OHDSI Home | Forums | Wiki | Github

Handling Increasing Disease Incidence during a Characterization Study?

Hi OHDSI folks :wave:

Suppose there exists some disease D.
It has been reported that D has been increasing in incidence over the last 20 years.
As part of my network characterization study, I have been focusing on characterizing patients with disease D.
I have access to multiple different datasets that contain patient data over varying year intervals throughout the 20 year window in which the incidence of D was increasing.
When I characterize these patients in these datasets, I look at variables such as their age, gender, race, etc.

Some questions I have concerning this trend in incidence during my characterization study is:

  1. How necessary is it for me to somehow control for this increasing trend in incidence of disease D?

  2. Across the datasets, would it be best to:

    1. Examine all patients across all years that are available and not give much care to incidence effects?

    2. Group patients by year for characterization to account for this increase in incidence?

    3. Only use patients from datasets who are in the same year windows for which data is collected?

I know that year does not have a causal effect on disease D, but I do know that the incidence of D does appear correlated with time in general.
Any thoughts on how I should best handle these concerns I have?


~ tcp :deciduous_tree:

P.S. I am not a epidemiologist by training so I apologize if I am using vocabulary wrong here.
Please ask questions and I am happy to clarify where needed! :smiley:

It depends why the incidence is increasing. Is it a real increase in incidence, e.g. due to trends in behaviors, exposures, preventive intervention etc, or it is an artifact, e.g., due to differences in diagnostic modalities (imaging or pathology advances) or even disease definitions (hypertension). The latter two in my experience could actually change the definition of who is diseased, and you may not want to group years with different definitions. I have never done a network characterization study so am not sure how you are going to use the info but in any case hope this is helpful

1 Like

Hi @retzioni - thanks for helping me clarify on this.
This was helpful in helping me frame my way of thinking around this problem as more a question of why which implies ideas around causal modeling.
I appreciate your input here and I hope you have a wonderful day!