I am doing an analysis of data collected in an intervention study seeking to improve antiretroviral treatment adherence using depression management (using a package of defined strategies ranging from psychoeducation, interpersonal psychotherapy, and pharmacotherapy depending on depression severity). My baseline data showed depression score to be significantly correlated to ART adherence score as well as major depression (when adjusted for other covariates) to be significantly associated with poor adherence. Also, for those who had depression or minimal symptoms (in a per-protocol analysis), the intervention significantly dropped their depression scores compared to controls. Now I want to get the real effect of this change in the depression on their change in adherence to treatment i.e. adjusting for the other covariates in the regression model at baseline. I have two questions; Is this the right approach to this analysis? If yes, how do I proceed in getting the change in adherence adjusted for the other covariates used in the model at baseline?
When you say the "real" effect, the word "real" is the problem. You have to take responsibility for defining the real effect. In trialling an intervention, especially one that consumes significant resources like this one, the effect needs to be measured on the basis of intention to treat.
Assuming that your treated and control groups are a sample from the relevant population, then you compare them head-to-head. Adjusting for covariates is a risky business. I can do no better than to point you to the discussion on Frank Harrell's perpetually interesting forum. There is a lot of discussion, and I would strongly recommend you read it all. For about 15 minutes reading you will learn a lot!
https://discourse.datamethods.org/t/guidelines-for-covariate-adjustment-in-rcts/2814/2
In 50 yrs. of clinical experience I offer these observations. Depression ranges on a spectrum of severity. I've seen persons presenting dysthymia, reactive depression, psychotic depression and situational depression. The 3rd manifestation is the riskiest. We find higher incidences of parasuicide and completed acts, especially in males abusing alcohol.
W/O efficacious interventions, it tends to recur. And I differentiate the treatment strategies when depression is a facet of bipolar d/o.
I wish you the best in untangling the syndrome of depression.
Rich
If you are interested in reactive depression you may get an idea from our paper:
I am a bit unclear regarding your question as you seem to be asking about both the effect of an intervention for depression and treatment adherence. From a methodology perspective, these are two different things. I agree with Ronan that if the idea is to assess the treatment effect, then intent to treat is certainly the way to go. However, if you are wondering about the role of treatment adherence (or even dosage), then this is somewhat different. In any intervention, not everyone will exhibit the same adherence. Some will be high and others will be low. It may be expected that the treatment should differ in effect by the degree of adherence (assuming dosage was the same). If you wish to look at how the intervention effect differed by treatment adherence, there are two common approaches. First, if the intervention has already been found to be effective, then you could simply look at the treatment group and how outcomes were moderated by treatment adherence. Second, and perhaps better, you could look at both conditions and model adherence as a moderator of treatment outcomes. I hope this helps, and please forgive me if I missed the purpose and intent of your question.
Cory
- Badges
- Science topic
- Similar topics
- Psychology