Methods Core Tip #3
The case-cohort design
It can be difficult to choose a study design that will answer the scientific questions of the research team within the confines of available data and research resources. Often, this tension between an “ideal” study and a practical study is not easily resolved by choosing one of the Big Three study designs. To test for associations between one outcome and several exposures, most researchers would choose a case-control study; but what if you don’t have an appropriate control group, or you want risk ratios, not odds ratios? A case-cohort study combines many of the advantages of a cohort design and the efficiency of a nested design.
This month, the Methods Core journal club discusses the increasingly popular case-cohort study design and how it can be used to study risk factors for hospitalization following a dog bite.
While this article focuses on how case-cohort studies can be used with administrative data and detailed records review, this type of study design is an especially powerful a priori tool for survival analysis of rare outcomes.
Methods Core Tip #2
Using “latent class analysis” to identify subgroups of individuals
“All poodles are dogs, but not all dogs are poodles” can be a handy way to think about the difference between groups and subgroups. But what if you aren’t looking for poodles, but instead are trying to identify all the nervous dogs, or all the dogs that might be good with kids?
Similarly, it’s relatively straightforward for researchers and clinicians to identify all the children or everyone with hypotension in a group of people, but it can be much more challenging to identify everyone with an eating disorder, which can involve a wide range of symptoms and varying degrees of severity. A latent class analysis can identify patterns of characteristics in a heterogeneous group of people, subdividing them into several homogenous groups, and then analyze the association between such patterns and various outcomes.
This month, the Methods Core journal club welcomes Dr. Elizabeth Parker to present on latent class analysis and how it can be used to analyze problem drinking and intimate partner violence.
While this article focuses on how latent class analysis can be used to relate subtypes of problem drinking to recent IPV, this type of analysis can be a powerful tool whenever you’re working on a topic with a loose definition and complex interrelationships between indicators.
Methods Core Tip #1
Using the “Difference in Difference” analytic approach to evaluate the effectiveness of policy changes
In honor of the mid-term elections, our last Methods Core journal club looked at how researchers can evaluate the impact of changes in law, using the example of a study on “Stand Your Ground” laws in the United States.
When considering how to evaluate the impact of a change in policy, you could compare the place where the policy changed (Place A) to a similar place that didn’t have the policy change (Place B), or you could compare what happened in Place A before and after the policy changed. However there could be important differences, not just the policy you want to study, between Place A and B or between the times before and after the policy change. A “Difference in Difference” approach combines these two comparisons to help overcome such confounding.
While the article we discussed last week focused on how this approach can be used to evaluate the effect of a change in law, it’s a good option for anyone interested in how a policy affects a large group of people, and can be done using preexisting data.