Hello! I am working on a proposal for creating an electronic health record phenotyping classification algorithm (mental health focus). I am having a hard time finding solid guidance re:cohort identification. Specifically, is there a gold standard ratio of patients with the identified phenotype:healthy controls that should be gathered? I would be very appreciative of any guidance toward gold-standard studies or systematic reviews on this topic. Thanks in advance for taking the time to answer this question.
The "gold standard" ratio for phenotype to healthy control in Electronic Health Record (EHR) phenotyping is not a fixed or universal value. The appropriate ratio depends on the specific context of the study, the characteristics of the population, and the goals of the phenotyping task.
In EHR phenotyping, researchers aim to develop algorithms or models to identify individuals with a specific phenotype or condition based on their electronic health records. The ratio of cases (individuals with the phenotype) to controls (healthy individuals without the phenotype) in the training dataset is an important consideration. However, there is no one-size-fits-all ratio that applies to all phenotyping tasks.
The gold standard ratio for phenotype to healthy control in electronic health record (EHR) phenotyping can vary based on the specific study, disease, and context. There isn't a universally fixed ratio that applies to all scenarios. The ratio depends on the research question, the prevalence of the condition being studied, and the characteristics of the population under investigation.
In general, the choice of the phenotype to healthy control ratio is influenced by factors such as:
It's common to see ratios ranging from 1:1 to 5:1 or even higher, depending on the factors mentioned above. However, researchers should carefully justify their choice of ratio in the context of their specific study objectives and characteristics of the population being studied.
Ultimately, there is no one-size-fits-all answer, and researchers should carefully consider the unique aspects of their study when determining the phenotype to healthy control ratio for EHR phenotyping. Consulting with statisticians, epidemiologists, and domain experts during the study design phase is often recommended to make informed decisions.
- Badges
- Science topic