Skip to main content
Figure 1 | Journal of Biomedical Semantics

Figure 1

From: Development and validation of a classification approach for extracting severity automatically from electronic health records

Figure 1

Example showing differences between ehr manifestations of severe (Myocardial Infarction or MI) and mild (Acne) phenotypes. Phenotype-level differences between severe and mild phenotypes are shown in Figure 1. Notice that there is very little difference between the two phenotypes if you only look at the number of procedures, comorbidities or prescribed medications. Therefore, if you use any of those three measures alone to identify severity, it would be difficult. However, if cost is used as a proxy for severity then the correct classification would be made (myocardial infarction is more severe than acne and also costs more). But if you use the treatment length then an incorrect classification of the phenotype-level severity will result (acne takes longer to treat as a result of chronicity, and therefore longer treatment length is not equal to increased phenotype-level severity). This underscores the importance of using multiple measures together as a proxy for severity, which is the approach employed by CAESAR.

Back to article page