorithm applied to identify 'Painful conditions' in CPRD GOLD data.
Correction applied to the selection of "Painful conditions" as follows:
Excludes all "Painful conditions" events related to an anti-epileptic code prescription from a patient that has ever been recorded with an Epilepsy diagnosis code. The anti-epileptic codes can be found in the MULTIPLY_Prodcode_painful_conditions_epilepsy.csv
Disclosure – to select "Painful conditions" we adopted the same criteria and codelists created by Dr Edwards and Dr Payne for the CPRD @ Cambridge – Codes Lists.
@finersarah @f-eto
As discussed in other papers looking at chronic pain associations with multimorbidity, this is a very difficult phenotype to define either using prescription data (due to overlap of medication with other conditions) or clincal coding. Despite it's importance!
The reason it may be tricky to create a clinical ‘chronic pain’ code list is the definition of ‘chronic pain’ is so broad - pain despite treatment for >12 weeks. (We can consider that 6 months for our purposes). But ‘pain’ is not well coded as it is usually the consequence of other underlying diagnoses. GPs would like enter discussion about pain in the free text – while treating chronic pain.
I have searched for chronic pain codes. But I don't think they will be use consistently in primary care. Lots of patients I see are on multiple analgesics for prolonged periods of time, but I haven’t seen that many with a ‘chronic pain’ code attached and I would not think to attach one myself most of the time. These codes may be used if someone has been referred to a pain clinic, or pain group (which will depend on local provision). But even then, there may not be a chronic pain diagnosis, rather the underlying condition (fibromyalgia, arthritis, neuropathy, etc). I wonder if this is why the Cambridge group just used prod codes?
With this in mind I suggest we: 1) Focus our phenotype on prodcodes using the same process as already defined in the Cambridge Group multimorbidity studies. 2) Consider extracting the codes in the attached excel file and using these in combination. These are codes suggesting chronic pain or referral to a pain management pathway. 3) With regards to these codes we may need to consider what we do with codes related to specific conditions that are included in the list (e.g. complex regional pain, diabetic neuropathy) so we don't give greater weight to pain from these conditions compared to pain from other conditions (e.g. arthritis).
Painful conditions codes.xlsx