A Note on Codelists
There is not a consensus definition of how to identify patients with multimorbidity. Different studies use different definitions for specific conditions, and the details of which are not always available. The increased availability and reliance on electronic health records (EHRs) for health services research prompted us to develop transparent methods for identifying specific conditions in this type of data.
We use Read codes and prescribing information to identify 37 long-term conditions from EHRs data. The list of conditions and corresponding definitions is adapted from the work of Barnett and colleagues in Scotland. (Barnett et al, 2012)
The code lists for specific conditions were either developed anew by Cambridge clinicians or adapted from available lists.
We began by reviewing all lists available on clinical codes.org up to January 2016 and obtained QOF lists in 2012. These were grouped when necessary, for example ischaemic heart disease is constructed by groups lists we had found for myocardial infarction, coronary artery intervention and angina Read codes. If a list was not found, a new one was constructed by a member of the team, using hierarchical groups of medical codes created by the Scottish ISD, hierarchical grouping of medical codes within the Read code system, and by using CPRD’s medcode browser to search relevant text words within the Read code system, for example “*diabetes*” would find multiple codes from different parts of the Read hierarchy. Likewise, if a disease definition was based upon a drug usage, CPRD’s productcode browser was used to search for relevant drugs both hierarchically using their BNF code, and using text words.
Finally, every list was checked by at least 2 clinicians to ensure it had face validity. The medcode and product code browsers were also used to check for additional medical codes or product codes not original included in the “found” lists, and new codes were added where necessary based on consensus.
Our codelists were developed for use in CPRD GOLD and reference Read codes, medcodes and prodcodes. Future work is needed to extend their use to CPRD Aurum referencing SNOMED IDs.
Cambridge Multimorbidity Scores (CMS)
Our team has developed and validated a set of multimorbidity scores for different clinical and health care outcomes. We have developed weights for the 37 conditions for primary care consultations, unplanned hospital admissions and mortality as well as a more general score (not outcome-specific). We have also developed a shorter version with only 20 conditions. We expect these scores to be particularly useful for clinical and health service researchers in studies that require the descriptive epidemiology of multimorbidity or the adjustment for multimorbidity (e.g. through matching or stratification). We also expect these scores to be useful for policy makers and clinicians for application at the population level (e.g. for a GP practice to identify a group of patients that would benefit from a different care approach). Our scores provide several advantages over other popular multimorbidity measures including being developed on contemporary data from primary care, using a set of conditions defined by clinical consensus with transparent definitions, and being tailored to multiple outcomes.
Rupert A Payne, Silvia C Mendonca, Marc N Elliott, Catherine L Saunders, Duncan A Edwards, Martin Marshall, Martin Roland. Development and validation of the Cambridge Multimorbidity Score (paper submitted)
|Code list V1.0||Code list V1.1||Number of Codes||New Codes||Deleted Codes||Rationale|
|428 codes added||92 codes removed||Newer products added, new search strategy
|2 [G3…00, G3…13]||–||Missed codes|
|2||–||Lewy Body codes added|
|1 [G6…00]||–||Missed top level|
Cassell A, Edwards D, Harshfield A, Rhodes K, Brimicombe J, Payne R, Griffin S. The epidemiology of multimorbidity in primary care: a retrospective cohort study. Br J Gen Pract. 2018 Apr 1;68(669):e245-51. https://bjgp.org/content/early/2018/03/12/bjgp18X695465/
See all publications here