I have a dataset that consists of health insurance members. For each member, I have individual air pollution exposure measures for 2009, 2012, 2015 and 2018 and individual depression (outcome) measures for those same years. All the people in my dataset have moved to a new address between 2009 and 2018 and I wanted to analyze whether changes in air pollution exposure due to moving lead to changes in the outcome (depression). My plan was to calculate the difference in exposure before vs after moving and calculate the difference in outcome before vs after moving and see whether those two variables are correlated. So I would effectively check the correlation between the change in exposure and the change in outcome due to moving to a new address.
I am looking for publications with a similar design and I wonder if there is a specific name for this kind of design/analysis, can anyone help me out?
Hello. It is an experimental study.
a similar design is: Article Changes in exposure to ambient fine particulate matter after...
Hi Arthur, I think you could apply a panel study design to the individuals that you have data for. Each year that you have data for forms one panel. So you have four panels for each of the years of data you possess (2009, 2012, 2015 and 2018). You could create a binary variable to represent whether the individual changed address during the study period. But I wouldn't use that alone to assess the impact of a change in air pollution exposure on depression because some individuals would moved to a less polluted place while some would have moved to a more polluted place. Instead, I would explore whether moving homes was an effect modifier of the association between air quality and depression. You can do this by fitting the binary term as an interaction variable between the air quality term and the outcome. The BMJ article actually depicts a cohort study because there is continuous follow-up of individuals recruited into the study. If you have continuous data of all individuals over time (from 2099 to 2018), then you could also use the cohort study design approach as well.
In terms of study design I see it primarily as a cross-sectional panel time series.
I think longitudinal descriptive study fits better your design since you repeatedly examine the same individuals to detect any changes that might occur over time.
- Badges
- Science topic