This is a complex clinical idea - and i have found conflicting ways in authoritative sources have defined it. This makes it difficult to phenotype. Can we phenotype Cardiomyopathy? A general principle that we follow - If we cannot come to shared understanding of what the phenotype is - by authoring a clear clinical description - then we probably cannot phenotype it?
I started by reviewing authoritative sources (Harrison’s text book, Uptodate) and put my notes here. Because of the wide variation, i struggled because the sub-types were truly different clinical ideas and considering them as one broad “cardiomyopathy” seemed not appropriate.
Because i felt like i was stuck - i went to Atlas and did a lexical search and found that there were many concept id’s for the idea of cardiomyopathy with two of them accounting for almost all of the concept id. These two were also non specific - and different from the contemporary classification of dilated cardiomyopathy, restrictive cardiomyopathy and hypertrophic cardiomyopathy.
I then investigated the trends in use of such codes in various data sources. see ARES reports for concept id and found that their utilization changed (increased dramatically) after 2016 (in US datasets) when the switch happened from ICD9-CM to ICD10-CM.
while in non US datasources (shown below JMDC) there was no such change in 2015/2016
After thought - i decided to make five different cohort definitions/phenotypes and evaluate them
- Cardiomyopathy (broad - includes all sub types including cardiomyopathy with known cardiac causes like ischemia) described here - this may be expected to have the highest sensitivity but lowest specificity.
- Cardiomyopathy (without known cardiac causes prior to index date) - this maybe considered less sensitivity and more specificity compared to 1
- Dilated Cardiomyopathy
- Hypertrophic cardiomyopathy
- Restrictive cardiomyopathy (probably not feasible because of low record count)
Since each is a different phenotype - they needed a different clinical description which i wrote here