Week 3 discussion

Week 3 discussion

by Francois Rerolle -

ARIC data: The ARIC data, following 4 different communities with different racial mixing is ideal to study effect modification by race of risk factors for atherosclerotic diseases (strong RQ). On the other hand, this study design with only a baseline and one follow-up period is inappropriate for any survival analysis (weak RQ). Longitudinal data would instead be necessary.

DHS: the demographic health survey, conducts cluster randomized cross-sectional surveys to gather health data in many countries around the world. DHS data from several rounds in 1 country could be used to study Age-Period-Cohort effect on malaria indicators such as mortality in children under 5 years old (Strong RQ). On the other hand, because of the cross-sectional design, the DHS dataset can’t be used to assess risk factors for malaria incidence (weak RQ). 

NHS: The nurse health study is a cohort study established in 1976 and included 121,700 female registered nurses aged 30 to 55 years. A baseline questionnaire and follow-up questionnaires (every 2 years) were self-answered by the cohort participants.

 Strong RQ: impact of anorexia/eating disorders on divorce rates.

  • This type of longitudinal data is very suited for survival analysis. Although several outcomes could be considered, because the follow-up data is gathered by self-reported answers, I think it is important to opt for an outcome as objective as possible and not too sensible to recall bias or self-reporting subjectivity. In particular that could be an issue with an alternative outcome I was considering: menopause age. The follow-up resolution of 2 years might not be precise enough but divorce processes are probably long enough (from time of decision to legal act) that we don’t need more than a 1-year resolution. If needed, we could still try to match up the data with divorce-certificate data.
  • The exposure is a lot less precise and could be very sensible to self-reporting biases. By age 30 though, maybe most women with a history of eating disorders have overcome it or at least reached a phase of acknowledgment. A medical evaluation at baseline could be necessary.
  • I am actually now wondering if that RQ is that strong after all… With self-reported data, an objective exposure should probably have been chosen as well, like BMI: impact of BMI on your divorce rates…

 

Weak RQ: Caregiving and CHD risk. I actually think the NHS data was not appropriate to answer the RQ from Lee’s article. First, if interested in caregiving as an exposure, I don’t think the study should be restrained to nurses who are professional caregivers. There are probably plenty of confounding variables associating caregiving at home and caregiving in their work and the effect on CHD risk has the potential of being heavily biases. Second, a 2 year resolution over a 4 year period (1992-1996) will only give 2 data point per individuals and the survival analysis might not be powered enough. Last, as detailed in the paper, missing data on caregiving and loss to follow-up seem to have suffered from selection bias.

Re: Week 3 discussion

by Maria Glymour -

Francois

One of the extremely frustrating aspects of the ARIC cohort was that although there were 4 sites, there were only 2 with non-trivial numbers of black residents, and the vast majority of the black participants in ARIC come from Jackson MS, where they sampled no whites (from the paper: "For Jackson, unlike the other three communities, only blacks are included in the cohort, although community surveillance covers all races in all communities").  Therefore any evaluation of race in ARIC intrinsically conflates race and geographic location.   It has actually been a major source of information on racial health disparities but in my view it is not possible to tell what results are about aspects of being black versus aspects of living in Jackson MS.   Although ARIC was originally funded for just one follow-up, follow-up continues to now, with varying degrees of frequency and changing measures as the cohort became older.  The 5 year funding cycle means that the original study designs hardly ever anticipate 20 year follow-ups.  

Note the difference between "cluster randomized" (ie, we randomly assigned treatment to people, and everyone in the same cluster would receive the same treatment) and "cluster sampled" (ie we divided the population into clusters and randomly selected some clusters from which to enroll our study). 

Re NHS: wouldn't the potential confounding of professional caregiving and CHD be a reason to match or restrict to professional caregivers?  This seems like a strength not a weakness as it could reduce confounding.  Re temporal resolution, in theory they could have defined specific dates of events, but I don't know if they did that (and they don't seem to discuss) however the power critique after the study has already been completed should really be based on the confidence intervals: are they so wide that you cannot learn anything useful from the study?  

Re anorexia and divorce: this is a really interesting problem.  I agree that NHS is a challenging setting to study this, but what other data source might be more appropriate?